Air Pollution Control - Arnulfo de Castro (Pada BAB I Saja) - FAJRI FAIZAL AINUL YAQIN

----------------------- Page 1-----------------------
22• Energy Conversion
Air Pollution Control Abstract | Full Text: PDF (185K)
Cepstral Analysis of Speech Abstract | Full Text: PDF (137K)
Compressed Air Energy Storage Abstract | Full Text: PDF (616K)
DC Machines Abstract | Full Text: PDF (222K)
Diesel-Electric Generators Abstract | Full Text: PDF (226K)
Diesel-Electric Power Stations Abstract | Full Text: PDF (188K)
Electric Machine Analysis and Simulation Abstract | Full Text: PDF (272K)
Electrochemical Electrodes Abstract | Full Text: PDF (169K)
Electrostatic Precipitators Abstract | Full Text: PDF (279K)
Electrothermal Launchers Abstract | Full Text: PDF (155K)
Engineering Workstations Abstract | Full Text: PDF (93K)
Fuel Cell Power Plants Abstract | Full Text: PDF (145K)
Fusion Reactors Abstract | Full Text: PDF (473K)
Gas Turbine Power Stations Abstract | Full Text: PDF (246K)
Geothermal Power Abstract | Full Text: PDF (250K)
Homopolar and Inductor Machines Abstract | Full Text: PDF (426K)
file:///N|/000000/0WILEY%20ENCYCLOPEDIA%20OF%20...RONICS%20ENGINEERING/22.Energy
%20Conversion.htm (1 of 3)17.06.2008 16:33:30
----------------------- Page 2-----------------------
HVDC Power Converters Abstract | Full Text: PDF (366K)
Hydraulic Turbines Abstract | Full Text: PDF (468K)
Hydroelectric Generators Abstract | Full Text: PDF (308K)
Hydrogen Energy Systems Abstract | Full Text: PDF (709K)
Hydrothermal Power Systems Abstract | Full Text: PDF (185K)
Magnetohydrodynamic Power Plants Abstract | Full Text: PDF (336K)

Multirate Filterbanks Abstract | Full Text: PDF (430K)
Ocean Thermal Energy Conversion Abstract | Full Text: PDF (191K)
Optical Waveguide Transitions Abstract | Full Text: PDF (199K)
Power System Control Abstract | Full Text: PDF (123K)
Primary Cells Abstract | Full Text: PDF (345K)
Quantum Storage Devices Abstract | Full Text: PDF (141K)
Short-Term Load Forecasting Abstract | Full Text: PDF (235K)
Solar Energy Conversion Abstract | Full Text: PDF (258K)
Solar Heating Abstract | Full Text: PDF (162K)
Steam Turbines Abstract | Full Text: PDF (173K)
Synchronous Generators Abstract | Full Text: PDF (204K)
Synchronous Motor Drives Abstract | Full Text: PDF (280K)
Thermal Energy Storage Abstract | Full Text: PDF (107K)
----------------------- Page 3-----------------------
Thermionic Conversion Abstract | Full Text: PDF (263K)
Thermoelectric Conversion Abstract | Full Text: PDF (190K)
Thermoelectric Power Abstract | Full Text: PDF (727K)
Turbogenerators Abstract | Full Text: PDF (246K)
Waste-to-Energy Power Plants Abstract | Full Text: PDF (158K)
Wind Power Abstract | Full Text: PDF (160K)
Wind Power Plants Abstract | Full Text: PDF (142K)
Wind Turbines Abstract | Full Text: PDF (157K)
----------------------- Page 4-----------------------
file:///N|/000000/0WILEY%20ENCYCLOPEDIA%20OF%20ELECTRICA...RONICS
%20ENGINEERING/22.%20Energy%20Conversion/W3010.htm
} { { } }
●
HOME
ABOUT US
CONTACT US
HELP
Home / Engineering / Electrical and Electronics Engineering
Wiley Encyclopedia of Electrical and

● Recommend to Browse this title
Your Librarian
Electronics Engineering
● Save title to My ● Search this title
Profile
Air Pollution Control
● Email this page Enter words or

Standard Article
● Print this page phrases

Arnulfo de Castro 1 1 1
, Saifur Rahman , Concha Reid
1Virginia Polytechnic Institute and State University, Blacksburg,
VA
❍ Advanced Product
Copyright © 1999 by John Wiley & Sons, Inc. All rights
Search
reserved.
❍ Search All Content
DOI: 10.1002/047134608X.W3010
❍ Acronym Finder
Article Online Posting Date: December 27, 1999
Abstract | Full Text: HTML PDF (185K)
Abstract
The sections in this article are
Global Electricity Generation and the Environment
Emissions From Fossil Fuels
Effects of Emissions
Pollution Control Technologies
International Agreements and Country Initiatives
Conclusions
About Wiley InterScience | About Wiley | Privacy | Terms & Conditions

Copyright © 1999-2008John Wiley & Sons, Inc. All Rights Reserved.
file:///N|/000000/0WILEY%20ENCYCLOPEDIA%20OF%20ELE...%20ENGINEERING/22.%20Energy
%20Conversion/W3010.htm17.06.2008 16:34:26
----------------------- Page 5-----------------------
AIR POLLUTION CONTROL 359
GLOBAL
ELECTRICITY GENERATION
AND THE
ENVIRONMENT
The Role
of Electricity in Supplying Clean Energy
Electricity has a large share in the energy system. It con-

sumes a
greater amount of primary fuels than any other sin-
gle
industry. Because of this, implementing emission control
measures
in this industry will result in considerable reduc-
tions in
overall emissions. It is also worth noting that al-
though
there are generation losses, electrification, particu-
larly in
industrial processes, usually contributes to overall
efficiency
improvements. Electricity likewise allows the eco-
nomical
use of nonfossil energy sources. Many nonconven-
tional
sources of energy, such as hydro, geothermal, wind
power,
tidal power, wave power, and ocean thermal energy
conversion, are site-specific and need the electrical transmis-

sion
system to make them available to consumers. Even nu-
clear and
solar power, due to safety and space considerations,
AIR POLLUTION CONTROL
respectively, are subject to site constraints.

In
terms of supply, conventional thermal efficiencies have
Electricity generation contributes significantly to the supply stabilized
at around 35% to 40%. However, electricity permits
of energy and therefore plays a major role in the production the use
of larger, more efficient plants due to economies of
and control of emissions. About two-thirds of the world’s elec- scale.
Also, new technologies are coming with higher efficien-
tricity is obtained by fossil fuel combustion, emissions from cies.
Combined-cycle plants generate electricity with more
which can produce urban ozone, bring about acid rain, and than 40%
efficiency. Ceramic-blade gas turbines can deliver
contribute to global climate change. power at
50% efficiency. Molten carbonate and solid oxide fuel
Urban ozone has been linked to respiratory problems and cells
(MCFCs and SOFCs) have close to 60% conversion effi-
eye irritation and is known to be harmful to many types of ciency.
trees and crops. Acid rain raises the acidity of streams and
Electricity storage systems allow the efficient operation of
lakes, causing harm to fish and invertebrates, and has been base load
plants. They permit the operation of thermal plants
blamed for damage to forests and structures. Elevated levels near
rated capacity by acting as a load in the system when
of sulfates and nitrogen oxides in the atmosphere have been the
demand is low and generating during the peak periods.
linked to respiratory ailments and increased mortality rates
Cogeneration systems also allow for higher conversion effi-
in humans. The possibility of global warming from so-called ciencies
by producing both heat and power at the same time.
The
existence of a strong electrical network ensures the opti-
greenhouse gases (GHG) could cause changes in crop produc-
mum use of
this capacity.
tivity, migration of forest ecosystems, rising sea levels, and
the extinction of some endangered species.
Electricity Projections
The continued growth in the demand for electricity raises
some challenges to the electric utility industry to develop At the
beginning of the 1990s, the world’s electricity genera-
technologies for efficient and environmentally benign electric- tion was
a little more than 12 billion megawatt-hours
ity generation and emission reduction. Currently available ca- (MW #h).
During the past decade, the world’s electricity gen-
eration
grew by an average of 3.6% per year. While the
pacity options are renewable resources and demand-side
growth in
electrical demand in the industrialized countries
management. Fuel switching and scrubber technologies are
was in
the neighborhood of 3% per year, the electricity re-
being used to reduce the emission of acid rain precursors.
quirements
of developing countries increased by 7% annually,
More efficient coal-generation technologies are being devel-
on
average. Developing countries will require around
oped and are expecting commercialization in the next two de- 1,300,000
megawatts (MW) of new generating capacity in the
cades. At the system level, operations planning and genera- next 15
years to serve their growing economies and popula-
tion planning activities include emission reduction among tions.
The world currently spends around $100 billion per
their objectives. year on
new generation capacity (1), with the developing
This article covers the role of electricity generation in the countries
accounting for some $50 to $60 billion. To meet the
production and prevention of emissions. It discusses some of growing
demand, developing countries will have to raise this
the mechanisms by which major pollutants from electricity figure to
$125 billion per year. The investments required for
generation are produced and gives a summary of the effects the
developed world are also substantial. If current trends
of these pollutants. The article also gives a description of continue,
the countries of the Organization for Economic Co-
technologies that are available and are being developed for operation
and Development (OECD) will need to add about
more efficient generation and for emission reduction. 400,000
to 500,000 MW of new generating capacity in the
Toward the end, the article also provides an overview of next 15
years.
current international activity addressing environmental is-
sues. It gives a summary of regulatory approaches and a Fuel Use
for Electricity Generation
background on the economic implications of environmental The

estimates in Table 1 show a projected increase in the
damage. global
electricity production from 12393 billion kilowatt-hours
J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering.

Copyright # 1999 John Wiley & Sons, Inc.
----------------------- Page 6-----------------------
360 AIR POLLUTION CONTROL
Table 1. World Electric Energy Projections
1993 1985–1993
1995 2000 2005
Generation Growth
Estimatea Projection Projection
Region/Country TW · h (%)
TW · h TW · h TW · h
OECD countriesb 7,483 3.1

7,967 8,475 9,382
North America 3,920 3.5
4,215 4,278 4,684
Europe 2,470 2.2
2,562 2,915 3,246
Pacificc 1,093 4.3
1,190 1,277 1,438
Non-OECD countries 4,910 3.9

5,847 7,545 9,752
Africa 343 3.8
396 497 624
Latin America 565 5.3
779 1,011 1,312
Asiad 635 9
762 1,116 1,784
China 839 9.3

870 1,200 1,510
India 356 8.7
436 670 1,030
Europe 423 0.4
447 471 496
Former USSR 1,460 0.7
1,849 2,083 2,347
Middle East 288 8
308 447 649
World 12,393 3.5
13,814 16,020 19,134
a Figures for OECD countries are estimates. Figures for non-OECD

countries are projections.
b Hungary and the Czech Republic are not included with OECD
countries.
c Australia, Japan, and New Zealand.
d Asia does not include China and India.
(kW #h) in 1993 to 19,134 billion kW #h in 2005. Table 2 indi-

the case of NO , are also possibly involved in the depletion of
cates that more than 60% of the world’s electricity is gener-

the ozone layer. These will be discussed in more detail later.
ated by burning fossil fuels (see Refs. 2–6). These combustion
The considerable amount of fossil fuel used in electricity gen-
processes produce carbon dioxide (CO2) and are major contrib-
eration, the tremendous potential for efficiency improvement,
utors to global warming. Coal combustion, which accounts for
and the possibility of substitution with environmentally be-
60% of total fossil generation, also emits sulfur dioxide (SO2)
nign technologies makes electricity an important factor in the
and oxides of nitrogen (NOx ), which produce acid rain and, in
supply of clean energy.
Table 2. Percentages of Electricity Generation by Fuel Type for

Selected Countries (1993)
Fraction (%)
Energy
Hydro/
Country/Region (TW · h) Coal
Oil Gas Nuclear Others
World 12,393 38.2

9.8 14.8 17.5 19.6
OECD 7,483 40.1
7.6 11.5 23.8 16.9
Australia 163 78.7
2.2 9.1 0.0 10.2
Canada 528 14.9
2.1 2.9 18.0 62.1
France 468 5.1
1.2 0.8 78.7 14.1
Germany 522 57.5
1.9 6.6 29.4 4.5
Italy 220 9.1
51.8 18.0 0.0 21.0
Japan 897 16.7
23.7 19.5 27.8 12.3
Norway 120 0.2
0.0 0.01 0.0 99.6
Spain 155 40.9
6.1 0.8 36.1 16.2
Sweden 145 2.1
2.1 0.6 42.3 53.0
UK 322 52.2
7.1 11.1 27.8 1.8
US 3,392 53.4
3.7 13.0 19.1 10.8
Non-OECD 4,910 35.3
13.3 20.0 7.7 23.8
Bangladesh 9 0.0
11.8 79.6 0.0 8.6
China 839 72.8
8.7 0.3 0.2 18.1
India 356 70.3
3.5 4.9 1.5 19.8
Indonesia 50 23.4
47.4 13.6 0.0 15.8
Nepal 1 0.0
6.5 0.0 0.0 93.5
Philippines 27 7.8
53.6 0.0 0.0 38.5
Poland 134 94.5
2.7 0.1 0.0 2.7
Saudi Arabia 82 0.0
60.7 39.4 0.0 0.0
Singapore 19 0.0
100.0 0.0 0.0 0.0
South Africa 175 95.0
0.0 0.0 4.2 0.9
Former USSR 1,460 16.6
9.0 43.2 14.1 17.2
----------------------- Page 7-----------------------

EMISSIONS FROM FOSSIL FUELS

emissions can therefore be calculated directly from the fuel’s
carbon and sulfur contents.

Emissions from the Fossil Generation of Electricity
The following example illustrates the calculation for car-
bon
dioxide emissions:
The exhaust stream consists mainly of gases derived from the
major elements of the fossil fuel. These include water vapor
atomic weight of carbon = 12
(H O), carbon dioxide (CO ), sulfur dioxide (SO ), and some of
2 2 2
the nitric oxide (NO). The remaining nitric oxide is produced
atomic weight of oxygen= 16
by oxidation of nitrogen molecules in the air used in the burn-

C + O2 → CO2
ers and not from nitrogen in the fuel. The other exhaust gases
include small quantities of hydrogen chloride (HCl), nitrogen The
molecular weight of CO2 is 12 # 2 # 16 # 44. Thus,
dioxide (NO ), nitrous oxide (N O), carbon monoxide (CO), and
2 2
complete combustion of one ton of carbon produces 44/12 #
sulfur trioxide (SO3). Carbon dioxide constitutes 12% of the 3##
tons of CO2 . For one ton of coal having 70% carbon content,
flue gas. Sulfur dioxide concentrations are estimated at 1000 the
amount of CO2 emissions would be 0.7 # 3.67 # 2.57 tons.
#L/L to 1700 #L/L and the nitrogen oxides at 400 #L/L to A
typical 400 MW coal plant uses more than 800,000 tons of
600 #L/L (7). These three constitute the bulk of emissions coal
per year. Applying this to Eq. (3) yields
from thermal generation, and we shall focus on these emis-
sions and their impacts on the environment from here on.
tons CO2 emissions = 800, 000 × 2.57 = 2,056,000
Calculation of Emissions The

calculation for annual SO2 emissions would be as follows:
Fuel Usage Calculations

atomic weight of sulfur = 32
Fuel Usage Calculation Based on Average Heat Rate. Calcula-
S + O → SO
tion of emissions begins with the calculation of the amount of
2 2
fuel used based on the amount of electricity generated. One
method of calculation uses the average heat rate of the plant, The
molecular weight of SO2 is 32 # 2 # 16 # 64. Thus, com-
electricity generation in gigawatt-hours (GW #h), and the heat plete
combustion of one ton of sulfur produces 64/32 # 2 tons
content of the fuel; thus, of
SO2 . For one ton of coal having 3% sulfur content, the
amount of SO2 emissions would be 0.03 # 2 # 0.06 tons.

fuel usage (tons) If
the same plant as above has a 90%-efficient sulfur
generation (GWh) × heat rate (MBtu/GW·h)
scrubber, the fraction of SO2 released would be 0.10. Using
= (1) Eq.
(3),
heat content of fuel (MBtu/ton)
tons SO2 emissions = 800,000 × 0.06 × 0.1 = 4800

Fuel Usage Calculation Based on Hourly Fuel Usage. Average
heat rate calculations are used for estimating longer-term
Nitrogen Oxide Emissions. Nitrogen oxide emissions come

(say annual) fuel consumption. Hourly operations use input–
from
two sources: (1) nitrogen found in the coal molecule
output and incremental heat rate representations for more
(fuel-bound NO ) and (2) nitrogen in the air (N ) oxidized dur-

accurate calculations of fuel consumption. The input–output
x 2
ing
the combustion process (thermal NO ). NO is a collective
curve is usually a quadratic function, relating heat input to
x x
term
for nitric oxide (NO) and nitrogen dioxide (NO2). Al-
power output. Hourly fuel usage can then be calculated from
though NO emissions are usually one order of magnitude

the power output P , where P is in terms of MW:
more
than those of NO2, these are usually oxidized also to
hourly fuel usage (tons) NO2 .
2
The amount of fuel-bound N2 converted to NOx during com-
= AP + BP + C (MBtu) (2)
bustion varies between 15% and 20% of the nitrogen in the
heat content of fuel (MBtu/ton) coal.
It depends on the characteristics of the coal, the firing
systems, furnace conditions, flame patterns and tempera-

Calculation of Emissions. The amount of emissions can be
tures, burning time, and furnace reducing or oxidizing condi-
calculated as the product of the fuel usage, the emissions con-
tions. The fuel-bound component of NOx emissions can be cal-
tent of the fuel and the emissions production rate:
culated using a similar method to that for SO2, this time
using
the atomic weight of 14 for nitrogen:
emissions (tons) = FU × EC × PR (3)
N2 + 2O2 → 2NO2
where FU # fuel usage (tons), EC # emissions content of fuel

(tons emissions/tons fuel), and PR # emissions production The
molecular weight of NO2 is 14 # 2 # 16 # 46. Thus,
rate (fraction released)
complete combustion of one ton of nitrogen produces 46/14 #
CO2 and SO2 Emissions. Practically all of the carbon and 3.29
tons of NO2 . For one ton of coal having 1% nitrogen
sulfur emissions are in the form of CO and SO . Carbon mon-
content, the amount of fuel-bound NO2 emissions would be
2 2
oxide emissions are very small and come only as a result of 0.01
# 3.29 # 0.0329 tons.
slight inefficiencies in the combustion process. Sulfur trioxide
For a plant with no nitrogen removal, the amount of fuel-
(SO3) emissions are much less than sulfur dioxide emissions. bound
NO2 emissions can be calculated using Eq. (3):
Carbon dioxide and sulfur dioxide are derived from the ele-
ments present in the fossil fuel. The amount of CO2 and SO2
tons NO2 emissions = 800,000 × 0.0329 = 26,320
----------------------- Page 8-----------------------
1200
atmospheric phenomenon as well. The terrestrial biosphere
1088
and
the oceans serve as sinks and sources for carbon dioxide,
800 907 each
exchanging with the atmosphere some 100 gigatons of
) carbon
(GtC) per year.
h 600
W The
OECD estimated anthropogenic CO2 emissions from
k
/ 590 energy
use in 1988 at 6.3 GtC per year, growing at the rate
g 400
(
2 of
2.5% per year (6). About 1.7 GtC was contributed by power
O 1200
generation (10). Half of the CO2 emitted appears to remain in
C
the
atmosphere and causes the rising trend in air concentra-
200 tions
of the gas.
Carbon dioxide is only one of several greenhouse gases pro-

0
Coal Oil Gas duced
by human activity (anthropogenic emissions). Also in-
cluded
in this category are methane (CH ), nitrous oxide
4
Figure 1. Average CO2 emissions in the US in 1990 from fossil-fired (N O),
ozone (O ), and the chlorofluorocarbons (CFCs). All of
2
3
power plants (8). these
have a greater warming effect than CO2 on a molecule-
for-
molecule basis. Calculations based on concentration, radi-
The amount of thermal NOx is dependent on combustion- ative
effectiveness, and lifetime, however, show that CO2 ac-
time–temperature factors, furnace combustion conditions (re-
counted for 66% of greenhouse gas contributions to global
ducing or oxidizing conditions), and the type of burner and
warming between 1880 and 1980 (11), and 55% between 1980
combustion air distribution. and
1990 (9).
Figures 1 and 2 show average CO , SO , and NO emissions
While models predict that the increase in greenhouse gas
2 2 x
from fossil-fired power plants in the US for each kWh of elec-

concentrations should already have caused global warming,
tricity output (8). there
is some doubt as to whether the experienced trend of
slightly less than half a degree Celsius (#C) [one degree Fahr-
enheit
(#F)] can be attributed to them. The models differ in
EFFECTS OF EMISSIONS
representation, particularly in the effects on cloud cover and

the
ocean surface, as well as in the assumptions and measure-
Table 3 gives estimated emissions for 1995 due to electricity
ment
methods used. As a result, there is much variation in
generation in selected countries, as well as those projected for
the
estimates of predicted temperature changes resulting
2000. The following sections show why there is much concern
from
the expected doubling of CO2 concentrations in the next
about the amount of these pollutants that are produced from
electricity generation.
century from as low as 0.2#C (0.4#F) to as high as 9.4#C
(16.9#F). A commonly accepted figure is 1.5 to 4.5#C (12).

Carbon Dioxide
Assuming that the expected increase in CO2 concentrations
will
result in global warming, the next question is what ef-
It is estimated that the earth would have a temperature of fects
global warming will have. Among the possibilities are
#18#C were it not for what are known as the greenhouse
worldwide changes in crop productivity, forest ecosystem mi-
gases (9). Carbon dioxide, water vapor, and ozone allow
gration, a rise in the sea level, and the extinction of some
higher-frequency solar radiation to pass to the earth but ab-
endangered species.
sorb the terrestrial infrared radiation. The mechanism is dif- A
direct effect of global warming on the utility industry is
ferent in a greenhouse, where heat is retained by preventing the
change in the demand for electricity. Once the uncertain-
convective flows out of the structure; but the similarity of the ties
of temperature changes are resolved, however, these
effect has resulted in the use of the term ‘‘greenhouse’’ for the would
be much easier to quantify. A more difficult problem is
the
assessment of the effects of emission control laws on the
cost
of operating electric utilities.
Sulfur
Dioxide
8 7.7
SO2
) 7 NOx Sulfur
dioxide is a colorless gas produced mainly from the
h 6
combustion of fossil fuels. It can react catalytically with such
W 5.4
k
oxidants as ozone, hydrogen peroxide, and organic free radi-
/ 5
g
(
s 4 4 cals
to produce sulfur trioxide (which hydrates quickly to sul-
n furic
acid), sulfuric acid, and sulfates:
o
i 3
s
s 2.1
i 1.9
m 2
OH + SO → HSO
E
2 3
1
0
HSO3 + O2 → HO2 + SO3
0
Coal Oil Gas
SO + H O → H SO
3 2 2 4
Figure 2. Average SO2 and NOx emissions in the US in 1990 from

fossil-fired power plants. Note: SO2 emissions from gas-fired plants Fog,
suspended particulate matter, and sulfur dioxide form
are negligible (8). smog,
which is known to have health effects particularly on
----------------------- Page 9-----------------------

Table 3. Carbon Dioxide, Sulfur Dioxide, and Oxides of Nitrogen
Emissions from Fossil Fuel
Generation of Electricity (Estimates for 1995 and Projections for
2000)
Emissions
1995
2000
Carbon SO2 NOx

Carbon SO2 NOx
(106 metric (103 metric (103 metric
(106 metric (103 metric (103 metric
Country/Region tons) tons) tons)

tons) tons) tons)
World 2,041 48,742 24,368

2,420 52,804 27,472
OECD 1,084 24,357 12,724
1,200 21,248 12,696
Australia 47 1,378 564
52 1,542 632
Canada 32 902 391
37 1,067 454
France 13 388 156
20 528 242
Germany 97 2,749 1,198
104 2,886 1,292
Italy 48 1,088 571
57 1,294 695
Japan 117 226 271
124 235 282
Norway 0 2 1
0 2 1
Spain 25 724 301
30 769 376
Sweden 2 36 19
3 52 30
UK 62 1,687 740
63 1,597 782
US 566 13,250 7,580
620 8,950 6,750
Non-OECD 957 24,385 11,643
1,220 31,555 14,776
Africa 85 2,474 1,018
107 3,106 1,278
Latin America 46 900 537
60 1,169 698
Asia 111 2,775 1,278
170 4,247 1,956
China 205 6,269 2,410
283 8,646 3,324
India 97 2,965 1,148
149 4,555 1,763
Europe 89 2,593 1,083
94 2,733 1,142
Former USSR 271 5,463 3,540
305 6,153 3,987
the elderly, the young, and those with respiratory ailments. cially
such textiles as nylon), leather, paper, electrical con-
There are also effects on the vegetation, on structural materi- tacts,
paints, and medieval stained glass. It accelerates the
als, and on the atmosphere. These effects are described in the corrosion
rates of metals such as iron, steel, zinc, copper,
following paragraphs. The reader can refer to Elsom (13) for and
nickel.
a more thorough discussion of the effects of SO2 and NOx Another
effect of SO2 emissions is the reduction in visibil-
emissions. ity due
to light absorption and scattering by sulfates. This
The best known example of the health effects of SO2 hap- suspended
particulate matter also enhances condensation and
pened in London in 1952, when 4700 deaths occurred above freezing,
leading to the formation of cloud and fog, increased
the expected number on account of respiratory failure. The
precipitation, and reduced sunshine levels.
largest single contributor was bronchitis, although death from
diseases involving the impairment of respiratory functions
Nitrogen
Oxides
also rose. The number of deaths due to heart disease in-
creased, possibly due to the strain on the heart caused by Oxides of
nitrogen account for 30% of acid rain precipitation
those respiratory problems. Sulfates suspended in the emis- in the US
(next to sulfur compounds, which account for 65%).
sions are suspect for increased asthma attacks, aggravation Acid rain
(precipitation having a pH lower than 5.6) is known
of heart and lung disease, and a lowering of resistance to re- to result
in the depletion of fish stocks, a reduction in forest
spiratory diseases.
productivity, human health problems, increased material cor-
Low added levels of SO2, have been seen to enhance plant rosion and
erosion, and reduced visibility.
growth by the addition of soil nutrients. At higher levels, how- The
most important oxides of nitrogen are nitric oxide and
ever, SO2 causes bleaching of plant chlorophyll and lowers soil nitrogen
dioxide; the other oxides are not known to be biologi-
pH values, resulting in reduced growth and yield. Effects are cally
significant. Anthropogenic nitrogen oxides are produced
more pronounced on plants such as lichen that contain rela- during
combustion when the temperature is higher than
tively little chlorophyll. Such information has been used to about
1000#C. Its principal sources are the combustion of fos-
map sulfur dioxide levels by experimental transplantation of sil fuels
in stationary sources and in motor vehicles. Like sul-
lichens and observing their fate. fur
dioxide, nitrogen oxide emissions have effects on health,
Sulfur dioxide also leads to corrosion of building stone. It on
vegetation, and on the atmosphere. In addition, they are
converts the calcium carbonate in limestone, sandstone, believed
to act as catalysts in the depletion of the ozone layer.
roofing slate, and mortar into soluble calcium sulfate (gyp- Aside
from promoting photochemical pollution, nitrogen
sum). The material increases in volume, resulting in scaling, oxides
have health effects of their own. High concentrations
of NO
3 3
blistering, and disintegration. The loose material is then x
(600 #g/m ) to 900 #g/m ) have been found to result in
washed away by rain. Sulfur dioxide also affects fabrics (espe- increased
susceptibility to respiratory infections, increased
----------------------- Page 10-----------------------
airway resistance, and decreased sensitivity to bronchocon-

pression, and lower blade temperatures enable the blades to
strictors.
withstand higher gas temperatures. In addition, precombus-
tion
catalytic reaction of natural gas with steam produces H2
and
CO, and results in chemical energy greater than that of
POLLUTION CONTROL TECHNOLOGIES
the
natural gas itself. This can result in efficiencies greater
than
52%.
Technologies for pollution control in the utility industry are
Integrated gasification combined cycle turbines, pressur-

available at different stages of electricity production and con-
ized
fluidized bed combustion turbines, and steam-injected
sumption. On the supply side, technologies for efficient com-
gas
turbines are expected to play major roles in future power
bustion are being developed, and techniques are available for
systems in view of their greatly reduced SO2 emissions and

capturing pollutants before, during, and after combustion. In-
their higher efficiencies, which essentially reduce CO2 emis-

directly, demand-side processes reduce emissions by lowering
sions per kilowatt-hour of electricity generated. However,

generation requirements. System level planning and schedul-
these technologies still produce a considerable amount of car-

ing activities determine the most effective use of these tech-
bon
dioxide emission because of the very nature of fossil fuel
nologies for overall economic operations in the context of reg-
combustion.
ulation and policy signals from the government.
The other available supply-side options also encounter ob-
stacles. The fusion reactor is still decades away from commer-

Supply-Side Technologies
cialization, and even the supposedly benign hydro is meeting
Efficient Coal Combustion Technologies. The efficiencies of
considerable opposition from environmentalists. Solar and
traditional coal combustion technologies average around 33%.
other renewable technologies hold considerable promise but
With advanced coal-fired power systems, average efficiencies are
yet to take off with large-scale applications, primarily due
are expected to increase to 50% and in specific technologies to
cost constraints.
can approach 85%. These technologies have the potential to
reduce carbon dioxide emissions by more than 40% and sulfur
Pre- and Postcombustion Control of Sulfur Dioxide Emis-
dioxide emissions to a fraction of the levels allowed by federal
sions. Compared with the above technologies for removing

air standards. Among these are integrated (coal) gasification
sulfur during combustion, those for sulfur removal before and

combined cycle (IGCC) technology, pressurized fluidized bed
after combustion are more mature. Some of these are de-
combustion (PFBC), and the steam-injected gas turbine
scribed below. The interested reader may refer to Refs. 16–17

(14,15). for
more information.
Integrated Gasification Combined Cycle. In the IGCC, coal is
Fuel Switching. One way of removing sulfur before combus-
transformed into synthesis gas composed mostly of hydrogen
tion
is by switching to low-sulfur fuels such as gas and low-
(H ) and CO, as well as methane, CO , and H S. The coal gas-
2 2 2
sulfur coal. The sulfur content of coal can vary between 0.5%
ifier partially combusts coal and limestone in an oxygen-rich
and
10% by weight, allowing the flexibility for immediate low-
environment to produce a coal-gas fuel of low or medium Btu
cost
implementation without producing extra waste.
content. Sulfur and particulates are removed from the gas be-
Coal Washing. This process involves the cleaning off of in-

fore its complete combustion in a gas turbine. A coal gasifier
organic sulfur, resulting in a reduction of around 30% of the

retrofit to a combined-cycle unit typically reduces the mega-
total sulfur content.
watt output by 5% and increases the heat rate by 1500 Btu/
Wet Limestone System Process. In the wet limestone system,

kW #h.
the
limestone received goes through a pulverizing mill to get
Pressurized Fluidized Bed Combustion. Fluidized bed com-
the
required size consistency. The ground limestone is then
bustors burn coal with limestone or dolomite while the mix-
combined with water. The resulting slurry is fed to the SO2

ture is suspended in jets of air. This allows the limestone to
scrubber–absorber vessel, where it is sprayed on the flue gas

take up about 90% of the sulfur that would otherwise be emit-
stream. This direct contact between the slurry and the SO2-
ted as SO2 . Since the constant churning facilitates the trans-
laden gas removes the SO2 from the flue gas through the fol-
fer of heat to the tubes, the temperature of combustion can be
lowing chemical reaction:

lower than in a conventional boiler. This reduces the forma-
tion of NO .
x
CaCO · 2H O + SO → CaSO · 2H O + CO
In the PFBC, the pressure is 6 to 16 times higher than in
3 2 2 3 2 2
the atmospheric fluidized bed combustor (AFBC). The pres-
surized hot gases in the combustion chamber of the steam A
demister section in the scrubber vessel removes water en-
boiler can then be used in a combined cycle to run a gas
trained in the flue gas. A flue gas reheater then evaporates
turbine. any
remaining water and superheats the flue gas prior to
Steam-Injected Gas Turbine. In the steam-injected gas tur-
stack inlet. This step minimizes corrosion of equipment and
bine, natural or synthesis (coal) gas is used to produce elec-
material downstream of the scrubber and provides plume
tricity. The exhaust gases from the turbine are then used to
buoyancy at the stack exit to ensure dispersion and mixing in
produce steam. Injection of this steam to the combustion the
atmosphere.
chamber increases the energy output and efficiency and re-
The scrubbing liquid stream (limestone slurry) containing
duces NOx emissions.
absorbed SO2 goes to a reaction tank beneath the scrubber
A variation called an intercooled steam-injected gas tur-
vessels. The reaction produces calcium salt (CaSO –CaSO )
3 4
bine system attains higher efficiencies by cooling the combus-

precipitates and sulfurous acid (H SO ). The limestone slurry
2 3
tion air between compression stages and diverting some of it

recycles, returning to the SO2 scrubber vessel, while makeup
to the turbine blades. Cool air requires less power for com-
slurry is fed to the reaction tank. The flue gas desulfurizer
----------------------- Page 11-----------------------
(FGD) sludge goes to waste storage tanks for further pro-

Absorption:
cessing or for subsequent disposal.
MgO + SO2 → MgSO3

Wet Lime System Process. The wet lime system process uses
MgSO + H O + SO → Mg(HSO )
lime (CaO) as reagent instead of limestone. Mixed with water,
2 2 2 3 2
this produces calcium hydroxide [Ca(OH)2] or slaked lime.
(primary absorption reaction)
The calcium hydroxide reacts with SO2 in the scrubber, ab-
MgSO3 + 1 O2 → MgSO4
sorbing it from the flue gas with the following chemical reac-
2
MgSO + 6H O → MgSO · 6H O (s)

tion:
3 2 3 2
MgSO + 7H O → MgSO · 7H O (s)
4 2 4 2
Ca(OH) + SO → CaSO + H O
Mg(HSO ) + M(OH) + 4H O → 2 (MgSO · 3H O) (s)
2 2 3 2
3 2 2 2 3 2
The rest of the process is similar to that of the limestone MgO

Regeneration:
system.
MgSO3 → MgO + SO2

Dual Alkali FGD System. This system uses sodium hydrox-
ide (NaOH) as the scrubber reagent. To provide this reagent,
MgSO4 + 1 C → MgO + SO2 + 1 CO2
2 2
soda ash (Na CO ) and lime (CaO) are fed to the system, pro-
2 3
viding the NaOH and calcium hydroxide [Ca(OH)2] for absor-

Dry SO2 Scrubber System. The dry SO2 scrubber system uses
bent regeneration, respectively. The chemical reaction in the an
alkali reagent such as soda ash (Na CO ) mixed with wa-
2 3
scrubber follows: ter.

The alkali solution is pumped to the SO2 absorber vessel,
where
spray atomizers mix fine droplets of the soda ash with
the
SO , forming sulfites and sulfates (Na SO and Na SO ).
2NaOH + SO → Na SO + H O
2 2 3 2 4
2 2 3 2
The
thermal energy in the flue gas evaporates the water. This
results in a dry powder mixture of sulfites and sulfates simi-

The absorbent is regenerated with the following basic reac-
lar
in size and characteristics to the fly ash in the flue gas.
tion:
The
fly ash, the dry powder mixture, and any unreacted soda
ash
are removed by the baghouse filter facility from the dry
Na SO + Ca(OH) → 2NaOH + CaSO SO
scrubber system.
2 3 2 3 2
The reaction between the soda ash and the SO2 is given
As in the wet limestone FGD process and for similar reasons, by
the following chemical balance equation:
flue gas leaving the scrubber is reheated.
Na CO + 2H O + SO → Na SO + CO + 2H O
Wellman–Lord Process. This consists of two stages. The
2 3 2 2 2 3 2 2
first stage consists of a wet absorption process wherein SO2

from the flue gas reacts with sodium sulfite (NaSO3) to form
Although the technology is already available for reducing
sodium bisulfite (NaHSO3) and some sodium sulfate SO2
emissions, it puts an additional financial burden on the
utility and, ultimately, the electricity consumer. Scrubbers

(Na SO ). The primary absorber reactions are
2 4 cost
around $175 to $200 per kilowatt to install, and they
penalize the plant efficiency. Lower-sulfur coals (those with

Na SO + H O + SO → 2NaHSO
2 3 2 2 3 less
than 1% sulfur content) can be bought only at a premium.
Na SO + 1 O → Na SO One
electric utility estimated that its electricity prices would
2 3 2 2 2 4
increase by about 5% on the average (and up to 20% for some
customers) due to acid rain compliance. Low-NOx burners

The absorber effluent liquor then goes through a filter to re-
likewise entail an additional cost.
move solids and divides into two streams, one for regenera-
tion and the other for purge treatment to reject the unreactive
Control of Nitrogen Oxide Emissions (17–19). Thermal NOx

sodium sulfate. Double-effect evaporator–crystallizers con-
is a
major source of the total NOx produced when low-NOx
vert the dissolved NaHSO to crystalline Na SO and liber-
3 2 3 fuels
are burned. However, it forms only a small fraction
ated SO . The regenerated Na SO crystals return to the
2 2 3 when
high-nitrogen fuels are used. Thermal NOx is produced
absorbers after dissolution, while the regenerated SO2 under
conditions of high temperatures, low heat extraction
stream is converted to elemental sulfur by reduction with
rates, high excess air levels, and high volumetric heat re-
natural gas. lease
rates.
Magnesia Slurry Process. This wet scrubber process uses a
Fuel-bound NOx is less dependent on temperature and is
magnesium sulfite slurry to absorb SO2 . This produces hy-
produced with fuel-rich mixtures in pyrolysis, a large evolu-
drate crystals of magnesium sulfite and sulfate precipitates, tion
of volatile nitrogen species, and low heat extraction rates.
which are then centrifuged and dried. Calcination of these
Control of the Combustion Process. Control of NO emissions
x
crystals with a small amount of coke under controlled condi- takes
advantage of the above characteristics of NOx produc-
tions produce solid MgO and an offgas of 7% to 9% SO2 . The tion.
The popular technique is to design combustion chambers
regenerated MgO goes back to the absorbent slurry prepara- and
burners for staged combustion—one under a fuel-rich
tion area, while the SO2 is converted to sulfuric acid in a con-
mixture and the other under a fuel-lean mixture. The initial
ventional contact acid plant. stage
involves partially burning coal in reduced atmosphere,
The following chemical reactions occur in the magnesia
allowing it to retain part of the fuel-bound nitrogen in the
slurry process:
residual unburned fuel. The burned fuel from the initial burn-
----------------------- Page 12-----------------------
ing stage then undergoes combustion in the fuel-lean mixture

There are also several ways of indirectly reducing CO2
using furnace secondary air supply. Of the 20% to 35% of fuel
emissions from electricity generation. These include the re-
nitrogen retained in the char, only 20% is converted to NOx ,
duction of demand for electricity, use of more nonfossil energy
regardless of stoichiometry. This allows good char burnout
sources, switching to fossil fuels with a lower carbon content
under high-excess-air conditions without a significant in- per
unit of energy, and lowering the rate of emissions from
crease in NOx levels.
fossil fuels through improvements in combustion efficiency
Two-stage combustion is achieved with tangentially fired and
electricity transmission and distribution (20). The next
furnaces, using corner firing and tilting burners with overfire
section describes demand-side technologies that can help re-
air nozzles. This makes it possible to fire the initial fuel-rich
duce coal burn for the same amount of energy service de-
mixture with reduced air flow. Reduced flame temperature
livered.
and extended burning time control NOx formation in the fur-
nace. A two-stage burner proportions secondary combustion
Demand-Side Technologies
air between the fuel-rich mixture in the initial zone and the
secondary combustion zone. This results in a reduction in the
Improved generation efficiencies reduce emissions by reduc-
range of 40% in conversion of fuel-bound nitrogen to NOx . ing
coal burn per unit of energy produced. In addition, the
Fuel Reburning. Another method being studied uses fuel re-
utility can avail itself of demand-side alternatives that delay
burning for chemically reducing the NO formed in the com- the
requirement for additional investments.
bustion stage. It involves burning the fuel in the lower fur-
Three demand-side management (DSM) load profile objec-
nace under slightly air-rich conditions. Additional fuel is then
tives utilities may consider are peak clipping, valley filling,
added downstream of the first zone to reduce most of the NO and
load shifting (21). Peak clipping reduces peak demand
generated in the first stage and convert it to N2 under slightly
without affecting off-peak demand. This strategy is effective
substoichiometric conditions. Tertiary air is then added to the in
reducing operating costs and postponing the need for addi-
fuel-rich products to complete the combustion under slightly
tional capacity. Valley filling increases off-peak demand,
air-rich conditions.
which improves the efficiency of utilizing existing generating
Other Options. The following options for limiting NOx
capacity, thereby reducing costs of delivering power. Finally,
emissions are also available: as
the name implies, load shifting moves energy consumption
from on-peak to off-peak periods, resulting in the combined

1. Use additives such as controlled NH injection in the
benefits of peak clipping and valley filling.
3
presence of a catalyst in the flue gas. This converts

If DSM strategies designed to achieve the three objectives
NOx compounds to elemental nitrogen and water vapor. are
applied successfully, theoretically the result will be a
2. Inject water into the furnace to reduce the flame tem-
lower peak and a smoother load profile (a higher load factor,
perature and increase combustion time. This reduces the
ratio from off-peak to peak load). With a smoother load
NO formation.
profile, utilities can operate their generating units more effi-
x
3. Use additives such as copper oxide sorbent and cata-
ciently. Combined with peak reduction, this can allow power
companies to postpone capacity additions, resulting in cost

lysts for scrubbing the flue gas to remove NOx .
savings and environmental benefits. Postponed capacity addi-

4. Use a catalyst in a fuel denitrogenation process.
tions not only delay the emissions of pollutants from power
plants, but also can potentially avoid them. A plant addition

Control of Carbon Dioxide Emissions. Carbon dioxide comes
postponed for a few years can be fitted with the latest technol-
from the oxidation of carbon in the fuel. Coal has the highest
ogies for emissions control, which will pollute less than the
carbon content among the fuels and produces about 32%
previous generation of power plants.

more CO2 than oil and more than two times as much as gas
Electric Power Research Institute (EPRI) studies show the

per unit energy. As yet, there are no known methods of reduc-
benefits of load shifting, peak shaving, and valley filling not

ing carbon dioxide emissions that are economically feasible.
only in the reduction of the load and in improving operating

Among the schemes for CO2 control are afforestation, which
efficiencies but also in emission reduction (22). Studies show

is the planting of forests to absorb CO2 from the atmosphere,
that considerable improvements can also be achieved in the

and sequestering of the gas before it exits to the air and then
efficiencies of residential appliances such as lighting, refriger-

piping it into the ocean or to such natural geologic formations
ators, air conditioners, and heaters (use of heat pump). Better

as natural gas reservoirs.
energy management control systems and variable-speed mo-

The proposed methods for CO2 gas recovery include the fol-
tors can also be used to reduce the total electric energy con-
lowing:
sumption (23).
1. Use a sorbent called selexol to recover CO from the gas-
Among the appliances that have enjoyed efficiency up-
2
ified coal in IGCC.
grades are refrigerators. The efficiency improvement poten-
tial for refrigerators of 100 to 300 L (the size normally used

2. Use molecular sieves that are permeable to some gases
in
developing countries) is in the neighborhood of 400 kW #h/
but not to others.
year, representing 30% to 40% of their total energy consump-

3. Use a monoethanolamine [MEA, HO(CH ) NH ]
2 2 2
tion. In extreme cases, this can even go as high as 50% to
scrubber to capture CO2 from the flue gas.
75%. In Japan, for example, the refrigerators are now about
4. Replace air with oxygen in the boiler to ease CO2 recov- 66%
more efficient than in 1973.
ery.
Another possible area for electric energy savings is electric
5. Use molecular sieves that are permeable to some gases
fans. Newer electric fans can save about 10% off the consump-
but not to others.
tion of standard fans.
----------------------- Page 13-----------------------
High-efficiency air conditioners have efficiencies 30% to or

products or process standards through a licensing and
40% higher than those of the average new air conditioners. In
monitoring system. Compliance is made mandatory for pollut-
regions requiring space heating, heat pumps are available ers, and
noncompliance results in sanctions on the polluters.
that use less than 50% of the electricity needed for electric
Uniform Performance Standard. Uniform standards that
resistance-based systems. High-efficiency heat pumps, which have
been used include
are 25% more efficient than standard heat pumps, can also
be used. These have been claimed to be more economical than 1.
Limitation on the maximum rate of discharge from a
gas heaters. With their introduction into the industrial sector,
pollution source
they are becoming more effective in reducing the overall con- 2.
Specification of the degree of pollution control required,
sumption of energy for heating.
such as percentage of removal of particles from the
Improved incandescent lamps save 5% to 20% of the elec-
emission
tricity use of standard incandescent lamps. Compact fluores-
3.
Limitation of the density of pollution discharged or
cent lamps, on the other hand, can save 60% to 75%. Im-
emitted
proved fluorescent lamps are now available with 5% to 15%
higher efficacy, and electronic ballasts can reduce energy con- 4.
Bans on discharges based on pollution concentration or
sumption by 20% to 25% relative to using ordinary ballasts.
damage costs
Combinations of rare earth fluorescent, low-temperature com- 5.
Discharge limits based on the use of specified inputs to
pact fluorescent lamps and high-frequency electronic ballasts
or outputs from the production process
have attained efficiencies that are 50% greater than those of
conventional fluorescent lamps.
Technology-Based Pollution Control Approaches. The 1970
In the US, motors are estimated to account for 57% of total Clean
Air Act was based on achieving air quality pollution
electricity use, 60% of which is in the industrial sector. In
standards through strict formulation of technology standards.
developing countries where the percentage of electricity use National
ambient air quality standards (NAAQSs) were to be
by industry is higher, this figure will also be higher. De-
implemented by states under state implementation plans
pending on motor size, high-efficiency motors can reduce en- (SIPs).
Ambient standards for ozone and other pollutants
ergy consumption by 2% to 15% in the US, at a cost of 10% to were
set together with new source performance standards
30% more than standard-efficiency motors. They normally (NSPSs).
pay for this additional investment within one year. Even

Between 1970 and 1976, SIPs fell behind schedule, and the
higher-efficiency motors are available in Japan. High-effi- EPA
started to formulate its offset policy. This allowed new
ciency motors are available at standard frame sizes, facilitat- and
modified sources in ‘‘nonattainment areas’’ when lowest
ing their use for new plant constructions and when existing
available emission rate technologies (LAERTs) were applied
motors need to be replaced at the end of their useful life. and when
any additional emissions were offset.
Prudent motor sizing and the use of variable-speed drives The
amendments of 1977 allowed the extension of dead-
allow savings of between 15% and 40% in energy use. Fric- lines
for achieving NAAQSs and the formulation of new tech-
tion, windage, and core losses are reduced by avoiding the nology
standards. Existing sources in nonattainment areas
tendency to oversized motors. Variable-speed drives modify were
allowed to use reasonably available control technology
the power going into the motor instead of allowing it to run (RACT)
which considered technological and economic feasi-
at full power and controlling fluid flow by throttling. bility.
There have also been reductions in the specific energy con-
Specifying a uniform performance standard instead of a
sumption of electric arc furnaces. Use of microwave and far-
particular technology allows more flexibility for the firms to
infrared rays is also promising. Household use of the micro- decide
how they will meet the goal, e.g., a limit on the amount
wave oven has reduced electricity consumption for cooking, of
pollutant that can be emitted per product output. Both of
and there is already talk of microwave clothes dryers. these
approaches, however, tend to impose high social costs.
For
example, the cost to control certain pollutants may vary
Environmental Policy Options by a
factor of 100 or more because of the differences in loca-
tion
and technology used for the different plants. If the gov-
The policy instruments available to the government for imple- ernment
desired to allocate the pollution control burden effec-
menting environmental policy can be classified into two major tively,
it would have to require all sources to control at the
groups—command-and-control approaches and economic in- same
marginal cost—something that would require detailed
centive systems (24,25). Earlier legislation leaned toward the
information on the operating cost of each individual source.
former, but more recent ones have taken advantage of the
market orientation of the latter.
Economic Incentive Systems. In order to allocate the control
burden
more effectively, the US government is now taking
Command and Control. A command-and-control approach
advantage of economic-incentive approaches. These policy op-
to pollution abatement involves the setting of standards and tions
use market forces to find the most cost-effective manner
direct regulation of polluters. This conventional regulatory of
pollution control. Economic-incentive approaches can be
mechanism either specifies the technology that must be used grouped
into four major categories: (1) pollution charges, (2)
(technology-based) or sets a cap on the emission rate that all
subsidies, (3) deposit-refund systems, and (4) market cre-
sources must meet (a uniform performance standard). The ation.
command-and-control approach has included the setting of
Pollution Charges. Pollution charges take several forms. Ef-
ambient air and water quality standards, objectives, and tar- fluent
charges can be collected on discharges into the environ-
gets, and the imposition of emission and discharge limits and/ ment
based on the quantity and/or quality of the effluent.
----------------------- Page 14-----------------------
These may (distributive effluent charges) or may not (incen-

emissions further than the 4.5 kg/Gcal rate can then sell or
tive effluent charges) be returned to the polluter in the form apply
their unused allowances to other facilities that cannot
of subsidies for new pollution control equipment. User meet
their limit. Title IV also provides extra allowances for
charges for the cost of collective or public treatment of efflu-
utilities using conservation and renewable energy resources
ents can also be imposed. Product charges can be added to and for
most of the affected sources in the Midwest (Illinois,
the price of products that are polluting or are difficult to dis- Indiana,
and Ohio) for each year from 1995 to 1999. A two-
pose of. Administrative charges in the form of control and au- year
deadline extension is given to plants that commit to in-
thorization fees can also be used. stalling
flue gas desulfurization (FGD) systems capable of
Subsidies. Subsidy instruments include grants, soft loans,
eliminating 90% or more of their SO2 emissions. The EPA will
tax allowances, and the elimination of subsidies, each of allocate
for these units allowances approximately equivalent
which is linked to the adoption of pollution abatement mea- to
their uncontrolled annual emissions during the extension
sures. Grants are forms of financial assistance that need not period.
In addition, qualified units using FGD also received
be repaid. Soft loans carry below-market rates of interest. Tax one
bonus allowance in 1997, 1998, and 1999 for each ton of
allowances include accelerated depreciation, tax exemptions,
reduction below 1.2 lb/MBtu (2.16 kg/Gcal).
or rebates. Subsidies that promote economically inefficient
Phase II, in effect January 1, 2000, sets an emissions limit
and environmentally unsound development can be elimi- for
utilities of 1.2 lb/MBtu. Bonus allowances are to be given
nated. to
states where utilities emit less than 0.8 lb/MBtu. Another
Deposit-Refund Systems. These systems impose surcharges 50,000
allowances will be given in phase II to plants in 10
on the price of the potential to pollute, and give a refund of states
(Illinois, Indiana, Ohio, Georgia, Alabama, Missouri,
the surcharge on the return of the product or its residuals.
Pennsylvania, West Virginia, Kentucky, and Tennessee, with
Market Creation. The government can also create artificial certain
plant exceptions) that meet phase I limits. Plants re-
markets where participants can buy and sell rights for actual ducing
SO2 emissions by 90% will receive allowances on a
or potential pollution. One example of this is emissions trad- two-for-
one basis.
ing (bubbles, offsets, netting, and banking) within a plant,
within a firm, or among different firms [e.g., Clean Air Act
Greenhouse Gases
Amendment (CAAA), 1990]. Another instrument is price in-
tervention to stabilize markets, typically secondary materials The
Framework Convention on Climate Change. The Frame-
(recycled) markets. Polluter liability has also led to develop- work
Convention on Climate Change (27,28) is an interna-
ment of a market for liability insurance. Still another instru- tional
agreement to stabilize levels of greenhouse gas con-
ment is the removal of barriers to allow more competition
centrations in the atmosphere to prevent dangerous
among firms and permit least-cost bidding to promote eco-
anthropogenic interference with the climate system. It directs
nomic efficiency [e.g., Public Utilities Regulatory Policies Act that
‘‘such a level should be achieved within a time frame
(PURPA) and EPA].
sufficient to allow ecosystems to adapt naturally to climate
By imposing absolute standards, command-and-control change,
to ensure that food production is not threatened, and
policy instruments attempt to ensure that certain levels of to
enable economic development to proceed in a sustainable
emissions will be met. The drawback to this type of policy is
manner.’’ It also suggests, as believed by most climatologists,
that unless the policy is strictly enforced, it is difficult or even that
some change is inevitable and calls for adaptive as well
impossible to achieve the desired results. Incentive methods, as
preventive measures.
on the other hand, are easier to enforce, since their flexibility The
parties to the Convention agreed to take climate
makes compliance less formidable. For example, SO2 tradable change
into account in such matters as agriculture, energy,
permits give power plants the opportunity to weigh the cost- natural
resources, and activities involving sea coasts and to
effectiveness of complying with emissions standards or to pur- develop
national programs to slow climate change. The Con-
chase allowances to postpone this cost to a later time. The vention
encourages the countries to share their technologies
limited number of total allowances available attempts to en- and
supports cooperative activities for reducing greenhouse
sure that emissions levels will remain below a particular gas
emissions, especially from energy, transport, industry,
threshold.
agriculture, forestry, and waste management sectors. It pro-
motes
scientific research on climate change and creates a
‘‘subsidiary body’’ for ‘‘scientific and technological advice’’ to

INTERNATIONAL AGREEMENTS AND COUNTRY INITIATIVES
assist
governments in decision making. It also calls for a
greenhouse gas inventory listing its national sources and

Acid Rain and the Clean Air Act Amendments of 1990
sinks
to allow the monitoring of changes in emissions and
Among the provisions of the Clean Air Act amendments of
evaluating the effects of measures for emissions control.
1990 (26), Title IV (Acid Deposition) has the greatest effects
Noting that the largest share of historical and current
on the utilities. This provision requires that sulfur dioxide
emissions originates in developed countries, and recognizing
emissions be cut 10 million tons below the 1980 levels to 8.9 the
right of poorer nations to economic development, the Con-
million tons. This is to be accomplished in two phases. vention
places the major share of the responsibility for bat-
The first phase began January 1, 1995, and affected the tling
climate change on the industrialized countries. Specific
111 dirtiest power plants. At that time, those plants would
commitments relating to financial and technological transfers
have had to reduce their SO2 emissions to 2.5 pounds per mil- apply
only to the 24 developed countries of the Organization
lion British thermal units (lb/MBtu), or 4.5 kilograms per gi- for
Economic Cooperation and Development (OECD). Com-
gacalorie (kg/Gcal). The EPA will issue allowances, each per- mitments
on efforts to limit greenhouse gas emissions and
mitting one ton of SO2 emissions. Facilities that cut their enhance
natural sinks apply to OECD and the so-called econ-
----------------------- Page 15-----------------------
omies in transition, which consist of 12 countries in Central The

program provides public recognition and limited tech-
and Eastern Europe and the former Soviet Union. nical
assistance to participants in approved projects. The
The Framework Convention asked OECD and transition
emissions reduced or sequestered by each project are tracked
countries to seek to return their greenhouse gas emissions to and
recorded. USIJI is administered by an interagency Secre-
1990 levels by the year 2000. With the adoption of the Berlin tariat
chaired by the Environmental Protection Agency and
Mandate in 1995, the policy focus shifted to establishing post- the
Department of Energy with significant participation by
2000 emissions targets. Participants in this first conference of the
Agency for International Development. Other participat-
parties (COP) to the convention acknowledged the inadequacy ing
agencies include the Departments of State, Agriculture,
of the provisions of the present convention in preventing
Commerce, Interior, and Treasury.
global warming. They agreed to start negotiations as soon as The
USIJI has the following objectives:
possible to prepare a protocol and complete a study of ap-
1.
To encourage the development and implementation of
proaches for the years beyond 2000 before the 1997 confer-
cooperative, cost-effective, voluntary projects between
ence of contracting parties.
US and foreign partners, especially projects that pro-
mote technological cooperation and sustainable develop-

Climate Change Programs in the US
ment
The Energy Policy Act of 1992. The Energy Policy Act of 2.

To promote projects to test and evaluate methodologies for
1992 (29) was legislated to promote more competition in the
measuring, tracking, and verifying costs and benefits
electric utility industry, provide tax relief to oil and gas drill- 3.
To contribute to the formulation of international crite-
ers, encourage energy conservation and efficiency, advance re-
ria for JI
newable energy and the use of alternative fuels on cars, facili- 4.
To foster private-sector investment and innovation in
tate the construction of nuclear power plants, and promote
the development and dissemination of technologies for
energy-related research and development through the infu-
reducing or sequestering GHG emissions
sion of funds. 5.
To encourage participating countries to adopt more
While not attacking the issue of greenhouse gases directly,
complete climate action programs, including national
it creates an office on climate protection and requires an ad-
inventories, baselines, policies and measures, and ap-
ministrative study on the methods and costs of curbing green-
propriate specific commitments
house gas emissions. It also requires the Energy Secretary
to develop a least-cost energy strategy that promotes energy Benefits
offered to US participants include the following:
efficiency and seeks to limit the emission of carbon dioxide

1.
Public recognition for their efforts to reduce the threat
and other greenhouse gases.
of climate change and contribute to sustainable develop-

The provisions of the Act that have a greater effect on the
ment
utilities include the creation of a category of exempt whole-
sale power producers. This change allows utilities to operate 2.
Limited technical assistance in the form of workshops,
guidance documents, papers examining specific issues,

independent wholesale plants outside their own service terri-
and hotline assistance

tories and encourages the operation of generating plants by
independent producers. The provision also allows wholesale 3.
Affiliation with a US government program in gaining
electricity generators to request that the Federal Energy Reg-
entry into new markets overseas
ulatory Commission order a utility to transmit their power. 4.
Official tracking and recording of the project’s emission
A range of standards and incentives on energy efficiency,
reductions, while the information gained can contribute
on renewable and alternative energy sources, and on energy
to the development of international criteria for joint im-
and coal research are also expected to have results in the in-
plementation
termediate and the far future.

Benefits
offered to foreign participants include the following:
The US Initiative on Joint Implementation. Joint implementa-
tion (JI) (30) was introduced during the negotiations leading 1.
Technology transfer through US private sector invest-
up to the 1992 Earth Summit in Rio de Janeiro. Since then,
ment in additional resources in the dissemination of in-
the term ‘‘JI’’ has been used for a wide range of possible ar-
novative technologies
rangements between interests in two or more countries for 2.
Reduction of transaction costs to facilitate investments
implementing cooperative development projects to reduce or
in technologies and projects that reduce greenhouse gas
sequester greenhouse gas emissions. Such projects are sig-
emissions and contribute to overall host country devel-
nificant in that while costs of reducing or sequestering green-
opment objectives
house gases vary among countries, all such gases affect the 3.
Generation of other local environmental and human
global climate in the same way, regardless of where they are
health benefits by preventing or reducing air, water, or
emitted. Joint implementation allows the reduction of emis-
soil pollution and/or contributing to more sustainable
sions at lower global cost than would be possible if each coun-
use of natural resources
try acted alone. 4.
Creation of local economic benefits through training,
The US Initiative on Joint Implementation (USIJI) was
construction of new or improved facilities, public partic-
launched in October 1993 as part of President Clinton’s Cli-
ipation in projects, and provision of new energy services
mate Change Action Plan. It is a flexible, nonregulatory pro-
gram that encourages private-sector participants in the US to The
first USIJI solicitation in the Autumn of 1994 gener-
engage in overseas projects that will reduce net GHGs. ated 25
proposals and resulted in seven projects representing
----------------------- Page 16-----------------------
more than $40 million in investments. Encompassing technol- 4.

Reduce greenhouse gas emissions by or to some other
ogies and practices ranging from sustainable forest manage-
specified level
ment to renewable energy technologies; these projects take 5.
Reduce or limit the rate of greenhouse gas emissions to
place in Belize, Costa Rica, the Czech Republic, Honduras, a
particular level
and Russia.
6.
Undertake specific projects or actions, or make specific
The second solicitation closed in July 1995 and resulted in
expenditures on projects or actions, to reduce green-
21 proposals. Countries participating in the second round are
house gas emissions
Costa Rica, Nicaragua, Russia, and Honduras, with projects
involving forestry management, renewable resources, and fu-
Utilities have two principal means of implementing their

gitive methane gas capture.
commitments under the Climate Challenge Program. The

The US Climate Challenge Program. The US Climate Chal- first is
through a number of individual activities that utilities
lenge Program (31) is a voluntary partnership between the
are
engaging in. These activities are catalogued in the Op-
US Department of Energy (DOE) and the electric utility in-
tions
Workbook, which was published by DOE and the electric
dustry. Under this program, nearly 500 electric utilities have
utility
industry in October 1994. This workbook has nearly
signed more than 100 participation accords with DOE to vol-
200
pages of options, ranging from supply-side and demand-
untarily reduce, avoid, or sequester greenhouse gases by the
side
management programs, to public and private partner-
year 2000. It has been one of the most successful programs
ships,
and to cross-sectional activities that involve utility cus-
under the US Climate Change Action Plan instituted by Pres-
tomers,
other sectors of the economy, and international
ident William Clinton in October 1993 in accordance with his
projects.
domestic commitment to return US greenhouse gas emissions
The
second approach is through a set of industry initia-
to 1990 levels by the year 2000. DOE estimates that these
tives
that utilities are engaging in collectively. These initia-
accords will reduce GHG emissions by 47 million metric tons
tives
currently include five Edison Electric Institute (EEI) in-
of carbon equivalent from expected year 2000 levels.
dustry
initiatives and five being conducted by other utility
Underlying the climate challenge program are two key trade
associations and utilities.
framework instruments. The first of these instruments is the
The
following are EEI’s five major initiatives (31):
Climate Challenge Program Memorandum of Understanding
(MOU) which was signed on April 20, 1994. This established
1.
The National Earth Comfort Program, which is a con-
the fact that the parties involved had agreed on a voluntary
sortium of more than 150 electric utilities and allied or-
partnership to reduce greenhouse gas emissions, and set forth
ganizations established to expand the geothermal heat

the guiding principles and key provisions for individual
pump market
DOE–utility agreements to follow. 2.
EnviroTechSM, which is an investment fund created to
The second instrument was the Model Climate Challenge
develop promising electrotechnologies and renewables

Participation Accord. Completed in July 1994, this document
to help reduce greenhouse gas emissions

provided the model from which the numerous individual par-
ticipation accords have subsequently been drafted, negoti- 3.
EV America, a multiphase market demonstration of 29
ated, and concluded. In order to suit those utilities with less
electric utilities established to promote and increase
than 50,000 customers, DOE also provided a letter
widespread use of electric vehicles
agreement approach. 4.
The Utility Forest Carbon Management Program, a
The MOU set forth three core elements of every participa-
group of more than 50 electric utilities which is forming
tion accord:
the Utilitree Carbon Company to develop projects for
forest management and carbon sequestration

1. A specific commitment or commitments to reduce, 5.
The International Utility Efficiency Partnerships
avoid, or sequester greenhouse gases
(IUEP) program, a group of 10 electric utilities estab-
lished to identify international energy project develop-

2. A commitment to report annually on activities and
ment opportunities and to sponsor workshops with host

achievements under the Climate Challenge Program;
and
country government officials to facilitate project invest-
ment and development

3. A commitment to confer at reasonable intervals to eval-
uate jointly the progress of the utility participant in The five
initiatives that are being conducted by other utility
achieving its Climate Challenge goals and to discuss trade
associations and utilities are as follows (31):
possible adjustments to its voluntary commitments
1.
Electric End-Use Efficiency Technology Initiative
One or more of the following specific commitments are ex- 2.
Tree Power
pected of utility participants:
3.
Partnerships in Joint Implementation Projects
4.
International Donated Equipment Initiative
1. Make a specified contribution to particular industry ini-
5.
Combined Purchasing Initiative
tiatives
2. Reduce greenhouse gas emissions by a specified amount The

International Utility Efficiency Partnerships. Among the
below the utility’s 1990 baseline level by the year 2000 more
successful activities in the Climate Challenge is the
3. Reduce greenhouse gas emissions to the utility’s 1990 IUEP

program (32). It is a separately funded activity within
baseline level by the year 2000 the EEI.
Formed in February 1995, it is designed to identify
----------------------- Page 17-----------------------
international energy project development opportunities to

Denmark. Denmark (34) was a major oil importer before
work with host country government personnel to facilitate the 1973
oil crisis. Having been greatly affected by the crisis,
project investment and to demonstrate US utility commit- the
country aggressively promoted other power generation op-
ment to voluntary approaches to global climate issues. tions.
It reduced dependence upon imported oil through en-
The IUEP identifies and supports international activities ergy
conservation programs, through development of indige-
to reduce, limit, or avoid emissions of greenhouse gases (for nous
oil, natural gas, and renewables, and through switching
example, energy efficiency, fuel switching, and renewable en- from
oil to coal as the primary fuel for electric power gener-
ergy projects). In addition, it monitors developments in the ation.
world’s fastest-growing emerging markets and provides ac- The
Danish government became dedicated to environmen-
cess to new business opportunities. US electric utilities then tal
reform in the 1990s. The country’s measures for reducing
evaluate and select proposed IUEP projects for financial and
dependence on oil imports resulted in an increase in the pro-
technical assistance. portion
of CO2 produced by the power plants to electricity gen-
Working closely with officials from the USIJI, the IUEP eration
due to the switch from oil to coal as primary fuel for
pursues official recognition for selected IUEP projects as JI.
electricity generation.
It also intends to coordinate the diverse technical and mana-
Denmark’s new energy policy, Energi 2000, was enacted in
gerial skills of its membership in management training pro- 1990.
Energi 2000 includes provisions for further develop-
grams held in the developing world. ment of
wind power and more research into solar and biomass
energies
in an effort to reduce global emissions. In 1991,
about
3% of Denmark’s total electricity generating capacity,
Climate Change Programs in Other Countries
or 273
MW, was produced by windmills. That number is ex-
Japan. The Japanese government instituted the Action
pected
to go up to 10% by the year 2000. The country is even
Program to Arrest Global Warming in 1990 in its effort to
attempting to produce wind power offshore to decrease the

find a systematic and comprehensive solution for global
local
environmental effects of wind generation, which include
warming. It calls for the stabilization of per capita CO2 emis-
visual
effects as well as the effects of audible and electromag-
sions in the year 2000 at 1990 levels. The Japan Action Re-
netic
noise. In the near future, the country expects to add 450
port on Climate Change submitted to the convention secretar-
MW of
biomass-fueled capacity. When combined with replant-
iat in September 1995 predicts per capita CO2 emissions at
ing,
the combustion of biomass has no detrimental effect on
about 2.6 tons for fiscal year 2000, which is almost the same
the
global environment.
as the 2.59 tons calculated in terms of carbon for fiscal year
The
Danish program has had high success with its re-
1990.
newable
energy program due to its implementation of high
However, total CO2 emissions are expected to increase by
energy
taxes on fossil fuels. High taxes on electricity have
10 million tons between 1990 and 2000 to 330 million. In com-
slowed
growth in power demand, further reducing local and
pliance with the Berlin Mandate, the Japanese government is
global
power plant emissions. In addition to the strides
considering new approaches to check the advance of global
Denmark
has already made, when signing the Framework
warming for the years beyond 2000.
Convention on Climate Change (FCCC) agreement in 1992,

One of Japan’s largest electric utilities is Tokyo Electric it
estimated that through the year 2005, it can reduce gross
Power Company (TEPCO), which therefore plays a major role energy
consumption by almost 15%, CO2 emissions by al-
in Japan’s efforts to control CO2 emissions. Among TEPCO’s most
30%, SO2 emissions by about 60%, and NOx emissions
thrusts are the following (33): by 50%.
These percentages are relative to the respective
1988
levels.
Promotion of a Best-Generating-Mix Effort that Centers on
Denmark expects to meet the goals outlined in the FCCC
Nuclear Power. This involves increasing the proportion through
legislation tailored to (1) improve end-use energy ef-
of nuclear power, new energy, and nonfossil energy ficiency
and conservation, (2) enhance efficiency in energy
sources in relation to others. It also includes promoting supply,
(3) increase use of environmentally benign energy
the introduction of liquefied natural gas (LNG) and sources,
and (4) promote research and development. The Dan-
other fossil fuels with a low carbon content. To this end, ish
Parliament also included a CO2 tax in its plan as partial
TEPCO targets an increase of two percentage points
incentive to decrease such emissions.
each in the share of nuclear, hydro, and LNG genera-
McGowan (35) suggests that the barrier of politics is the
tion for the year 2000 over that of the year 1994. most
important one to surmount in the case of renewables.
The
Danish energy policy program is an example of how poli-
Improvement of the Thermal Efficiency of Thermal Power
tics,
legislation in particular, can spur progress toward envi-
Generation. TEPCO successfully improved the thermal
ronmental reform by making provisions for the use of renew-

efficiency of thermal power generation from 38.3% in
ables.
1970 to 39.2% in 1994.
France. Like Denmark, France (34) was able to signifi-

Effective Use of Hydroelectric Power. Between 1979 and cantly
reduce its power plant emissions beginning in the early
1994, TEPCO increased its hydroelectric power capacity 1980s.
France accomplished this by implementing an energy
by nearly 35,000 kW by remodeling or renewing its wa- policy
that heavily promoted nuclear power and energy effi-
ter turbine generators. ciency.
As a result, by 1991, the country’s CO2 emissions due
Promotion of the Effective Use of Resources. This effort in- to

energy transformation (including power plants, heat
cludes development of geothermal power and purchas- plants,
and refineries) were about half of what they were in
ing surplus power from waste-to-energy, photovoltaic, 1980,
despite growth in population and higher electricity gen-
and other new energy-based power generation facilities. eration
every year.
----------------------- Page 18-----------------------
Although France has already significantly reduced its CO2

Commission as the most promising measures available, 17
emissions below 1980 levels, the country has dedicated itself had
taken effect by 1994, which included the following:
to decreasing such emissions to below two tons per capita per
year by the year 2000. (France’s energy-related CO2 emission 1.
A wind energy promotion program
per capita was 6.8 tons in 1991.) The country intends to 2.
A photovoltaic promotion program
achieve these goals by imposing thermal efficiency regula- 3.
Tax benefits for heat–power cogeneration
tions on buildings, promoting wood-burning furnaces for
4.
An energy diagnosis program for buildings
space heating in buildings, expanding its public transporta-
tion, establishing GHG reduction programs for GHG intensive 5.
A credit–loan program for efficient energy use and for
industries, labeling energy consumption of products in new
the use of renewable energy by small and medium-sized
housing, and possibly imposing a tax on the carbon content of
enterprises
fossil fuels. 6.
A thermal insulation ordinance
France’s commitment to the goals outlined above is contin- 7.
An ordinance on heating installations
gent upon the establishment of similar goals in other indus-
trialized countries. Since GHGs affect the global environ- Other
measures include an ordinance on small firing installa-
ment, France would like to ensure that all countries that are tions,
an ordinance on heat use, an energy management act,
economically able to control their global power plant emis- an
emissions-based motor vehicle tax, and a criterion to limit
sions contribute their fair share. The country recognizes that CO2
emissions from motor vehicles.
a global emission control commitment is necessary in order to
achieve significant results.
Expected Results of Climate Change Programs. At the Rio
Germany. Germany (28) attended the 1992 Earth Summit Summit
of 1992, the world’s leaders agreed to draft the
in Rio de Janeiro and signed the Convention. It deposited the
Framework Convention on Climate Change (FCCC) to ad-
instrument of ratification at the United Nations in December dress
the challenge of global warming. Because of the signifi-
1993, and hosted the first conference of parties (COP) in 1995. cant
role played by electric utilities in the energy balance and
German efforts at international cooperation started much the
subsequent production and potential control of green-
ˆ house
gases, the FCCC has possible major repercussions in
earlier, however, with the establishment of the Enquete-Com-
mission in 1987 to recommend ‘‘preventive measures to pro- the
electric utility industry and the whole energy sector in
general. Compliance with the FCCC can take several ap-

tect the earth’s atmosphere.’’ Upon the commission’s recom-
proaches: demand reduction, use of nonfossil resources, fuel

mendation, the federal government decided to work towards
switching, and efficiency improvements. Other techniques are

international agreements for the protection of the climate.
ˆ
afforestation and CO2 sequestration, which is not economi-
Among the measures recommended by the Enquete-Com-
cally
viable at present. The use of these technology options
mission were a total stop to the production and use of CFCs
can
be promoted with economic-incentive or command-and-
and halons, restoration of tropical forests to 1990 levels be-
control policy options. In either case, compliance with FCCC

tween 2010 and 2030, region-specific reductions in per capita
entails high costs. There is also the problem of uncertainties

emissions of CO2 of between 10% and 30% by 2005, and adop-
on
the relationships between human activities and GHG
tion of an overall strategy on the protection of the earth’s at-
emissions, the resulting climate effects, and the effects of cli-

mosphere by the year 2000 (36).
mate
change on environmental and economic systems. To bet-
The official government position supports joint implemen-
ter
inform the process, EPRI has taken an active role in sup-
tation contingent on the stipulation of clearly defined reduc-
porting research on these relationships. It has likewise

tion commitments by the parties to the convention towards
initiated integrated assessment efforts for evaluating global
attainment of the stabilization goal. It provides detailed in-
climate change management proposals.
ventories of the country’s greenhouse gas emissions and re-
Several US government programs also address CO2 reduc-
ports on the state of implementation of the CO2 reduction tion
through its Initiative on Joint Implementation and other
program.
programs such as the Clinton Administration’s Climate
Germany’s own plans, unveiled in August 1993, included Change
Action Plan and the DOE’s Climate Challenge
cuts in carbon dioxide emissions of between 25% and 30% and
Program.
a reduction in all greenhouse gas emissions (excluding CFCs, As
a result of the FCCC, industrialized countries have
which are already banned, and including methane, nitrous
committed to reductions in CO2 emissions mostly to 1990 lev-
oxide, nitrogen oxides, volatile hydrocarbons, and carbon els
by the year 2000. This indicates, however, that CO2 will
monoxide) of 50% from 1987 emission levels by 2005 (36). Its
continue to accumulate in the atmosphere. For industrialized
CO2 reduction program combines regulatory measures, mar-
countries to cut emissions to 20% below 1990 levels by the
ket-based elements, and other supporting instruments such year
2005 as proposed during the first COP would cost about
as consulting, information, and training. It favors the in- $2.7
trillion, according to one estimate. Relaxing short-term
creased use of economic instruments to incorporate environ-
targets but meeting longer-term objectives, however, can
mental costs into energy prices and supports a revenue-neu-
drastically reduce costs to around $500 billion.
tral energy tax that does not distort competition. In addition,
The first round US Initiative on Joint Implementation has
the government supports specific energy-saving measures and
generated $40 million in private investments in GHG reduc-
the use of alternative sources of energy. tion.
Second-round proposals were also approved in December
The CO2 reduction program proposes actions in the sectors 1995.
Together, these proposals have the potential to reduce
of energy, buildings, industry, households, agriculture, for- CO2
emissions by tens of millions of metric tons per year. At
ˆ
estry, and R&D. Of the 29 measures identified by the Enquete the
same time, the voluntary government–utility partner-
----------------------- Page 19-----------------------
ships to mitigate GHG emissions called the US Climate Chal- 11. R

L. Kane et al., Global climate change: A discussion of major
lenge program is expected to reduce GHG emissions by 47
uncertainties, in Proc. Amer. Power Conf., 53-I, April 29–May 1,
million metric tons of carbon equivalent (MMTCE) from the
Chicago, 1991, pp. 646–652.
baseline year 2000 level. 12.
J. Douglas, The cost of greenhouse insurance, Elect. Power Res.
Inst. J., pp. 26–33, December 1992.
13.
D. M. Elsom, Atmospheric Pollution: Causes, Effects and Control
CONCLUSIONS
Policies, New York: Basil Blackwell, 1987.
14.
W. Fulkerson, R. R. Judkins, and M. K. Sanghvi, Energy from
Air pollution from electricity generation can be addressed at
fossil fuels, in Readings from Scientific American: Energy for
all stages of the energy cycle—from resource planning
Planet Earth, New York: W. H. Freeman, 1991, pp. 83–94.
through operations. Pre- and postcombustion technologies are 15.
H. G. Stoll, Least-Cost Electric Utility Planning, New York: Wi-
sufficiently mature, and the industry is nearing transition to
ley, 1989.
highly efficient next-generation combustion technologies. 16.
Organization for Economic Cooperation and Development, Envi-
However, the electric utility industry remains dependent on
ronmental Effects of Electricity Generation, OECD, 1985, chap. 2,
fossil fuel for electricity generation.
pp. 42–53.
As the consumption of energy, electrical or otherwise, in- 17. R
H. Shannon, Handbook of Coal-Based Electric Power Genera-
creases, the growth of CO2 emissions continues irrespective of
tion, Park Ridge, NJ: Noyes Publications, 1982.
conventions, treaties, and debate in the public media. Even 18.
R. Clement and R. Kagel, Emissions from Combustion Processes:
after many of the industrialized countries committed to re-
Origin, Measurement and Control, Chelsea, MI: Lewis Publish-
duce their CO2 emissions under the FCCC, most of them
ers, 1990.
failed to achieve their own goals. There is scientific evidence 19.
The Atmospheric Research and Information Centre (ARIC), Tech-
that global warming is unavoidable. However, the degree of
nologies for Reducing Acidic Emissions from Power Stations,
this warming can be slowed down if prudent action is taken
http://www.doc.mmu.ac.uk/aric/techno.htm, June 24, 1997.
by all parties, but most importantly, by the ones who can af- 20.
US Congress, Office of Technology Assessment, Changing by De-
ford the most and have some historical obligations due to
grees: Steps to Reduce Greenhouse Gases, OTA-O-482, Washington,
their past emission practices.
DC: US Government Printing Office, February 1991.
It is true that a few developing countries will gradually 21.
T. W. Keelin and C. W. Gellings, Impact of Demand-Side Manage-
surpass some of the highest CO2 emitters of the late twentieth
ment on Future Customer Electricity Demand, EPRI EM-4815-SR,
century, and the level of atmospheric CO2 concentration af-
Palo Alto, CA: Electric Power Research Institute, October 1986.
fects all countries. It is, however, judicious for the industrial- 22.
Customer Systems Division, Technical Brief: Load Management
ized countries to take unilateral actions to reduce their CO2
and the Environment, Report RP2788, Electric Power Res. Inst.,
emissions, at least so that they can take the moral high
1991.
ground in the debate on who has been responsible for global 23.
M. D. Levine et al., Electricity end-use efficiency: Experience with
warming and who will make how much sacrifice to reverse
technologies, markets, and policies throughout the world, Energy
the trend.
Int. J., 20 (1): 37–61, 1995.
24.
Project 88—Round II, Incentives for Action: Designing Market-
Based Environment Strategies, a public policy study sponsored by

BIBLIOGRAPHY
Senator Timothy E. Wirth, Colorado, and John Heinz, Pennsylva-
nia, Washington, DC: US Senate, May 1991.

1. K. E. Yeager, Electric vehicles and solar power, IEEE Power Eng. 25.
D. W. Pearce and R. K. Turner, Economics of Natural Resources
Rev., 12 (10): 13–18, 1992.
and the Environment, Baltimore: Johns Hopkins University
2. International Energy Agency, Energy Statistics of OECD Coun-

Press, 1990.
tries: 1990–1991, Paris: OECD, 1993. 26.
A. Pytte, Clean Air Act amendments, Congressional Quart., pp.
3. International Energy Agency, Energy Policies of IEA Countries,

3934–3963, November 24, 1990.
1991 Review, Paris: OECD, 1992. 27.
UNEP, Understanding Climate Change, UNEP/WMO Informa-
4. International Energy Agency, Energy Statistics and Balances in

tion Unit on Climate Change (IUCC), December 1994.
Non-OECD Countries: 1990–1991, Paris: OECD, 1993. 28.
International Energy Agency, Climate Change Policy Initiatives:
5. International Energy Agency, Climate Change Policy Initiatives:
1994 Update, Vol. 1, OECD Countries, Paris: OECD, 1994.
1994 Update, Vol. 1, OECD Countries, Paris: OECD, 1994. 29.
H. Idelson, National energy strategy provisions, CQ, pp. 3722–
6. Organization for Economic Cooperation and Development, OECD

3730, November 1992.
Environmental Data: Compendium 1991, Paris; OECD, 1991.
30.
United States Initiative on Joint Implementation, description of
7. L. E. J. Roberts, P. S. Liss, and P. A. H. Saunders, Power Genera-
the USIJI, Joint Implementation World Wide Web Server, http://
tion and the Environment, London: Oxford University Press, 1990.
www.ji.org/usiji/descrip.shtml, June 18, 1997.
8. Acid Rain Division, Energy Efficiency and Renewable Energy: Op- 31.
W. L. Fang, The US climate challenge program voluntary govern-
portunities for Title IV of the Clean Air Act, Washington DC: Envi-
ment-utility partnerships to mitigate greenhouse emissions,
ronmental Protection Agency, February 1994.
IEA/OECD Workshop on Voluntary Approaches for Mitigating
9. N. Sundararaman, Energy and global climate change issues, in
CO2 Emissions, Bonn, Germany, October 30–31, 1995, JI Online
Electricity and the Environment, Proc. Senior Expert Symp., Hel-
World Wide Web Server, http://www.ji.org/iuep/eabonn.shtml, Oc-
sinki, Finland, May 13–17, Vienna: Int. Atomic Energy Agency,
tober 1995.
1991, pp. 203–217. 32.
International Utility Efficiency Partnerships, Inc., The Interna-
10. S. Rahman and A. de Castro, Environmental impacts of electric-

tional Utility Efficiency Partnerships, Inc., Joint Implementation
ity generation: A global perspective, IEEE Trans. Energy Con-
World Wide Web Server, http://www.ji.org/brochure.shtml, June
vers., 10: 307–314, 1991.
18, 1997.
----------------------- Page 20-----------------------
374 AIR TRAFFIC
33. Tokyo Electric Power Company, Environmental Action Report:

TEPCO’s Commitment to Energy and Environmental Issues,
July 1995.
34. C. M. Callwood, The role of legislative processes and electric utili-

ties in effecting global environmental goals, Master’s Thesis, Vir-
ginia Polytechnic Institute and State University, February 1996,
pp. 36–38.
35. F. McGowan, Controlling the greenhouse effect: The role of re-

newables, Energy Policy, pp. 110–118, March 1991.
36. W. Bach, Global warming climate protection measures of the En-

ˆ
quete-Commission of the German Parliament, in External Envi-
ronmental Costs of Electric Power: Analysis and Internalization,
Proc. German–Amer. Workshop, Ladenburg, FRG, October 23–25,
1990, Springer-Verlag, 1991, pp. 36–58.
Reading List
1. P. W. F. Reimer, A. Y. Smith, and K. V. Thambimuthu (eds.),

Greenhouse Gas Mitigation: Technologies for Activities Imple-
mented Jointly, 26th–29th May 1997, Oxford: Elsevier, 1998.
2. M. M. Schorr, A 1992 update on legislation and regulations af-

fecting power generation, TMI Handbook, 1992, pp. 49–69.
3. J. Edmonds, J. Dooley, and M. Wise, Atmospheric stabilization

and the role of energy technology, paper presented to Climate
Change Policy, Risk Prioritization, and U.S. Economic Growth
convened by the American Council for Capital Formation Center
for Policy Research, The Pacific Northwest Laboratory, Washing-
ton, D.C., September 11, 1996.
4. Official Site of the Third Conference of the Parties to the

UNFCCC, http://www.cop3.de/, May 25, 1998.
5. J. T. Houghton, G. J. Jenkins, and J. J. Ephraums, Climate

Change: The IPCC Scientific Assessment, Cambridge: Cambridge
University Press, 1990.
6. R. A. Hinrichs, Energy: Its Use and the Environment, Fort Worth,

TX: Saunders College Publishing, 1996.
7. International Energy Agency, Electricity Information 1995, Paris:

OECD/IEA, 1996.
8. J. Edmonds, M. Wise, and D. W. Barns, Carbon coalitions: the

cost and effectiveness of energy agreements to alter trajectories
of atmospheric carbon dioxide emissions, Energy Policy (Special
Issue: Integrated Assessments of Mitigation, Impacts and Adap-
tation to Climate Change), April–May 1995, vol. 23, pp. 309–336.
9. P. Read, Responding to Global Warming: The Technology, Econom-

ics, and Politics of Sustainable Energy, London: Zed Books, 1994.
10. IEA Greenhouse Gas Programme Home Page [v4], http://

www.ieagreene.org.uk/, May 25, 1998.
ARNULFO DE CASTRO
SAIFUR RAHMAN
CONCHA REID
Virginia Polytechnic Institute and
State University
AIR POLLUTION CONTROL. See ELECTROSTATIC PRE-

CIPITATORS.
----------------------- Page 21-----------------------
} { { } }
●
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Cepstral Analysis of Speech
Standard Article

Yoh’ichi Tohkura 1 and Kiyoaki Aikawa2
1
NTT Laboratories, Kyoto, Japan

2NTT Laboratories, Kyoto, Japan
Search
reserved.
DOI: 10.1002/047134608X.W6702
❍ Acronym Finder
Abstract

Cepstral Analysis
Liftering
Temporal Analysis of Cepstrum

%20Conversion/W6702.htm17.06.2008 16:34:48
----------------------- Page 22-----------------------
CEPSTRAL ANALYSIS OF SPEECH 147
CEPSTRAL ANALYSIS OF SPEECH
The
most popular model of speech production views speech
signals as consisting of two components, representing an exci-
tation source (either quasiperiodic pulses or random noise)

and
vocal tract resonance. In order to study the nature of
speech and to develop speech processing technologies in vari-

ous
ways, it is desirable to separate these two components.
Cepstral analysis (or homomorphic analysis) is a procedure

which
can satisfy this demand. The word cepstrum was cre-
ated
by reversing the first four letters of the word spec-trum.
In
general, if two spectrally different components are com-
bined
additively, it is more or less possible to separate them
by
linear filtering. The two components of speech, the exci-
tation source and the vocal tract response, are sufficiently
different in their spectral features (i.e., rapidly varying com-
ponent vs. slowly varying component) that these two
components can be separated as follows. Let S(#) and H(#) be

the
spectra of the excitation source and the vocal tract reso-
nance, respectively. The speech spectrum represented by

their
product X (#) # S(#)H(#) is transformed into a sum by
a
logarithmic transformation:
log(|X (ω)|) = log(|S (ω)|) + log(|H (ω)|) (1)
Because of the sufficient difference in the spectral features of

the
two components, log(S(#)) and log(H(#)), are linearly sep-
arable.
CEPSTRAL ANALYSIS
The
inverse discrete Fourier transform (IDFT), c , of
log#X
(#)# is called the cepstrum of speech signal x(t) and is
represented as
cn = 1 # π log(|X (ω)|) cos(ωn) dω
2π −π (2)
for n = 0, 1, . . .,N

----------------------- Page 23-----------------------
148 CEPSTRAL ANALYSIS OF SPEECH
The cepstrum is normally real because the spectrum #X (#)# is

4 Speech wave
× 10
symmetrical against the origin of the frequency axis.
2
In the cepstrum, the cepstra of the excitation source and
1.5
the vocal tract resonance are additively combined since the
IDFT is a linear operation. Thus, as is illustrated in Fig. 1,
1
the cepstral analysis of a speech signal is a series of opera-
tions consisting of windowing, DFT, absolute operation, loga- e
0.5
d
u
t
rithmic transformation, and IDFT. An input speech segment i
l
0
p
(Fig. 2), a log spectrum [Fig. 3(b)], and the resulting cepstrum m
(Fig. 4) are depicted. When plotting the cepstrum, the ordi-

– 0.5
nate is called the quefrency (created from frequency ) instead
–1
of time. The slowly varying component of the log spectrum,
corresponding to vocal tract resonance, is represented by the
– 1.5
low-quefrency component of the cepstrum. In contrast, the
–2
rapidly varying component, corresponding to the excitation
0 10 20 30 40
source, is represented by the high-quefrency component. Note
Time (ms)
that a strong peak component is observed at a quefrency
Figure 2. Speech wave.

equal to the pitch period of the input speech signal.
Spectral Envelope and Pitch Extraction
A low-
order cepstral coefficient represents a slowly varying
component of the log

spectrum. Taking only low quefrency
Speech wave
components of a cepstrum yields the spectral envelope.
In
the case of voiced speech sounds, the excitation source
is
quasi-periodic with the fundamental frequency (pitch fre-
quency
or F0) of the vocal cord vibration. Since the excitation
Data window source
component can be separated from the vocal tract reso-
nance
in the cepstrum, cepstral analysis is a valuable tool
for
pitch extraction as well as formant analysis. Again the
excitation source corresponds to the rapidly varying compo-

DFT nent,
i.e., the high-quefrency cepstrum component. As is
shown
in Fig. 4, the fundamental frequency component ap-
pears
as a strong peak in the high-quefrency component. The
peak
location in terms of quefrency is equal to the pitch pe-
I Spectrum I riod
(1/F0). Therefore, automatically picking the peak of the
cepstrum within the possible pitch range of quefrency is a
Logarithmic transform
Spectrum of a vowel
40
IDFT
20
(a)
)
B
d
(
0
e
d
u
(b)
Liftering t
i
n
g
–20
a
M
(c)
DFT Peak detection

–40
(d)
–60
0 1 2 3 4 5 6
Frequency (kHz)
Spectral Fundamental
envelope frequency Figure
3. Spectrum of a vowel sampled at a frequency of 12 kHz:
(a) LPC
spectral envelope (the order of LPC analysis is 16); (b) FFT
Low quefrency region High quefrency region
spectrum; (c) FFT spectrum obtained by truncating cepstrum at que-
frency
# 48; (d) LPC spectrum obtained by truncating cepstrum at
Figure 1. Cepstral analysis.
quefrency # 12.
----------------------- Page 24-----------------------
Cepstrum If
Eq. (3) is represented by
0.4
H (z ) = (7)
A (z )
0.2
A (z ) = 1 − # α z −n (8)
n= 1
0
= # (1 − q z −1) (9)
k
–0.2
k = 1
the
log-spectrum can be represented as
–0.4
log(1/A (z )) = Rnz −n (10)
n
–0.6
n= 1
0 5 10 15
p
Quefrency (ms)
Rn = # qn (11)
l
Figure 4. Cepstral coefficients.
l = 1
The
variable denoted as Rn is called the root-power sum. By
viable pitch extraction method. Such strong cepstrum peaks
equating equal powers of z #1 between Eq. (10) and Eq. (4), the
are indicators of voiced speech sounds and they are not ob-
simple relation
served in the case of unvoiced portions of speech. The value
1
of the peak indicates the periodicity of the speech signal and
cn = Rn (12)
can be used for voiced-unvoiced decision purposes.
n
Linear Predictive Coding Cepstrum is

derived (1).
A
direct relation between predictor coefficients and cep-
Linear predictive coding (LPC) analysis is an alternative to strum
is derived as
cepstral analysis. The LPC cepstrum is a parameter which
#
has basic properties similar to the cepstrum and can be de-
p p
# #
(( k ) − 1)!
i= 1 i k
rived in a computationally efficient manner.
cn = #p (−α ) i (13)
(k !) i
In the linear prediction model, the transfer function H(z)
i= 1 i i= 1
of the vocal tract is represented by an all-pole transfer func-
tion with p poles as where
the sum is taken over all combinations of ki to meet
the
condition
1
#
H (z ) = p −n (3)
1 − α z
p
#
n= 1 n
iki = n (14)
where # , n # 1, 2, . . ., p are LPC coefficients and z is the
i= 1
n
usual z-transform variable. By considering the power-series

expansion of the logarithmic transfer function with powers The
root-power sums can be computed nonrecursively by the
z #1, log(H(z)) is described by the LPC cepstrum coefficients
direct relations between cepstrum and LPC coefficients (1).
c as
n
Spectral Distance Measure

p
log H (z ) = # c z −n (4) A
variety of spectral distance measures and distortion mea-
n
n= 1 sures
have been proposed for speech processing. These are
used
for measuring the spectral difference between two spec-
All the poles of H(z) must be inside the unit circle. After sub- tra.
The spectral distance measures have been widely applied
stituting H(z) from Eq. (3) into Eq. (4), derivative operations in
template matching-based speech recognition.
for both sides of Eq. (4) eventually lead to the following simple
The Euclidean distance of two spectra S(f )(#) and S(g )(#) is
relationship between the cepstral coefficients cn and LPC coef- given
by
ficients #
n
c1 = −α1 (5)
DCEP = 1 # π (log(S(f ) (ω)) = log(S(g ) (ω)))2 dω (15)
2π
−π
 #n −1 k

−αn − 1 − α c 1 < n ≤ p
∞
#
k = 1 n k n −k
c = (6)
(f ) (g ) 2
#
n p k
= (cn − cn ) (16)

− k = 1 1 − n αk cn −k p < n
n=−∞
Thus, the cepstrum coefficients can be recursively derived DCEP

is called the cepstral distance, and c0 displays a wider
from the LPC coefficients.
dynamic range than other coefficients. Therefore, the follow-
----------------------- Page 25-----------------------
ing weighted distance measure can be used onances

can be well determined by applying an automatic for-
mant
tracking algorithm.
N
D = r (c(f ) − c(g ) )2 + 2(1 − r) #(c(f ) − c(g ) )2 (17)
CEP 0 0 n= 1 n n Truncation
of Cepstrum
Truncating
cepstrum cn at n #
, is equivalent to multiplying
where r is a balancing weight. Spectral distance between it by the
lifter weight
spectra can be used for representing spectral difference, how-
ever, the log spectral distance provides better speech recogni-
#1 |n | ≤ ν
tion performance when applied as a spectral matching mea-

ln = (24)
sure in speech recognition.
0 |n | > ν
The
spectrum corresponding to the truncated cepstrum is
LIFTERING
smoothed.
As is shown in Fig. 3, the LPC spectrum (a) has
sharper
peaks than that produced by the truncated LPC cep-
An operation to separate slowly varying spectral components
strum (d).
from rapidly varying ones begins with linear filtering of the

spectrum. The filter in the spectral domain is called the lifter.
The word lifter was created by reversing the first three letters Spectral
Slope Distance
of the word filter. The

frequency derivative of the log spectrum is given by
Let S(#) be a spectrum and H(#) be a lifter. A liftered spec-
#
trum Q(#) is given by
d log(S (ω))
= ncn cos(nω) (25)
#
dω
∞
n=−∞
log(Q(ω)) = log(S (ω − λ))H (λ)dλ (18)

−∞ A
distance measure can then be defined to measure the Eu-
clidean
distance between two spectral slope functions (2,3).
In digital speech processing, a speech signal is obtained not
(f ) (g )
The
spectral slope distance between two spectra S and S
as a continuous function of time, but as a sampled time series.
is given
by
When the sampling frequency is f s , the bandwidth of a speech
signal is limited to
#
2
1 π d log(S(f ) (ω)) d log(S(g ) (ω))

DSS =
− dω (26)
f = f /2 (19)
2π dω dω
max s
−π
∞
If Q(#) and S(#) are represented in a logarithmic magni- =
# #nc(f ) − nc(g )#2 (27)
n n
tude scale, these spectra are represented by cosine expansions
n=−∞
∞
log(Q(ω)) = # bn cos(ωn) (20) Weighted
Cepstrum
n=−∞
To
measure distance in a
multidimensional space, variance-
∞
log(S (ω)) = # cn cos(ωn) (21) normalized
distance measures such as the Mahalanobis dis-
n=−∞ tance
provide good performance in general pattern recogni-
tion. A
weighted cepstrum is such a variance-normalized pa-
where bn and cn are cepstral coefficients. The lifter impulse rameter
if each cepstral coefficient is regarded as an
response can also be represented by a cosine expansion.
independent parameter (4,5).
∞
cn
#
c(W ) = (28)
H (ω) = ln cos(ωn) (22)
n σn
n=−∞
where #2
is the variance of the nth ceptral coefficient. In case
By using the orthogonality of cosine functions, the following n
of actual
speech cepstrum data, 1/#n is approximately propor-
equation is derived:
tional to
order n, so the weighted cepstrum is approximately
b = l c (23) given by
n n n
c(W ) = w c (29)
Lifter weights ln correspond to the transfer function of each
n n n
cosine component in the log spectrum.
#
n if n < ns
Applying a low-pass lifter yields the slowly varying cep-
wn = (30)
ns otherwise
strum component corresponding to the vocal tract resonance.
DFT computation of the low-pass liftered cepstrum produces
the so-called spectrum envelope [Fig. 3(c)] consisting of the If ns #
#, the weighted cepstrum is equal to the root-power
slowly varying spectrum component. This operation is called sum shown
in Eq. (11). The weighting function wn saturates
cepstral smoothing. Since the spectral envelope is sufficiently at n # ns
to suppress excess weighting of high-order cepstral
smooth, local spectral peaks characterized by the formant fre-
coefficients. The Euclidean distance between two weighted
quencies and bandwidths corresponding to the vocal tract res- spectra
can be defined as a spectral distance measure.
----------------------- Page 26-----------------------
Group Delay Spectrum ing

that function of the auditory system. Delta-cepstrum is
one such
dynamic feature parameter (8) and is defined as the
The group delay spectrum provides a spectrum wherein peaks
slope of
the linear fitting curve of the cepstral time series:
are emphasized more than the power spectrum (6). Let H(z)
be the transfer function of an all-pole filter representing a
L
speech spectrum as
c# (i) = l =−L cn (i + l )lw (l ) (41)
n L 2
l =−L l w(l )
1
#
H (z ) = p −n (31)
1 + α z
n= 1 n where i
denotes frame number of the cepstral time series.
p
#
1 Delta-
cepstrum picks up the trend in the window #L
l
L
= n= 1 1 − (zn /z ) (32) of a
cepstral time series. The original delta-cepstrum is de-
rived
based on linear regression against the scatter of cep-
p
#
stral
values in the time axis.
= Hn (z ) (33) The
formulation of the delta-cepstrum in Eq. (41) can be
n= 1
translated into a filter for cepstral time series. The delta-

Let the gain and the phase be A (#) and (#), respectively; cepstrum
shows bandpass filtering characteristics. A triangu-
n n
each term of the all-pole model is given by lar
weighting function
j ω j φn (ω)

L − l
Hn (e ) = A n (ω)e (34)

 if l ≥ 0
G
w(l ) =  L (42)
The group delay spectrum T (#) is defined by the frequency


n
L + l
if l < 0
derivative of the phase
L
p
TG (ω) = # −dφn (ω) (35) gives a
bandpass transfer function that is better in eliminat-
n dω
n= 1 ing
side-lobes than a uniform weighting expression derived
from the
original formulation of regression. The frequency of
Cepstral coefficients for a group delay spectrum are de- the
temporally varying component in the cepstral time series
rived as is
called the modulation frequency.
A
delta-cepstral distance can be defined by a Euclidean
g n = ncn (36)
distance
like the cepstral distance. Let DCEP be a cepstral dis-
tance
and D#CEP be the delta-cepstral distance, the combina-
The original group delay spectrum formulation excessively
tional
distance can be defined as
emphasizes high-order cepstral coefficients. Thus, the follow-
ing generalized weighting function is used for practical
DCEP+#CEP = rDCEP + (1 − r)D #CEP (43)

speech recognition.
g = w c (37) where r
is a balancing weight between the cepstral distance
n n n
2 and the
delta-cepstral distance. A typical value of r for auto-
n
wn = ns exp −2τ 2 (s ≥ 0) (38) matic
speech recognition is 0.05.
This representation is called the smoothed group delay spec-

Dynamics-Emphasized Cepstrum
trum. # and s are parameters that control the smoothness of

Dynamics-emphasized cepstrum is a spectral representation
the spectrum. composed
of instantaneous and transitional feature parame-
ters. As
discussed in Reference (9), the formulas for calculat-
Bandpass Lifter ing
dynamics-emphasized cepstrum are given by
The above approaches can be generalized as liftering by lifter
2
weights represented as
dn (i) = cn (i) + 8#cn (i) − 8# cn (i) (44)
3
b = w c (39)
l =−3 lcn (i + l )
n n n
#cn (i) = (45)
3 l2
l =−3
The lifter weight
3 (l2 − 4)cn (i + l )
l =−3
2
ν n
# cn (i) = (46)
#
wn = 1 + sin π (1 ≤ n ≤ ν) (40)
3 (l2 − 4)2
2 ν
l =−3
is a good choice to increase speech recognition accuracy (7). where

#2c (i) denotes the n-th coefficient of the second-order
delta-
cepstrum at time i.
TEMPORAL ANALYSIS OF CEPSTRUM
RASTA
Delta-Cepstrum
RASTA is
another filter for cepstral time series and performs
The auditory system is sensitive to temporal changes in as a
bandpass filter (10). RASTA achieves a lower resonance
sound features. Research effort has been focused on simulat-
modulation-frequency by multiplying a temporal integration
----------------------- Page 27-----------------------
term by a delta-cepstrum. The z transform of RASTA is given

elapsed time after the masker is given to the auditory system.
by A
masked cepstrum is derived from the masked spectrum.
Let
M(#, #) be thetwo-dimensional impulse response that
−2 2 −1 −2
z (2z + z − z − 2z )
simulates the auditory masking function, the masked spec-
H (z ) = 0.1 (47)
z −4 (1 − 0.98z −1) trum
is given by
L ∞
#
The numerator is a delta-cepstrum. The denominator con-
tributes to temporal integration. The terms z #2 and z #4 do not
log(Q(ω,t)) = log(S (ω − λ, t + i#t))M (λ, i#t) dλ (52)
affect the gain of the modulation-frequency transfer function.
i=0 −∞
Bandpass and Highpass Filters where

i#t is the elapsed time after the end of the masker
sound,
#t is the frame shift, and # denotes frequency dif-
BPF (bandpass filter) and HPF (highpass filter) formulations
ference.
have also been proposed for filtering the modulation-fre- The
masked cepstrum is represented by the cosine expan-
quency component in the cepstral time sequence (11). A band- sion
coefficients of the masked spectrum. Let ln and cn be the
pass filter can be formulated for speech recognition as
inverse Fourier transform of the masking impulse response
and
the log spectrum, respectively, the inverse Fourier trans-
#
κ L (l − L −1 )z −l form
of Eq. (52) is given by
H (z ) = l =0 2 (48)
1 − ρz −1
N
and a highpass filter as
bn (t) = # ln (i)cn (t + i#t) (53)
i=0
#
L l −l
l = 1 β z
H (z ) = 1 − (49)
This lifter has two different aspects as an order-dependent
#
L βl
l = 1
temporal filter for cepstral time series, and an elapsed-time-
dependent lifter. The former acts as a high modulation-fre-

where #, #, and # are constants for determining the transfer
quency-pass filter for cepstral time series. The latter acts as
functions. a low
quefrecncy-pass lifter for the spectrum at a given
elapsed time.
FM Neuron Model
A biological FM neuron detects unidirectional frequency
BIBLIOGRAPHY
change. A model has been proposed to simulate the function
of the FM neuron (12). The FM neuron model is formulated
as a time and frequency derivative of a spectral time series 1.
M. R. Schroeder, Direct (nonrecursive) relations between ceps-
trum and predictor coefficients, IEEE Trans. Acoust. Speech Sig-

and extracts formant movement. When a spectral sequence is
nal Process., ASSP-29 : 297–301, 1981.

given by S(#, t), the output of the FM neuron model can be
given by 2.
D. Klatt, Prediction of perceived phonetic distance from critical-
band spectra: A first step, Proc. ICASSP82, pp. 1278–1281, 1982.

∂2 log(S (ω,t)) 3.
B. A. Hanson and H. Wakita, Spectral slope distance measures
F (ω,t) = (50)
with linear prediction analysis for word recognition in noise,
∂ω∂t
IEEE Trans. Acoust. Speech Signal Process., ASSP-35 : 968–
973, 1987.
The cepstrum coefficients for the FM neuron output are given
by 4.
Y. Tohkura, A weighted cepstral distance measure for speech rec-
ognition, IEEE Trans. Acoust. Speech Signal Process., ASSP-35 :

∂cn (t)
1414–1422, 1987.
f n = n ∂t (51) 5.
K. K. Paliwal, On the performance of the frequency-weighted cep-
stral coefficients in vowel recognition, Speech Commun., pp. 151–

A delta-cepstrum can be used for practical time-derivative op-
154, 1982.
eration on a cepstral time series. When the center frequency 6.
F. Itakura and T. Umezaki, Distance measure for speech recogni-
of a spectral peak decreases with time, the FM neuron out-
tion based on the smoothed group delay spectrum, Proc.
puts a negative response. When the center frequency in-
ICASSP87, pp. 1257–1260, 1987.
creases with time, the FM neuron outputs a positive response. 7.
B.-H. Juang, L. R. Rabiner, and J. G. Wilpon, On the use of band-
pass liftering in speech recognition, IEEE Trans. Acoust. Speech

Masked Cepstrum
Signal Process., ASSP-35 : 947–954, 1987.
8.
S. Furui, Speaker-independent isolated word recognition using
The filter coefficients are identical among all cepstral orders
dynamic features of speech spectrum, IEEE Trans. Acoust. Speech
in the above cepstral analysis. A spectrotemporal spectral
representation is achieved by a two-dimensional matrix lifter 9.
S. Furui, Speaker-independent isolated word recognition based
based on the auditory masking effect (13). In an auditory sys- on
dynamics-emphasized cepstrum, Trans. IECE Jpn., E 69 (12):
tem the target sound is suppressed by a masker sound. The
1310–1317, 1986.
masked spectrum is given by the spectrum reduced by the 10.
H. Hermansky et al., Compensation for the effect of the commu-
masking level. The masking level is a function of the fre-
nication channel in auditory-like analysis of speech (RASTA-
quency difference between the masker and the signal and the
PLP), Proc. Eurospeech 91, pp, 1367–1370, 1991.
----------------------- Page 28-----------------------
CEREBELLAR MODEL ARITHMETIC COMPUTERS 153
11. B. A. Hanson and T. H. Applebaum, Subband or cepstral domain

filtering for recognition of Lombard and channel-distorted speech,
Proc. ICASSP93, 2: 79–82, 1993.
12. K. Aikawa and S. Furui, Spectral movement function and its ap-
plication to speech recognition, Proc. ICASSP88, pp. 223–226
15.S5.11, 1988.
13. K. Aikawa et al., Cepstral representation of speech motivated by

time-frequency masking: An application to speech recognition, J.
Acoust. Soc. Am., 100 (1): 603–614, 1996.
YOH’ICHI TOHKURA
KIYOAKI AIKAWA
NTT Laboratories
----------------------- Page 29-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Compressed Air Energy Storage
Standard Article

Ioan-Iulian Irimie1 1 1
, Sabina Irimie , Ildiko-Camelia Tulbure
1University of Petrosani, Petrosani, Romania
reserved.
Search
DOI: 10.1002/047134608X.W3012
❍ Acronym Finder
Abstract
Energy Conversion and Storage Systems
Means of Energy Storage by Compressed Air
Compressed Air—Energy Storage Medium and Energy Transfer Medium
Conclusion

%20Conversion/W3012.htm17.06.2008 16:35:09
----------------------- Page 30-----------------------
J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering

c
COMPRESSED AIR ENERGY STORAGE
Energy Conversion and Storage Systems
Energy conversion paths are either direct or indirect (Fig. 1). With
indirect technologies the line of energy
conversion is the following: the primary source is exothermally oxidized
(burnt), the chemical energy being
converted into thermal energy. By a thermal agent (motor fluid), the thermal energy
is used as such, or by using
a thermal machine (turbine, engine) it is converted into mechanical energy
actuating an energy-generating
working engine or generator. With direct technologies thermal
intermediation is eliminated along with its
thermodynamic imperfections, and electrical energy is directly gained, for
instance, from the chemical energy
of fuel. Wind energy falls into this category.
There are several competitive possibilities for converting primary energy
into electrical energy: hydraulic,
wind-driven, electrochemical, and nuclear (including fusion); solar energy may be
directly utilized on a photo-
voltaic line.
Of all primary energy forms suitable for sustainable development, wind energy
is the most advantageous:
it is practically inexhaustible, nonpolluting, available everywhere, and free of
charge, since no primary extrac-
tion is required. In converting the wind energy to other forms, the following
stages are to be noted: extraction,
conversion, storage, and consumption. For wind, energy storage is necessary both
because of the limitations
of wind (it is irregularly distributed in time and space, and it is dilute, that
is, it has low concentration per
unit area), and because of the variety of its consumers. Figure 2
presents a general diagram of extraction,
conversion, storage, and utilization of wind energy.
In the first stage (extracting energy from wind), mechanical energy is gained
as rotation, translation, or
oscillation, depending on the type of the extraction installation. This motion may
be used in the second stage
for conversion into another form of energy by liquid pumping, gas compression, or
electrical or heat generation.
The energy gained may be stored by using pneumatic or hydraulic systems, electric
batteries, heat, hydrogen,
flywheels, etc. This is the third stage. The fourth stage, that of consumption,
consists in supplying energy to
consumers in the form of heat, electricity, water, etc.
A comparison between different energy storage systems is presented in Fig. 3.
For compressed air, starting
with the data from the literature, which takes into consideration only the
mechanical component of the energy,
we have also taken into consideration the possibilities of utilizing the thermal
component.
Means of Energy Storage by Compressed Air
Energy storage by compressed air can be accomplished using storage tanks,

hydropneumatic accumulators,
and pneumatic network elements.
Compressed Air Storage Using Storage Tanks. Figure 4
shows the diagram of compressed air
storage. The installation is made up of the following elements: wind
turbine 1, hydrostatic transmission 2,
----------------------- Page 31-----------------------
2 COMPRESSED AIR ENERGY STORAGE
Fig. 1. Energy conversion paths.
compressor 3, storage tank 4, safety valves 5, manometers 6, purging valves 7,

pressure line 8, valve 9, supply
duct 10, safety valve 11, contact manometer 12, solenoid valve 13, pressure
regulator 14, pneumatic valve 15,
and cutoff valve 16.
Compressed air delivered by the compressor is stored in tanks. The
diagram provides for successive
supply of the two tanks. An automatic filling scheme for the two tanks is shown,
made up of a manometer with
contacts, giving the signal for solenoid valve to be opened, actuating in its turn
a normally closed pneumatic
valve through a pressure regulator. When a certain pressure is reached in the first
tank, the contact manometer
commands the opening of the supply to the second tank, which fills up from the first.
When the pressure falls
below a certain limit, the contact manometer will close off the supply, the
pressure in the first tank will rise
again, and the cycle will be repeated.
Figure 5 gives a diagram of the storage of compressed air generated by a wind
compressor, where the
compressed air is used to drive a gas turbine to produce electricity during
maximum-load periods of a power
system. For this purpose the traditional gas turbine type may be modified
by kinematic separation of the
compressor and turbine stages using gears (couplings), as shown in Fig. 6. In one
of the separation stages the
wind turbine actuates the compressor. The compressed air generated is stored in an
accumulator, which may
be a cavity or an exhausted natural-gas seam. In another stage, when electricity
demands exceed the system
----------------------- Page 32-----------------------
COMPRESSED
AIR ENERGY STORAGE 3
Fig. 2. Diagram of wind energy extraction, conversion,

storage, and utilization.
capacity on basic load, the compressor is switched off and the generator is
switched to the wind turbine. The
burner is supplied with gas fuel and compressed air from the storage system.
On compression without heat loss (adiabatic compression), the air temperature
increases. If compressed
air is stored in a thermally well-insulated accumulator, we speak of adiabatic
storage. When compressed air
at this temperature is taken out and used to actuate a turbine, as in the
previously mentioned case, less heat
is required than if the storage were at ambient temperature. Therefore, adiabatic
storage is more economical
than isothermal.
Compressed Air Storage with Hydropneumatic Accumulators.
Description of the Accumulator. The hydropneumatic accumulator (Fig. 7)
is made up of a pneumocham-
ber (1) and a hydrochamber (2). Between the two chambers there is a 65 m to 70 m
vertical level difference,
corresponding to a 6.5 bar to 7.0 bar pressure, joined by a connecting incline (4).
Both chambers are equipped
with closing dams (5). At higher storage pressures, the level difference
increases. The pneumochamber is
----------------------- Page 33-----------------------
Fig. 3. Energy density for various energy

accumulators.
Fig. 4. Storage system for compressed air.
----------------------- Page 34-----------------------

COMPRESSED
Fig. 5. Compressed air energy storage system.
Fig. 6. Adiabatic storage of compressed air.
connected to the compressor station (6) by a compressed air supply duct (7), and
to the main network by a
distribution line (8). Both ducts are dimensioned to suit the average consumption.
Water is supplied to the
hydrochamber by a supply duct (9).
A hydraulic connection (11) exists between the pneumochamber and the
hydrochamber, made up of metal
ducts, with mining workings forming a hydraulic arrester in the lower part (12).
All ducts are fitted with shutoff valves (14), permitting the hydropneumatic
accumulator to be isolated for
inspection and maintenance as necessary. Between the compressed air supply duct and
the distribution line
there is a bypass (15), which may be used to short-circuit the hydropneumatic
accumulator.
The pneumochamber must be carefully designed to avoid water or compressed air
losses. Proper sealing
is decisive for good operation and for the economic efficiency of the accumulator.
With fissured rocks or with
medium-hard and soft rocks, consolidation and sealing are effected by cement-milk
or synthetic-resin injections.
The resistance dam closing the pneumochamber (1) must be specially
constructed for durability and
sealing, as it is crossed by the compressed air supply and distribution
ducts, transducers for working and
maximum levels, and an access opening where personnel can enter for control and
maintenance. Optionally
the resistance dam may be fitted with a water-gauge glass, which, during
test periods, may provide data
----------------------- Page 35-----------------------
Fig. 7. The hydropneumatic accumulator.
about the adjustment or functioning of the signaling and monitoring

equipment. The latter is essential for
the protection of the accumulator, and its operation is programmed in synchrony
with the compressor station
working periods.
According to the general features presented, several types of accumulators
may be devised, adapted to the
local geological and mining conditions, to existing or planned mine workings, and
to the compressor station’s
capacity.
----------------------- Page 36-----------------------
COMPRESSED
In order to cut down investment costs, in building a hydropneumatic

accumulator it is recommended to
reuse decommissioned mining or geological research workings.
Operating Principle. A hydropneumatic accumulator is a complex of mining
workings providing for the
3 3
storage of a large volume of compressed air (1500 m to 8000 m ) at
a constant pressure of 6.5 bar to 8.5
bar, which can be utilized by consumers in case the compressor station is out of
service. Thus, idle running of
compressors in low-consumption periods is avoided, and optimum parameters of
compressed air in peak periods
are ensured when the consumption required is higher than the capacity of the
compressors in operation. Two
distinct stages occur in the operation of the hydropneumatic accumulator:
• Accumulation of compressed air in the pneumochamber, in periods when the flow of

the compressor station
exceeds the flow absorbed by consumers, during which water is evacuated into the
hydrochamber.
• Return of compressed air, while water from the hydrochamber flows into
the pneumochamber and air
consumption continues. In this stage the compressor station may run
at a reduced flow rate or may be
completely stopped.
To summarize, the hydropneumatic accumulator is a compressed air reservoir

with variable volume and
constant pressure, ensured by the potential energy of the water column
joining hydrochamber and pneu-
mochamber.
By the monitoring and signaling installation the current state of the
pneumochamber is reported to the
compressor station so as to stop or start the compressors. The moments at which the
compressors are started
and stopped being thus clearly determined, idling is completely eliminated, leading
to a valuable reduction in
energy consumption.
At the end of the first stage in the operation of the
hydropneumatic accumulator, if the compressors
are not stopped in time, there is a risk that the compressed air in
the pneumochamber will pass into the
hydrochamber through the hydraulic connection, where it can spray or push water out
of the hydrochamber,
causing unbalanced operation and possible severe damage to the surrounding mining
workings. To prevent
such accidents the accumulator is fitted with a hydraulic arrester made up of an
additional 10 m to 13 m water
column, for the evacuation of which the pressure in the accumulator is increased
over the working pressure of
the compressors by 1 bar (10 m H2 O).
Dimensioning Criteria for the Accumulator. The main operational
parameters to be established for a
hydropneumatic accumulator are the pneumochamber volume, the hydrochamber volume,
the pressure in the
pneumochamber, and the height between chambers.
The pneumochamber volume may be assessed according to several criteria:
• When the active capacity of the compressor station has been determined as a
function of the maximum
consumption of the users, we have
where
V = pneumochamber volume (m3 )
3
Qmax = maximum compressed air consumption (m /min)
Q 3
st = actual flow of active compressors (m /min)
τ = time of maximum-consumption requirement (min)
• One may use the following empirical function of the active power of
the compressor station: for each
kilowatthour, 1.5 m3 of accumulation buffer reservoir volume is needed at 7
bar pressure (1).
----------------------- Page 37-----------------------
• The volume may be chosen according to the minimum capacity of

operation of the compressor station
determined by the minimum primary energy (wind) availability. In periods of
minimum availability, only
a few of the compressors may be operated or the entire station may
cease to function. If Cmf is the
minimum capacity for operation in minimum-availability hours, Qm
is the average consumption, τ is the
longest duration of the primary energy deficit, and p is the working pressure
of the accumulator, then the
pneumochamber volume in cubic meters will be
The hydrochamber volume is an established function of the pneumochamber

volume, being approximately
10% larger to compensate possible water losses and to accommodate water
infiltration.
Thepneumochamber pressure is function of the rated pressure of the compressor
station and is determined
with the relationship:
where
p p = pressure in pneumochamber (bar)
p s = pressure in compressor station (bar)
#p l = sum of linear losses (bar)
#p ε = sum of local losses due to supports, bends, or cross-section
alteration (bar).
The condition for the accumulator to be able to operate is that the

compressed air pressure supplied by
the compressor station at the entrance to the pneumochamber should be higher (by
0.2 bar to 0.3 bar) than the
pressure within. If this condition is not met, hydraulic compensation is no longer
possible and the accumulation
volume is thus reduced to zero due to the water flooding in.
The height between chambers is an established function of the pressure in the
pneumochamber, according
to the relationship
where
H = height between the pneumochamber and the hydrochamber (m)
2
p p = compressed air pressure in the pneumochamber (N/m )
2
p 0 = air pressure above water in the hydrochamber, (N/m )
γ = volumetric weight of water (N/m3 ).
The height determined as above is a maximum, and it refers to the working

height of water; the geometrical
height is lower by the differences between minimum and maximum water levels in the
chambers in the two
working stages.
Besides the previously presented storage systems, two other systems applied
in the case of natural gases
are considered by us as particularly efficient. These are:
• Storing compressed air in the main network

• Storing compressed air in a high-pressure annular collector–distributor
Storing Energy by Compressed Air for Thermal Power Stations.

To produce low-cost electric
energy, thermal power stations must be operated continuously and, as far as
possible, at optimum load with
respect to energy (maximum efficiency, minimum specific consumption, and low
pollution). The performance
----------------------- Page 38-----------------------
COMPRESSED
Fig. 8. Diagram of daily load (P =power;

t=time).
of distribution networks is also favored by these conditions. Unfortunately,

however, they are rarely met in
practice. In many cases, most of the time, the electricity consumption in an area
is almost twice as high in the
daytime as at night. The diagrams presented in Figs. 8 and 9 show that there is an
unbalance both in daily
and in weekly loads.
In most cases, peak loads are covered by the use of gas turbines. Gas-turbine
installation costs are low,
but the fuels used in them (natural gas or petroleum) are expensive and liable to
be depleted, with the further
disadvantage of not being able to use the basic stations’ nighttime power reserves.
For large storage capacities (≥ 1500 MWh) and economical operation, hydraulic
and pneumatic storage
have proved to be the most advantageous. Hydraulic storage is recommended for areas
with uneven topography.
In flat regions pneumatic storage is better.
Thermoelectric power stations with air reservoirs have the following
advantages:
• Improvement of daily and weekly load balance and a more rational use of
existing stations
• The possibility of using secondary energy resources at the stations
• Fuel savings compared to other peak stations
• The possibility of coupling systems for the use of renewable energy resources
to thermoelectric stations.
The most advantageous economic and functional conditions for compressed air
storage reservoir arrange-
ments are offered by underground saline rock formations. Air storage containers can
also be excavated in hard
rocks such as granite, gneiss, sandstone, or limestone. Since these are
expensive operations, air reservoirs
should be hydraulically compensated (water column) to minimize the volume necessary
for the reservoir. In
some cases, natural or existing artificial cavities (abandoned mine workings,
emptied petroleum or natural-gas
reservoirs) can be used as air reservoirs.
Various schemes for pneumatic energy storage at thermoelectric stations are
now compared with respect
to the storage cycle and choice of reservoir.
----------------------- Page 39-----------------------
Fig. 9. Diagram of weekly load.
Adiabatic Storage Installations. An adiabatic installation (Fig.

10) is the pneumatic equivalent of a
hydraulic pumping and turbocompressing station, representing the ideal
case. No heat exchange with the
exterior takes place (except in the cooler F, intended to removing humidity from
the compressed air).
During the loading stage, compressed air without intermediary cooling yields
heat to a heat accumulator
before passing to the reservoir. At discharge and commissioning of the
turbine, the accumulator warms up
the air coming from the reservoir before its detention in the turbine. The input
temperature in the turbine is
determined by the final compressor temperature, which is lower (t
<t , Fig. 10). The turbine input temperature
11 4
is therefore a function of the reservoir pressure.
Nonadiabatic Storage Installation Fitted with a Combustion Chamber.
In practical cases, pneumatic-
storage stations use nonadiabatic storage installations (Fig. 11) with
low-pressure and high-pressure com-
bustion chambers. In the case of nonadiabatic installations, turbine input
temperatures are independent of
reservoir pressure, being determined by the fuel injected upstream of the turbine.
Comparison Criteria for Different Storage Cycles. Four
characteristic values for comparing various
thermodynamic cycles of air-resevoir stations are suggested in the literature (2):
----------------------- Page 40-----------------------
COMPRESSED AIR ENERGY STORAGE 11

Fig. 10. Adiabatic storage installation. (a) Schematic diagram: A, alternator–
motor; B, coupling; C, low-pressure com-
pressor; D, high-pressure compressor; E, heat accumulator; F, cooler; G,
air reservoir; H, turbine. (b) Entropy diagram:
T=temperature; s=specific entropy.
The first two characteristic ratios (α and β) are related to

construction expenses, and the last two (γ
and δ) to operation expenses, of an air-reservoir station. The ratio α determines
the turbine and compressor
dimensions, and β the size of the air reservoir. The ratio γ determines
the amount of fuel needed in the
combustion chamber, and δ the necessary pumping energy.
To appreciate the cooperation between the air-reservoir station and the
electric network or basic station,
RT is relevant. It equals
which supplies the energy necessary for the compressor drive, the total
efficiency η
the total fuel energy (kJ) required in the basic station and in the
air-reservoir station to obtaining 1 kJ at
RT the loss coefficient ε can be derived,
expressing the total energy lost to the
the alternator terminals. From η
surroundings as heat. The expressions for the two indicators are
where
EGen = energy delivered at alternator terminals at the air-reservoir
station
EMot = energy supplied to compressor motor at the air-reservoir station by
the basic station
η =E /E = basic station efficiency versus motor terminals.
B Mot f1
----------------------- Page 41-----------------------
Fig. 11. Nonadiabatic storage installation fitted with low-pressure and high-
pressure combustion chambers. (a) Schematic
diagram (see also Fig. 10): J, recuperator; K, high-pressure combustion chamber; L,
high-pressure turbine; M, low-pressure
combustion chamber; N, low-pressure turbine; O, intermediary cooler. (b) Entropy
diagram.
Ef1 = fuel energy for basic station

Ef2 = fuel energy for air-reservoir station
EAmb = heat lost to the surroundings
The efficiencies of the adiabatic and nonadiabatic schemes are shown in Table
2.
A few remarks should be added on the financial aspects of these
comparative values. The total costs
associated with an air-reservoir station can be analyzed as follows (for
ten years of turbine operation with
reservoir load):
Machines . 40% to 50%

----------------------- Page 42-----------------------
COMPRESSED AIR
ENERGY STORAGE 13
Fig. 12. Representation of compressed air expansion in a T–

s diagram.
Air reservoir . 30% to 15%

Annexes . 30% to 35%
The low values for the reservoir refer to reservoirs in saline rocks. If
several groups of machines were
connected to the same reservoir, its share in the overall cost would be even
smaller.
In conclusion one may say that the energy storage installations described in
this section are economic.
----------------------- Page 43-----------------------
Fig. 13. Energy storage and delivery by compressed

air.
Compressed Air—Energy Storage Medium and Energy Transfer Medium
Advantages and Disadvantages of Compressed Air for Use in Energy Transfer and
Storage.
Compared to other energy storage systems, compressed air shows the following
advantages:
• It is an elastic, flexible, and comparatively safe medium.

• It has a reasonable capacity for energy per unit volume or weight,
and it becomes competitive at high
pressure when mechanical storage and thermal storage are combined.
• It uses part of the energy lost to irreversibility (xl in Fig. 12) as kinetic
energy during its flow.
• By a comprehensive assessment it is less polluting than other energy storage
systems.
• It is safe in explosive atmospheres.
• In underground mine workings, when exhausted from pneumatic equipment, it
cleans the atmosphere, and
in an emergency (collapses, people trapped underground) the pneumatic network
allows trapped staff to
be supplied with air, water, and food.
• It allows a large number of accumulation–discharge cycles without diminishing
rated capacity.
• It offers simple construction and high operating reliability.
• It allows multiple use of pneumatic energy.
Low energy efficiency is its main disadvantage, both in industrial uses and in
traditional storage systems.
The losses within the process of compressed air generation, transport, storage, and
use are as follows: 18% loss
by friction in the compressor, 10% air losses due to compressor leakage, 18%
internal energy losses by cooling
in storage, and 34% losses in adiabatic expansion. These losses sum to 80%, so that
20% remains available.
Ignoring constructive improvements of the compressor (losses due to friction
and leakage diminished) and
focusing on recovering the thermal component in compressed air generation and on as
complete as possible
utilization of the adiabatic expansion (pneumatic expansion engines,
refrigeration plants with air), we can
state that the available energy may be increased by about 40%, so that the
utilization coefficient for pneumatic
energy should be competitive with that for electricity.
----------------------- Page 44-----------------------
COMPRESSED
Fig. 14. Classification of compressors. In volume compressors the flows do

not depend on the exhaust pressure; in
centrifugal compressors the exhaust flow depends on the compression.
General Scheme for Compressed Air Energy Storage. The

diagram in Fig. 13 represents a sum-
mary of the possible developments by which energy storage with compressed air could
become economically
attractive.
Types of Compressors and Fields of Use. The various
types of compressors used to generate
compressed air are presented in Fig. 14. In the selection of the compressor type,
the conditions required by the
primary energy source, the consumer, and the economic circumstances must be allowed
for.
The limits of use for various types of compressors in terms of
pressure and flow are presented in
Fig. 15.
Exergetic Analysis of Processes from the Component Elements of Energy
Storage.
General View. The notions of heat exergy and thermal agent exergy in
continuous and stationary flow
are natural consequences of the Gouy–Stodola law and are, as will be seen, useful
in determining the losses
caused by external irreversibility of thermal processes (3,4).
In a series of cases found in engineering (e.g., compressors and detentors
with pistons), separate iden-
tification of the influence exerted by various physical state values
on the exergy of a thermal agent is
----------------------- Page 45-----------------------
Fig. 15. Domains of use for various types of compressors in

terms of pressure and flow.
necessary. To develop this idea, the case of the exergy e of a perfect

gas, characterized by the parameters
p >p 0 and T >T0 , is considered. Denoting the isobaric specific heat of the gas by
cp and the gas constant by R ,
we have
where
i = specific enthalpy
s = specific entropy and subscript 0 refers to the surroundings.
The exergy e is seen to consist of two components: (a) one that depends only
on the temperature of the
fluid, called the thermal component eT , and (b) one that depends only on its
pressure, called the mechanical
----------------------- Page 46-----------------------
COMPRESSED AIR
ENERGY STORAGE 17
Fig. 16. Representation in p –v and T–s diagrams of the two

exergy components.
component ep (4). Consequently,
and thus
A thorough analysis of the structure of the two components leads to the

following interpretations:
• The thermal component represents the absolute value of heat exergy exchanged by
the gas with the exterior
in an isobaric temperature change from T to T0 .
• The mechanical component represents the isothermal work exchanged by
the gas with the exterior at
temperature T0 during a pressure change from p to p 0 .
The p–v [Fig. 16(a)] and T–s [Fig. 16(b)] graphs clarify the significance of
the two exergy components.
Examining the expressions for the components eT and ep of exergy, we
see that the thermal fluid can be
brought from state 1 to equilibrium with the environment in the following stages:
• Isobaric cooling 1–γ at pressure p 1 from temperature T1 to T0

• Isothermal expansion γ–0 at temperature T0 from pressure p 1 to p 0 .
----------------------- Page 47-----------------------
Noting that [Fig. 16(a)]
and taking into consideration the graphical significance of the exergy e1, we have
In the T–s graph [Fig. 16(b)] one may notice that
Considering that
where k is the adiabatic exponent of the gas, and representing graphically the
dimensionless values
the curves shown in Fig. 17 are obtained. It is to be noticed that unlike the
component eT , which is always
positive [Fig. 17(a)] irrespective of the temperature range (T >T or T <T
), the component e is positive in the
0
0 p
range of pressures p >p 0 and negative in the range p <p 0 [Fig. 17(b)].
Exergetic Method of Analysis of Processes within Pneumatic Networks.
Exergetic analysis that in-
cludes the first and second laws of thermodynamics allows quantitative definition of
the qualitative term energy
degradation. Consequences of various thermodynamically irreversible phenomena can
thus be calculated, lead-
ing to correct understanding of thermodynamic losses within a system. Exergetic
analysis contributes to the
definition of a thermodynamic efficiency, expressing the degree of perfection
or thermodynamic quality of a
system. The concept of exergy stands at the basis of this type of analysis: a
nonconserved quantity with the
dimensions of energy but with unlimited transformation capacity. In the case where
the system is made up of a
flowing fluid, the exergy of the system is equal to the difference between the
initial value of the Gibbs function
(free enthalpy) of the system and the value of the Gibbs function of the system
when it cannot generate useful
work by interaction with thermal reservoir, that is, when the system is in thermal
and mechanical equilibrium
with the reservoir (3).
In the particular case of compressed air flow through the sections
of a pneumatic network, the main
causes of internal and external irreversibility are:
(1) For internal irreversibility:
• Nonstatic character of expansions

• Internal friction between various parts of the fluid and between fluid and
duct walls; friction and shocks
due to eddies;
----------------------- Page 48-----------------------
COMPRESSED AIR
ENERGY STORAGE 19
Fig. 17. Graphical representation of the dimensionless

quantities eT and ep .
• Moisture condensing and partially obstructing the duct

• Mixing and homogenization of humid air, water, oil, dust, and rust
• Finite pressure and temperature differences generated between various parts
of the fluid
(2) For external irreversibility:
• Heat exchange at finite temperature differences between fluid and environment
• Material transfer at finite pressure differences from fluid to environment

due to flow losses caused by
leakage.
Let Mit be the compressed air mass flow into a section; M et the mass flow out
of the section; M mt the average
mass flow through the section; #M =M et −Mit the leakage; Tac the average
compressed air temperature; p m
the average compressed air pressure; i0 and s0 the specific enthalpy and
specific entropy of the environment,
respectively; im and sm the average specific enthalpy and specific entropy
of compressed air; lfr the specific
frictional work; and qext the heat transfer to the environment. Then we have for
the equation of exergy balance
of a duct section
----------------------- Page 49-----------------------
Inserting the exergy expression Eq. (5) for a flowing fluid in Eq. (9), we have
Let us move the term representing the exergy of the fluid coming out of the
section to the left side of Eq.
(10) and suppose that losses due to leakage take place through adiabatic laminar
flow, kinetic energy variations
being negligible, so that i =i . Then the exergy losses for compressed air
flow through the considered duct
m 0
section will be given by the equation
From Eq. (11), substituting s −s =c ln(T /T )−R ln(p /p ),

replacing l with (1−x )l from Fig. 8, noting
m 0 p ac 0 m 0
fr fr
that T# ac =(T +T#2 )/2, and recognizing the mechanical component of exergy
dissipation by friction, we obtain
2
the following relationship for the exergy losses along the analyzed section:
From the terms giving the exergy loss due to irreversible transfer thermal
between the system and the
thermal source,
one may notice that if all values entering the expression are assumed constant,
then the exergy loss #Eq is
the same irrespective of the evolution of the system to T >T or T
<T —an explicable fact, since the return
ac 0 ac
0
of the system to thermal equilibrium with the thermal source, from temperatures
equally spaced above and
below the temperature of the source, implies the same loss of exergy. The terms on
the right side of Eq. (12)
represent exergy losses due to irreversible phenomena accompanying compressed air
flow along the sections
of the pneumatic mining network. The analysis of Eq. (12) may lead to a series of
technical and organizational
measures to reduce exergy losses to a technical and economic minimum.
The exergy losses due to leakage in pneumatic networks,
represent the largest share (65% to 75%) of the total losses. To reduce
them, the following measures are
advisable:
----------------------- Page 50-----------------------
COMPRESSED
• Continual monitoring and elimination of leaks

• Providing a compressed air temperature (Tac ) along the section as close as
possible to the ambient temper-
ature (T ), on condition that T >T
0 ac 0
• Providing a compressed air pressure along the section as close as possible to
the ambient pressure, consis-
tent with the pressures required by the consumers, as well as with the upper
limit on #Ep .
The exergy losses induced by the irreversibility of thermal transfer between

compressed air and environ-
ment are given by
They may be reduced by the following measures:
• Reducing the lateral area of the section ( l ) (also useful in lowering

investment costs and reducing #EM ,
but leading to an the increase in #Ep , so that a compromise is needed
regarding the optimum diameter)
• Reducing the global thermal transfer coefficient K
• Providing a compressed air temperature along the section as close as possible
to ambient temperature.
The energy losses due to flows with friction,
may be reduced by:
• Providing an as high as possible pressure for the compressed air along the
section
• Reducing pressure losses due to friction (#p ) by increasing the duct diameter
• Reducing pressure losses resulting from partial obstruction of the duct (#p
obt ) by removing humidity from
the compressed air and cleaning the insides of the ducts when they are mounted
• Conveying compressed air along the section at the highest possible
temperature, thus cutting down exergy
losses corresponding to work done against friction. (not advisable for
pneumatic mining networks, as it
leads to considerable increase of other loss categories)
The exergetic analysis is completed with a synthetic indicator suitable for

characterizing the quality of
a thermal installation in the sense of thermodynamic energetics—exergetic
efficiency. Exergetic efficiency is
a measure of the degree of irreversibility of the transformations taking place in a
thermal installation, being
quantitatively expressed as the ratio of useful work (useful exergetic effect) to
the exergy difference between
entrance and exit of the installation (available exergy):
----------------------- Page 51-----------------------

In the particular case of pneumatic mining network sections, since
compressed air does not generate
useful work while in them, exergetic efficiency is expressed by the relationship
Taking into account the relationship in Eq. (12), the expression for the
exergetic efficiency for a pneumatic
network section becomes
Exergetic efficiency represents the effectiveness of a thermal process. As it

refers to the thermodynamic
potential corresponding to the state of the environment, it constitutes measure of
actual process imperfections.
Study of Storage-Tank Filling Process. For an exergetic approach

to the filling and emptying of storage
tanks we use a method perfected by Bejan (5) and Radcenko (4). We model the storage
tank as a cylinder with
a piston and a valve. The diagram in Figure 18 shows the complex process of filling
the cylinder. When the
intake valve opens, a residual air quantity me is in the cylinder, taking up a
dead volume Vc at parameters p c
and T [Fig. 18(a)]. The mass of air entering the cylinder during admission, m
=m −m , occupies the volume
c
ad a c
V 1 in the intake duct at parameters p 1 and T1; it is separated from the rest of
the air by an imaginary piston
under pressure p 1. At the end of the filling stage at V =const, the cylinder
contains the mass of air ma >mc at
parameters p a =p 1 −#p a and Ta >Tc [Fig. 18(b)]. When the intake valve is
closed, the mass of air in the cylinder
becomes ma , occupying volume Va at parameters p a and Ta [Fig.
18(c)].
Let eqa be the exergy exchanged by the air in the filling process, ea the
exergy when filling is over, e1 the
exergy when admission starts, lta the mechanical work of the isochoric
filling stage, and πira the exergy loss
due to the internal irreversibility of the filling process. Then the following
exergetic balance equation of the
filling process is obtained:
where
Denoting
----------------------- Page 52-----------------------
COMPRESSED AIR
ENERGY STORAGE 23
Fig. 18. Description of the tank filling

process.
it is seen that
is the exergy loss due to laminar air flow in the cross section of the intake valve.
Therefore
Denoting by
----------------------- Page 53-----------------------
Fig. 19. Description of the tank

evacuation process.
the exergy losses induced only by finite pressure and temperature differences during
the filling process, we
conclude that
Study of Storage-Tank Evacuation Process. Figure 19

shows a diagram of the evacuation process.
When the release valve opens, the mass of gas md in the cylinder occupies the
total volume Vd at parameters
p d and Td [Fig. 19(a)]. When evacuation is over, V =const, and the air
mass md −m#d leaving the cylinder
occupies a volume V#2 at parameters p 2 , T2 in the exhaust duct, being separated
from the rest of the exhausted
air by an imaginary piston under a counterpressure p 2 [Fig.

19(b)]. When the exhaust valve is closed, the
residual air mass m in the cylinder occupies the volume V at
parameters p and T . The evacuated air mass
r r
r r
m =m −m , equal to the admitted mass m , occupies a volume V
in the exhaust duct at parametersp and T
ev d r ad
2 2 2
[Fig. 19(c)].
Let eqe be the exergy exchanged by the air in the evacuation process, e2
the exergy when evacuation is
over, ed the exergy when evacuation starts, lt2 the mechanical work of
evacuation, lTod the isothermal work
of expansion, and πire the exergy loss due to internal irreversibility of
evacuation. Then the exergetic balance
----------------------- Page 54-----------------------
COMPRESSED AIR
ENERGY STORAGE 25
equation for the evacuation process is
where
We denote
In the last relationship, the second term gives the exergy loss due to
laminar gas flow in the section of the
release valve:
Consequently, observing that eq0e <0 and denoting

it results that
Denoting by
the exergy losses induced by finite pressure and temperature differences

characterizing the evacuation process,
we have
Combining with the above balances the exergetic balances of other components
in the energy conversion
and storage scheme, a thermal–economic calculation of the performance of the
energy-conversion and -storage
installation may be made.
----------------------- Page 55-----------------------
Determination of the Accumulation Capacity of the Transport Section in

Pneumatic Networks.
An efficient energy storage system using compressed air is the pneumatic network .
Increasing the accumulation
capacities of its transport section provides temporary energy storage and
availability to cover consumption
peaks. In order to calculate the accumulation capacity of a transport
section, the pressure and mass-flow
variation laws for it have to be known. For this purpose the balance
equation for nonstationary fluid flow
through long ducts will be applied, and will be solved with the help of an
approximate method suggested by I.
ˆ
A. Ciarnıi (6).
The momentum conservation equation is written in the form
where
w =
flow velocity
τ =
time
x =
spatial coordinate
g =
gravitational acceleration
ρ =
density
p =
pressure
λ =
fluid-dynamic resistance ratio
D = duct internal diameter;
δ= boundary-layer thickness of flow section
d= D −2δ. The fluid load loss was approximated with the loss corresponding
to the steady flow
expressed by the Weissbach–Darcy relation.
For the continuity equation we use
From Eqs. (32) and (33) we have
2
On neglecting the terms due to kinetic energy variation [∂(ρw )/∂x]
and gravitational forces (ρgx ) in
comparison with the term due to friction [ρ(γ/d) w2/2], Eq. (34) becomes
Equation (35), together with the continuity equation in Eq. (33) and the
state equation ρ=ρ (p ), forms a
determinate system of equations.
----------------------- Page 56-----------------------
Considering the definition of the isothermal compressibility ratio and

Newton’s relationship for the sound
propagation velocity in fluids, we obtain for barotropic fluids
The equation of continuity will be written as
We linearize the nonlinear equation in Eq. (35) by approximating the

expression λw/2d with its average
ˆ
value 2c=(λw/2d)m , and we use the relationship suggested by I. A.
Ciarnıi (6), assuming a constant fluid-
dynamic resistance coefficient λ in the expression for c, so that 2c=(λ/3d)
(winitial +wend ). Then the equation
becomes
For long ducts at times τ>L/a the term ∂(ρw)/∂τ may be neglected, and from
Eqs. (36) and (37) the following
system is obtained:
Eliminating the product ρw from the obtained equation, differentiating Eq.

(38) with respect to x , substi-
tuting the result in Eq. (39), and introducing the notation b=a2/2c, we obtain
The same technique is used to remove the pressure from Eqs. (38) and (39).
Differentiating Eq. (38) with
respect to time and Eq. (39) with respect to x , and equating them, we obtain
Solving Eqs. (40) and (41) allows the pressure and mass-flow variation in
nonstationary fluid flow through
a long duct to be determined.
----------------------- Page 57-----------------------
The pressure variation law can be obtained by applying the Fourier

transformation (7):
For the mass-flow variation law a relationship having the same form is found:
with the conditions
The functions f (x ), g (x ), F (τ), F (τ), G (τ), and G (τ) are

calculated using regressions based on sets of
1 2 1 2
experimental data.
The pressure and mass-flow values at different points of a pneumatic network
are calculated with Eqs. (42)
and (43) using numerical methods of integration. We use the following notation for
experimentally determined
pressures: p s =adjusted pressure at buffer reservoir discharge valve p n
=compressed air pressure behind buffer
reservoir in normal functioning regime (the flow supplied by the compressor meets
the requirements of the con-
sumers)p d = compressed air pressure behind buffer reservoir when the flow
supplied by the compressor is less
than the flow required by consumers at the downstream end of the transport sections
p cs =compressed air pres-
sure corresponding to pressure p s p cn =compressed air pressure corresponding to
pressure p n p cd =compressed
air pressure corresponding to pressure p d
When the conditions
are met, the trunk line operates in accumulation regime. When
----------------------- Page 58-----------------------
COMPRESSED AIR
ENERGY STORAGE 29
the trunk line compensates for the extra flow rate required by the consumers. When
the flow required by the consumers exceeds the flow supplied by the compressor and a
second compressor must
be coupled to the network to provide the required supplementary flow. The
accumulated flow, compensated
flow, and additional flow required are determined by Eq. (43) from the
differences #M =M (0, τ)−M (L , τ)
corresponding to each operating regime of the trunk line.
Possibilities for Compressed Air Storage with the Help of a
Collector–Distributor Section.
Pneumatic equipment is supplied with compressed air from a pneumatic distribution
network at points rel-
atively close to one another. Nevertheless, the distribution network is
connected to two trunk lines coming
from different compressor stations, to provide better service to pneumatic
consumers. For large systems it is
advisable to transform the pneumatic network from a branched network into an
annular one, by developing an
annular collector–distributor made up of one or several loops resulting from
parallel connection of distribution
ducts.
The collector–distributor will be supplied at the two ends of the
arragement. Flexible supply is thus
assured, leveling the load curve, consumption peaks being taken over by the
collector–distributor using com-
pressed air accumulated during low consumption periods. Supplying consumers with
required flows at neces-
sary pressures by this means involves certain restrictions on the pressure
variations in the collector–distributor
due to uneven consumption.
At the branching points of the consumers’ supply ducts to the collector–
distributor, the pressure must
be higher than a value p 2min determined by the pressure required for
pneumatic equipment and the losses
accompanying compressed air flow along the distribution ducts. Collector–distributor
supply points will be at
a lower pressure than the value p 1max set at the buffer reservoir discharge valve
in the ecompressor station.
The compressed air required for pneumatic equipment loads must be supplied
over a 24 h cycle by meeting
the following conditions:
where
p 1(L , τ) = compressed air pressure variation at the branching points of
the collector–distributor from
trunk lines
p 2 = compressed air pressure variation at consumer–distributor consumption
points
Two types of issues are raised in practice:
• Calculation of the needed geometrical size of the collector–distributor to

cover variations in consumption
• Determination of time variations of flow and pressure at the supply
points of the collector–distributor
as functions of their variations at the consumption points, for certain
geometrical sizes of the collector–
distributor.
The required volume of the collector–distributor results from the condition

of compensating momentary
flow deficiencies during a 24 h period:
----------------------- Page 59-----------------------
where
3
Q =Q +#Q −Q = flow deficit during the time τ (N·m /min)
di ci i sci di
3
Qci = flow required by consumers during the time τdi (N·m /min)
3
Qsci = flow conveyed to compressor station during the time τdi (N·m /min)
τdi = life span of flow deficit Qdi (min)
p 1, p 2 , p N = absolute pressures of compressed air before and after
filling up the collector–distributor,
and pressure corresponding to the normal physical state (bars)
T1, T2 , TN = absolute temperatures of compressed air before and after
filling up the
collector–distributor, and absolute temperature corresponding to the normal
physical state (K)
The time taken to fill up the collector–distributor is calculated with the

formula:
3
where Q =Q −(#Q+Q ) is the available flow for filling (N·m /min).
u sc c
The flow and pressure variations in time in the collector–distributor supply
are determined by Eqs. (42)
and (43) when their variations at the consumption points are known.
Turning the collector–distributor into a pneumatic-energy “flywheel” provides
the following benefits in
the operation of pneumatic installations:
• It facilitates adjusting the pressure and flow of compressed air conveyed to

consumers.
• It allows pneumatic energy storage so that energy can be supplied at the right
moment;
• It levels the pneumatic installation load curve, providing for
consumption peaks, so that no additional
compressor has to be coupled to the network.
• Compressor stations may be sized to required average loads.
• It provides a stationary working regime for compressor stations.
• It increases the fluid stability of the pneumatic network.
• It represents a pneumatic energy supply in case of emergency at the compressor
stations.
• It removes humidity from the compressed air.
Conclusion
The evolution of mankind is currently focused on harmonization with nature.

Sustainable development will
thus be a goal, especially in the energy sector, and energy costs will include
externalities as well. In view of
ecological restrictions and externality costs, some primary energy sources
will become very expansive, and
wind and solar energy will gain ground. Except for local utilization of wind or
solar energy, considering the
abundance, degree of dispersion, and intermittence of these sources, energy storage
solutions must be found.
The favorable characteristics of storage systems based on compressed air and
the necessity of providing
autonomy in time for wind and solar energy conversion justify efforts at
development and proliferation of
energy storage by compressed air.
Because compressed air is an medium of energy transmission that is capable of
simultaneously accumu-
lating thermal and mechanical energy, a possible combination of wind and solar
systems might be developed,
using compressed air both as an energy transfer medium and as an energy storage
medium.
----------------------- Page 60-----------------------
BIBLIOGRAPHY
1. A. Bacu A. A. Bacu Indrumator pentru Utilizarea Aerului Comprimat in Minerit

(Guide to Using Compressed Air in
Mining Engineering) , Bucharest: Technica, 1972.
2. P. Zaugg Z. S. Stys H. Hoffeins Air storage power plants with special
consideration of USA–conditions, presented at 11th
¨
World Conf. of Energy, Munchen, 1980.
3. A. Bejan Advanced Engineering Thermodynamics , New York: Wiley, 1988.
4. V. Radcenco Criterii de Optimizare a Proceselor Termice (Optimization Criteria
of Thermal Processes), Bucharest: Tehnica,
1977.
5. A. Bejan Method of entropy generation minimization, or modeling and
optimization based on combined heat transfer
and thermodynamics, Rev. Gén. Therm., 35: 637–646, 1996.
6. H. Pascal Curgerea Nestat¸ionar˘a a Gazelor prin Conducte Magistrale
(Unsteady Gas Flow through Central Ducts),
Bucharest: Academy Publishing House, 1967.
7. I. I. Irimie I. Matei Gazodinamica Ret¸elelor Pneumatice (The Gasdynamics of
Pneumatic Networks), Bucharest: Tehnica
Publishing House, 1994.
READING LIST
B. Andersen Finite-time termodynamics and thermo-dynamic length, Rev. Gen. Therm. ,

35: 647–650, 1996.
¨ ¨
S. Balkan Warmespeicherantriebe fur einen Schadstoffreien
Fahrzeugbetrieb, Dissertation an der TU Berlin,
1974.
N. Bradeanu Instalatii Pneumatice Miniere (Mining Pneumatic Installations),
Bucharest: Tehnica, 1976.
H. Bruntland Our Common Future, New York: Oxford University Press, 1987.
´
P. Chesse J. F. Hetet J. P. Frayret Determination par simulation d’une zone
critique des champs compresseur situee entre
´
le pompage et la stabilite, Entropie , 201: 9–11, 1996.
D. Coiante Uno Sguardo al futuro dell’Energia, Studi e Ricerche, ENEA, Italy: 1996.
Al. Danescu (ed.), Termotehnica si Masini Termice (Thermotechnics and Heat

Engines), Bucharest: Didactic and Pedagogic
K. Euler Methods d’Accumulation de l’Energie Electrique , Francfort sur le Main:
Revue VARTA AG, 1984.
V. I. Evenko An exergetic appraisal of the thermodynamic perfection of
compressors, Thermal Eng., 44 (3): 233–238,
1997.
C. Flavin N. Lensen Power Surge , London: Norton, 1994.
S. Frandsen On the wind speed reduction in the center of large clusters of wind
turbines, presented at EWEC’ 91, Amsterdam.
P. Glansdorff I. Prigogine Structure, Stabilité et Fluctuation, Paris: Masson,
1971.
A. Houberechts Thermodynamique Technique, Paris: Dunod, 1962.
V. Ilie Utilizarea Energiei Vântului (Utilization of Wind Energy), Bucharest:
Tehnica, 1984.
I. I. Irimie S. Irimie I. C. Tulbure Solutions for turning to account the thermal
energy of compressed air, Study, University
of Petros¸ani, 1994.
M. F. Jischa Herausforderung Zukunft , Heildelberg: Spektrum Akademischer Verlag,
1993.
A. R. I. Kanal F. B. L. Janaina M. A. Zanardi Models of liquid storage tanks,
Energy , 22 (8): 805–815, 1997.
D. Kanellopoulos Cluster efficiency model. An application to the seven wind farms of
the HORS–QUOTA programme, Proc.
1991 Thirteenth BWEA Wind Energy Conference, 1991.
A. Kodra B. Fortunato A. Dadone Prestazioni e ottimizzazione di centrali
eoliche, Energia Mater. Prime , 121: 10–15,
1996.
A. Leca Principles of Energy Management , Bucharest: Tehnica, 1997.
T. Oroveanu (ed.) Colectarea, Transportul, Depozitarea si¸ Distribut¸ia
Gazelor (Collection, Transport, Storage, and Distribu-
tion of Gases), Bucharest: Didactic and Pedagogic Publishing House, 1985.
¨
J. Schroder K. Gawron Latent heat storage Energy Res. , 5: 103–109, 1981.
T. Svenson et al. Environmental Consequences of New Energy
Technology , Swedish Council for Building Research,
1991.
----------------------- Page 61-----------------------
I. C. Tulbure Zustandsbeschreibung und Dynamik Umweltrelevanter Systeme ,

Clausthal–Zellerfeld: CUTEC-Schriftenreihe,
No. 25, 1997.
´
P. Zaugg Z. S. Stys Centrales avec reservoir d’air, specialement pour les
conditions US, Rev. Brown Boveri 12 , 67: 723–733,
1980.
IOAN-IULIAN IRIMIE
SABINA IRIMIE
ILDIKO-CAMELIA TULBURE
University of Petrosani
----------------------- Page 62-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
●
HELP

Your Librarian
Profile
DC Machines

Standard Article

Andreas J. F. Binder 1
1Darmstadt University of Technology, Darmstadt, Germany

reserved.
Search
DOI: 10.1002/047134608X.W3001
❍ Acronym Finder
Abstract
Basic Description of DC Machines
Basic Equations of DC Machines
Motor–Generator Sets
Solid-State Drives
Control
Armature Coils
Temperature Control
Time-Varying Loads
Limits of DC Machines

%20Conversion/W3001.htm17.06.2008 16:35:29
----------------------- Page 63-----------------------
DC MACHINES 63
within
slots along the circumference of the cylindrical rotor
iron
core near the surface; its coil ends are connected to the
mechanical commutator, which consists of copper segments

that
are separated by electrically insulating thin mica sheets.
The
rotor, with its bearings mounted into bearing end shields,
is
rotating within the stator, which consists of several mag-
netic
pole pairs (at least one pole pair is necessary). The sta-
tionary
magnetic field of these poles in the airgap between
stator
and rotor induces an ac voltage (back emf) into the ro-
tating
armature winding according to Faraday’s law (1,2).
Carbon
brushes, sliding on the commutator cylindrical sur-
face,
along with the commutator itself rectify the ac voltage,
thus
creating a dc voltage between two brushes of opposite
electric potential. All positive brushes are connected to the
positive terminal, and all negative brushes to the negative
terminal likewise. If, for example, an ohmic resistor is con-

nected
to these terminals, a dc current will flow, and electric
power
will be dissipated as heat in the ohmic resistor. The
machine
is acting as a generator and needs mechanical input
power
to prolong the described power conversion. Inside the
armature winding the current flow is—like the induced ac

voltage
—an ac current flow, which, along with the previously
noted
magnetic air-gap field, generates a constant braking
torque
according to Lorentz’s law (1,2). Thus the mechanical
input
torque at the shaft, produced, for example, by a driving
diesel
engine, has to overcome the electromagnetic braking
torque
in order to generate energy flow from the diesel engine
into
the resistor.
By
reversing the direction of power flow by reversing the
direction of dc current flow outside the dc machine, one

gets
motor operation. For example, an external battery sup-
ply
with a certain dc voltage sufficiently larger than the
induced
back emf to overcome the unavoidable voltage drop
of
internal end external electrical resistances will feed a dc
current
of opposite direction into the dc machine. The gen-
erated
electromagnetic torque therefore also changes its di-
rection
and accelerates the rotor instead of the previously
mentioned braking. Thus, the dc machine is able to drive

other
mechanically coupled machines such as pumps or ele-
vators.
So, both motor and generator operation are two
modes
of the same machine, depending on the direction of
power
flow.
The
sliding brush–commutator mechanical contact (3) and
DC MACHINES the
mechanically sensitive commutator itself put electrical
and
speed limitations to dc machines. Furthermore, brush
Direct-current machines (dc machines) are used both as gen- wear
under normal operation conditions, along with carbon
erators and motors. In the generator mode, mechanical input dust
production, requires maintenance such as changing the
power at the rotating shaft is converted into electric output brushes
that are too short and cleaning the machines to avoid
power at the electric terminals of the machine with an effi-
electrical flashover within the windings due to the electrically
ciency less than unity due to inevitable losses in the machine.
conducting carbon dust layers. Therefore dc generators are
Electrical output quantities are current and voltage. Both of
currently replaced in many applications by ac synchronous
them are dc values and may be used to feed electrical devices.
generators with silicon rectifiers such as the electric genera-
In the motor mode, the energy flow is reversed: Electrical dc tors in
cars or rotating dc exciter units for large turbo genera-
input power at the terminals is converted into mechanical tors in
power plants (so-called ‘‘brushless’’ dc generators). As
output power at the rotating shaft, thus enabling the machine speed
and torque of dc machines can be influenced separately
to drive a pump, an elevator, or a similar mechanical appara- by
changing the applied voltage, thus changing the speed, or
tus, again with an efficiency of less than unity. by
changing the armature current, thus changing the torque,
these
machines are preferred as variable-speed drives. The
BASIC DESCRIPTION OF DC MACHINES control
methods are simple, the produced torque is smooth,
and
dynamical response is very quick, as the electromagnetic
These dc machines consist of a laminated iron core, fixed on time
constant of the armature circuit is low. For many years
a rotating shaft. An electrical armature winding is placed these
machines have been used as variable-speed drives in

----------------------- Page 64-----------------------
64 DC MACHINES
numerous industrial applications such as steel mills, paper

are necessary, whereas with ac synchronous drives there is
production, centrifugal drives, elevator drives, cranes, trac-
only one machine needed. Still there is ample application of
tion motors for electric railways, and motor–generator sets
dc drives now, mainly in the power range of 100 kW up to
in submarines. These machines are both used in motor and
1000 kW due to the previously noted benefits. Their technical
generator mode, feeding back electric power to the grid during
performance has steadily improved with benefits such as ex-
regenerative braking. Variable armature and field voltage is
tended brush life, improved winding insulation, optimized
supplied by silicon-controlled rectifiers (thyristor bridges),
cooling for a more compact build, and the use of on-line ma-
thus enabling the machines to be operated also with weak-
chine diagnostics to reduce maintenance (4,5).
ened magnetic field to extend the speed range when the arma-
ture voltage limit is already reached. These power converter
units are simple and therefore cheap, resulting in rather low
acquisition costs for the combined machine–converter–control
BASIC EQUATIONS OF DC MACHINES
unit in spite of the fact that the machine itself is rather ex-
pensive due to the mechanically complicated rotor structure.
In Fig. 1(a) a simplified cross section of a two-pole dc machine
Nevertheless, the decrease of costs for ac inverters, their im-
is depicted (number of pole pairs p # 1). This simple machine
proved switching performance due to improved power elec-
has only one armature coil lying in two opposite slots at diam-
tronic devices such as insulated gate bipolar transistors
eter d, with its ends connected to a simplified commutator
(IGBT), and the advent of improved motor control for ac in-
with two copper segments. The machine—with its rotor turn-
duction machines (so-called vector control) due to improved
ing with rotational speed n— acts as a dc generator at no-load
microcomputer technology led to a change from ac to dc
condition, as there is no current flow in the armature, as no
drives. The induction motor is more robust, having only a
load is connected to the terminals. The north and the south
squirrel cage in the rotor instead of the mechanical commuta-
pole air-gap magnetic field with peak flux density B induces
tor and the wound armature winding, thus aiming for higher
an ac voltage (emf) ui into the armature coil due to the move-
speed limits and resulting in higher dynamic performance
ment of the coil sides with air-gap velocity v (length l of each
due to the lower moment of inertia, at reduced maintenance
coil side corresponding to the iron stack length and number
costs (1). In the high-power region above 1 MW inverter-fed
of turns per coil Nc) [Fig. 1(b)]:
synchronous machines are also replacing large dc machines,
as they allow a more compact build due to the lacking commu-
v = dπn (1)
tator. Larger ratios output power versus motor volume are
achieved and higher mechanical speed is reached. For big dc
drives, for example, two coupled machines (tandem motors)
The rectified ac voltage u yields a dc voltage U , which for the
i i
N Blv
N
c
B
Main pole Armature l d/2
(“north”) coil
ui
Rotor
Slot n
Nc Blv
Brush
Armature
B
coil Brush
Commutator Ui
N
Main pole
(“south”)
(a)
(b)
Figure 1. Simplified dc generator under

no-load conditions with one pole pair, one e Ui
g
armature coil, two commutator segments, a
t
l
and two brushes. (a) Cross section and o
e
V
g
air-gap field B . (b) Induced ac voltage ui u
a
i
t
l K
and rectification by the commutator U . (c)
o 2N Blv
i 0
V 1 2 c
The nonuniform air-gap flux distribution 0 T 2N
Blv T Time t TC = nK
causes a dc terminal voltage U with con- 2 c
i
0
siderable ripple. (d) The voltage ripple ui
0 Time t
gets smaller if more than one armature
coil is connected in series (K/2 # 1). (c)
(d)
----------------------- Page 65-----------------------
DC MACHINES 65
U
U
Uf
Ia Ia
If
If
A A
F
C C
(a) (b)
U U
I = I
f a
If
A F A
F
Ia Figure 2. Different possibilities of connecting ar-

mature (A), compole (C), and field winding (F) to
C C
the dc grid. (a) Separate excitation. (b) Shunt exci-
tation. (c) Series excitation. (d) Compound exci-

(c) (d)
tation.
‘‘one-coil machine’’ shows a considerable ripple with armature

The equivalent pole arc ratio #e (typically 0.7) considers the
frequency f a [Fig. 2(c)]:
decrease of the air-gap field in the neutral zone,
U = 2N Blv (2)
dπ
i c
# = αe lB (6)
1
2p
f a = np = (3)
T
In
the same way the expression for the electromagnetic
torque M is derived when a current Ia (dc value outside the

In order to get a smooth dc voltage, usually many series
machine current) flows in the armature:

connected coils are used instead of only one. These K
coils—being placed in Q slots as a two-layer winding, re-
k
sulting in u # K/ Q coils per slot and layer—are connected
M = Ia # (7)
2π
with both ends to K commutator segments, resulting in two
parallel paths of K/2 series connected coils between the
The terminal voltage U has to balance the emf and the volt-
plus and minus brush, when a two-pole machine is consid-
age drop due to the armature resistance Ra and the brush
ered. The induced ac voltages of adjacent armature coils
voltage drop U , which typically is about 2 V for plus and
b
are phase-shifted according to the displacement of one com-
minus brush, nearly independent of load current. Depending
mutator pitch, yielding a superimposed ac ripple on the on
machine operation mode (motor or generator) and thus po-
rectified voltage with commutator frequency f C [Fig. 1(d)].
larity of armature current (positive or negative) one gets two
This ripple of typically 1% to 2% is lower the higher the
expressions for the voltage equation (6,7).
ratio K/(2p ) is chosen (typically 30 to 40), yielding a nearly
Motor operation:
ideal dc voltage
U = U + I R + U (8)
i a a b
1
f C = Kn = (4)
T
C
Generator operation:
Usually dc machines are manufactured with two or more pole

U = U − I R − U (9)
i a a b
pairs; thus the number of parallel paths 2a increases likewise

when lap windings are chosen, Eq. (28). So one derives for the
induced dc voltage equation (5), which depends on rotary
MOTOR–GENERATOR SETS
speed n and the magnetic flux per pole
,
Different possibilities of connecting the armature winding (A)

U = 2KNc p n# = kn# (5)
and the stator field winding (F), which is exciting the electro-
i
a
magnetic poles via the field current If , to the dc grid are possi-
----------------------- Page 66-----------------------
66 DC MACHINES
by the loaded armature winding. At the main-pole tips the
main flux density is either strengthened or weakened by
this so-called armature reaction field. If there is no iron

U , I = I const
i a a1
saturation, the average main-pole flux would remain un-

U , I = 0
0 a
changed. With a nonlinear iron characteristic there is addi-
tional load-dependent flux saturation in the teeth beneath

e Ia1Ra + Ub
the pole tips, where main flux and armature reaction flux
g
a
t
are added with the same polarity. Therefore under load the
l
o
V
resulting flux per pole is lower at high armature current
U
values, yielding a lower induced voltage at load U than at
no-load U (Fig. 3). This armature reaction effect can be
eliminated by using a compensating winding, placed in slots

UR
in the main-pole shoes opposite the armature rotor slots
0
(Fig. 12).
Field current If
0
Generator Characteristics
Figure 3. No-load characteristic of a dc generator for fixed rotational
speed n: no-load voltage U , residual voltage U , induced voltage U
An external torque (generated by a turbine, a diesel engine,
0 R i
under load with influence of an armature reaction, and terminal volt-

an electric motor, etc.) is driving the dc generator with fixed
age U for variable field current I and fixed armature current I .
f a
rotational speed n.
The load characteristic of the separately excited genera-
tor [Fig. 4(a)] is given for fixed field current, yielding a no-
load voltage U , Eq. (5). According to Eq. (9) the terminal

ble and in practical use (Fig. 2). Auxiliary windings (C) such
0
voltage U drops with rising load current Ia slightly due to

as compole and compensating windings (see the section enti-
the previously mentioned ohmic voltage drop. If the arma-

tled ‘‘Armature Coils’’) are always connected in series to the
ture reaction is not compensated, an additional voltage drop

armature winding. For clarification, in Fig. 2 no field rheo-
occurs at high load values due to load-dependent iron satu-

stats or field shunts in the electric field circuit or starting
ration.
resistors in the armature circuit are depicted. The field resis-
The decline of the load characteristic of the shunt genera-

tors are necessary to change the field current under fixed volt-
tor [Fig. 4(b)] is larger than that of the separately excited

age-supply conditions, whereas the starting resistors are nec-
generator, as the decreasing terminal voltage U is at the

essary to limit the armature current at motor standstill and
during motor starting.
same time also the decreasing voltage at the field windings
One distinguishes four different main-pole excitation
terminals, which leads to decreasing field current with in-
methods both in motor and generator operation:
creasing load. Below a certain value of field excitation the iron
parts are unsaturated, giving two terminal voltage values for
one certain armature current value. Thus a maximum current

1. Separate excitation [Fig. 2(a)]
output exists.
2. Shunt excitation [Fig. 2(b)]
From Fig. 3 one easily derives the load characteristic of
3. Series excitation [Fig. 2(c)]

the series generator [Fig. 4(c)], as armature and field current
4. Compound excitation [Fig. 2(d)], which is a combination
are identical. As the armature current is usually much larger
of methods 2 and 3
than the field current of the separately excited generator, the
series-connected field winding consists of few turns Ns,

Equations (4) and (5) yield the corresponding external volt-
whereas with separate excitation this number Nf is large in
order to reach the same magnetomotive force (mmf) I N #

age–current characteristics of the generator mode and the
a s
speed–torque characteristics of the motor mode for the four
IfNf with a low field current. Practical use of this type of gen-
different excitation modes. As a basic understanding of all
erator is only under regenerative braking conditions of series
these characteristics one has to consider first the open-
motors, typically in electric vehicles and railway dc locomo-
circuit curve (I # 0), which gives the no-load voltage U of
tives.
a 0
a separately excited generator for fixed rotational speed n
By combining both series and shunt excitation an almost
(Fig. 3). With increasing field current If the main flux per
load-independent output voltage between no-load and rated
pole
and thus the emf U increases, too, but due to load
can be achieved. Therefore this type of generator—along
i
saturation of the iron parts (mainly the rotor teeth) beyond
with the separately excited generator—are in practical use.
an air-gap flux density of about 1 T (corresponding to a
Note that with the shunt generator it is possible to convert
tooth induction of about 1.8 T to 2 T) the curve bends,
mechanical into electrical energy without any external elec-
limiting the magnetic air-gap field. A certain iron hysteresis
tric source. The remanence induction, due to the rotation of
effect, resulting in two slightly different branches for in-
the armature in this air-gap field, induces already at the no-
creasing and decreasing field current, usually can be ne-
load condition a certain residual voltage U (Fig. 3), which
glected. Only the effect of the stator iron remanent flux

drives a small initial field current through the circuit resis-
density has to be considered, yielding a small residual ar-
tance. Thus an electrically excited magnetic flux density
mature voltage U even at the open-circuited excitation cir-
buildup starts, amplifying the armature voltage, which again
R
cuit. With load (Ia # 0) an additional air-gap flux is excited

increases the field current up to a certain value at which the
----------------------- Page 67-----------------------
DC MACHINES 67
U0
U0 I R
U
a a
e
U
g
t
e
l
g Voltage drop
o
a
V
t
l due to armature
o
V I = const reaction
f
0
0
0 Armature current Ia
0 Armature current Ia Iam
(a)
(b)
Voltage drop
due to armature Ui (Ia = 0)
reaction
U (I > 0)
U i a
e
U0
g U
a I R + U
t a a b
U
l
o
e
V
g
UR
I N
0
0
0 Current Ia = If
0 Armature current Ia
(c)
(d)
Figure 4. External characteristics of dc generators

for fixed rotationalspeed n. (a) Separate
excitation. (b) Shunt excitation. (c) Series excitation.
(d) Compound excitation.
voltage drop along the circuit resistance and emf are equal.
smaller shaft torque (typically 1% to 2%). The decrease of
The discovery of this so-called self-excitation (dynamoelectric
speed with increasing torque is rather small [Fig. 5(a)], yield-
principle ) by Werner von Siemens in 1866 was a revolution to ing
a nearly constant speed with an upper value n0 at the no-
electric power generation and was one of the main reasons for load
condition:
the breakthrough of electricity in industrial application. The
same effect is now used for self-excitation of large synchro-
U
n0 = (11)
nous power generators during startup after blackout, and also
k#
for self-excitation of synchronous and induction stand-alone
generators, for example, in wind turbine application or emer- Flux
drop at high torque (or at high armature current),
gency generating sets.
caused by uncompensated armature reaction, will increase
speed with increasing load, which might cause unstable motor

Motor Characteristics
operation for certain external load characteristics such as
loads with speed-independent torque (for example, elevators

In the following the motor terminal dc voltage U is assumed
and
cranes). In these cases either a compensating winding or
to be constant as well as the dc supply voltage U for the field
f
a
motor-speed control is necessary.
circuit. Therefore, there is no basic difference between the
Considering Ia # If and thus

# Ia (constant saturation
separately excited motor and shunt motor. By combining Eqs.
assumed), one gets from Eqs. (5), (7), and (8) the following
(5), (7), and (8), one derives the speed–torque characteristic
steady-state characteristic for the series motor [Fig. 5(b)] with

of both types of motors (brush drop U for simplification ne-
b
a
strongly load-dependent speed decrease and a considerably
glected):
large starting torque:

2πRa
n = n0 − 2 M (10)
U
(k#)
n ∝ √ − const ( 12)
M
For simplification, the electromagnetic torque M is assumed
to be the same as the shaft torque. Losses in the active rotor Thus
series motors are recommended drives for traction, ei-
iron, eddy current losses in the rotor conductors due to the ther
in battery-supplied electric vehicles and dc railway loco-
coil ac current, and inevitable friction losses cause a slightly
motives or in ac railway locomotives as single-phase commu-
----------------------- Page 68-----------------------
68 DC MACHINES
π
2 Ra
2
n
k
d
l
e
a n
e
n U = const
p U = const
o d
s
i
t e
l
a e
a
t p
n
o s
Speed rise due to o
R
i
armature reaction t
R
0
0
0
Torque M 0 Torque M
(a)
(b)
n0
a n
o d
t e
a e
t p
R
Figure 5. Speed–torque characteristics of dc motors
0
for fixed armature voltage U. (a) Separate and shunt
0 Torque M
excitation, fixed field current If . (b) Series excitation.
(c) Compound excitation.
(c)
tator motors, usually with series compensation winding. The is

rectified by six-pulse silicon rectifiers [Fig. 6(a)] to gener-
reduction of field current has to be limited to prevent ate a
dc terminal voltage U. According to Eqs. (5) and (8)
overspeeding with low loads. Series motors are well suited for this
voltage should be variable to get variable speed drives.
being operated in parallel, supplied by one common electric
Depending on the firing angle # (7) one gets a dc voltage
source. For example, each of the four wheel sets in a four- U of
variable amplitude with a superimposed ac voltage
axis locomotive is driven by a separate motor. Slight speed ripple
[Fig. 6(b)]. As a result, the dc armature current also
differences at the different wheel sets due to slip on wet rails shows
a considerable ripple with the frequency 6f (300 Hz
results in only a slight difference in armature current con- or 360
Hz), which can be reduced by an additional series
sumption of the different motors due to the strongly load-de-
armature choke. Further improvements are possible by us-
pendent characteristic, Fig. 5(b). Separately excited motors ing
12-pulse converters (6). The power-converter arrange-
would show large differences in armature current consump- ment
of Fig. 6(a) allows only one direction of current flow,
tion due to the nearly load-independent characteristic, Fig. as the
thyristors are blocking the reverse direction, but
5(a), leading to quick overheating of the motors with large both
positive and negative polarity of voltage U:
current input. Therefore this type of motor is used in railway
traction only with speed and current control.
U = U cos (α) (13)
0
Combining series and shunt excitation again yields a
3
rather strongly load-dependent speed characteristic [Fig.
U0 = 2UL (14)
5(c)], but with the advantage of avoiding overspeeding with
π
low loads. The shunt excitation ensures that speed cannot

surpass the no-load speed n0 .
Rectifying the line voltage without any delay angle (# # 0)
yields
the maximum voltage U . Usually the range for the
thyristor delay angle is limited:
SOLID-STATE DRIVES
30◦ ≤ α ≤ 150◦ (15)
Usually there is no dc grid available to operate dc ma-

chines, as the power supply of today is performed by a The
upper limit 150# is given by the inverter stability limit.
three-phase ac line sinusoidal voltage u (amplitude #2U ) At
180# the inverter would shoot through (9). By defining a
L L
with frequency f (50 Hz or 60 Hz). Before the advent of minimum

angle of 30#, a voltage reserve between 87% and
power electronics the necessary variable dc voltage was gen- 100% of
U , corresponding to angle values between 0# and 30#,
erated by a separately excited dc generator with variable remains

for the speed controller. Thus the rated motor volt-
excitation, driven by an induction motor, which was sup- age is
U # 0.87U . For motor operation in a clockwise rota-
N 0
plied by the ac line (so-called Ward–Leonard system) (1). tion

direction the delay angle is smaller than 90#, yielding a
Nowadays, silicon-controlled rectifiers (usually six-pulse thy-
positive motor terminal voltage. With positive armature cur-
ristor power converters) are used (8,9). The ac line voltage rent
flow direction, positive dc input power Pi
----------------------- Page 69-----------------------
DC MACHINES 69
uL
Fourth quadrant—motor operation, converter 1 active:
n < 0, M > 0, U < 0, Ia > 0, α1 > 90◦
By
changing the delay angle #, the armature voltage U can
be
varied continuously, thus making itpossible to run the
Uf
drive with any selected pair (n,M) under steady-state condi-
U, u
tions within the limits of rated torque MN and maximum
speed n . Rated torque is defined by the rated current I ,
m N
which is a thermal limit to the armature, and the rated field

Ia
current IfN , which is both a thermal limit to the field coils and
a
saturation limit of the iron core. At rated speed nN the maxi-
If mum
armature voltage, called the rated voltage U , is
reached. Beyond this speed a further increase in speed is only

(a)
possible, if—according to Eq. (5)—the main flux
is reduced
by
decreasing the field current If (field-weakening operation,
Fig.
7). The armature current is kept constant, but the torque
decreases with 1/n, Eq. (7). The slope of speed decrease with
a
increasing load is larger with weakened flux than with rated

u
e flux,
Eq. (10).
g
a
Two different operation modes are possible and in use for
t
l
o U ÷ 2
antiparallel converter supply.
V U L
0
1. Both converters are active and produce the same aver-
0 Time t
age output voltage U, but only one of them provides the
1
6f
(b)
Rotational speed n
Figure 6. Converter-fed, separately excited dc machine. (a) Six-pulse
converter for the armature circuit and two-pulse converter for the
field circuit (first-quadrant operation). (b) Average voltage U is de-
nM
creased by increasing the delay angle #.
Φ 2
Second quadrant First quadrant
P = UIa (16)
i
Φ > Φ U = const
1 2
is transferred from the grid to the motor and converted into

nN
mechanical output power Po :
Po = 2π nM (17)
U1
U > U
2 1 Φ = const
With a firing angle # larger than 90# the voltage U becomes
negative and the direction of speed will reverse from clock-
U = 0
wise (cw) to counterclockwise (ccw) direction. As current
M N
and torque are still positive, the dc power becomes nega-
tive, too, thus becoming a power flow from the dc machine
–M N 0 Torque M
back to the grid. The dc machine acts as a regenerative

brake. Motor operation with both cw and ccw speed is only
possible if a second antiparallel controlled rectifier is added
in Fig. 6(a). In that case the following values of speed,
torque, voltage, and current are possible, while the main
flux
is kept constant.
–n
N
First quadrant—motor operation, converter 1 active:
M ~1/n
n > 0, M > 0, U > 0, Ia > 0, α1 < 90◦
Third quadrant Fourth quadrant

Second quadrant—generator operation, converter 2 active:
–n
n > 0, M < 0, U > 0, Ia < 0, α2 > 90◦
m
Figure 7. Speed–torque characteristic of separately excited dc ma-

Third quadrant—motor operation, converter 2 active:
chine, fed with variable armature voltage U and fixed flux
for n #
nN
and with fixed armature voltage U and weakened flux
for nN #
n < 0, M < 0, U < 0, Ia < 0, α2 < 90◦ n #
n .
----------------------- Page 70-----------------------
70 DC MACHINES
i
of current flow (first quadrant operation). By varying the
T a
duty cycle T1/ TP (9) from zero to unity [Fig. 8(b)], the aver-
I
age output voltage is varied between 0 and U (pulse-width
f
0
U0 D
modulation technique):
T
U, u
U = 1 U0 (19)
Thus again a variable-speed drive is realized. The high
switching frequency and the inductance of the armature lead

(a)
to a rather smooth dc current with only a small superimposed
current ripple with the basic frequency f P . The ripple reaches
its maximum peak-to-peak value at T1/ TP # 0.5. During the
time T1 the transistor (T) is active, whereas during T2 #

t
n u
TP # T1 the free-wheeling diode (D) is active. Four-quadrant
e
r
r
operation is possible if a B2 bridge, consisting of four transis-
u
c
d
tors and four free-wheeling diodes, is provided.
n
a
i
e a
g U0
a
CONTROL
t
l T U
o 1
V
A speed-controlled dc drive needs a tachometer or speed sen-

0
0 Time t
sor coupled to the rotating machine to measure the actual
1/f P = TP
speed value (7). The speed controller determines the differ-
ence n* # n between the speed

setpoint value n* and actual
(b)
value n, producing as an output signal the setpoint value I *
Figure 8. Separately excited dc machine with pulse-width-modu-

for the armature current controller (Fig. 9). The actual cur-
lated chopper supply. (a) Chopper circuit for first-quadrant operation.
rent value is usually measured on the ac side by current
(b) By increasing the duty cycle T / T the average voltage U is in-
transformers and a rectifier bridge to get a voltage signal,
1 P
creased.
which corresponds to the dc current value I . In the case of a
controlled silicon rectifier supply, the current controller deter-
mines the necessary firing angle #, depending on the differ-

armature current. Then the firing angles of both con-
ence I * # I . In case of a pulse-width modulated dc chopper

verters must be correlated due to Eq. (13) according to
a a
supply, the current controller determines the duty cycle of the

α2 = 180◦ − α1 (18)
chopper. Thus, the dc machine is supplied with the necessary
armature voltage U, Eq. (13), to adjust the desired speed n*

2. Only one converter is active, supplying U, whereas the
and the armature current Ia*. The structure of control is the
other one is blocked.
so-called cascaded controller with the speed controller acting
on the current controller. The time constant for the change of

Solution 1 has the advantage of very quick changes of direc-
armature current is determined mainly by the time constant
tion in current flow and torque. Note that only the average
of the armature winding Ta, Eq. (23). Change of speed is in-
values of the rectified voltage of both converters are equal,
fluenced mainly by the mechanical time constant Tm , which
whereas the time functions u (t) and u (t) are different. Their
1 2
increases with increasing moment of inertia J of the rotating
difference u (t) # u (t) drives an additional circular current
1 2
flow through both converter bridges, causing additional losses

and temperature rise in the power electronic devices. So, for
big drives, solution 2 is preferred at the expense of reduced
dynamic performance, when change of torque direction is re-
u ~i
quired (5 ms to 10 ms elapsing time for the changeover be-
I a
tween the two converters).
i *
Smaller dc machines (for example, permanent magnet
n* + Speed a – Current α
excited machines) are often fed by transistor choppers, be-
– controller + controller
ing operated with a fixed switching frequency f # 1/T .
P P
i
a
Larger machines in dc railway drives often use gate turn-
off thyristor (GTO) choppers. With GTO choppers for larger
currents only low switching frequencies are possible, typi-
cally 200 Hz to 500 Hz, whereas with transistors (used for
smaller currents) higher switching frequencies of 1 kHz to
I u ~n
f T
3 kHz are typically realized. These choppers operate from

a dc link, which is fed, for example, by a six-pulse rectifier
Figure 9. Speed and current control loop of a converter-fed, sepa-
from the grid and stabilized by a big capacitance, acting as
rately excited dc machine with tacho generator T (output voltage u )
energy storage and smoothing the 300 Hz (360 Hz) voltage

for speed sensing and current transformers with rectifier bridge (out-
ripple. Figure 8(a) shows a simple chopper for one direction
put voltage u ) for dc sensing.
----------------------- Page 71-----------------------
DC MACHINES 71
masses and decreases with increasing magnetic flux, Eq. (24). the
period of the current ripple. This time is needed to calcu-
As T is much shorter than T , the settling time of the current late
the average current value in order to eliminate the influ-
a m
controller is much shorter than that of the speed controller. ence

of the superimposed current ripple on the control perfor-
This is necessary, because for proper function of a cascaded
mance. Therefore the minimum current rise time of the
controller the settling time of the underlying controller must
controller output is 1/6f . With chopper control this limit is
be considerably shorter than that of the master controller. not
given. Drives with pulse-width modulation of sufficient
From Eqs. (5), (7), and (8) one derives the dynamic equations high
switching frequency therefore have a higher dynamic
(20) and (21), considering now also time-varying armature
performance. Thus servo drives for tooling machinery are
voltage u, current i , speed n, and load torque m :
preferably permanent magnet dc machines with chopper con-
a l
trol.
These drives are
now replaced by permanent magnet
La dia + Raia + kn# = u (20)
synchronous machines with chopper control, which is trig-
dt gered
by a rotor position sensor. With these machines the
dn k
pulse-width-modulated inverter along with the rotor position
J2π dt = 2π ia # − ml (21)
sensor may be regarded as the replacement of the mechanical
commutator, and the three-phase stator armature winding of

2 1 1 k# 1
d ia + dia + ia = 1 m + du (22) these
synchronous machines corresponds to the rotor arma-
dt2 T dt T T 2π JL l L dt
a a m a a ture
of the dc machines. The advantage of the synchronous
La servo
drives is the lower rotor moment of inertia due to the
Ta = Ra (23) lack
of the mechanical commutator, yielding better dynamical
performance, as Tm is lower, and the low maintenance due to

JRa the
lack of brushes. Therefore these drives are called brush-
T = (24)
m
# #
k# 2 less
dc drives (7).
2π
The cascaded speed control is usually extended by an emf
controller (field-weakening controller). If the measured speed
Combining Eqs. (20) and (21) yields Eq. (22), where u(t) is the n
surpasses rated speed nN , this controller—by using the no-
manipulated variable of the control circuit, whereas ml(t) de- load
characteristic (Fig. 3) as a numerical table—generates a
notes the disturbance. The response characteristic of a dc ma- new
setpoint value If , which is lower than the rated value
I ,
for the underlying I -current controller. This controller
chine with constant main flux, for example, to an increase of fN
f
load torque Ml may be described by the time constants Ta and
generates the necessary firing angle #f for the controlled recti-
Tm . The response of the armature current to a step of arma- fier
bridge of the field-current circuit [Fig. 6(a)], thus applying
a
voltage U to the field winding, which drives the wanted
ture voltage U at t # 0 is derived from Eq. (20). With ia(t) #
f
I for t # 0 and ia(t) # I
i for t # 0 one gets field current If . The
electromagnetic time constant Tf of the
field
circuit
# #
#U # t #
i(t) = 1 − exp − (25)
L
Ra Ta
T = f (27)
f R
In the same way the increase of speed due to a sudden reduc-

tion in load torque (at t # 0 load step- Ml) is determined by
defined by the inductance Lf and the resistance Rf of the field
Eq. (21), and—considering T # T — one gets for the change
winding, is much larger than the armature time constant T
a m
a
of speed n, if n was the speed for t # 0 and n

n is the due to the big main flux linkage, as the number of turns per
speed for t # 0, field
coil Nf of separately excited dc machines is much larger
than
that of the armature winding. Therefore speed control is
#M T
#n (t) = l m [1 − exp(−t/T )] (26) done
preferably by the armature circuit control due to its bet-
m
2πJ ter
dynamical performance than with the rather slow field
circuit, which is only used to extend the speed range when

With Eqs. (20) to (24) controller design is accomplished (for the
armature voltage limit is reached.
example, PID controller for the speed loop and PI controller
For mathematical modeling of the drive to get an accurate
for the current loop).
parameter design of the controller, not only electromagnetic
Control used to be realized by analog circuits, but is now but
also mechanical properties of the motor–drive set have to
done digitally to benefit from improved possibilities in incor- be
considered in detail. So, if necessary, elastic coupling (10),
porating mathematical sophisticated models for so-called
torsional vibrations of long shaft connections, which might be
technology controllers, for example, for cranes, winders, and
excited by the ac ripple of the torque due to the motor current
mills. The influence of drifts and offsets in digital controllers
ripple, and other influences have to be taken into account.
is much smaller than that in analog controllers, and there
is good reproducibility. Computing time and analog-to-digital
conversion time are steadily decreasing with the progress of
ARMATURE COILS
microcomputers and electronic devices so that dynamic per-
formance of digital control is now sufficient for many techni- The
single armature coil of Fig. 1 is of the lap winding type.
cal applications. Moreover, with six-pulse converter-fed dc In
Fig. 10(a) an example with K/(2p ) # 14 commutator seg-
drives the cycle time of the digital control is not the limit for ments
between two brushes of opposite electric potential is
dynamical performance of the control. The limit is given by
depicted, resulting in 14 series-connected lap-type coils be-
the time 1/6f (3.3 ms at 50 Hz or 2.8 ms at 60 Hz), which is tween
plus and minus brush. In reality the number of series-
----------------------- Page 72-----------------------
72 DC MACHINES
τ
yielding in general the following law for simplex lap armature
p
windings (a is the number of parallel paths per half arma-

Lap armature coil
ture):
South pole North pole
2a = 2p (28)
l
Magnetic asymmetries of different main poles cause slight
differences between the induced voltages of different parallel
paths. These differences can

be equalized by connecting
Lower layer
points of the same electric potential of different parallel paths
Upper layer Commutator
with so-called equalizers, usually one connection per rotor
#####
v
slot. As one parallel path of simplex windings is usually de-
c
signed for currents up to 500 A with limits due to the avail-
able slot cross section and the current density in the slot cop-
Brush
per bars, all larger machines are designed with lap windings
– +
(13). Very big machines used to have duplex lap windings,
U
consisting of two independent lap windings, each having a
(a)
commutator pitch YQ # 2. Parallel connection of both lap
windings is accomplished by the brushes and additional
equalizers to compensate voltage differences due to asymmet-

τ p
ries. Thus the number of parallel paths is twice that of sim-
Lap armature coil
plex windings.
For smaller output power the simplex wave armature

South pole North pole
winding is preferred, which has only 2a # 2 parallel paths,
independent of the number of pole pairs p of the machine
[Fig. 10(b)]. This is accomplished by a waveform shape of the

l
coils, which connect commutator segments, that are distanced
#####
by nearly two pole pitches (11,12). The coil span is given by
Upper layer Lower layer

K − 1
YQ = (29)
2p
With only two parallel paths it is sufficient to use only one
brush pair (one plus and one minus brush). In practice in

Brush
most cases p brush pairs are used, just as in the case with
– +
lap armature windings. The main reason is that the steady-

U
4
state brush current density is limited to typically 10 # 10

2 4 2
(b)
A/m to 12 # 10 A/m (14). Thus the brush coverage ratio of
commutator segments with only two big brushes would be too

Figure 10. Schematic sketch of armature winding with connection
large. Two positive parallel-connected brushes, distanced
to the commutator segments (stator pole shoes depicted as shaded
with two pole pitches 2# , are short-circuiting only one coil if
areas). (a) Simplex lap winding. (b) Simplex wave winding (four-pole
p
the brush coverage ratio is 1, or two coils if the ratio is two

machine assumed).
and so on, so it is possible to arrange the brushes in the same
way as with lap windings. With smaller dc machines wave

connected coils between the brushes is 30 to 40 for a medium-
armature windings are well suited, because with only two
power dc machine to achieve a low-voltage ripple of 1% to 2%
parallel paths the number of turns in series of the armature
of rated voltage [Fig. 1(d)] (11,12). The commutator pitch of
is large enough to reach the rated voltage of typically 400 V
the coil Yc, expressed in number of segment pitches, is 1, as
to 600 V even with a small air-gap flux per pole
.
adjacent segments are connected to one coil. As the brush cov-
Apart from very small dc machines with rated steady-state
erage ratio of commutator segments, defined by the ratio of
power below typically 1 kW, usually an auxiliary winding is
brush width versus segment width, is usually designed be-
needed, the so-called compole winding. It is necessary to en-
tween 2 and 5, two to five of the K/(2p ) series-connected coils
sure ‘‘dark’’ commutation of the armature coil current. This
are always short-circuited by the brushes. During this short-
means that reversing the polarity of coil current when the coil
circuit time T the armature coil current has to reverse its
is short-circuited by the brushes [Fig. 11(a)], is possible with-
s
direction of current flow. The coil voltage that is short-cir-
out heavy sparking at the trailing edge of the brush, when
cuited is low, as during short circuit the coil passes the neu-
the brush–commutator segment mechanical contact opens.
tral zone, where the inducing air gap field is nearly zero [Fig.
The change of the ac coil current ic during the short-circuit
1(c)]. A four-pole dc machine has four adjacent brushes dis-
time Ts, given by the brush width b and the commutator cir-
tanced along the commutator circumference by one pole pitch,
cumferential speed vC,
two of positive and two of negative electric polarity. Brushes
b
of the same polarity are connected in parallel. Therefore the
Ts = (30)
v
armature winding consists of four parallel electric circuits,
C
----------------------- Page 73-----------------------
DC MACHINES 73
This goal is reached by feeding the coils of the compoles (Fig.

Coil current i
c
12) with the armature current Ia . As the armature reaction
mmf NAIa (NA is the armature windings per pole) has to be
overcome by the compole mmf NCIa (NC is the number of turns
per compole) to generate a resulting air-gap compole flux den-

I
a
sity B , the ratio N /N is usually chosen as 1.12 to 1.25. Ad-
C C A
2a
justment of BC is accomplished by the proper choice of NC and
by adjusting the compole air gap #C with additional iron

0 Time t
sheets during the final test in the
manufacturer’s shop (15).
–I
a
It is necessary, that the magnetic path for BC is unsaturated
2a
to obtain the relationship I # B ; this should be ensured up
a C
T
to twice the rated armature current.
s
(NC −NA )Ia

(a)
BC ∝ (34)
Big machines with an output power higher than typically 200
kW to 300 kW are equipped in addition with a series-compen-
sating winding, which is fixed in slots in the pole shoes of
the stator poles (Fig. 12). This winding is also loaded by the
i
armature current, thus exciting an air-gap field, which op-
c
poses the armature reaction field. By adjusting the number of

I I
turns of the coils of the compensating winding in relation to
a a
2a 2a
those of the armature coils and by choosing a slot pitch that
is similar to the rotor slot pitch, but not identical to avoid
cogging, it
is possible to extinguish the armature reaction
v
field with its negative influence on the machine (16). The aux-
c
b
iliary windings do not play an active part in electrical energy
conversion, but cause additional ohmic losses. Thus, along

I /a
a
with the rotor iron losses, the electrical brush losses, the
brush friction losses, the eddy current losses in the armature
(b)
coils due to the ac coil current, and the converter-induced cur-
Figure 11. Commutation of armature coil current i . (a) The coil cur-
rent ripple (17), these losses yield a slightly lower machine
c
rent has to reverse its polarity, while the coil is short-circuited by the
efficiency when compared with ac induction and synchronous
brush. (b) The commutating lap coil is short-circuited by the brush.
machines (18).
TEMPERATURE CONTROL
from
I /(2a) to #I /(2a) and vice versa, causes a reactance
a a
There exist many different cooling systems for electrical ma-
voltage ur due to the inevitable coil stray inductance Lc [Fig.
chines (19); dc machines are manufactured often with open-
11(b)]:
circuit cooling. The cooling air flow enters the dc machine
usually through openings in theend shield at the nondrive

dic∼ Ia
ur = Lc = Lc ∝ nIa (31)
end side, where the commutator is situated, and passes
dt aTs
through axial and radial ventilation ducts of the rotor and
through the axial gaps in the stator between the main poles,
This voltage causes ignition of sparks at the brush’s trailing
leaving the machine through openings in the drive-end end
edge, causing wear of the brush and the commutator, thus
shield (20). For variable-speed drives an external fan to gen-
reducing the brush life drastically. The ideas of the compole
erate a constant, speed-independent air flow is necessary. A
winding is to excite an additional air-gap field BC in the neu-
fan mounted on the rotor shaft of the machine itself would
tral zone between two adjacent main poles, which opposes the
deliver an insufficient air flow at low motor rotational speed.
armature reaction field. Thus ur can be compensated by an
The different windings of the dc machine (armature, field,
additional voltage uC, which is induced in the commutating
compole, compensation) and the surface of the commutator
coil due to the movement of the armature with speed v in the
with the brushes are cooled intensively with the cool air en-
compole field BC :
tering on the commutator side, passing closely to the surface
of the windings. This method ensures a high utilization of the

u = 2N B lv (32)
C c C
machine, resulting in a compact build, but causes a rather
large temperature gradient of typically 20 K to 40 K between
The compensation must be accomplished at any speed n and
the cool-air inlet and the hot-air outlet, where the heated air
load Ia to ensure dark commutation for any arbitrary load:
flow leaves the machine. The hot spots in the coils are there-
fore located on the hot outlet side, so the temperature sensors

uC = ur ∝ nIa (33)
are fixed to the windings also on the hot side. In separately
----------------------- Page 74-----------------------
74 DC MACHINES
Compole coil Stator yoke
Field coil
####
Main pole
##
Compensating
winding
Main airgap δ
Rotor tooth
c Rotor slot
####
###
Rotor iron yoke
###
Shaft
Figure 12. Cross section of a four-pole dc machine with

rotor armature winding, stator field, compole, and com-
pensating winding (main air gap # and compole air gap
Armature winding (lower layer)
#C enlarged for clarity).
(upper layer)
excited dc machines usually one sensor is necessary for the and
copper wire is necessary to ensure a low thermal time
compole winding and one for the field winding. The rotating constant
of the sensor measuring system (several seconds)
armature winding would need additional sliding contacts or a and a
similar temperature rise of winding and sensor. So, the
wireless signal transmission if a temperature sensor were sensors
are usually put between the wires of the windings.
placed into this winding. So the compole temperature sensor After
this, the winding is impregnated with epoxy resin to get
is also used to protect the armature winding, as both wind- good
electric insulation and protection against humidity and
ings are loaded with the same armature current, which carbon
dust. In addition to temperature sensors, often the air
causes the copper losses Pd as dissipated heat along with the flow
itself is controlled directly by an air-pressure sensor at
temperature rise #: the air-
flow inlet.
#ϑ ∝ P = RI 2 (35)
d a
TIME-
VARYING LOADS
One has to consider that the heating of the armature winding
is caused not only by load losses, Eq. (35), but also by main As
already stated, the resulting inductivity of the armature
flux-dependent losses in the active iron core. By careful de- L is
rather low, yielding a low electrical time constant T ,
a
a
sign of armature and compole winding one tries to obtain giving
the dc drive good dynamical performance, which is nec-
nearly the same steady-state temperature rise in both wind- essary
especially under time-varying load conditions. The rea-
ings. So it is possible to protect the armature winding with son for
the low inductance is that the compole winding (or
the compole sensor and yet utilize both windings to their tem- compole
and compensating winding) with inductance LC #
perature limit. This limit is given by the thermal limit of the N2 on
one hand and the armature winding with inductance
C
utilized insulation material of the electrical winding (21). For L # N2
on the other hand excite air-gap fields, which oppose
A A
example, insulation class F material sustains a steady-state each

other. Thus the inductive coupling of this pair of wind-
temperature rise of 105 K in addition to 40#C ambient tem- ings,
given by the mutual inductance M # N N , yields a
AC A C
perature plus 15 K additional temperature rise in hot spots, low
resulting inductance La for the whole armature circuit
resulting in a maximum temperature of 160#C, for at least (Fig.
13).
20,000 h, de facto for a much longer period of time. The corre-
sponding temperature control is designed for insulation class
LA = LAσ + LAm (36)
F usually with an operating temperature of 105#C
40#C #
145#C. Tripping miniaturized bimetal temperature relays and
LC = LCσ + LCm (37)
semiconductor components such as negative temperature co-
efficient (NTC) and positive temperature coefficient (PTC) ele-
La = LA − MAC + LC − MAC (38)
# #
ments are used as protective sensors. For continuous temper-

2
N
ature measurement thermocouples or platinum resistance
La = LAσ + LCσ + 1 − C LAm (39)
N
elements are utilized. Good thermal contact between sensor
A
----------------------- Page 75-----------------------
DC MACHINES 75
Armature winding LA
forced open-circuit cooling will have an armature circuit
equivalent thermal time constant of typically 15 min to 30

MAC magnetic coupling
min. The temperature rise in dc machines with intermittent

Ia
load with a load-cycle duration longer than only a few seconds
Compole is
determined by the thermal time constants of the machine.
winding L C
Such a situation is typical for elevators and cranes (for exam-
ple, 10 min cycle time, 6 min load, and 4 min no-load period).
Thus with intermittent periodic duty higher motor currents

are
permissible than under steady-state conditions, as the
Field winding Lf
temperature decreases during the no-load period. As an
equivalent measure for the temperature of the windings the

If
temperature rise in the middle of the load period is taken

Figure 13. The compole (and compensating) winding excites a mag-
(18).
netic field that opposes the armature reaction field, yielding a low
resulting inductance. The field winding is not magnetically coupled
LIMITS OF DC MACHINES
with these windings (besides saturation cross-coupling).
Torque Limit
As NC/NA is typically 1.12 to 1.25, La represents mainly the The

delivered electromagnetic torque M for a given current
stray inductance of the rotating armature and of the compole
load A and air-gap induction B is determined by the main
winding LA# and LC# , thus explaining its low value.
geometric data (11,16) d [diameter of the armature at the sur-
face, Fig. 1(b)] and l [iron stack length of the core, Fig. 1(b)],
Change of Load Current Eq.
(40), which is derived from Eqs. (6) and (7). Therefore a
machine designed for big torque usually has a large diameter.

Modern dc machines are designed usually to sustain sudden
load peaks of twice the rated armature current for several
π 2
M = αe d lAB (40)
seconds (22). One limit is given by avoiding saturation of the
2
compole magnetic circuit when the linearity between arma-
ture current and compole field, Eq. (34), would be lost with
Modern dc machines are designed for high utilization to keep
large armature currents, yielding too weak a compole field the
ratio of investment costs to installed power as low as pos-
BC . A second limit is the brush current density, for brushes
sible. A high output torque per motor volume can be achieved
4 2 by
increased current load and by increased air-gap induction.
can sustain current densities of about 20 # 10 A/m only for
several seconds without being overheated. The rate of change The
latter is limited due to the saturation of the rotor teeth
of the armature current can be very large if the compole mag- to
about 1 T. High current load with limits of typically 50 kA/
netic paths consist completely of laminated iron parts (dia/ m
(11) is accomplished by optimized cooling and by designing
dt 600IN per second, typically), as eddy currents in the iron the
machines for a higher steady-state temperature rise
core are suppressed. Older machines with massive iron yokes
through use of improved materials.
suffer from a delay of the compole flux change due to the in-
trinsic magnetic field of the eddy currents in the massive iron
Speed Limit
parts, causing too weak a compole field under transient load
A
high output power can be obtained by an extended speed
conditions (16). Thus with these machines the rate of current
range according to Eq. (17). Whereas A and B set nearly the

change is limited to 20IN per second.
same limits also on ac machines, dc machines are more sensi-

tive to high speed than ac machines. Speed limits are caused
Change of Speed
by
the electrical commutation of the armature current and by
At standstill the load current has to be reduced to stay well the
mechanical speed limit of the commutator.
4 2
below the 6 # 10 A/m brush current density to avoid burn-
ing-in on the commutator surface.
1. The critical mechanical speed due to the centrifugal
forces of the commutator copper segments is an upper

Load Cycles
speed limit for dc machines (23).
Thermal time constants of the machine depend on the ma-

2. At high speed the brushes no longer slide smoothly on
chine size and on the cooling condition. Small machine and
the commutator surface. Brush chatter, which is a high-
machines with a forced open-circuit cooling usually have
frequency bending vibration of the brush, and brush
lower time constants than large machines or machines with
jumping may occur with a rapid increase of mechani-
surface cooling. The stator windings may be described by one
cally induced brush wear (14).
thermal time constant for each winding system, whereas for
3. The reactance voltage ur rises linearly with speed, Eq.
the armature winding a shorter and a longer time constant
(31). So, sparking will occur preferentially at high
have to be taken into account, as two different loss sources
speed, causing increased electrically induced brush
have to be considered, namely the current-dependent copper
wear. Therefore above a critical speed nf , corresponding
and additional losses and the flux-dependent iron losses. Usu-
to a maximum value of urm of about 8 V to 10 V, the
ally one tries to evaluate one single equivalent thermal time
armature current has to be reduced, Eq. (41). Therefore
constant. For example, a four-pole 100 kW dc machine with
modern variable-speed dc drives have a steady-state
----------------------- Page 76-----------------------
76 DC MACHINES
load characteristic as depicted in Fig. 14. For example,

deeper insight into the physical nature of the brush–
the maximum speed nm of a 100 kW dc machine (shaft
commutator contact, further improvements are possible (26).
height 160 mm, nN # 2000/min) lies typically between
4500/min and 6000/min, depending on the mechanical
Voltage Limit
rotor design, corresponding with an overspeed test of
The mechanical commutator puts also a limit to the armature
120% of nm for 2 min.
voltage. The average bar-to-bar voltage between to commuta-
urm
tor segments us must stay below 18 V (uncompensated ma-
Ia ∝ (41)
chines) and 20 V (compensated machines) to avoid flashover
n
already at no-load conditions. With a load there is a distortion

Highly utilized dc machines with considerable high magnetic
of the air-gap flux density distribution due to armature reac-
flux densities in the stator yoke (Fig. 12) suffer from satura-
tion with uncompensated machines, yielding a maximum bar-
tion of the stator yoke due to the main flux. As both main and
to-bar voltage um , which is considerably higher than the aver-
compole flux pass through the same parts of the stator yoke,
age value especially at considerable flux weakening, thus at
the compole flux is also saturated to a certain extent. Beyond
high speed. This maximum value should stay below 35 V.
rated speed, when the main flux is weakened, the yoke re-
With a minimum commutator pitch of about 3.5 mm, needed
turns to the unsaturated state, resulting in an increase of the
for connecting the armature coil ends to the segment, and a
compole air gap flux density BC . The compole voltage uC then
minimum mica insulation thickness of about 0.8 mm to en-
overcompensates the reactance voltage ur at high speed, caus-
sure the just noted voltage limits, maximum armature volt-
ing sparking of the brushes.
ages of typically 1000 V are reached. Larger voltages would
All these negative influences on commutation can be en-
yield large commutator diameters due to the large number of
countered partially by designing improved armature wind-
segments between two brushes, which again would be critical
ings, which are more insensitive to commutation distur-
to centrifugal forces.
bances. Examples of measures to improve commutation are
These limits are the reason why there is now a strong
the design of a low coil inductance Lc to get a low reactance
trend to change from dc drive technology, especially with
voltage. Reduced slot height to get low slot leakage induc-
high-speed application, to ac drives with the robust induction
tance and the use of split-throw windings (24,25) (necessary
motor and the ‘‘brushless’’ dc motor. Nevertheless, for me-
for rated power above typically 200 kW) yield low coil induc-
dium speeds up to typically 4000/min in the power range of
tances. The last coil per slot has the highest coil inductance
100 kW to 1000 kW dc drives are still a solution of high tech-
Lc due to lack of magnetic coupling with other short-circuited
nical maturity with a reasonable ratio of investment cost to
commutating coils of the same slot. Therefore its commutat-
output power and an excellent dynamical performance for
ing condition is worse than that of the other coils. Equalizers
many technical applications, and will remain of economical
connect the last coil per slot with coils of the same electric
importance for years to come (27).
potential. Thus those coils, which are distanced by two pole

pitches, produce unavoidable magnetic asymmetries of the
BIBLIOGRAPHY
machine, which cause different sparking of certain brushes.
For machines with both rotation directions, the first and the
1. J. Hindmarsh, Electrical Machines and their Applications, 4th ed.,
last coil per slot must be connected to equalizers, of course,
Oxford: Pergamon, 1991.
yielding 2Q/p equalizers (p # 1). Moreover, by obtaining
2. J. Hindmarsh, Electrical Machines and Drives—Worked Exam-
ples, 2nd ed., Oxford: Pergamon, 1991.
3. N. Carver, Aspects of carbon brush performance, Elevator World,
(Feb.): 60–69, 1990.

M, Ia P, U, Φ
4. F. J. Bartos, DC drives still have miles to go before they sleep,
Control Eng., (Aug.): 48–50, 1991.

U
UN
5. R. Robbins, DC will continue as viable drives technology option,
Drives Controls, (Mar.): 60–61, 1997.

P N
6. A. Kusko and S. M. Peeran, Application of 12-pulse converters to
reduce electrical interference and audible noise from DC motor

P ~1/n
drives, IEEE Trans. Ind. Appl., 29: 153–160, 1993.

I N
7. W. Leonhard, Control of Electrical Drives, 2nd ed., Berlin:
I ~1/n
Springer, 1996.
a
M N M ~1/n
8. B. K. Bose (ed.), Modern Power Electronics—Evolution, Technol-
ogy, and Application. New York: IEEE Press, 1992.

Φ N M ~1/n2
9. J. M. D. Murphy and F. G. Turnbull, Power Electronic Control of
AC Motors, Oxford: Pergamon, 1988.

Φ ~1/n
10. W. Hulsbosch et al., Influence of the Elastic Coupling on the Con-
0 0 nN nf nm
trol of Converter Fed DC Motor Drives, Proc. 3rd Eur. Conf. Power
Electron. Appl. (EPE), Aachen, Germany, 1989, pp. 849–854.

Rotational speed n
11. K. Vogt, Berechnung rotierender elektrischer Maschinen, Wein-

Figure 14. The dc motor as a variable-speed drive, being fed by vari-
heim, Germany: VCH, 1996.
able armature voltage U and variable main flux
. At high speed 12. J. Schneider, Elektrische Entwurfsberechnung,
in W. Boning
¨
n # n the output power P has to be reduced to limit the reactance
(ed.), Hutte-Elektrische Energietechnik, Band 1: Maschinen, 29.
¨
f
voltage of commutation.
ed., Berlin: Springer, 1978.
----------------------- Page 77-----------------------
DC TRANSFORMERS 77
13. A. Binder, Comparison of the electromagnetical performance of

simplex wave and lap windings, IEEE Trans. Energy Convers. 8:
698–703, 1993.
14. E. I. Shobert, Carbon Brushes, New York: Chemical Publishing,

1965.
15. G. M. J. Parsley, A. S. Meyer, and C. F. Landy, Factors affecting

the prediction of commutating limits for a DC machine under
varying speed and load conditions, Trans. South African Inst.
Electric. Eng., (Sept.): 171–176, 1992.
16. W. Dull, Gleichstrommaschinen, in W. Boning (ed.), Hutte-Elek-

¨ ¨ ¨
trische Energietechnik, Band 1: Maschinen, 29. ed., Berlin:
Springer, 1978.
17. A. Binder, Additional losses in converter fed uncompensated dc

machines-their calculation and measurement. Electric. Eng. (Ar-
chiv fur Elektrotechnik) 74: 357–369, 1991.
¨
18. Rotating electrical machinery, Part 2: Methods for determining
losses and efficiency of rotating electrical machinery from tests
(excluding machines for traction vehicles), IEC Publication IEC
34-2, Geneva: IEC Central Office, 1972.
19. W. Liebe, Entwarmung elektrischer Maschinen, in W. Boning

¨ ¨
(ed.), Hutte-Elektrische Energietechnik, Band 1: Maschinen, 29.
¨
ed., Berlin: Springer, 1978.
20. R. Schamberger and E. Thum, DC Machines of Rectangular Con-

struction for Torques from 300 to 3000 Nm, Siemens Power Eng.,
3: 17–21, 1981.
21. Rotating electrical machinery, Part 1: Rating and performance,

IEC Publication IEC 34-1, Geneva: IEC Central Office, 1994.
22. J. D. White, M. Zeller, and W. Goss, DC Motors for Variable Speed

Drives, Catalog DA12, Erlangen, Germany: Siemens AG, 1992.
´
23. I. Huszar, Bemerkungen zur Wahl des Gewolbedrucks von Kom-
¨
mutatoren, Elektrische Bahnen 36: 291–296, 1965.
24. W. J. Pratt, The influence of design parameters on the sparkless

zones of dc machines, GEC J. Sci. Technol., 45: 51–55, 1978.
25. A. Binder, Measures to raise brush life of modern dc machines,

Eur. Trans. Electric. Power Eng., 3: 193–200, 1993.
26. G. N. Friedman and T. I. Fomicheva, Effect of commutator con-

ductor materials on commutation of electric machines, Elektro-
tekhnika 60: 32–36, 1989 (reprinted by Allerton Press, 1989).
27. E. Auernhammer et al., Kompakte Gleichstromantriebe durch

Leistungssteigerung, Elektrotechnische Z., 113: 1342–1349, 1992.
ANDREAS J. F. BINDER
Darmstadt University of Technology
DC OPERATING POINTS. See HIGH DEFINITION TELEVI-
SION; HOMOTOPY METHODS FOR COMPUTING DC OPERATING
POINTS.
DCT. See HADAMARD TRANSFORMS .
----------------------- Page 78-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
●
HELP

Your Librarian
Profile
Diesel-Electric Generators

Standard Article

Graham Reader1 1 1 1
, Ian Potter , Caroline Bowen , Ming Zheng , J.
Gary Hawley2
1University of Calgary, Calgary, Alberta, Canada

2University of Bath, Claverton Down, Bath, United Kingdom
Search
reserved.
❍ Acronym Finder
DOI: 10.1002/047134608X.W3002
Abstract
Generator Applications
Engine Size, Classification, and Selection
Engine and Generator Set Ratings
Basic Concepts and Working Cycles of Engine Operation
Practical Aspects of Engine Operation
Environmental Aspects
Operations with Non-Air Intake Mixtures
AC Generators

%20Conversion/W3002.htm17.06.2008 16:35:51
----------------------- Page 79-----------------------
368 DIESEL-ELECTRIC POWER STATIONS
DIESEL-
ELECTRIC POWER STATIONS
Worldwide,
interest in today’s electricity supplies is focused
on
centralized systems. Whereas electricity can be supplied
almost
everywhere in industrialized countries (ICs), thanks
to
national or international integrated grids, the developing
countries
(DCs) are far behind. Only 23% of global electricity
consumption
is generated in DCs, where it supplies about
43% of
humanity. A further 35% of the population in DCs has
no access
to electricity. Although integrated grids are increas-
ing in
global importance, they are not feasible in many DCs
for
technical and economic reasons. Decentralized electricity
supply is
thus of great interest in such places as well as, for
different
reasons, in ICs.
Autonomous electric power stations are used wherever we

need to
have electric energy and there is no power system
or when the
available grid is not reliable enough (i.e., utility
companies
cannot provide electric power without any failure
or clear
and free of disturbances to users); in such cases, to
prevent
equipment problems, a good solution is for users to
provide
their own generating units as a supplement to the
power
supply. A generating set can be installed to have an
emergency
or standby power source. The emergency system
operates in
parallel to ensure no loss of continuity in the elec-
tric power
supply. Standby systems are mainly intended as
alternative
sources (1).
Large
megawatt-range generator sets are frequently com-
bined with
other types of power plants both in large isolated
grids and
in interconnected grids because of their ease of dis-
patch.
Power generating sets for decentralized electricity sup-
ply usually
consist of several aggregates in the low-to-medium
power
range. About 37% of the power capacity installed
worldwide
for local electricity generation is for decentralized
use.
Generating units driven by internal combustion engines

consist of
a synchronous generator mechanically coupled to a
power
source, both installed on a rigid frame. In some cases
the engine
and the generator may be connected by a clutch,

----------------------- Page 80-----------------------
DIESEL-ELECTRIC POWER STATIONS 369
frame
which supports the entire system; (4) the synchronous
generator with the electric connection box and automatic volt-

age
regulator (above it); (5) the diesel engine.
The
ability to connect in parallel several generating groups
permits us to reach installed power up to tenths of a mega-

watt.
Power stations using multiple generators may obtain
greater energy conversion efficiency by turning on only those
generators which are sufficient to supply the required load,

thus
allowing each generator to work at a power rate as close
as
possible to the rated power. Multiple diesel units, working
in
parallel, may ensure a higher energy availability and a
lower
probability of transient or steady-state instabilities. A
generating set can supply emergency power in a few seconds

and
when associated with uninterruptible power systems, can
ensure
long-term operation free of disturbances to a critical
load.
Figure 1. Typical water-cooled diesel generating set in the power
range of some tenths of a kilowatt. (Courtesy of Ausonia Generating
Set, Marsala, Italy.)
INTERNAL COMBUSTION ENGINES
which permits the electric machine to be available as a syn-

Engines used in generating units burn diesel, gasoline, or
chronous motor connected in parallel to the electric grid with-
natural gas (2,3). Diesel engines, supplied by several compa-
out the presence of the engine. In this case the electric ma- nies
in different models, are rugged and provide safety
chine will operate by absorbing a little active power (there are
against fire and explosions compared with other engines.
no mechanical loads connected to the synchronous motor) and
Smaller sizes use a natural air supply, while larger sizes are
delivering reactive inductive power to the grid, acting as a so- turbo-
charged. Their rated power ranges between a few kilo-
called synchronous capacitor. The rated powers of generating volt-
amperes to thousands of kilovolt-amperes. Gasoline en-
sets, equipped with internal combustion engines, range be- gine
generating units are available with rated power less
tween a few kilovolt-amperes to several thousand kilovolt-am- than
100 kV-A. They offer the advantages of having a quick
peres, which represents the size above which it is convenient run-
up, having a low cost, and being lightweight. However,
to use different prime movers such as steam or gas turbines. the
use of gasoline prevents long-term storage of the fuel and
Smaller sizes are suitable for small isolated users (home ap-
increases the hazard of fire and explosion. Moreover, the cost
pliances) while larger sizes are used for villages, islands, and of the
energy produced is higher. Gas engine generating units
industrial and commercial applications. Figure 1 shows a typ- can
burn natural gas or a blend of gas, stored as liquid, ob-
ical diesel engine generating set, illustrating an entire system tained
from crude oil. Gasoline engines can be, by minor ad-
consisting of a synchronous generator, driven by a diesel en-
aptation, used as natural gas engines. Major advantages of
gine, with the cooling system at the opposite side of the gener- such
an operation are the reduced maintenance due to the
ator. Figure 2 shows an exploded view of the main compo-
limited residue from combustion. On the other hand, a gaso-
nents of a diesel generating set: (1) the cooling system of the line
engine supplied by gas provides reduced output power.
diesel engine; (2) the built-in small fuel tank; (3) the rigid Table
1 presents features of commercial generating units
available on the market.
4 CHOICE
OF A DIESEL GENERATING SET
1
The
choice of a diesel generating set revolves around several
parameters, such as: kind of service to which the generator

must
be devoted, electrical characteristics of the loads, envi-
ronmental conditions, the allowed frequency and voltage vari-

ations
(particularly during transient conditions), and failures
2
which
can happen on the main power systems. The precise
evaluation of the aforementioned parameters is fundamental

to
realizing reliable electric power plants, which can produce
electric energy at a convenient cost.
3 Class
of Service
Figure 2. Exploded view of the diesel generating set shown in Fig. A

diesel generator set may be used to serve several kinds of
1. (1) Cooling system of the diesel engine. (2) Build-in small fuel tank.
applications: (1) main source of energy for isolated users (con-
(3) Rigid frame supporting the whole system. (4) Brushless synchro-
tinuous service), (2) backup systems (service on demand), and
nous generator. (5) Diesel engine. (3)
emergency systems. Some common failures and distur-
----------------------- Page 81-----------------------
Table 1. Technical Characteristics of Generating Units
50 Hz, 1500 rpm 60 Hz, 1800 rpm
Dimensions
Emergency Duty Continuous Duty Emergency Duty
Continuous Duty
Length Width Height Weight

kW kVA kW kVA kW kVA kW
kVA (mm) (mm) (mm) (kg)
1.5 1.2 1.5 1.2 1.8 1.44

1.8 1.44 510 340 400 35
10 8 10 8 11 8.8
11 8.8 1,250 500 700 160
100 80 92 73.6 115 92
105 84 2,400 1,050 1,340 1,270
550 400 500 400 640 512
585 468 3,350 1,600 1,850 3,700
1,000 800 910 728 1,250 1,000
1,100 880 4,600 1,900 2,400 8,400
2,000 1,000 1,825 1,460 2,500 2,000
2,250 1,800 5,900 2,200 2,550 13,200
Courtesy of Ausonia Generating Set, Marsala, Italy.
bances regarding electric plants, which can create the need to

Derating
install generating units, are as follows: (1) interruption (out-
Rated
powers of generating units refer to precise ambient
age, that is, a loss of power that may range from less than a
conditions, which are defined by standards for diesel engines

second to several hours and affect all equipment; (2) over- or
and
synchronous generators. As an example, typical rated
undervoltage: overvoltage (surge) and undervoltage (dip) are,
data
for ambient conditions of engines are: altitude above the
respectively, increases or decreases in the rated value of the
sea
level 300 m, relative humidity 60%, pressure 100 kPa,
line voltage that range between a few milliseconds and a frac-
and
temperature 29#C. When there are different conditions,
tion of a second, which can cause shutdowns (dip) or can dam- some
derating must be applied. In fact, pressure, tempera-
age electrical equipment (surge); (3) voltage spikes, which are ture,
and humidity influence the air mass introduced to the
short and sharp impulses that appear momentarily from line
engine during intake as well as during heat exchange. As a
to line or line to neutral (normal mode) or between neutral
consequence of the reduction of the inlet air, the fuel quantity
and ground (common mode); (4) electrical noise, which is
entering the combustion chamber decreases, which means de-
high-frequency interference that ranges between a few kilo-
rating the engine power. Such a derating is different for natu-
hertz and tenths of a megahertz; and (5) harmonics, which rally
aspirated and turbo-charged diesel engines. For natu-
are frequencies that are multiples of the fundamental fre- rally
aspirated engines, a rule of thumb gives a power
quency.
derating of 2% for each temperature step of 5.5#C over the
User needs vary, which affects the type of generator se-
reference temperature, while for turbo-charged engines, the
lected. For example, isolated users principally need a system
derating increases to 2.5%. Of course, the ambient tempera-
which is as reliable as possible, even if the quality of service ture
in the generator room is normally not allowed to exceed
may not be the best. Efficiency may be quite important be- a
threshold value since it is controlled to remain in the range
cause it is linked to the cost of the energy produced. A backup 20#C
to 50#C. Thus the effective temperature derating is
system permits integration of the power from the main source
bounded by a small percentage.
to meet peak load requirements as needed. An emergency sys- A
derating factor also needs to be applied as the altitude
tem must be absolutely reliable; efficiency does not have to be of
the installation increases. Consequently, the rated power
considered. The system will probably operate for only a few
should be decreased, as a rule, by 3.5% and 2% for each 300
hours in a year, but for those hours it must be able to start m
increase of altitude, respectively, for naturally aspirated
suddenly and operate correctly. To increase reliability, emer- and
turbo-charged engines. The humidity derating is consid-
gency systems may be doubled using two identical generating ered
as a combined action of relative humidity and tempera-
sets or using two systems based on different sources of ture
of the air since the two quantities are strictly tied. In
energy. any
case, combined effects of high humidity and temperature
give
a derating of about 5% to 6%. In effect, standards dictate
the
conditions of derating for the cumulated effect of tempera-
Environmental Conditions
ture,
humidity, and altitude (atmospheric pressure). Figure 3
Environmental conditions are often critical to diesel station
shows, for turbo-charged engines, the curves calculated at a
design. The parameters to be considered are mainly (1)
relative humidity of 60% of the derating coefficient versus
ambient temperature, (2) humidity, and (3) dust. The temper-
temperature and atmospheric pressure.
ature affects the power of the generator. The humidity also
influences the deliverable power due to its effect on combus-
Efficiency
tion, while an excess of dust may require a special air filter The
efficiency of the generator in fuel conversion depends on
for the engine. In some cases, temperature and dust problems
several factors, but mainly on the size of the machine and on
may necessitate the use of a special ‘‘filter-container’’ includ- the
percentage of the load applied to the generator with re-
ing all the equipment and using particular cooling systems spect
to the rated power (the efficiency is zero at no load and
which prevent the external air from entering into the con-
increases almost linearly reaching the highest value at a load
tainer. close
to the rated power). The rated fuel consumption is de-
----------------------- Page 82-----------------------
Height above sea level (m)
0 200 400 600 800 1000 200 400 600 800 2000 200 400
600 800 3000 200 400 600 800 4000 200
100%
98
96
94
92
90
88 1
0°
86 C
2
) 84 0°
% C
( 82
t 3
0°
n 80 C
e
i 78
c 4
i 0
f
°
f 76 C
e
o 74 5
c 0°
g 72 C
n
i
t 70
a
r 68
e
D
66
64
62
60
58
56
54
52
50%
Height above sea level (m)
0 200 400 600 800 1000 200 400 600 800 2000 200 400
600 800 3000 200 400 600 800 4000 200
780 740 720 700 680 660 640 620 600 580 560
540 520 500 480 460
Atmospheric pressure (mm Hg)
Figure 3. Derating coefficient of turbo-charged engines

versus temperature and atmospheric
pressure calculated at relative humidity of 60%.
fined per produced kilowatt-hour when the engine runs at damage the
engine by corrupting the lubricant or, as a result
rated power. It ranges from 130 g/kW-h to 150 g/kW-h, for of passing
through the exhaust gas, may damage the turbine
big units, to 250 g/kW-h to 300 g/kW-h for small generat- (for the
turbo-charged engines), thereby increasing mainte-
ing sets. The lower fuel consumptions are typical of big nance costs
and lengthening the time that the equipment is
turbo-charged diesel units, and the smallest consumption out-of-order.
rate (130 g/kW-h) is characteristic of huge diesel units
(several MW of rated power) operating with a two-stroke cycle
with turbo-charger and low rotational speeds (as low as 100
rpm).
The diesel generator should operate at a level of power as )
100
n
f o
close as possible to the rated power, in order to achieve a o i
t
higher efficiency. Other operating conditions cause higher % p
( m
75
specific fuel consumption both for low load and for overload. n u
o s
i n
t
Figure 4 shows a typical curve of fuel burned even at no load, p o
c
m l
50
which represents the worst condition with respect to effi- u e
s u
n f
ciency. Particular care must be taken not to oversize the die- o d
sel set because both diesel engines and synchronous genera- c e
25
l t
e a
tors are characterized by low efficiencies at low load, thus u r
F e
h
increasing the specific fuel consumption (g/kW-h) and conse- t
0
quently the cost of the produced kW-h. The specific fuel con-
0 25 50 75 100
sumption is not critical when the generating set is used for
Electric power output
an emergency service while it is very critical when the plant
(% of the rated power)
provides continuous service. Furthermore, at low load, the en- Figure 4.

Typical variation of the fuel flow versus output power for
gine does not reach an optimal temperature condition, and a diesel
generating set. The fuel consumed without delivering electric
the combustion may not completely consume the fuel injected power is due
to the losses to maintain the system rotating at nomi-
into the combustion chamber. The fuel not burned may then nal speed.
----------------------- Page 83-----------------------
PARALLEL OPERATIONS OF MULTIPLE istics,

we must be careful when connecting the neutral con-
DIESEL GENERATING SETS ductors.
These connections can produce third harmonic cur-
rent
circulation which may be greater than the first harmonic
A diesel generating set may operate in parallel to have (1) an rated
current, causing the parallel operations to fail. The
increase in available power, (2) a backup system, and (3) an neutral
of the generator may be connected to the earth in
increase in reliability. If we consider a generator already run- one of
the following ways: (1) solidly by direct wiring to a
ning and delivering power to the user as a ‘‘grid’’ and consider grounding
electrode, (2) by a grounding resistor inserted
another generator to be paralleled to the first one, to obtain a between
the terminals of the system and the grounding elec-
correct parallel connection and operation the following condi- trode,
(3) ungrounded by leaving the winding potential float-
tions apply: (1) The output voltages must be equal in magni- ing.
Detailed alternatives to grounding systems can be found
tude both on the grid and on the generator terminals, (2) the in Ref.
1.
frequency of the generator must be as close as possible to the
grid frequency, (3) the phase sequence of the two systems
must be the same and almost without angle displacement,
SIZING OF
THE GENERATING GROUPS
and (4) each generating unit must be equipped with an auto-
matic power sharing device. The diesel engine can furnish
only the ‘‘active power’’ (the mechanical power through its Sizing of
the generating groups to cover the needs of a power
shaft) to the synchronous electric generator, while the genera- plant
requires a careful analysis of the nature of the loads
tor delivers such an active power subtracted of the total losses
(induction machine, lamps, electric oven, etc.) in order to com-
plus an amount of reactive power because of the power factor ply with
the transient currents, the effective power due to
(cos #) of the load which normally ranges between 0.8 and
nonsimultaneous presence of loads, and the power factor
0.9. Since a low power factor causes inefficient utilization of which
changes depending on the operating conditions. More-
the diesel-generator (low power request from the diesel and over, we
must be careful during transient stability, when a
high apparent power request from the generator), in some ap- step
application of load may cause a severe transient on the
plications this can be useful to produce a supplemental reac-
generator. In fact, in such a case, a momentary decrease of
tive power by using static devices (stepped back of capacitors). both
frequency and voltage occurs. Some limitation may be
required
on the maximum applicable load step for it to re-
Active Load Sharing in Local Grid main
within a tolerable percentage of voltage and frequency
drops.
There are standards that set limits and also give some
The active load sharing between different generators in local
constraints depending on the mean effective pressure of the
grid connections is controlled by varying the fuel actuator of
engine
driving the generator. We must use care during induc-
the diesel engines. If several generators are delivering power
tion
motor starting. In fact, induction motors are designed for
on the same load, each generator takes a load angle which
line
starting capability with a stiff source, and in such a case
depends on the amount of load it is supplying. An efficient
the
inrushing current can be five to eight times the rated cur-
load sharing will cause each generator to deliver power pro-
rent at
full load. The torque developed by induction motors is
portional to its rated power. On an autonomous grid, powered
proportional to the square of the terminal voltage. Depending

by multiple diesel generators, to perform load sharing the
on the
characteristic of the load torque and the robustness of
generators must be able to change their output frequency by
the power
source, the startup may fail. We must calculate the
realizing a frequency drop between the full load (lower fre-
maximum
tolerable decrease of the line voltage, which must
quency) and no load (higher frequency). If one of the genera-
tors delivering power on the load has a different frequency comply
with the motor starting needs. The line starting of
drop characteristic, it can be easily overloaded or un- motors
also requires particular attention to the evaluation of
derloaded. Consequently, the diesel engines of the units in the power
required at low power factor. An induction motor
parallel must have the same frequency drop characteristic with a
full load power factor of 0.8 can have a 0.2 power factor
(identical speed variation for each load condition from no load at no
load. Assuming the case of a motor of 100 kW, which
to rated load), which normally is approximately 4%. For lower has an
efficiency of 0.90, the active power at full load ab-
values of frequency drop, imbalance and instability of load sorbed
from the line will be 100/0.90 # 111 kW, and the ap-
sharing may occur. parent
power will be 111/0.8 # 138 kV-A. At startup with no
load,
the line current can reach up to eight times the rated
Reactive Load Sharing value and
the same applies to the apparent power, which will
be 138 #
8 # 1104 kV-A with an active power equal to 1104
The reactive load sharing depends on the excitation systems # 0.2 #
220 kW. The active power absorbed for a few seconds,
of the synchronous generators. The voltage regulator of the during
the startup transient, can be two times the rated
synchronous generators must allow the same voltage drop
power.
(identical voltage variation for each load condition from no
load to rated load). If the regulator does not have the same
voltage drop, current circulation may occur between the gen- Rated
Power of Synchronous Generators
erators.
Rated
power is calculated according to standards at the fol-
lowing
ambient conditions: ambient temperature 40#C; alti-
Neutral Links and Grounding
tude less
than 1000 m; power factor equal to 0.8. For different
For multiple diesel stations with two or more generating sets
altitudes, temperatures, or power factors a derating must be
in parallel having generators with different electric character-
calculated. A suitable method to evaluate the power derating
----------------------- Page 84-----------------------
1
3
k 1.10
k 1.00
t
t
n
n
e
e
i
i 0.90
c
c
i
i
f 1.00
f
f
f
e
e
o
o 0.80
c
c
g 0.90
g
n
n 0.70
i
i
t
t
a
a
r 0.80
r
e 30 40 50 60
e 0.60
D
D
1 0.8 0.6 0.4 0.2 0

Ambient temperature (°C)
Power factor
Figure 5. Derating coefficient for temperature variation of a syn- Figure
7. Derating coefficient for a synchronous generator loaded
chronous generator at temperature above 40#C. with a
power factor different from 0.8.
modern
generators for diesel sets have a static or brushless
based on k coefficients can be applied. Figures 5, 6, and 7,
excitation system. In brushless alternators the excitation
give the k , k , and k derating coefficients, respectively, refer- power is
fed by an auxiliary alternator, coaxial with the main
1 2 3
ring to variations of temperature, altitude, and power factor.

generator, which has stationery fields and a rotating arma-
The total derated power can be calculated as the product of ture.
The rotating armature supplies the rotor winding
the rated power for reference conditions times the coefficients through
a three-phase rotating bridge rectifier as shown in
k , k , and k . Fig. 8.
The excitation power of the field winding of the auxil-
1 2 3
iary
machine is fed by the main generator. The auxiliary sta-
tionary
field is fed by a power supply controlled by an auto-
ELECTRIC PERFORMANCE AND CONSTRAINTS
matic
voltage regulator.
Regulating the fuel flow to the diesel engine, which generates
Frequency
electric power at constant speed, implements the power con-
trol. The control also regulates the current of the excitation The
voltage frequency of the electric power delivered by the
field of the synchronous generator in order to maintain a con- diesel-
generator is usually 50 or 60 Hz depending on the local
stant output voltage at both variable load and power factor.
standard. Such a frequency is determined by the rotational
The terminal voltage of synchronous generators can be con- speed
and by the number of pole pairs of the diesel generator;
trolled by altering the current in the field winding. An auto- its
value is maintained fixed by means of a speed regulator
matic voltage regulator (AVR) compares the terminal voltage when the
generator works alone. The frequency in this case
of the generator with a voltage reference, and the error is does not
depend on the amount of power (torque on the main
processed to drive a rotating or static exciter. A need for accu- axis)
delivered to the load. If more than one generator is par-
rate modeling arises from the fact that the excitation system alleled
on a common bus to carry a load, it is necessary to
forms a feedback loop around the generator and we must be allow
each generator to vary its frequency in order to share
careful not to introduce instability in the generator system.
correctly the power between all generators.
An IEEE Committee has developed models to represent exci-
tation systems (4,5).
Frequency and Voltage Regulations
The
speed regulator of the diesel engines can be mechanical
Excitation Systems
or
electronic. Regulators can work in droop mode or isochro-
Very old types of excitation systems contain auxiliary direct- nous
mode. Droop mode regulators allow a frequency regula-
current (dc) machines with lower power than the generator; tion of
about #2% between no-load to full load in steady-state
these machines feed the dc power to the rotor windings of the
conditions. Isochronous regulators guarantee a narrower reg-
synchronous machine by two copper rings fitted on the shaft. ulation
bounded by #0.5% for the same load variation. Volt-
The current flows through a system of brushes. Almost all the age
regulation in steady-state conditions allows a variation of
#1.5%
for loads ranging from 0% to 100% and power factor
between
0.8 and 1. In transient conditions the voltage drop is
prevented from exceeding 15% in the least favorable case of

2 1.10 sudden
application of a full load. In this case, the time needed
k
t to
return to within #3% of the steady-state voltage ranges
n
e
1.00
i from a
fraction of a second to several seconds.
c
i
f
f
e
o 0.90 Power
Factor
c
g
n
i Rated
powers (kilovolt-amperes) of generating units corre-
t 0.80
a
r spond to
a power factor of 0.8. At this power factor the active
e
D 0.70 rated
power of the synchronous generator corresponds to the
0 1000 2000 3000 4000
rated
power (kilowatts) of the diesel engine. The limit of
Height above sea level (m) power
of the diesel engine is the active power delivered to
Figure 6. Derating coefficient for a synchronous generator installed the load
while the limit of the synchronous generator is the
at altitude higher than 1000 m above sea level. apparent
power, as the limit is imposed by the current circu-
----------------------- Page 85-----------------------
Stator Rotor
Stator
Main armature windings Main field Rectifier Auxiliary
armature windings Auxiliary Detection and control
a windings
field
windings
Automatic
voltage
regulator
Figure 8. Simplified electric schematic of a three-phase

‘‘brushless’’-type generator, showing the
rotating bridge rectifier used to realize the excitation of
the synchronous machine, the automatic
voltage regulator, and the auxiliary excitation generator
delivering the excitation power.
lating in the stator winding. In case of variation between 0.8 cooling,

(4) exhaust pipeline, (5) noise reduction, (6) fuel sup-
and 1 the engine can always deliver the same rated power ply
systems, (7) electric wiring.
(which corresponds to the active power delivered by the gen-
erator). If the power factor is lower than 0.8, the apparent Building
Characteristics
power must be the same but the diesel will be underloaded
with reduction of the diesel efficiency. Consequently, as the A direct
connection by a large door of the generator room to
power factor decreases, the active power delivered by the syn- the
outside will facilitate the installation of the diesel set,
chronous generator must also decrease, with only the appar- while two
opposite walls should be used for intake of fresh air
ent power remaining constant. Conversely, if the power factor and to
exhaust hot air coming from the cooling system. Care
is greater than 0.8, as the power delivered by the diesel en- must be
taken to ensure good air circulation inside the gener-
gine must not remain higher than its rated power, the appar- ator room.
The sizing of the generator room will consider also
ent power (volt-amperes) which may be delivered to the load, the needs
for the fuel tank, control switch-gear, and aisles for
maintenance service.
must decrease. At a power factor equal to 1 the apparent
power of the generator will equal the active power delivered
at a power factor equal to 0.8. At low power factor values Foundation
Block and Noise Reduction
with respect to the rated value, the active power delivered is Internal
combustion engines, because they are reciprocating
reduced, while the excitation system is overcharged, which machinery,
are a typical source of vibrations. Thus we must
sets a limit owing to the thermal conditions of the field wind- isolate
the generating sets by noise-reducing mufflers. High
ing. If the diesel engine is undercharged (the active power is levels of
noise insulation are normally expensive, and a
less than the rated value), this yields too low a temperature trade-off
is generally accepted between cost and acceptable
in the combustion chamber where poor combustion may cause level of
noise, depending on the specific application and
the formation of liquid particles. These phenomena are partic- according
to the noise reduction standards. In case of
ularly dangerous in turbo-charged units, because the particles ground
installation, a heavy block of concrete having two
mixed with the exhaust gas may damage the turbine. So it is to three
times the weight of the generating set will ensure
preferable to provide an automatic power factor regulator in a reliable
foundation. The maximal horizontal dimensions of
order to avoid low-power-factor operating conditions. the
generating set may be used to establish the horizontal
dimensions
of the foundation block, and consequently the
Single-Phase Loads block
height can be easily calculated because the block
Single-phase loads can be fed by the generating set if the weight is

known.
phase current does not exceed the rated value. However, such
an imbalance will increase the voltage drop, which can exceed Air Supply
and Generating Set Cooling
the value guaranteed by the three-phase load regulator.
Fresh air
supply must be ensured in the generator room to
maintain
a comfortable temperature and to provide a fresh
GENERAL CHARACTERISTICS FOR INSTALLATION charge of
air for engine combustion. Good rules of practice
are as
follows: (1) Intake of fresh air should be far from
Successful installation of a generating set requires careful heat
sources; (2) possibly provide a direct connection of the
consideration of the following: (1) building characteristics, (2) engine
radiator to the outside by a duct, (3) fresh air should
foundation block design, (3) air supply and generating set flow
through a path with surfaces at increasing tempera-
----------------------- Page 86-----------------------

tures, (4) air flow should be at speeds less than or equal

SYNCHRONOUS GENERATOR MODELING
to 1.5 m/s.
Prediction of electromechanical behavior of synchronous gen-
erators can be performed by a system of equations describing

Exhaust Pipeline
the
synchronous machine (6,7). The rotors of synchronous
The exhaust pipeline must ensure low resistance to gas expul-
generators used in a diesel set are fabricated as salient pole
sion in order to avoid high back pressure on the engine ex-
structures, with magnetic anisotropy, carrying a field winding
haust system, which can cause (1) loss of engine power, and F
and a damper winding k. The voltage frequency of the gen-
(2) increase of both the combustion temperature and specific
erating groups is 50 Hz or 60 Hz and is given by f #
fuel consumption, which yields smoky exhaust and acceler- (NP/
120), with N the rotational speed per minute and P the
ated wear. We must isolate the pipeline in order to avoid the
number of poles of the machine. Diesel groups have at least
presence of very hot surfaces and exhaust leakage in the gen- four
pole machines, and for large power groups a higher num-
erator room. Rules of good practice for the design of exhaust ber
of poles are more suitable in order to lower the rotational
pipelines are as follows: (1) Avoid small diameter pipes; (2)
speed of the diesel engines. A suitable methodology used to
reduce as much as possible changes in pipe diameter; (3)
represent the electromagnetic phenomena in such machines
avoid pipe elbows with radius less than two times the pipe is
to transform the three-phase stationary winding, located on
diameter; (4) minimize the whole length of the pipeline; (5) the
stator, into a reference frame rotating at synchronous
include some flexible pipes which absorb noise and vibration
speed with two winding axes aligned with the direct axis and
and which also compensate for length increase due to thermal
quadrature axis in the rotor. A simplified electric equivalent
variation and mechanical inaccuracies.
circuit of a two-pole three-phase synchronous generator is
shown in Fig. 9. The stator windings in the actual machine

are
uniformly distributed in slots and have a mechanical dis-
Noise Reduction
placement of 120#. The rotor has a field winding and a damper

Diesel generating sets are strong sources of noise, and stan-
winding (squirrel-cage type). The angle # represents the dis-
dards give mandatory admissible acoustic levels to protect the
placement between the stator axis a (fixed) and the rotor axis
health of operators. Several methods of noise reduction can q
(rotating), and thus it is a function of speed # # d#/dt. The
be applied, depending on the goal, as a single means or as
voltage equations of the machine expressed in stator quanti-
associated means: (1) shielding by placing physical barriers ties
are much too complicated. In fact, the explicit time deriv-
between the sources and operators; (2) absorption by porous
atives of the flux linkages contain many terms that are func-
materials, which avoid reflection and reverberation close to
tions of the angle #, which is a function of time. The voltage
the source of the noise; (3) damping by flexible panels which
equations result in a system with time-varying coefficients.
cover the source of the noise; (4) isolation of the source of
Transformation in a rotating reference frame simplifies the
noise by elastic supports between the noise source and the
voltage equations and also results in equations which have
constant coefficients in time. The algebraic complexity, with-

frame rigidly connected to the ground; (5) suitable separation
out
applying such a transformation, can be seen by writing,
of the noise sources aiming to reduce the whole noise. The
for
the electric circuit depicted in Fig. 9, the voltage equation
main sources of noise are (1) the diesel engine, (2) the cooling
relative to phase a:
fans, (3) inlet and outlet air windows, (4) synchronous genera-
tor, and (5) pumps, compressor, and other auxiliary ma-
v
= −r i + p λ (1)
chinery.
a s a a
λ
= −(L i + L i + L i ) + L i + L i + L i (2)
a aa a ab b ac c af f akd kd akq kq
Fuel Supply System
where p # d/dt is the derivative operator, and the self-induc-

A small tank (50 to 100 L) is normally installed on the frame,
tance L aa and the mutual inductances (L ab, L ac, L af , L akd,
while an auxiliary tank should be located as close as possible and
L akq) are periodic functions of #; thus when the rotor is
to the generating group to ensure the daily stock of the fuel.
moving, L aa are functions of time, expressed by
The main tank should have a capacity adequate to power the
whole diesel station. Position and accessibility, as well as dis-
Laa = L# + LA − LB cos 2θ (3)
tance from heat sources, should be carefully evaluated. The
daily filling of the tank is performed by a level switch control-
1 L # 2π #
Lab = − A − LB cos 2θ − (4)

ling an electric pump. Correct operation of diesel engines re-
2 3
quires unspoiled fuel. The stored fuel may release solid parti-
1 L # 2π #
cles while allowed to fit undisturbed inside the tank, or it may
Lac = − 2 A − LB cos 2θ + 3 (5)
be filtered. The water content of the fuel must be separated

by a water separator device. It is necessary to ensure the fol-
Laf = Laf sin θ (6)
lowing in the fuel supply system: (1) A valve to stop fuel flow
Lakd = Lakd sin θ (7)
must be placed outside the generator room; (2) include ade-
quate pipeline junctions to prevent fuel leakage; (3) the fuel
Lakq = Lakq cos θ (8)
pipeline must be separated from electrical cable ducts; (4) fab-

ricate the fuel tanks with steel-based materials and shape with
both the leakage inductance L# of the windings and other
them to allow easy removal of the bottom residuals (water,
inductances (L , L , etc.), depending on the specific
A B
solid particles, etc.).

machine.
----------------------- Page 87-----------------------
vb ib
b
+
ikd
v i
c c
c rs rs
nkd
+
nkq
ns ns
rkd
rkq
nf
ns
ikq
rf
r
s
if
v i
a a
a
vf
+
+ –
Figure 9. Simplified electric equivalent circuit of a

two-pole three-phase synchronous generator.
The layout represents the stator windings, the
rotor field excitation, and the rotor damper
winding.
From inspection of Eqs. (1) to (8), it is clear that the ex-

The mechanical equation is
pression of the phase a voltage is complicated. In contrast, by
transforming the stator variables to the rotor reference frame,
Te = −J 2 p ω + TD (22)
all the time-varying inductances reduce to constant terms and
P
the voltage equations (Park’s equations) simplify noticeably,
resulting in dq0 quantities:
with # # p #.
By considering a steady-state sinusoidal behavior with a

v = −r i + ωλ + p λ (9)
terminal voltage Vt , equations that have been written for a
q s q d q
generic behavior reduce to

v = −r i − ωλ + p λ (10)
d s d q d
V = V sin δ = r I + X I (23)
d t s d q q
v = −r i + p λ (11)
0 s 0 0
V = V cos δ = −r I −X I + E (24)
# # # #
q r s q d d f
v = r i + p λ (12)
f f f f
0 = r# i# + p λ# (13)
The torque equation, expressed in terms of the terminal volt-
kq kq kq
age, is
0 = r# i# + p f λ# (14)
kd kd kd
where the prime on the variables and parameters of rotor
Te = 3 P EfVt sin δ + 1 # 1 − 1 # Vt2 sin 2δ (25)
2 2 X 2 X X
d d q
quantities means that they have been referred to the stator
by properly multiplying by the turns ratio of the windings.
where # is the rotor angle defined as the electric angular dis-
The flux linkages are expressed by
placement of the rotor with respect to the peak value of the ter-
minal voltage; Ef is the internal electromotive force; and X d and

λq = −L iq + Lmq (−iq + i# ) (15)
# kg
Xq are, respectively, the direct and quadrature reactances.
# #
The detailed model of synchronous generators, described
λ = −L i + L (−i + i + i ) (16)
d # d md d kd f
by Eqs. (9) to (22), contains two state variables relative to the
λ0 = −L# i0 (17)
mechanical part (position and speed of the rotor) and six state
# # # #
variables relative to the electromagnetic part. This model is
λ = L i + L (+i − i ) (18)
kq #kq kq mq kq q
used when there are large variations of the variables. In cer-
# # # # #
tain circumstances it would be unnecessary to adopt this
λ = L i + L (−i + i + i ) (19)
kd #kd kd md d f kd
model, and some approximations can be applied. One of the

# # # # #
λ = L i + L (−i + i + i ) (20)
common assumptions is that the rate of change in the stator
f #f f md d f kd
flux linkages is negligible, thus reducing the voltage equa-

where L # ###(L # L ), and L ###(L # L ) are the q- and
tions to algebraic relations. The same can be applied to the
mq A B md A B
d-axis magnetizing inductances, and L , L , L , and L

are flux linkages of the damper windings. As a further approxi-
#kq #kd # #f
leakage inductances.
mation the rate of change in the field flux linkage can be ig-
The electromagnetic torque can be calculated by
nored. Reduced models (8) can be used for electric transients
and electromechanical transients. Electric transients are de-

3 P
fined as behaviors which have fast constant times and do not
T = (λ i − λ i ) (21)
e 2 2 d q q d
involve speed variations of the generators. In the worst case,
----------------------- Page 88-----------------------
these can be used as a fourth-order model by neglecting the ies,

and to expel exhaust gases and induct fresh charge. The
time derivatives of the flux linkages and the mechanical equa-
mechanical efficiency can be defined as
tion. Electromechanical transients are defined as behaviors
which neglect only the time derivatives of fluxes in the volt-
ηm = Pb = 1 − Pf (28)
age equations. In such cases the model can be reduced to
P P
sixth order or lower. A complication in modeling of electrical
machines is caused by saturation in the magnetic paths of the where
Pb is the power measured at the brake, and Pf is the
machine. Modeling of transient (effect of the field winding) or power
equivalent to all losses. The mean effective pressure
subtransient (effect of the damper winding) behavior can be (mep)
is defined as the ratio of the work per cycle divided by
performed neglecting saturation. It is, however, essential to the
volume displaced for cycle:
include the saturation issue in the steady-state model to ob-
Wc
tain accurate initial conditions. Various methods for dealing
mep = (29)
V
with saturation exist, and the final choice is often governed
d
by the availability of reliable machine data.
The
capacity of an engine to convert the energy contained in
fuels
into mechanical energy can be evaluated by the fuel con-
DIESEL ENGINE MODELING
version efficiency, given by the ratio of the work per cycle di-
vided
by the energy of the fuel mass:
Diesel engine modeling is very complex due to the difficulties
in describing the engine internal processes by equations (2,3).
Wc P
Two basic types of models are used to predict performance
ηf = m QHV = pm QHV (30)
f f
and emissions of engines. Models based on equations derived
by the energy conservation principle are known as thermody-
where
QHV is the heating value of the fuel, mf is the mass of
namic-based models, while models based on equations de-
fuel,
and pm f is the rate of mass per unit time. The intake
scribing the fluid motion are named fluid-dynamic models.
system
efficiency can be evaluated by the volumetric effi-
Basic assumptions of the former models are knowledge of the
ciency
, defined as the ratio of the volume of the mass of air

mass transfer into (intake) and out (exhaust) of the cylinders,
V
inducted in the cylinder per cycle divided by the volume dis-

the energy release from the fuel, and the heat transfer be-
placed
by the piston:
tween the gases inside the cylinders and the inner surfaces of
the chamber. The calculated quantities are the work transfer
ma
and the pressure on the piston surface. To use the thermody-
ηv = ρa,iVd (31)
namic-based model, the data of the pressure p of the gas in-

side the cylinder and the volume V of the cylinder during a
where
#a,i is the inlet air density.
cycle can be plotted as a p–V diagram, in order to calculate
Substituting Eqs. (30) and (31) into Eq. (27), we get

the area enclosed within the cycle trajectory. Such an area is
the work transfer Wc per cylinder:
mf
η η nV Q ρ
f v d HV a,i
m
#
Wc = p dV (26)
P = 2 a (32)
From knowledge of the work, the masses of fuel and air in- and
the torque
ducted, and the engine speed, the performance parameters
m
can be derived. The power P developed per cylinder by a four-
f
η η V Q ρ
stroke diesel engine is expressed by
TD = P = f v d HV a,i ma (33)
2πn 4π
Wcn
P = (27)
2
Finally, substituting Eqs. Eqs. (30) and (31) into Eq. (29), we
get
where n is the rotational speed (rev/s) of the crankshaft.
m
The above is gross power, which differs from the brake
mep = η η Q ρ f (34)
f v HV a,i
power by the power required for friction, for engine accessor-
ma
e
Speed Governor
Engine
reference
Ki K2 pm f
–s τ T – 1 ω
K + K
e 1 K
+ – p s 1 + sτ 2 1
mep T + Js
PI regulator Actuator
Figure 10. Block diagram of the diesel engine and of its

governor. The generator electric torque
(Te) is represented as a disturbance.
----------------------- Page 89-----------------------
378 DIFFERENTIAL AMPLIFIERS
The fuel actuator is commonly represented as a first-order

DIFFERENCE SETS, THEORY. See THEORY OF DIFFER-
system characterized by a gain K and a time constant
. The
2 2
ENCE SETS.
command to the actuator yields a flow of fuel mass m , which

f
is linked to the mep by a time delay

.
1
Diesel Engine Governor
A governor controls the intake fuel flow in the engine, which

regulates the speed. The regulator is normally realized by a
PI controller which yields a proportional action to correct the
speed according to the operating conditions and an integral
action to nullify the steady-state error (9–11). In recent years,
some proposals have been presented to utilize adaptive regu-
lators instead of PI regulators in order to obtain faster re-
sponses, in case of load variation or disturbances (9). When
called into service, diesel prime movers are requested to oper-
ate at almost constant speed to obtain a constant frequency,
even in the case of transients caused by load insertions or
faults. The diesel model described above can be represented
by the block diagram shown in Fig. 10, which gives a simple
representation of engine behavior.
BIBLIOGRAPHY
1. IEEE Std 446-1995, IEEE Recommended Practice for Emergency

and Standby Power Systems for Industrial and Commercial Appli-
cations, New York: IEEE, 1996.
2. J. B. Heywood, Internal Combustion Engine Fundamentals, New
York: McGraw-Hill, 1988.
3. C. F. Taylor, The Internal-Combustion Engine in Theory and Prac-

tice, 2nd ed., Cambridge, MA: MIT Press, 1985.
4. IEEE Committee report, 1968, Computer representation of
excitation systems, IEEE Trans. Power Appar. Syst., June: 1968.
5. IEEE Committee report, 1973. Excitation system dynamic

characteristics, IEEE Trans. Power Appar. Syst., Jan./Feb:
1973.
6. P. C. Krause, Analysis of Electric Machinery, New York: McGraw-

Hill, 1995.
7. A. E. Fitzgerald, C. Kingsley, Jr., and S. D. Uman, Electric Ma-
chinery, New York: McGraw-Hill, 5th ed., 1990.
8. IEEE Std 399-1990, IEEE Recommended Practice for Industrial
and Commercial Systems Analysis, New York: IEEE, 1997.
9. S. Roy, O. P. Malik, and G. S. Hope, Adaptive control of speed
and equivalence ratio dynamics of a diesel driven power-plant,
IEEE Trans. Energy Convers., 8: 13– 19, 1993.
10. A. J. Tsitsovitis and L. L. Freris, Dynamics of an isolated power
system supplied from diesel and wind, IEE Proc., part A, 130:
587–595, 1983.
11. G. S. Stavrakakis and G. N. Kariniotakis, A general simulation
algorithm for the accurate assessment of isolated diesel-wind tur-
bines systems interactions, part I: A general multi-machine
power system model, IEEE Trans. Energy Convers., 10: 577–
583, 1995.
12. General Reference Manual of Standard Diesel Generating
Groups, AUSO MAN. 01 C. 1.000, 1996, Ausonia S.p.A., Mar-
sala, Italy.
ANTONIO LEOTTA
UBALDO NOCERA
Alternative Energies and Energy
Saving
ANGELO RACITI
University of Catania
----------------------- Page 90-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Diesel-Electric Power Stations

Standard Article

Antonio Leotta 1 1 2
, Ubaldo Nocera , Angelo Raciti
1Alternative Energies and Energy Saving, Catania, Italy
2University of Catania, Catania, Italy
Search
reserved.
DOI: 10.1002/047134608X.W3003
❍ Acronym Finder
Abstract
Internal Combustion Engines
Choice of A Diesel Generating Set
Parallel Operations of Multiple Diesel Generating Sets
Sizing of the Generating Groups
Electric Performance and Constraints
General Characteristics for Installation
Synchronous Generator Modeling
Diesel Engine Modeling

%20Conversion/W3003.htm17.06.2008 16:36:15
----------------------- Page 91-----------------------
} { { } }
HOME
ABOUT US
●
CONTACT US
HELP

Your Librarian
Profile
Electric Machine Analysis and Simulation
Standard Article

T. A. Lipo1
1University of Wisconsin, Madison, WI

reserved.
Search
DOI: 10.1002/047134608X.W3013
❍ Acronym Finder
Abstract
d–q–0 Representation of Three-Phase Quantities
d–q–0 Representation of Three-Phase Induction Machines
Simulation of Induction Machine Using Flux Linkages as State Variables
Modeling of Saturation
Simulation of Deep Bar Effect
Saturation Model with Currents as State Variables
Simulation of Synchronous Machine

Saturation Model of Wound Field Synchronous Machine
Other Machines

%20Conversion/W3013.htm17.06.2008 16:36:33
----------------------- Page 92-----------------------
ELECTRIC MACHINE ANALYSIS AND SIMULATION 313
d–q–0
REPRESENTATION OF THREE-PHASE QUANTITIES
The
highly coupled nature of induction and synchronous ma-
chines
had led to the use of artificial variables rather than
actual
(phase) variables for the purpose of simulation as well
as for
visualization. The essence of the nature of the transfor-
mation
of variables that is utilized can be understood by ref-
erence
to Fig. 1 which shows three-dimensional orthogonal
axes
labeled a, b, and c. Consider, for instance, the stator cur-
rents
of a three-phase induction machine which is, in general,
made
up of three independent variables. These currents
(phase
variables) can be visualized as being a single three-
dimensional vector (space vector) existing in a three-dimen-

sional
orthogonal space, that is, the space defined by Fig. 1.
The
projection of this vector on the three axes of Fig. 1 pro-
duce
the instantaneous values of the three stator currents.
However, in most cases, the sum of these three currents adds

up to
zero since most three-phase loads do not have a neutral
return
path. In this case, the stator current vector is con-
strained to a plane defined by
ia + ib + ic = 0 (1)
This
plane, the so-called d–q plane, is also illustrated in Fig.
1.
Components of the current vector in the plane are called
the d–q
components, while the component in the axis normal
to the
plane (in the event that the currents do not sum to
zero)
is called the zero component. When the phase voltages
and
phase flux linkages also sum to zero, as is the case with
most
balanced three-phase loads (including even a salient
pole
synchronous machine), this same perspective can be ap-
plied
to these variables as well. The components of the phase
current, phase voltage, or phase flux linkage vectors in the

d–q–0
coordinate system in terms of the corresponding physi-
cal
variables are
# 1 − 1 − 1#
# # # # 2 2 # # #
f f
q 2 # 3 √3 # a
# # # # # #
f = f (2)
# #
d 0 − b
# #
ELECTRIC MACHINE ANALYSIS AND SIMULATION

2 2
f f
0 # 1 1 1 # c
While an electrical machine exists for the bulk of its time in
√2 √2 √2
the steady state, it is during the brief period of transient, non-

stationary, behavior that most of the stresses occur which where f
denotes the current variable i, voltage v, or flux link-
limit the life of the machine. Because the differential equa- age .
tions of an electrical machine are nonlinear, a closed form In
the dominant case where the three-phase variables sum
solution for many of these transient conditions is impossible, to
zero (i.e., the corresponding vector is located on the d–q
and it is necessary to resort to time domain simulation of the
relevant differential equations. The modern era of electrical
machine simulation had its beginnings largely through the
d-q plane
efforts of Dr. Vannevar Bush of M.I.T. Over 70 years ago,
c -axis
Bush described a device called the integraph which realized
continuous integration by a principle related to that of the
b-axis
watt-meter (1). Within a year, Bush’s integraph was used in

the analysis of the pulsating torques of a synchronous motor–
compressor set (2). Hence, simulation techniques for modeling
transient behavior of ac machines was under development
a -axis
even before the classic papers of Park (3) and Stanley (4)
which developed the basic d–q model of the synchronous and
induction machine, respectively. Development of simulation
techniques has been ongoing since that time with almost 200 Figure
1. Cartesian coordinate system for phase variables showing
papers identified in a 1974 publication (5).
location of the d–q plane.

----------------------- Page 93-----------------------
314 ELECTRIC MACHINE ANALYSIS AND SIMULATION
plane), this transformation reduces to vectors.

For example, if axes are defined which rotate with
the stator
voltage vector, one realizes the synchronous voltage
# 3 # reference
frame. In general, it is not necessary to define rotat-
# # 0 0 # #
f f
q # 2 # a ing axes
to rotate synchronously with one of the vectors but
# # # # # #
# #
f = 1 1 f (3) simply to
define a general rotating transformation which
# # # #
d b
# √ √ #
0 −
f 0 # 2 2 # f c transforms
the phase variables rotating axes on the d–q
0 0 0 plane,
# #
where the last row is now clearly not necessary. Figure 2

# 2π # # 2π #
cos θ cos θ − cos θ +
# #
# #
shows the location of the various axes when viewed from the
# #
f (θ )
3 3 f
# # # #
q
a
d–q plane. Note that the projection of the a-phase axis on the
2 # 2π 2π #
# #
# # # #
f (θ ) =
sin θ sin θ − sin θ + f
d–q plane is considered to be lined up with the q-axis (the a- d
3 # # b
# #
3 3
f
f
# #
phase axis corresponds to the magnetic axis of phase a in the 0
1 1 1 c
case of an electrical machine). The other axis on the plane is,
√2 √2 √2
by convention, located 90# clockwise with respect to the q-
(5)
axis. The third axis (necessarily normal to the d–q plane) is
chosen such that the sequence d, q, 0 forms a right hand set.
Other notation, using symbols #, # (Clarke’s components), is For
completeness, the zero component is also given. The
sometimes used to denote these same variables. Also, with scale
change of #2/3 has also been included. The angle # is
the transformation shown, the d-axis is located 90# counter- the angular
displacement of the vector on the d–q plane mea-
clockwise with respect to the q-axis. These two axes are some- sured with
respect to the projection of the a-axis on this
times interchanged so that the reader should exercise caution plane.
Since the same scale change has been made for all
when referring to the literature. three
components, the zero component uses somewhat uncon-
When balanced sinusoidal three-phase ac voltages are ap- ventional
scaling. Specifically, Fortesque’s scaling for this
plied to such a load, it can be shown that the phase voltage component
was selected as
vector traces out a circle on this d–q plane with radius #3/2
V where V is the amplitude of the phase voltage. The vector
1
pk pk
f = (f + f + f ) (6)
rotates with an angular velocity equal to the angular fre-
0 3 a b c
quency of the source voltage (377 rad/s in the case of 60 Hz).

The current and flux linkage vectors, being a consequence of and is also
widely used.
applying the voltage to a balanced load, also trace out circles Note
that the zero axis does not enter into the rotational
on the d–q plane in the steady state. The fact that the length
transformation. Hence, the zero axis can be considered as the
of the vector differs from the amplitude of the sinusoidal vari- axis about
which the rotation takes place, that is, the axis of
able has prompted methods to correct this supposed defi- rotation.
Because of the scaling, the power (and subsequently,
ciency. Specifically, if the transformation of Eq. (3) is the torque)
is different in d–q–0 components than a–b–c vari-
multiplied by #2/3, a scale change is made in moving from ables, and
a 3/2 multiplier must be added to the power in
a–b–c to d–q–0 variables. The transformation becomes the
transformed system of equations since both current and
# #
voltage
variables have been scaled by #2/3.
# # # #
1 0 0
f q f a In vector
notation, Eq. (5) can be written as
# #
1 1
# # # √ √ # # #
f d = 0 − f b (4)
# 3 3#
f T f
f 0 0 0 0 f c
ff qd 0 = TTqd 0 (θ )ff abc (7)
The visualization of vector rotation on the d–q plane has where

also led to transformations which serve to rotate with these
# #
# 2π # # 2π #
cos θ cos θ − cos θ +
# #
3 3
b-axis
2 # # 2π # # 2π # #
T
# #
TT (θ )
= sin θ sin θ − sin θ + (8)
qd 0
3 # 3 3 #
# #
# #
1 1 1
√2 √2 √2
0-axis
(Normal to q-axis a -axis The
transformation Tqd0(#) can, for convenience and for com-
paper) putational
advantage, be broken into two portions, one of
which takes
variables from physical phase quantities to non-
rotating d–
q–0 variables (stationary reference frame) and
then from
nonrotating to rotating d–q–0 variables (rotating
reference
frame). In this case, one can write
d-axis
c -axis
f T f R T f
Figure 2. Physical and d–q–0 axes when viewed on the d–q plane.
ff qd 0 = TTqd 0 (θ )ff abc = RR (θ )TTqd 0 (0)ff abc (9)
----------------------- Page 94-----------------------
where
( – )ω ω λ′
ωλ L L ′ r qr
r qs ls lr r ′
s r
– + + –
# 2 − 1 − 1#
+ i i′ +
# #
3 3 3
ds dr ′
v
L v
ds
# #
1 1
m dr
# #
√ √
T 0 −
TT (0) = (10)
# #
qd 0 3 3
– –
# #
√ √ √
# 2 2 2 #
d -axis equivalent circuit
3 3 3
( – )ω ω λ′
ωλ L L ′ r dr
r ds ls lr r ′
and
s + – – + r
+ +
# #
i i′
cos θ − sin θ 0
v qs L qr v′
qs m qr
# #
R sin θ cos θ 0
RR (θ ) = (11)
– –
0 0 1
q-axis equivalent circuit
Note that Tqd0(0) is obtained by simply setting # # 0 in Eq.

(8). The inverse transformation is
r Lls Llr′ r ′
s r
+ +
−1 −1 −1
+ +
f T f T R f
ff abc = TTqd 0 (θ ) ff qd 0 = TTqd 0 (0) RR (θ ) ff qd 0 (12)
i0s No i′0r ′
v0s v 0r
coupling
where
– –
0-axis equivalent circuit

3
T −1 T T
TTqd 0 (θ ) = 2 TTqd 0 (θ ) Figure
3. d–q–0 equivalent circuit of an induction machine repre-
# #
1 sented
in a rotating reference frame.
cos θ sin θ √
# #
2
# # # #
(13)
# 2π 2π 1 #
# #
= cos θ − sin θ − √
# 3 3 #
# #
2
Except fornotational differences, the parameters in this
# # 2π # # 2π # 1 #
circuit are essentially the same as the conventional per phase
cos θ + 3 sin θ + 3 √
equivalent circuit. That is, r , L , r#, L #, and L correspond to
2
s ls r lr m
the
per phase stator resistance, stator leakage inductance, ro-
# 1 # tor
resistance, rotor leakage inductance, and magnetizing in-
1 0 √
# #
2
ductance, respectively [typically labeled as R , L , R , L , and
1 1 2 2
# √ #
# #
−1 3 T 1 3 1 L#
(or Lm) respectively]. In most cases, the impressed rotor
# #
T T − −
√
TTqd 0 (0) = TT (0) = (14)
2 # 2 2 2 #
voltages are identically zero (squirrel cage machine) and will
# √ #
# #
1 3 1 be
assumed henceforth herein as zero. The zero sequence cir-
−
√
2 2 2 cuits
are included for completeness but are seldom necessary
and
will now also be omitted from further consideration. The
# #
cos θ sin θ 0 primes

used for the rotor variables are included as a reminder
−1 T
# #
R R − sin θ cos θ 0 that

the physical variable has been referred to the stator by
RR (θ ) = RR (θ ) = (15)
0 0 1 the
stator/rotor turns ratio in much the same manner as for
a
transformer. The use of these primes is often dropped for
convenience.
d–q–0 REPRESENTATION OF THREE- The
voltage generators in the circuit represent speed volt-
PHASE INDUCTION MACHINES ages
which appear due to the fact that the circuit is being
solved
in a rotating reference frame. This term is to be ex-
The benefits of visualizing three-phase variables as a vector pected
of any physical system represented in a rotating
represented in a nonphysical coordinate system become ap- frame.
(Recall from mechanics the # # r term representing
parent only when the system equations of a coupled three- the
relative velocity of a stationary point in rotating system.)
phase magnetic component such as a reactor, transformer, or The
speed voltages in the stator portion of the circuit are pro-
motor are represented in terms of these newly defined vari-
portional to the reference frame angular velocity #(# #
ables. The equivalent circuit of an induction motor repre- d#/dt)
since the circuits, themselves, are stationary. The
sented in a rotating reference frame is shown in Fig. 3. Here, speed
voltages in the rotor portion are proportional to # # #r
the second subscripts ‘‘s’’ and ‘‘r’’ are used to denote ‘‘stator’’ since
the rotor circuits, themselves, are rotating at an electri-
and ‘‘rotor’’ quantities, respectively. The enormous simplicity cal
angular velocity of # (# # d# /dt, where # is the angular
r r r r
afforded by this equivalent circuit can be better appreciated

rotation of the rotor in electrical degrees). That is, the relative
if it is mentioned that the original circuit defined in physical
angular velocity appears in this case. The electrical angular
variables involves the mutual coupling among all six circuits
displacement is related to the actual physical angular rotor
(three stator and three rotor) with 36 consequent mutual and
displacement #rm by #r # (P/2)#rm , where P is the number mag-
self inductance terms. netic
poles of the machine.
----------------------- Page 95-----------------------

The differential equations corresponding to the circuits of
written
Fig. 3 are (neglecting the zero components),
T − T = # 2 # J dωr (30)
dλ
em load P dt
v = r i + qs + ωλ (16)
qs s qs dt ds
dλ
where Tload is the load torque, and J is the inertia in SI units.
ds
v = r i + − ωλ (17)
ds s ds dt qs
dλ#
SIMULATION OF INDUCTION MACHINE USING
# # # qr + (ω − ω #
v = 0 = r i + r )λ (18) FLUX
LINKAGES AS STATE VARIABLES
qr r qr dt dr
dλ#
# # # dr − (ω − ω #
Since the differential equations of the machine, Eqs. (16–19),
v = 0 = r i + r )λ (19)
dr r dr dt qr
contain mixed variables (i.e., flux linkages and currents), ei-
ther
of these two quantities could be eliminated from the dif-
where the flux linkages are defined by
ferential equations by means of the algebraic relations, Eqs.
(20–
25). The traditional approach to simulation is to consider
λqs = Lls iqs + λmq (20) the
flux linkage as the state variables and currents as depen-
dent, algebraically related variables (6). Proceeding in this

λds = Lls ids + λmd (21)
manner, the currents can be solved in terms of the flux link-
λ# = L# i# + λmq (22) ages

as
qr lr qr
# # #
λ − λ
λdr = Llr idr + λmd (23)
iqs = qs mq (31)
ls
and
λ − λ
ids = ds md (32)
ls
λmq = Lm (iqs + i# ) (24)
qr
λ# − λ
#
i# = qr mq (33)
λmd = Lm (ids + idr ) (25)
qr L#
lr
λ# − λ
Note that while not necessary to be defined explicitly, the mu-
i# = dr md (34)
dr L#
tual (air gap) flux components mq and md have been included
lr
to aid the simulation process.
1 # λ# #
λ
The torque produced by the machine can be identified as
λmd = ds + dr (35)
1 1 1 L L#
the power consumed by the voltage generators in Fig. 3 di-
L + L# + L ls lr
vided by the actual rotor speed. Multiplying these voltage
ls lr md
generators by their respective currents,
1 # λ# #
λmq = qs + qr (36)
1 1 1 L L#
# # # #
T = 3 P 1 [ω(λ i − λ i )
Lls + L# + Lmd ls lr
em ds qs qs ds
lr
2 2 ωr (26)
+ (ω − ωr )(λ# i# − λ# i# )]
dr qr qr dr
These results can be inserted into the differential equations.
# # # #
3 P Upon
solving for the time derivative terms and integrating,
= (λ# i# − λ# i# ) (27) the
result is
2 2 qr dr dr qr
R # rs #
λqs = vqs + (λmq − λqs ) − ωλds dt (37)

The 3/2 term occurs because of the scale change taken during
Lls
the d–q–0 transformation. The equation has two useful
equivalent forms,
R # rs #
λds = vds + (λmd − λds ) + ωλqs dt (38)
ls
# # # #
# # #
3 P
r
T = (λ i − λ i ) (28)
# r # #
em ds qs qs ds
λ = (λ − λ ) − (ω − ω )λ dt (39)
2 2
qr L# mq qr r dr
lr
R # r# #
or
λ# = r (λ − λ# ) + (ω − ω )λ# dt (40)
dr L# md dr r qr
lr
# # # #
# #
3 P
T = L (i# i − i# i ) (29)
P 1
em m dr qs qr ds
ω = (T − T ) dt (41)
2 2
r 2 J e load
Finally, the machine must be physically tied to an external

Equations (28, 31–34) form the necessary equations to
load/prime mover in order to achieve energy conversion. In its
simulate a squirrel cage induction machine. A block diagram
simplest form (neglecting mechanical damping), the equation of
showing the flow of information which can be arranged in
which couples the electrical to the mechanical world can be a
suitable simulation language such as MATLAB or ACSL is
----------------------- Page 96-----------------------
vqs 1/p λ qs
λ ds x (Eq. 37)
i
qs
(Eq. 31)
λ
mq
ω (Eq. 36)
i′
qr
(Eq. 33)
x
λ dr′ λ ′
1/p qr
(x – y )*z
z (Eq. 39)
y
λ
qs
T
i em
ω qs
r (Eq. 28)
λ ds
ω
1/p r
ids
(Eq. 41)
vds 1/p λ ds Tload
λ (Eq. 38)
qs x
ids
(Eq. 32)
λ md
ω (Eq. 35)
idr
′
(Eq. 34)
x
λ ′
qr 1/p λ dr′
(x – y )*z
z (Eq. 40)
y
Figure 4. Flow of signals for simulation of a
squirrel cage induction machine in a rotating

ω
r
d–q–0 representation.
shown in Fig. 4. The term 1/s denotes integration with re-

quantities, the stator phase voltages across the machine are
spect to time.
While Fig. 4 forms the simulation model of the induction
vas = eag − vsg (42)
machine, the external inputs, namely vqs and vds , must be de-
vbs = ebg − vsg (43)
fined. These inputs vary, of course, from problem to problem
but can be represented in a general way by the circuit shown
vcs = ecg − vsg (44)
in Fig. 5. The voltages eag , ebg, and ecg are assumed to be known
from another portion of the overall system simulation. For
Upon adding these three voltages,
example, these three voltages could correspond to the phase
to negative dc pole voltages of a three-phase PWM inverter,
vas + vbs + vcs = eag + ebg + ecg + 3vsg (45)
the output voltages of a generator, or any of a variety of other
waveforms obtained either implicitly through simulation or as
If we let Z(p ) denote an arbitrary load impedance which can
explicit functions of time. Assuming these voltages as known
even be nonlinear provided that it does not vary with stator
----------------------- Page 97-----------------------
even if the
load is a symmetrical three-phase induction ma-
a a
chine or even a salient pole synchronous machine.
+
The phase voltages can now be solved in terms of the
eag Z (p )
known source voltages as
–
– +
g s
2 1 1
– ebg b Z (p ) Z (p )
vas = 3 eag − 3 ebg − 3 ecg (48)
e
cg
1 2 1
c + c
vbs = − 3 eag + 3 ebg − 3 ecg (49)
1 1 2
Figure 5. Three-phase wye connection having source voltages deter-
vcs = − 3 eag − 3 ebg + 3 ecg (50)
mined external to the motor. The point ‘‘g ’’ is at an arbitrary (not

necessarily ground) potential. or
equivalently,
# 2 − 1 − 1#
current then, assuming equal impedances in all phases,

# # # 3 3 3 # # #
v e
as ag
# 1 2 1 #
# #
# # # #
v e Se
= v = = SeSe (51)
# #
v = Z (p )i
bs abcs − − bg abcg
# #
as as
3 3 3
v e
vbs = Z (p )ibs
cs # 1 − 1 2 # cg
3 3 3
vcs = Z (p )ics
A block diagram illustrating the procedure for developing

where p denotes the differential operator p # d/dt, and
1/p # #(
) dt so that the motor
d–q voltages is shown in Fig. 6. The two blocks
denoted by Eq. (51) and Eq. (10) can be readily solved to form
vas + v + vcs = Z (p )(ias + i + ics ) (46)
# #
bs bs
2 1 1
− −
3 3 3
Hence, when the sum of the three load currents equals zero,
# #
T S # 1 1 #
TT (0)SS = (52)
the sum of the phase voltages also sum to zero. Equation (45)
qd 0 # #
0 −
becomes
# √3 √3 #
0 0 0
1
vsg = 3 (eag + ebg + ecg ) (47)
The last row of zeros show that the zero sequence component
of
voltage is zero. That is, the zero component of voltage is
While this result has been illustrated for simple passive im-
impressed across the open circuit between points s and g and
pedances, it can be shown that the same conclusion is true
not across the zero sequence circuit of the machine itself
i s , i s
qs ds
i , i , i
as bs cs
i i
qs, ds
–1 –1
Tqdo (0) R ( )θ
(Eq. 14) (Eq. 15)
External
s s
vqs vds vqs vds

system
Induction
S Tqdo (0) R ( )θ machine
(Eq. 51) (Eq. 10) (Eq. 11)
(Fig. 4)
sinθ cosθ
e , e ,
e v , v , v
ag bg
cg as bs cs
– +
1/p x 1/p x
Figure 6. Typical simulation of a wye connected

ω
squirrel cage induction machine including modeling
of source voltages.
----------------------- Page 98-----------------------
(shown in Fig. 3). In the event that the three source voltages
Air gap line
ω ]
also sum to zero, we have, finally,
[slope = L m (unsat)
# #
ω L l
1 mag
1 0 0
Vl-n
# #
# #
1 1
T S
TT (0)SS = (53)
# #
0 −
√ √
qd 0
# 3 3 #
Vl-n, Vgap V
0 0 0
(V, rms) gap R 1 j Lω 1 j Lω 2
+ +
In general, the reference frame velocity can be selected to
Vl-n Vgap j Lω m R /S
be any explicit or implicit function of time. The speed of the
2
reference frame is typically chosen to best suit the problem
– –
under investigation. For example, if the simulation requires
I mag (A , rms)
modeling piecewise linear or nonlinear elements such as
semiconductor switches, then the reference frame must be
(a)
constrained to rotate either with the stator or the rotor de-
pending upon where the switches are located. When a simple
balanced three-phase sinusoidal operation is investigated, a
synchronous reference frame can be used and often adds in-
Air gap line ∆λ
sight into the problem being investigated. In motor control
m
problems such as field orientation, it is possible to fix the ref-

erence frame on a vector corresponding to a variable such as
λ m(sat)
the stator current or rotor flux vector. In the large majority
(V/s , pk )
of cases, a simulation in the stationary reference frame, how-
ever, suffices in which case # is constant. If # # 0, R becomes
the identity matrix and can be eliminated since the signals
45°
pass directly through the block without modification. In Fig.
6, the reference frame velocity # is shown as coming from the
= L I (V/s , pk )
λ
external system as would be the case if the synchronous volt-
m(unsat) m (unsat) mag
age rotating reference frame were used. Krause and Thomas
(b)
(7) give an excellent treatment of simulation techniques to be
employed when series connected semiconductor switches open
Figure 7. (a) No-load saturation curve. (b) Derived curve.
and close, producing temporary open circuit conditions in the
phases.
abscissa of Fig. 7(a) is multiplied by Lm(unsat) and the ordinate

MODELING OF SATURATION by
1/#e , the normalized curve of Fig. 7(b) results in a new plot
in
which the abscissa remains proportional to MMF (but
Probably the most important effect absent from the model de-
scaled in terms of the unsaturated value of flux linkages) and
veloped thus far concerns saturation of the magnetic core. In the
ordinate equal to the saturated value of flux linkages. The
most cases, the saturation of the teeth dominate, in which
slope at the air gap line is now clearly unity. The difference
case saturation can be taken into account accurately by ex-
between the saturated and unsaturated values of flux linkage
can
be defined as # .
pressing the air gap flux linkage as a nonlinear function of
m
the air gap MMF. While the air gap MMF is difficult to deter-
The quantity #m can now be plotted as a function of the
mine under a loaded condition, the required relationship can
unsaturated value of air gap flux linkages m(unsat) . Since satu-
be established if the motor is operated under a unloaded con-
ration does not result in a phase shift in the fundamental
dition in which case the MMF is clearly proportional only to
component of flux linkages and only decreases the amplitude,
the stator current since the rotor current is, in this case, zero.
both the d- and q-components of saturated air gap flux should
If the no-load voltage is plotted versus the no-load current, be
decreased by the same value. Thus,
the saturation curve of Fig. 7(a) can be established. Neglect-
λ
ing stator resistance, the slope of a line drawn from the origin
#λmd = md (unsat) #λm (54)
to a point on the curve is proportional to the sum of the stator
λm (unsat)
leakage plus magnetizing reactance # (L # L ), or (# (L #
e 1 m e ls
λ
Lm) in d–q notation) where #e is the angular frequency of the
#λmq = mq (unsat) #λm (55)
λ
source voltages. If the leakage reactances of the machine have
m (unsat)
been measured by locked rotor test or calculated, the voltage
drop due to magnetizing current flow in the stator leakage
Saturation in the q-axis can be incorporated if Eqs. (20) and
inductance branch can be subtracted from the terminal volt-
(22) are modified to form
age to obtain the voltage at the air gap. The slope of the air
gap voltage versus magnetizing current is clearly the mag-
λqs = Lls iqs + λmq (sat) = Lls iqs + λmq (unsat) − #λmq (56)
netizing reactance # L . The slope of the linear portion line
e m
λ# = L# i# + λmq (sat) = L iqr + λmq (unsat) − #λmq (57)
(air gap line) yields the unsaturated value of #eLm(unsat) . If the
qr lr qr lr
----------------------- Page 99-----------------------
λ
md(sat)
– ∆ λ md λmd (unsat)
m (unsat)
* 1 1
L +
m (L L ′ )
ls lr
*
L
λ ds m
Lls
+
+
(⋅)2
2
L * +
λ λ m (unsat)
λ ′ m
md (unsat)
dr ′
Llr
∆ λ m
+ λ
m (unsat) ∆ λ
+
*
λ
L
m (unsat)
λ m
qs Lls +
λmq (unsat)
(⋅)2
+
*
L
λ ′ m
qr L ′
1 1
lr
*
L +
m (L L ′ )
ls lr
λ
+
Figure 8. Block diagram for the proce- mq(sat)
∆ λ md λmd (unsat)
dure to calculate saturated air gap flux

– λ m (unsat)
linkages mq(sat) and md(sat) .
When combined with Eq. (24), the q-axis portion of unsatu-

layers. The machine can then be simulated by modeling the
rated value of flux linkage is
equations defining the circuit of Fig. 9. A treatment of model-
ing of machines with deep bar effect is given in (9). A method

# 1 1 #
to establish the parameters of the equivalent circuit of Fig. 9
λmq (unsat) = + 1 + 1
is given in (10).
Lm Lls L# (58)
lr
# λ# # 1 # #
λ 1
qs qr
SATURATION MODEL WITH CURRENTS AS STATE VARIABLES
L + L# + L + L# #λmq
ls lr ls lr
In recent years, numerous papers have been written concern-

A similar result holds for the unsaturated value of d-axis air
ing the simulation of saturated induction machines proposing
gap flux linkage. A block diagram showing the overall flow of
the use of currents as the model state variables [e.g., (11–
signals to model induction motor saturation is given in Fig. 8
(8) where, for convenience, we have defined
ω λ L L
1
r ds mr 1 mr2
L∗ = (59)
m
1 1 1
+ – +
+ +
rs Lls
Lm Lls L#
Llr 1 Llr2 Llr3
lr
ds L
SIMULATION OF DEEP BAR EFFECT
rr 1 rr2 rr3
–
Another very important phenomenon in squirrel cage induc-
d -axis
tion machine concerns the uneven distribution of currents in
ω λ L L
the rotor bars, termed deep bar effect. Because a filament of
r qs mr 1 mr2
current experiences a greater inductance at the bottom of the
+ + –
bar than on the top portion, the current tends to rise to the
rs Lls
top of the bar facing the air gap, resulting in greater torque
Llr 1 Llr2 Llr3
v
as well as higher losses at a given slip frequency. This phe-
qs Lm
nomenon is frequently used to improve the starting perfor-
rr 1 rr2 rr3
mance of a squirrel cage machine since the effect is greatest
–
under the starting condition due to the fact that the bar re-
actance is greatest at this point.
q-axis
Simulation of the deep bar phenomenon is readily accom-

Figure 9. Simulation of a squirrel cage induction motor with deep
plished by breaking up all of the bars of the rotor into equal
bar effect modeled with three bar sections—rotor reference frame.
----------------------- Page 100-----------------------
13)]. In the process of analysis, a so-called cross-saturation

solved for the currents and air gap flux linkages as
phenomenon has been identified which was supposedly ne-
λ − λ
glected prior to this time. In reality, prior to 1981, flux link-
iqs = qs mq (70)
ages rather than currents were used to model core saturation
Lls
primarily to avoid the difficulties addressed in these refer-
λ# − λ
ences. It has been demonstrated that the solution of the two
i# = qr mq (71)
qr L#
methods are identical (8). Because of the complexity of the
lqr
simulation (involving inversion of a 4 # 4 matrix every time
λ − λ
step), the method is not recommended.
ids = ds L md (72)
ls
λ# − λ
i# = dr md (73)
SIMULATION OF SYNCHRONOUS MACHINE
dr L#
ldr
λ# − λ
Wound field and permanent magnet synchronous machines
i# = f r md (74)
f r L#
can be modeled by use of the same d–q–0 transformation
lfr
# #
used for induction machines. However, in this case, since the

i i# i#
rotor is not symmetric, wound field machines must be mod-
λ = L∗ ds + dr + f r (75)
md md L L# L#
eled in a reference frame rotating with the asymmetry (i.e., a
ls ldr lfr
reference frame rotating with the rotor) in order to simplify
# iqs i# #
the coupled equation which exists in phase variable form.
λmq = L∗ + qr (76)
mq L L#
The d–q–0 differential equations depicting behavior of a
ls ldr
wound field synchronous machine are
where
dλ
v = r i + qs + ω λ (60)
1
qs s qs dt r ds
L∗ = (77)
md 1 1 1
dλ
+ # + #
ds
L L L
v = r i + − ω λ (61)
md ldr lfr
ds s ds dt r qs
1
dλ#
L∗ = (78)
0 = r# i# + qr (62)
mq 1 1
qr qr dt
+
L L#
#
mq lqr
dλ
0 = r # i# + dr (63)
dr dr dt
The equation for the electromagnetic

torque is the same as
dλ#
for the induction machine, Eq. (28) but not Eq. (29), because
v# = r # i# + f r (64)
of the nonsymmetrical rotor. A block diagram showing flow of
f r f r f r dt
data for purposes of simulation is shown in Fig. 11.
The dr and qr circuits are called the amortisseur windings

(‘‘killer’’ or damper windings) and are physically realized by a
ω λ L′ r′
shorted squirrel cage constructed in much the same manner
r ds lfr fr
as the squirrel cage of an induction machine (often labeled as
+ – + +
kd and kq). The last equation corresponds to the excited rotor
rs Lls
field winding. Primes are again used as a reminder that the
L′
rotor circuits have been referred to the stator by the appro-
v ldr v′
ds fr
priate turns ratio. Note that Eqs. (60–63) are identical in
Lmd
form to the induction motor equation except that the speed of
r′
dr
the reference frame # has been set equal to the speed of the
– –
rotor #r , and that the rotor has asymmetry (rqr # rdr ).
d -axis
The flux linkages are related to the currents by
ω λ
r ds
λqs = L iqs + Lmq (iqs + i# ) = L iqs + λ # (65)
ls qr ls mq
+ + –
# # # # # #
rs Lls
λqr = Llqr iqr + Lmq (iqr + iqs ) = Llqr iqr + λmq# (66)
L′
λ = L i + L (i + i# + i# ) = L i + λ (67)
v lqr
ds ls ds md ds dr f r ls ds md
qs
mq
λ# = L# i# + L (i + i# + i# ) = L# i# + λ (68)
dr ldr dr md ds dr f r ldr dr md
r′
qr
λ# = L# i# + L (i + i# + i# ) = L# i# + λ (69)
–
f r l f r f r md ds dr f r l f r f r md
q-axis
An equivalent circuit of this machine can be established from
Figure 10. d–q–0 equivalent circuit of a wound field synchronous
these equations as shown in Fig. 10. These equations can be
machine.
----------------------- Page 101-----------------------
v
qs
λ
1/s qs
λ ds
x (Eq. 60)
qs
(Eq. 70)
mq
(Eq. 76)
i ′
qr
(Eq. 71)
ω
r
λ ′
1/s qr
(Eq. 62)
qs
i em
qs (Eq. 28)
λ ds 1/s ω r
ids (Eq. 30)

vds
1/s λ ds Tload
λ
(Eq. 61)
qs
x
ids
(Eq. 72)
λ md
(Eq. 75)
i ′
dr
(Eq. 73)
λ ′
1/s dr
(Eq. 63)
i ′
fr
(Eq. 74)
λ ′
1/s fr
v ′
fr
(Eq. 64)
Figure 11. Simulation flow diagram for wound field syn-

chronous machine.
It should be mentioned that the circuit of Fig. 11 is ade-

field poles of a salient pole synchronous machine. Hence, sat-
quate only when predicting stator currents but fails when ac-
uration becomes primarily determined by the flux in only one
curate portrayal of the rotor currents is desired. In this case,
of the two magnetic axes (d–q axes), namely the d-axis. Deri-
more detailed models are required which include the fact that
vation of the unsaturated flux linkage versus saturated flux
there exists a flux component which links both the d-
axis linkage characteristic is done in the same manner
by ob-
damper winding and field winding which does not enter the
taining first the open circuit saturation curve (see Fig.
7).
air gap and, therefore, does not link the stator windings. In
Derivation of the equations expressing saturation in this case
this case, the reader is referred to (14).
is very simple but follows the induction machine example ex-
plained earlier. A block diagram of the resulting equations is

SATURATION MODEL OF WOUND
shown in Fig. 12.
FIELD SYNCHRONOUS MACHINE
In cases where saturation occurs in the stator rather than,
or in addition to, the rotor, the effect must be modeled by

Because of the heavy excitation current (ampere turns) usu-
several saturation functions. A good discussion of this prob-
ally employed, saturation normally occurs first within
the lem is given in (15). High-speed synchronous motors and tur-
----------------------- Page 102-----------------------
λ
md (sat) – ∆ λ md
* 1 1
L +
L * m (L L ′ )
ls l dr
λ ds m
Lls ∆ λ md
+ + λ md(unsat)
∆ λ m
+
*
L +
λ ′ m
dr λ
L ′
md(unsat)
ldr
*
L
λ ′ m
fr
L ′
Figure 12. Saturation model for a salient pole
lfr
wound field synchronous machine.
bogenerators are constructed with a round rotor which is typi-

the machine in a nonrotating (stationary) frame of reference.
cally not laminated. Because of the eddy currents which flow
Reference 18 is a good place to begin concerning simulation
in the rotor body, the saturation phenomenon is complicated.
of permanent magnet machines.
While saturation is often still modeled as in Fig. 12, a more
Finally, various types of reluctance machines are also re-
detailed model is needed for good correlation with physical
lated to the synchronous machine. The synchronous-reluc-
tests. Reference 16 is a good beginning point.
tance machine has a conventional three-phase stator struc-
ture similar to an induction or synchronous machine but has
a special rotor to enhance the reluctance torque produced by

OTHER MACHINES
a salient pole rotor structure. The equivalent circuit and its
modeling is the same as a wound field synchronous machine

While emphasis has been placed on the three-phase squirrel
except that the field winding is omitted. The variable reluc-

cage induction and salient pole wound field synchronous ma-
tance machine is unusual in that it has both a salient pole

chines, many of the other common machines are close rela-
stator and a rotor thereby invalidating the common develop-

tives of these two basic machine types. For example, a single-
ment of all of the other ac machines up to this point. The

phase induction machine is, effectively, a two-phase induction
subject is a special one to which the reader is referred to the

machine with an unequal number of turns in the two phases.
literature. If simulation techniques for this machine are of

Simulation of this machine closely follows the d–q model pre-
particular importance, (19) is a recommended starting point.

sented in Fig. 4 with a turns ratio correction on the auxiliary
Three books that treat the overall subject of analysis and sim-
winding to yield the equivalent of a balanced two-phase in-
ulation of ac machines in some detail are listed as (20–22).

duction motor. Details of the derivation are given in (17). Per-
manent magnet motors can be modeled by using the synchro-
nous machine d–q model with the d-axis circuit modified, as
BIBLIOGRAPHY
shown in Fig. 13. In this case, since a ferrite or rare-earth
magnet has a relative permeability nearly that of air, Lmag
1. V. Bush, F. D. Gage, and H. R. Stewart, A continuous integraph,
represents the inductance due to the magnet itself. The induc-
Franklin Inst. J., 203: 63–84, Jan. 1927.
tance Llmag accounts for the leakage flux produced by the mag-
2. L. Teplow, Stability of synchronous motors under variable-torque
net but can be neglected since the magnet, itself, is modeled
loads as determined by the recording product integraph, Gen.
as a current source. If the rotor of the machine is not
Elec. Rev., 31 (7): 356–365, July 1928.
equipped with a cage, the cage windings can be simply re-
3. R. H. Park, Two-reaction theory of synchronous machines, Part
moved from the circuit and their corresponding circuit equa-
1—Generalized method of analysis, AIEE Trans., 48: 716–727,
tions eliminated. In this case, it is even possible to simulate
1929.
4. H. C. Stanley, An analysis of the induction motor, AIEE Trans.,
57 (Supplement): 751–755, 1938.
r ω λ L L
5. P. C. Krause, T. A. Lipo, and D. P. Carroll, Applications of analog
s r ds ls lmag
and hybrid computation in electric power system analysis, Proc.

+ – +
IEEE, 62: 994–1009, 1974.
L L′ L
6. C. H. Thomas, Discussion of ‘‘Analogue Computer Representation
md ldr mag
of Synchronous Generators in Voltage Regulation Studies’’ by M.

v I
ds mag
Riaz, Trans. AIEE (Power Apparatus and Systems), 75: 1182–
r′
1184, 1956.
dr
–
7. P. C. Krause and C. H. Thomas, Simulation of symmetrical in-
duction machinery, IEEE Trans. Power Appar. Syst., PAS-84:

d -axis
1038–1053, 1965.
Figure 13. Equivalent d-axis circuit for three-phase permanent mag-

8. M. Osama, K. Sakkoury, and T. A. Lipo, Transient behavior com-
net machine.
parison of saturated induction machine models, IMACS-TC1-93,
----------------------- Page 103-----------------------
324 ELECTRIC NOISE MEASUREMENT

(Computational Aspects of Electromechanical Energy Converters
and Drives), 577–584, 7–9 July, 1993.
9. E. A. Klingshirn and H. E. Jordan, Simulation of polyphase in-

duction mchines with deep rotor bars, IEEE Trans. Power Appar.
Syst., PAS-89: 1038–1043, 1970.
10. W. H. Creer, D. W. Novotny, and T. A. Lipo, Determination of

equivalent circuits for induction machines with skin effect using
terminal characteristics, Electric Machines and Power Systems,
10 (4–5): 379–394, 1985.
11. P. Vas, Generalized analysis of saturated a.c. machines, Archiv

fur Elektrotechnik, 64: 57–62, 1981.
12. P. Vas, K. E. Hallenius, and J. E. Brown, Cross-saturation in

smooth air gap electrical machines, IEEE Trans. Energy Convers.,
EC-1: 103–109, 1986.
13. J. A. A. Melkebeek, Magnetising-field saturation and dynamic be-

haviour of induction machines, Part 1: Improved calculation
method for induction-machine dynamics, IEE Proc., 130, Pt. B
(No. 1): 1–9, Jan 1983.
14. I. M. Canay, Causes and discrepancies on calculation of rotor

quantities and exact equivalent diagram of synchronous ma-
chines, IEEE Trans. Power Appar. Syst., PAS-88: 1114–1120,
1969.
15. G. R. Slemon, Analytical models for saturated synchronous ma-

chines, IEEE Trans. Power Appar. Syst., 409–417, 1971.
16. R. P. Schultz, W. D. Jones, and D. N. Ewart, Dynamic models of

turbine generators derived from solid rotor equivalent circuits,
IEEE Trans. Power Appar. Syst., PAS-92: 926–933, 1973.
17. P. C. Krause, Simulation of unsymmetrical 2-phase induction

machines, IEEE Trans. Power Appar. Syst., PAS-84: 1025–
1037, 1965.
18. P. Pillay and R. Krishnan, Modeling, simulation and analysis of

permanent magnet motor drives, Part I: The permanent magnet
synchronous motor drive, IEEE Trans. Ind. Appl., 25: 265–273,
1989.
19. T. J. E. Miller, Switched Reluctance Motors and Their Controls,

Hillsboro, OH: Magnaphysics Publishing, 1993.
20. C. M. Ong, Dynamic Simulation of Electric Machinery Using

Matlab/Simulink, Upper Saddle River, NJ: Prentice-Hall, 1998.
21. P. C. Krause, O. Wasynczuk, and S. Sudhoff, Analysis of Electric

Machinery, Piscataway, NJ: IEEE Press, 1995.
22. D. W. Novotny and T. A. Lipo, Dynamics and Vector Control of

Induction Motor Drives, London: Oxford Press, 1996.
T. A. LIPO
University of Wisconsin
----------------------- Page 104-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Electrochemical Electrodes

Standard Article

D. Traub Hooie1
1US DOE, Federal Energy Technology Center

reserved.
Search
DOI: 10.1002/047134608X.W3014
❍ Acronym Finder
Abstract
Fuel Cell Components
The Electrode Role
Polymer Electrolyte Fuel Cell Electrodes
Alkaline Fuel Cell Electrodes
Phosphoric Acid Fuel Cell Electrodes
Molten Carbonate Fuel Cell Electrodes
Solid Oxide Fuel Cell Electrodes
Summary

%20Conversion/W3014.htm17.06.2008 16:36:51
----------------------- Page 105-----------------------
ELECTROCHEMICAL ELECTRODES 383
ELECTROCHEMICAL ELECTRODES of
the fuel and oxidant gases is generally counter or cross flow
rather than concurrent flow.

A fuel cell consists of a series of basic building blocks that
The fuel can be any substance capable of being utilized
incorporate an electrolyte layer in contact with a porous
through chemical oxidation (at the fuel cell operating temper-
anode and porous cathode on either side. There are five main
ature) that can be supplied continuously. PEFC’s generally
types of fuel cells: alkaline (AFC), proton exchange membrane can
only use hydrogen, but the higher temperature fuel types
(PEFC), phosphoric acid (PAFC), molten carbonate (MCFC), can
utilize virtually any hydrogen rich stream (liquid or gas-
and solid oxide (SOFC). Each fuel cell uses different materials
eous) and sometimes can utilize other materials compatible
for all the components and, as a result, has different op-
with the electrolyte such as carbon dioxide (CO2) or ammonia
erating temperatures. In addition, some variations exist in
(NH3). Similarly, the oxidant can be any fluid or gas that can
the materials within each fuel cell type. However, the basic
provide an oxygen-rich stream at the appropriate rate. How-
materials’ properties are consistent within each fuel cell type.
ever, the oxidant is generally air.
A schematic of a typical fuel cell, along with the gas and
ion conduction flow through the cell for PAFCs, PEFCs,
MCFCs and SOFCs, is shown in Fig. 1. Ion specie and trans-
FUEL CELL COMPONENTS
port direction are different for each fuel cell type, which in-
fluences the site where water production will occur. In addi- The
functions of electrodes in fuel cells are to: (1) provide a
tion, each fuel cell type has different sensitivity to impurities
surface site where gas/liquid ionization of deionization reac-
and alternate fuels. The fuel and oxidant gases flow past the
tions can take place, (2) conduct ions away from or into the
surface of the anode and cathode and generate electrical en-
three phase interface once they are formed (so an electrode
ergy by the electrochemical oxidation of fuel, usually hydro-
must be made of materials that have good electrical conduc-
gen, and the electrochemical reduction of oxygen. The electro-
tance), and (3) provide a physical barrier that separates the
lyte provides the environment for this reaction to occur. Flow
bulk gas phase and the electrolyte (1). In order to provide the
surface site and to increase the rates of reactions, the elec-
trode material should be catalytic as well as conductive, po-
rous rather than solid.

e′′
Ionization reaction rates increase with temperature.
Therefore, the catalytic function of electrodes is more impor-
tant in lower temperature fuel cells and less so in high tem-

Depleted fuel Depleted oxidant
perature fuel cells. In addition, the porous electrodes must be
and product and product
permeable to both electrolyte and gases, but not such that the
gases out gases out
media can be easily ‘‘flooded’’ by the electrolyte or ‘‘dried’’ by
the
gases in a one-sided manner.
In addition to transporting dissolved reactants to the elec-
trode, theelectrolyte also conducts ionic charge between the

SOFC H2 0′′ O2
electrodes and thereby completes the cell electric circuit, as
H O
illustrated in Fig. 1. It also provides a physical barrier to sep-
2
arate the fuel and oxidant gas steams. In low temperature
fuel cells (PEFC, AFC, PAFC), protons or hydroxyl ions are

PAFC + the
major charge carriers in the electrolyte, whereas in the
and H2 H O2
high temperature fuel cells, MCFC and SOFC, carbonate ions

PEFC H2O
and
oxygen ions are the charge carriers, respectively.
Individual fuel cell packages or sandwiches (anode, electro-
lyte, and cathode)

generally do not provide sufficient power
H CO ′′ O
2 3 2 for
most applications. Therefore, individual fuel cells must be
MCFC CO2
combined to produce
appreciable amounts of electricity and
H2O CO2 are
joined by interconnects. Because of the configuration of a
flat
plate cell, Fig. 2, the interconnect becomes a separator
plate which has two functions: (1) to provide an electrical se-
ries connection between adjacent cells, and specifically for flat

Fuel in Oxidant in
plate cells, and (2) to provide a gas barrier that separates
the
fuel and oxidant of adjacent cells. The interconnect of the
tubular solid oxide fuel cell is a special case and has a slightly
altered function. All interconnects must be an electrically con-
ducting and impermeable to gases.

Anode Cathode
Depending on the type of fuel cell, other parts of the cell
could include (1) the structure for distributing the reactant

Electrolyte
gases across the electrode surface and which serve as me-
(ion conductor)
chanical support, shown as ribs in Fig. 2, (2) electrolyte reser-

Figure 1. Schematic of typical gas and ion conduction flow through
voirs for liquid electrolyte cells to replenish electrolyte lost
fuel cells.
over life, and (3) current collectors (not shown) which provide

----------------------- Page 106-----------------------
384 ELECTROCHEMICAL ELECTRODES
The nature of this interface plays a critical role in the electro-
chemical performance of a fuel cell, particularly in those fuel
cells with liquid electrolytes (i.e., AFC, PAFC, MCFC). In

such fuel cells, the reactant gases diffuse through an electro-
lyte film that wets portions of the porous electrode and react
Separator plate
electrochemically on the electrode surface. The balance of fuel
cell and open porosity is critical to the performance of the fuel
cell. If the porous electrode contains an excessive amount of
electrolyte, the electrode may ‘‘flood’’ and restrict the trans-
port of gaseous species in the electrolyte phase. The conse-
quence is a reduction in the electrochemical performance of

Composite anode
the porous electrode.
substrate:
Much of
the recent effort in the development of fuel cell
### Porous acid
technology has been devoted to reducing the thickness of cell
reservoir +
components while refining and improving the electrode struc-
# Catalyst layer +
ture and the electrolyte phase, with the aim of obtaining a
# Half electrolyte
higher and more stable electrochemical performance, while si-
### matrix
multaneously reducing the cost of the materials and system.
Composite
# cathode
Electrode Reactions
# substrate
###
In a fuel cell, the anodic and cathodic reactants are fed into
their respective chambers, and an electrolyte layer provides a

#
region for the gases to move through the system. The half cell
#
reaction at the anode yields electrons, shown in Table 1.
###
These electrons are transported through the external circuit
and reach the cathode where they are transferred to the ca-
# Separator plate
thodic reactants. The cycle is completed by the transport of
#
ions from one electrode to the other through the electrolyte.
The typical electrochemical reactions that occur with dif-
ferent fuels and oxidants for the various fuel cell types are
also summarized in Table 1. CO and CH4 are shown in the
table as undergoing anodic oxidation, but in most cases, direct
oxidation may not

occur. Instead, these reactants are poten-
Figure 2. Typical planar fuel cell configuration.
tial fuels because they undergo chemical reaction with water
(H O) to produce hydrogen (H ), which is the oxidizable fuel.
2 2
In MCFCs, CO, and CH , combined with H O in the proper
4 2
environment, are sources of H2 from water gas shift and

a path for the current between the electrodes and the separa-
stream reforming reactions, respectively. The direct oxidation

tor of flat plate cells.
of CO and CH4 in high temperature SOFCs is feasible because

The physicochemical and thermomechanical properties of
of the high temperature.

materials used in the cell components (i.e., electrodes, electro-
Low temperature fuel cells (AFC, PEFC, PAFC) require

lyte, interconnect, current collector, etc.) determine the prac-
separate reformers containing noble metal electrocatalysts to

tical operating temperature and useful life of a fuel cell. Aque-
achieve practical reaction rates at the anode and cathode, and

ous electrolytes are limited to temperatures of about 200#C or
hydrogen is the only acceptable fuel. Impurities in the fuel

lower because of their high water vapor pressure and/or rapid
rapidly decrease the performance of the system. Carbon mon-

degradation at higher temperatures. The operating tempera-
oxide ‘‘poisons’’ the noble metal anode electrocatalyst such as
ture also plays an important role in dictating the type of fuel
platinum (Pt) in low temperature fuel cells. However, with

that can be utilized in a fuel cell. The low temperature fuel
high temperature fuel cells (MCFC, SOFC), the requirements

cells with aqueous electrolytes are, in most practical applica-
for electrocatalysis are relaxed, and the number of potential

tions, restricted to hydrogen (which includes reformed meth-
fuels is increased. Carbon monoxide serves as a potential

ane) as a fuel. In high temperature fuel cells, carbon monox-
source of H2 in high temperature fuel cells where nonnoble

ide (CO) and even methane (CH4) can be used directly because
metal electrocatalysts such as nickel (Ni) are used.

of the inherently rapid electrode kinetics and the lesser need
The overall electrochemical reactions corresponding to the

for high electrocatalytic activity at high temperature (2).
individual electrode reactions listed in Table 1 along with the
appropriate form of the Nernst equation, are shown in Table

THE ELECTRODE ROLE
2. The Nernst equation provides a relationship between the
standard potential (E 0) for the cell reaction and the equilib-
The chemical and physical interactions that occur at the elec-

rium potential (E) at various temperatures and partial pres-
trode interface are critical to the performance of the fuel cell.
sures (activities) of reactants and products. According to the
A three-phase interface is established in the region of the po-
Nernst equation, the equilibrium cell potential at a given
rous electrodes, the electrolyte, and the gaseous reactants.
temperature can be increased by operating at higher reactant
----------------------- Page 107-----------------------
Table 1. Typical Electrochemical Reactions in Fuel

Cells
Fuel Cell Anode

Reaction Cathode Reaction
Proton exchange H # 2H# # 2e#

##O # 2H# # 2e# # H O
2
2 2
Alkaline H # 2(OH) # 2H O
# 2e# ##O # H O # 2e# # 2(OH)
2
2 2 2
Phosphoric acid H # 2H# # 2e#
##O # 2H# # 2e# # H O
2
2 2
Molten carbonate H # CO# # H O
# CO # 2e# ##O # CO # 2e# # CO#
2 2 2
2 2 2 3
CO # CO# # 2CO
# 2e#
3
2
Solid oxide H # O# # H O #
2e# ##O # 2e# # O#
2 2
2
CO # O# # CO2
# 2e#
CH # 4O# # 2H O
# CO # 8e#
4
2 2
pressures. In fact, improvements in fuel cell performance

sity (current per unit area), a substantial increase in elec-
have been observed at higher pressures.
trode area has an important effect on cell performance. Large
The fuel (generally hydrogen) enters the pores of the anode
three-phase zones can be achieved by the use of metal pow-
2
and reaches the reaction zone where gas, the electrolyte, and
ders with a very high specific area (#100 m /g) or
carbon
the solid conducting structure meet. This is the ‘‘three-phase
(1000 m2/ g).
interface.’’ Hydrogen diffuses to the electrochemically active

site where it is adsorbed and dissolved by the electrolyte. Dis-
Surface Area Effect. Porous electrodes can have surface
sociation and ionization then occur.
areas which are orders of magnitude larger than their geo-

The electrochemical reactions of H2 and O2 in fuel cells pro-
metric area. Fillers and binders are sometimes mixed with
duce H2O. When a carbon containing fuel is involved in the
the electrode components during manufacture and removed

anode reaction, CO2 is also produced. In the case of
the
at the end of the manufacturing process. This leaves voids

MCFC, CO2 is required in the cathode reaction to maintain
that enhance the porosity of the structure and permit

the
an invariant carbonate concentration in the electrolyte. Since
electrode to have a hydrophobic or hydrophilic surface. Hy-
CO2 is produced at the anode and consumed at the cathode in
drophobic electrodes are made of fine carbon powder bonded

MCFCs, and the concentrations in the anode and cathode feed
with a plastic material. The porous electrodes that are used

streams are not necessarily equal, the Nernst equation in Ta-
in low temperature fuel cells consist of a composite structure

ble 1 includes the CO2 partial pressure for both electrode reac-
that contains Pt electrocatalyst on a high surface area carbon

tions.
black and a polytetrafluoroethylene (PTFE) binder. Carbon

In some cases, the validity of the mechanism depends on
powders are light, have a large surface area, are suitable for
the surface structure of the electrodes and the current op-
the deposition of active catalysts, such as platinum, and are

erating conditions. Cell potential/current density and charge
readily manufactured on a commercial scale. Such electrodes

curves allow determination of the mechanism. The data for
for acid and alkaline fuel cells are described by Kordesch et

these special potentiostatic and potentiodynamic analyses can
al. (3). In these porous electrodes, PTFE is hydrophobic (acts

be obtained on rotating and stationary ring-disc electrodes, as
as a ‘‘wet-proofing’’ agent) and serves as the gas permeable

well as impulse methods.
phase. In addition, carbon black is an electron conductor that
provides a high surface area to support the electrocatalyst.

Porous Gas Electrodes
Platinum serves as the electrocatalyst which promotes

the
Since reactants for the fuel cells are usually gaseous, the most
rate of electrochemical reactions (ionization/deionization) for
significant breakthrough in fuel cell technology has been the
a given surface area. The carbon black also has a certain de-
development of porous gas diffusion electrodes. The higher
gree of hydrophobicity, depending on the surface properties of
temperature fuel cells can provide power densities
greater the material.
than those produced in combustion engines.
Two layers can be identified in these PTFE bonded carbon
The principle function of the gas diffusion electrode is to
electrodes: a highly hydrophobic porous gas diffusion
layer
provide a large reaction zone area with a minimum of mass
and an electrolyte-wettable thin layer. The composite struc-
transport resistance (for the reactants and removal of byprod-
ture of PTFE and carbon establishes an extensive three phase
ucts). Since electrode potential changes with the current den-
interface in the porous electrode, which is the benchmark of
Table 2. Nernst Equations for Fuel Cell

Reactionsa
Cell Reaction
Nernst Equation
H # ##O # #H O E #
E 0 # (RT/2F )ln[P /P ] # (RT/2F )ln[P1/2]
2 2 2
H H O O
2 2 2
H # ##O # CO (c) # E #
E 0 # (RT/2F )ln[P /P (P ) ]
2 2 2
H H O CO a
2 2 2
H O # CO (a)
# (RT/2F )ln[P1/2](P ) ]
2 2
O CO c
2 2
CO # ##O # CO E #
E 0 # (RT/2F )ln[P /P ] # (RT/2F )ln[P1/2]
2 2
CO CO O
2 2
CH # 2O # 2H O # CO E #
E 0 # (RT/8F )ln[P /P 2 P ]
4 2 2 2
CH H O CO
4 2 2
# (RT/8F )ln[P 2 ]
a (a) # anode; (c) # cathode; E #

equilibrium potential; E 0 # standard potential; P # partial
pressure of gas; R # universal gas constant;

T # temperature.
----------------------- Page 108-----------------------
PTFE bonded electrodes. The electrochemical reaction takes steps

involved in the overall electrochemical reaction in a po-
place at the interface, where the catalytically active material rous fuel
cell electrode. With the exception of the dissolution
must be present. of the
reactant in the electrolyte and the transport of the re-
It should be noted that the conductivity of carbon is not actant
through the electrolyte to the electrode surface, each
sufficient for high currents (#200 mA/cm2). To improve their of the
other steps can be associated with a polarization that
conductivity, a wire screen as a current collector or a bipolar could

contribute to an increase in the inefficiency of the elec-
plate material is required. trode
reaction. These include physical, chemical, and electro-
Sintered metal powders are used for hydrophilic elec- chemical
reactions in the fuel cell. Not all of the steps contrib-
trodes. The gas diffusion layer of these electrodes has larger ute
significantly to the total polarization, and those steps can
pores than the reaction layer. Although these electrodes are be
disregarded. Thus, efforts to improve the performance of
heavy, they have a high conductivity, which is a great advan- porous
fuel cell electrodes can be directed at reducing the po-
tage in monopolar plate electrodes, where the current is col- larization
associated with adsorption of the electroactive spe-
lected from tabs at the edges of the electrodes. cies onto
the electrode, the surface migration of the absorbed
species,
the electrochemical reaction of the electrically
Current Density. The current densities that are obtained charged
species, and the transport of gaseous and liquid prod-
from smooth electrodes are usually in the range of single ucts away
from the surface and interface.
units mA/cm2 or less because of rate limiting processes such
Transport processes involving diffusion, convection, and
as the available area of the reaction sites. Porous electrodes migration
can take place in the gas phase in the pores of the
are commonly used in fuel cells to achieve much higher cur- electrodes
or in the liquid phase in the pores of the electrode.
rent densities. These high current densities are possible be- The
relative contribution of these transport processes to the
cause the electrode has a high surface area which signifi- overall
electrode polarization is affected by the porous elec-
cantly increases the amount of reaction sites, and the trode
structure, electrolyte composition and temperature, and
optimized electrode structure has favorable mass transport reactant
gas composition and pressure.
properties. In an idealized porous gas fuel cell electrode, high The
maximum rate of transport of reactant species to the
current densities at reasonable polarization are obtained electrode
surface provides an upper limit to the rate of elec-
when the liquid (electrolyte) layer on the electrode surface is
trochemical reaction. Under these conditions, a limiting cur-
sufficiently thin so that it does not significantly impede the rent is
reached. In a fuel cell with porous electrodes con-
transport of reactants to the electroactive sites, and a stable taining a
liquid (electrolyte) layer, the diffusion of the
three phase (gas/electrolyte/electrode surface) interface is es- reactant
species through the electrolyte is usually the rate
tablished. limiting
transport process.
In MCFCs, which operate at relatively high temperatures,

there are no known materials that can serve to wet proof a
Irreversible Thermodynamics. Useful amounts of work (elec-
porous structure against permeation by molten carbonates. trical
energy) are obtained from a fuel cell only when a rea-
Consequently, the technology used to obtain a stable, three sonably
large current is drawn, but the cell potential will be
phase interface in MCFC porous electrodes is different from decreased
from its equilibrium potential because of irrevers-
that used in PAFCs. In the MCFC, the stable interface is ible
losses. There are several sources that contribute to irre-
achieved in the electrodes by carefully tailoring the pore versible
losses in a practical fuel cell. The losses, which are
structures of the electrodes and the electrolyte matrix often
called polarization, overpotential, or overvoltage, origi-
(LiAlO2) so that the capillary forces establish a dynamic equi- nate
primarily from three sources: (1) ohmic polarization
librium in the different porous structures. (#ohm),
(2) concentration polarization (#conc), and (3) activation
SOFCs, on the other hand, do not have a liquid electrolyte
polarization (#act). These losses result in a cell voltage (V) for
present that is susceptible to movement in the porous elec- a fuel
cell that is less than its reversible potential.
trode structure, and electrode flooding is not a problem. Con- Ohmic
Polarization. The ohmic losses occur because of re-
sequently, the three phase interface that is necessary for effi- sistance
to the flow of ions in the electrolyte and resistance to
cient electrochemical reaction involves two solid phases (solid flow of
electrons through the electrode materials. The domi-
electrolyte/electrode) and a gas phase. A critical requirement nant
ohmic losses through the electrolyte phase are reduced
of porous electrodes for SOFC is that they are sufficiently thin by
decreasing the electrode separation and enhancing the
and porous to provide an extensive electrode/electrolyte in- ionic
conductivity of the electrolyte. Because both the electro-
terfacial region for electrochemical reaction. lyte and
fuel cell electrodes obey Ohm’s law, the ohmic losses
can be
expressed by the equation,
Transport Processes
ηohm = iR
The transport processes involving the mass transfer of

reactants/products play a prominent role in the performance where i
is the current flowing through the cell, and R is the
of porous electrodes in fuel cells. Transport processes involv- total cell
resistance, which includes electronic, ionic and con-
ing heat transfer, and thermal management are important in tact
resistances.
fuel cell systems. The slow transport of reactant and product
Concentration Polarization. As a reactant is rapidly con-
species through the porous electrode gives rise to the polariza- sumed at
the electrode by electrochemical reaction, concentra-
tion. A sequence of steps, involving both transport and rate tion
gradients will be established. Several processes may con-
processes, occurs when a reactant species undergoes electro- tribute to
concentration polarization: (1) slow diffusion in the
chemical reaction in a porous electrode. One sequence sug- gas phase
in the electrode pores, (2) solution/dissolution of
gested by Liebhafsky and Cairns (4) indicates there are 13
reactants/products into/out of the electrolyte, and/or (3) diffu-
----------------------- Page 109-----------------------
sion of reactants/products through the electrolyte to/from the

drives the occurrence of exchange current, would decrease
electrochemical reaction site. At practical current densities,
further.
slow transport of reactants/products to/from the electrochem-
ical reaction site is a major contributor to concentration polar-
POLYMER ELECTROLYTE FUEL CELL ELECTRODES

ization.
Activation Polarization. Activation polarization is present
The
electrolyte in this fuel cell is an ion exchange membrane
when the rate of an electrochemical reaction at an electrode
(fluorinated sulfonic acid polymer or other similar polymers)

surface is associated with sluggish electrode kinetics. In other
which is an excellent proton conductor. The only liquid in this

words, activation polarization is directly related to the rates
fuel
cell is water; thus, corrosion problems are minimal. Wa-
of electrochemical reactions. There is a close similarity be-
ter
management in the membrane is critical for efficient per-
tween electrochemical and chemical reactions in that both in-
formance; the fuel cell must operate under conditions where

volve an activation barrier that must be overcome by the re-
the
byproduct water does not evaporate faster than it is pro-
acting species. Processes involving absorption of reactant
duced because the membrane must be hydrated. Because of

species, transfer of electrons across the double layer, desorp-
the
limitation on the operating temperature imposed by the
tion of product species, and the nature of the electrode surface
polymer and problems with water balance, usually less than

can all contribute to activation polarization.
120#C, a H2-rich gas with little or no CO is used, and higher

Electrode Polarization. Activation and concentration polar-
catalysts loadings (Pt in most cases) than those used in

izations can exist at both the positive (cathode) and negative
PAFCs are required in both the anode and cathode.

(anode) electrodes in fuel cells. The net result of current flow
The electrodes for this system are typical gas-diffusion

in a fuel cell is to increase the anode potential and to decrease
electrodes. The backing is a porous carbon cloth with a hy-

the cathode potential, thereby reducing the cell voltage. The
drophobic coating. In order for efficient electrochemical reac-

cell voltage includes the contribution of the polarization and
tions to take place, the electrodes must be catalyzed. Plati-

the anode and cathode potentials. Current flow in a fuel cell
num
is generally considered the best catalyst for both the
results in a decrease in the cell voltage because of losses by
hydrogen oxidation and the oxygen-reduction reactions. To re-

electrode and ohmic polarizations. The goal of fuel cell devel-
duce
the amount of catalyst required and increase the surface
opers is to minimize the polarization so that the cell voltage
area, the use of supported platinum catalysts, similar to those

(Vcell) approaches the difference between the anode and cath-
used
in the phosphoric acid cell, have been adopted. These
ode potential (#Ee). This goal is approached by modifications
consist of 2 to 5 nm diameter platinum particles on the sur-

to the fuel cell operating conditions (i.e., higher gas pressure,
face
of fine carbon particles. Unfortunately, Ticianelli et al.
higher temperature, change in gas composition to lower the
(5)
determined that only about 10 to 20% on the platinum is
gas impurity concentration, etc.), improvement in electrode
electrochemically active in the fuel cell reaction. Research ef-

structures, better electrocatalysts, more conductive electro-
forts are focused on increasing the catalyst activity.

lyte, etc. However, for any fuel cell, trade-offs exist between
To function, the catalyst must contact the gas as well as

achieving higher performance by operating at higher temper-
both
the electrical and protonic conductors. This is done by
ature or pressure and the problems associated with the
impregnating the supported-catalyst electrode with protonic

stability/durability of cell components encountered at the
conducting material. The most common method is to cover the

more severe conditions.
surface with a solution of solubilized membrane material of

Although ohmic, concentration, and activation polarization
Nafion or its equivalent. Newly developed materials have re-
are the classic fuel cell losses, others exist in the electrodes.
2
sulted in catalyst loadings down to 0.25 mg Pt/cm (with

Two worthy of consideration are contact resistant losses and
0.716 V), which greatly improves the commercial viability of

losses due to exchange currents.
this
concept.
Contact Resistance. Contact resistances at electrodes and
separators account for losses due to spatial variation in mate-
rial properties, fluxes, reaction rate densities, etc. Contact or
ALKALINE FUEL CELL ELECTRODES
terminal resistance is actually a form of ohmic resistance.
Exchange Current. This is the exchange of electrons The
first successful, commercially utilized fuel cell was the
through the electrolyte rather than through the external load. H2-
O2 alkaline cell, which was used in the space program.
This is the only form of losses that decreases with an increase
However, this system has found little other commercial appli-
in current output. Exchange current is directly related to the
cations.
chemical potential difference, so even at zero external load
The electrolyte in this fuel cell is concentrated potassium
current, there are electrons delivered to the cathode. Once
hydroxide (85 wt% KOH) in fuel cells operated at 250#C or
oxygen ions are formed, they migrate through the electrolyte less
concentrated (35 to 50 wt%) KOH for lower temperature
to the anode where they deionize to release an electron. The
(#120#C) operation. The electrolyte is retained in a matrix
electron released migrates back to the cathode to continue the
(usually asbestos). The thermodynamics of this system and
process or ‘‘exchange.’’ With the ionization/deionization reac- the
use of pure hydrogen as the fuel, permit the use of a wide
tions proceeding at a low (but non-zero equilibrium) rate, the
range of electrocatalysts (e.g., Ni, Ag, metal oxides, spinels,
cell voltage is diminished below that of reversible potential, and
noble metals).
even though there is no external current flowing. This form of
The alkaline fuel cells used for the space shuttle orbiter
loss, however, would become less significant once the external are
a significantly advanced verison of the early Bacon cells
current is increased beyond a certain value. Moreover, the and
the Apollo service module. They were developed by a
magnitude of the exchange current actually decreases with
Pratt and Whitney Corporation division which is now known
an increase in external current since the cell voltage, which as
International Fuel Cells. The early orbiter module oper-
----------------------- Page 110-----------------------
5 5 2
ated at a pressure of 4.0 10 to 4.4 10 N/m and a tempera-
the PAFC is capable of operating at the high end of the acid
ture of about 90#C. The high performance at low temperature
temperature range (100 to 220#C). In addition, the use of con-
was achieved using Teflon-bonded electrodes with pure, high-
centrated acid (#100%) minimizes the water vapor pressure,
loading noble metal catalysts. The anode consisted of pure so
water management in the cell is not difficult. The matrix
noble metal black (80% platinum, 20% palladium) on a silver-
universally used to retain the acid is silicon carbide (10), and
plated nickel screen. The cathode had 90% gold with 10% the
electrocatalyst in both the anode and cathode is Pt.
platinum on a gold-plated nickel screen. Later versions, which
The porous electrodes used in this system are described
are proprietary, of this system had improved materials, differ-
extensively in the patent literature (11). These electrodes con-
ent operating parameters, less noble metal loading, and elec-
tain a mixture of the electrocatalyst supported on carbon
trolyte reservoirs. However, this system is still far from being
black and a polymeric binder such as PTFE (30 to 40 wt%).
cost-effective for everyday applications (2). The
polymer binds the carbon black particles together to form
More commonly, the systems are built with catalyzed car- an
integral porous structure. A carbon paper substrate is also
bon on nickel screen current collectors or porous nickel sub-
used and serves a dual role as a structural support for the
strates with a hydrophobic backing. Platinum is the key in-
electrocatalyst layer as well as the current collector. A typical
gredient in the catalyst. Because these systems operate at
carbon paper used in PAFCs has an initial porosity of about
such low temperatures, plastics can be used for the cell
90%. This wet-proof carbon paper should contain macropores
frames. of
3 to 50 #m diameter, with a median pore diameter of about
More recent advances involve the use of a porous gas diffu-
12.5 #m. It should also have micropores with a median pore
˚
sion layer with 60 wt% PTFE on carbon, a nickel wire mesh,
diameter of about 34 A for gas permeability. The composite
and an active layer of 12.24 wt% cobalt tetraphenylporphyrin
structure consisting of a carbon black/PTFE layer on carbon-
(CoTPP) and 12 wt% PTFE on carbon for the cathode. The
paper substrate forms a stable, three-phase interface in the
catalyst for the anode consists of 8.3 wt% platinum, 8.3 wt%
fuel cell, with H3PO4 electrolyte on the electrocatalyst side
palladium and 17 wt% PTFE on charcoal (6). Additional re- and
the reactant gas environment on the other side of the
search is being conducted to improve the catalyst and elec-
carbon paper (12).
trodes (7).
The individual cells are physically separated by a bipolar
plate which also serves to connect the cells electrically in the
fuel cell stack. This plate also contains the gas channels for
PHOSPHORIC ACID FUEL CELL ELECTRODES
introducing the reactant gases to the porous electrodes and
removing the products and inerts. Heat-treated bipolar plates

The phosphoric acid fuel cell is commercially available today are
made from carbonized graphite-resin plates. The heat
in both stationary and on-site cogeneration plants. Interna-
treatment, which occurs at about 2700#C, improves the corro-
tional Fuel Cells Corporation/ONSI is the recognized technol-
sion resistance. That is, the corrosion current is reduced by
ogy leader in this field and manufactures the only commer- two
orders of magnitude at 0.8 V in 97% H3PO4 at 190#C and
cially available units. (These units are also sold by Toshiba 486
kPa. This is important for the goal of 40,000 h of life for
and Ansaldo.) The electrochemical reactions for this system the
stack (13).
occur on highly dispersed electrocatalyst particles supported
Electrode performance may degrade rapidly by flooding
with carbon black. Catalysts are used on both the anode (Pt) due
to carbon corrosion and loss of PTFE. This is sometimes
and cathode (Pt or an alloy containing Pt).
observed in cells with high initial performance. On the con-
In the early 1960s, the conventional porous electrodes were
trary, an electrode having low initial performance may have
PTFE-bonded Pt-black, and the platinum loadings were ex-
long life; it has high wetproofing ability to the electrolyte,
tremely high—about 9 mg Pt/cm2 . The hardware was gold-
which improves the formation of the three-phase boundary
plated Tantalum. During the past two decades, the cost of the
(14).
materials and the requirement for Pt has decreased. A major
Despite the progress in the development of cell components
breakthrough was the development of carbon blacks and
over the past decade, there is still significant research under-
graphite for cell construction materials (8,9). At that time, it way
on alternative and component improvements that have
was shown that carbon black and graphite were sufficiently the
potential of cost reductions and performance improve-
stable to replace the more expensive gold-plated tantalum cell
ments. This includes improved catalysts, catalyst supports,
hardware. Because the carbon black had high surface areas, and
electrolyte management. Alternative support materials
a dramatic reduction in Pt loading could be obtained with lit-
with improved physicochemical properties are desired to in-
tle or no reduction in performance. Appleby (8) reports that
crease the flexibility and remove the current temperature and
‘‘without carbon, a reasonably inexpensive acid fuel cell would
electrical constraints on PAFCs. A large number of materials
be impossible, since no other material combines the necessary
offer potential (15).
properties of electronic conductivity, good corrosion resis-
tance, low density, surface properties (especially in high-area
form) and above all, low cost.’’ Table 3 shows the evolution of
MOLTEN CARBONATE FUEL CELL ELECTRODES
the cell component technology for these fuel cells.
Concentrated phosphoric acid is used for the electrolyte in
MCFC is an intermediate temperature fuel cell that operates
this fuel cell, which operates between 150 and 220#C. At lower
between 600 and 700#C. The chronology of component devel-
temperatures, phosphoric acid is a poor ionic conductor, and
opment is shown in Table 4. Since the mid-1970s, the materi-
CO poisoning of the Pt electrocatalyst in the anode becomes als
for the electrodes and electrolyte structure have remained
more severe. The relative stability of concentrated phosphoric
essentially unchanged. The primary developments have been
acid is high compared to other common acids; consequently, the
modification of the electrolyte and improvements in the
----------------------- Page 111-----------------------
Table 3. Evolution of Phosphoric Acid Fuel Cell

Technology
Component 1965
1975 Current
Anode PTFE-bonded Pt
PTFE-bonded Pt/C PTFE-bonded Pt/C
black
Furnace black Furnace black

Anode catalyst 9 mg/cm2 Pt
0.25 mg Pt/cm2 0.25 mg Pt/cm2
Cathode PTFE-bonded Pt
PTFE-bonded Pt/C PTFE-bonded Pt/C
black
Furnace black Furnace black

Cathode catalyst 9 mg/cm2 Pt
0.5 mg Pt/cm2 0.5 mg Pt/cm2
Electrode support Ta mesh screen

Carbon paper Carbon paper
Electrolyte support Glass fiber paper
PTFE-bonded SiC PTFE-bonded SiC
Electrolyte 85% H PO
95% H PO 100% H PO
3 4
3 4 3 4
fabrication processes. Developments in the fuel cell compo-

ing mechanism:
nents have been reviewed by many people including Maru
(16,17), Petri (18), Selman (19), and others.
NiO + CO2 → Ni2+ + CO2 −
2
The electrolyte in this fuel cell is usually a combination of
alkali (Li, Na, K) carbonates, which is retained in a ceramic
The dissolution of NiO has been correlated to the acid-base
matrix of LiAlO2 . The fuel cell operates where the alkali car-
properties of the electrolyte. The basicity of the electrolyte is
bonates form a highly conductive molten salt, with carbonate
directly proportional to the CO2 partial pressure (log PCO ). In
ions providing ionic conduction. At the high operating temper-

addition, the solubility of NiO in the electrolytes is compli-
atures in MCFCs, Ni (anodes) and nickel oxide (cathodes) are
cated by its dependence on several additional parameters:
adequate to promote reaction, and noble metals are not re-
carbonate composition, H2O partial pressure, and temper-
quired. However, the major problems with Ni-based anodes
ature.
and NiO cathodes are structural stability and NiO dissolu-
The bipolar plates used in MCFC stacks are usually fabri-
tion. The dissolution of the NiO in molten carbonate electro-
cated from thin sheets of an alloy (such as Incoloy 825, 310S,
lyte became evident when thin electrolyte layers were used.
or 316L stainless steel) that is coated on the inlet side with
Although NiO has a low solubility (10 ppm) in the electrolyte,
Ni. The Ni layer is stable in the reducing gas environment
Ni ions diffuse in the electrolyte toward the anode, and metal-
of the anode compartment and provides a conductive surface
lic Ni can precipitate in the reducing environment by the
coating with low contact resistance. To minimize corrosion at
anode. This phenomenon becomes worse at high CO2 partial
the wet-seal area, a thin Al coating is usually applied. This
pressures (20,21) because dissolution may involve the follow-
layer reacts with the Al and electrolyte and becomes LiAlO2 .
Table 4. Evolution of Molten Carbonate Fuel Cell

Technology
Component 1965
1975 Current
Anode Pt, Pd, or Ni

Ni-10 wt% Cr Ni-10 wt% Cr
3–6 #m pore
50–60% porosity
0.5–1.5 mm thick
0.1–1 m /g
Cathode Ag2O or lithiated NiO
Lithiated NiO Lithiated NiO
7–15 #m pore
70–80% porosity
0.5–0.75 mm thick
0.5 m2/g
Electrolyte support MgO

Mixture of #-, #-, #-LiAlO2
and #-LiAlO2
2 2
10–20 m /g 0.1–12 m /g
Electrolyte 52% Li, 48% Na
62% Li, 38% K 62% Li, 32% K
43.5% Li, 31.5%
60–65 wt% Li 50% Li, 50% Na
Na, 25% K
50% Li, 50% K
50 wt% Li
Fabrication process Paste
Hot press ‘‘tile’’ Tape-cast
1.8 mm thick Dry press
Printing process
0.5 mm thick
----------------------- Page 112-----------------------
However, it is important that LiAlO2 not be present in due

to poor point contact at the three-phase boundaries as
those areas where electronic conduction is required since this
well as poor gas edge seals. A tubular configuration was
is an insulating material. Thus, it is not suitable for providing
adopted which had bell-shaped cells that overlapped. This
corrosion resistance to the cell current collectors. The typical
also had a three-phase boundary problem as well as problems
materials used for this application are stainless steels (316,
with thermal expansion coefficients.
310, and 446) and chromium-plated stainless steels. However,
materials with improved corrosion resistance need to be de-
Tubular Solid Oxide Fuel Cells. After many iterations, ‘‘test
veloped for long-term operation (22).
tube’’ configuration was developed. This was partially an out-
Significant research has focused on the development of al-
growth of the oxygen sensor industry which often used a
ternative materials for both the anode and cathode. Numer-
closed-end tube. These systems operate in the 800 to 1000#C
ous studies (23–25) have examined the corrosion of materials
range. In this design, the cathode, cell interconnection, elec-
and have served as the foundation for the research being con-
trolyte, and anode are sequentially deposited on a closed-end
ducted today.
porous tubular support material which provides a mechani-
Several metal and alloy materials have been proposed as
cally strong structure for the thin cell components. In the cur-
alternative anodes. The Ni–Cr anodes currently in use yield
rent configuration, the cathode has a dual purpose: serving as
acceptable performance but are high cost components and are the
porous support tube as well as the air electrode. This de-
susceptible to mechanical deformation. The Cr in the alloys
sign allows relatively large single tubular cells to be con-
reacts to form LiCrO2 which aids in stabilizing the porous
structed with the successive active layers deposited with
anode structure in sintering resistance.
chemical or materials interference with the previously depos-
However, these materials are susceptible to creep at the
ited layers. In addition, the manifolding of the oxidant and
high compressive loads encountered in the stack. An anode of
fuel gases is simplified, and no gas seals are required.
Ni–10Cr alloy in a MCFC exhibited a creep of about 4% after
Current designs make use of thin film wall concepts where
a 10,000 h test, which is about four times higher than the
films of material are deposited one on another and sintered
goal (18). Metal oxides (such as Al O , LiAlO , and ZrO ) and to
form the cell structure. The fabrication technique differs
2 3 2 2
metals (such as Ti, Co, Cu, and Al) have been incorporated for
each component. The air electrode provides the inner tube
in anode structures (16,18,26–28). Copper-based materials for
the cell package. This tube is generally extruded and sin-
have also been evaluated as a lower cost alternative
tered with one end closed. The anode is generally deposited,
(26,29,30). to
a thickness of about 150 #m, using a slurry process. It is
An acceptable cathode material must have good electrical
important that these electrodes permit only electronic conduc-
conductivity and structural strength. It must also have a low
tion and interdiffusion of ionic species at operating tempera-
dissolution rate in molten alkali carbonates to avoid precipi-
ture (1000#C). In addition, they must be resistant to thermal
tation of the metal in the electrolyte structure. Over 50 ma-
cycling effects. That is, they must have closely matched ther-
terials have been examined to identify suitable materials. mal
expansion coefficients and little or no grain growth.
Perovskite-type compounds (Ln M M
O where Ln is a The resistivities for the anode are
critical to the perfor-
1#x x 3
lanthanide element, M is an alkaline earth or transition

mance of the cell. At 1000#C, the anode and cathode should
metal, a M
is an transition metal) and mixed metal oxides have
resistivities of 0.001 #
cm and 0.01 #
cm, respectively
have shown promise. Studies of lithium ferrates have been
(34). The electrolyte is the least conductive component in the
shown to be a viable candidate for cathodes despite its high
cell.
solubility (78 ppm Fe) in the Li/K electrolyte at 650#C. In
contrast to NiO, this material is independent of the partial
Planar Solid Oxide Fuel Cells. Planar cells, on the other
pressure of the CO and O . In addition, LiCoO and LiCoO -
hand, generally operate at lower temperatures (650 to 800
2 2 2 2
coated NiO have also been shown to have low solubility in the
#C). This enables the use of metallic interconnects, relaxes
electrolyte. However, it has a very high resistivity (about 300
some cell construction constraints, reduces internal electrical
#
cm). Unfortunately, this has a high material cost which
resistance, and potentially can reduce the cost of materials in
may prevent large scale implementation (31–33). the
balance of plant. Planar cells require gas-tight seals as
part of the manifolding to separate the fuel streams. The
seals are a critical technological hurdle for this technology.

SOLID OXIDE FUEL CELL ELECTRODES
Bipolar Configuration. The bipolar configuration is the most
common configuration for the planar systems. These systems

The SOFC technology has emerged from single cell tests in can
either be high or intermediate temperature. This consists
the 1980s to prototype multicell arrays up to 100 kW in the of
a simple series electrical connection between cells without
1990s. These fuels cells can operate at intermediate and high the
need for external cell interconnections. Perpendicular cur-
temperatures (650 to 1000#C) and differ from the other fuel
rent collection in a cell stack has a lower ohmic polarization
cells in that they have a solid ceramic electrolyte. As a result,
than the tubular configuration. However, gas leaks with edge
the electrolyte management problems that are common to the
seals and poor triple phase interface contacts generally result
other fuel cells are not an issue. In addition, they do not rein
lower performance. Overall, the flat plate design offers im-
quire CO2 recycle from the anode to the cathode.
proved power density relative to the tubular design. Fabrica-
Two basic configurations of SOFCs are being developed:
tion and assembly appear to be simpler for the flat plate de-
planar and tubular. In addition, the planar has variations
sign. Tape casting, hot pressing, and dry pressing are the
such as radial and monolithic. This is shown in Fig. 3. Origi-
most common fabrication methods.
nally, experimental SOFCs had a planar configuration similar
Monolithic Configuration. The monolithic configuration is a
to the PAFCs. This configuration was generally unacceptable
more complex bipolar design. The structure resembles the
----------------------- Page 113-----------------------
Tubular
Interconnection
Electrolyte
Air
electrode
Fuel
flow
Air flow
Fuel electrode
Monolithic Flat plate

Interconnect
Anode
Cathode ### ###
Interconnect
Anode
### ###
Electrolyte
Cathode
###
### ###
###
### Fuel ###
### ###
###
Oxidant
###
Fuel ###
### Cathode
###
Anode
###
Oxidant
Electrolyte
###
Cathode
Figure 3. Various solid oxide fuel cell

configurations.
corrugated assembly seen in cardboard boxes.The electrodes both planar and

tubular cells. However, the intermediate
for this system are taped or calendered and then formed into temperature cells
can use different materials, and they are
corrugated structures. These corrugations also serve as the noted
separately. The evolution of materials for this technol-
passages for the fuel and air flows. The electrodes are ogy is shown in
Table 5. Much of the current research focuses
‘‘stacked’’ with the other components, and the entire ‘‘fuel cell on the
intermediate temperature cells. The materials used in
stack’’ is then sintered into a rigid structure. The thermal cell components
are limited by (1) chemical stability in gas-
expansion of the components are critical during heat treating eous (either
oxidizing and/or reducing) environments, (2) con-
and sintering as well as during stack operation. Residual ductivity, and (3)
thermomechanical compatibility. The elec-
stresses caused by the differences in thermal expansion coef- trolyte in this
fuel cell is a solid, nonporous metal oxide,
ficients often lead to mechanical failure (35). In addition, it is usually Y O
-stabilized ZrO . The cell operates between 650
2 3
2
difficult to cofire metals and oxides in a single atmosphere and 1000#C

where ionic conduction by oxygen ions takes place.
without either reducing or oxidizing one or the other. The tri- Currently, the
anode is a Co–ZrO2 or Ni–ZrO2 cermet con-
ple-phase boundary of these systems has been shown to have taining about 30
mole% Ni or Co. Various modifications to
poor contact and show less performance than comparable sys- this include partial
or complete replacement of the ZrO2 with
tems. Lastly, another major challenge is the manifolding for CeO2 .
Takahashi (36) and Steele (37) have shown that there
the fuel and oxidant gases against the rigid, uneven, edge is enhanced
catalytic behavior for the direct oxidation of
surfaces. Calendaring, a variation on tape casting, is the most methane using
Ce0.9Y0.1O1.95. Unfortunately, they also appear
common fabrication method for the electrodes. to have poor
electronic conductivity and large thermal ex-
pansion at higher
temperatures. Mixed protonic and ionic
Solid Oxide Component Materials. In general, the materials conductivity has
also shown promise as anode materials (38).
used for the manufacture of the comonents are similar for In addition, a
variety of acceptor doped Perovskite (ABO3)
----------------------- Page 114-----------------------
Table 5. Evolution of Solid Oxide Fuel Cell Technology

for Both Planar and Tubular
Configurations
Component 1965
1975 Current
Anode Porous Pt Ni–ZrO2

cermet 30 mole% Ni–ZrO2
cermet
12.5 10#6 cm/cm

#Ca
Deposit slurry, dry press,
calendar, tape cast

Cathode Porous Pt
Stabilized ZrO2 10 mole% Sr-doped
impregnated with lanthanum manganite
praesodymium oxide
12 10#6 cm/cm
#Ca
Deposit slurry/dry press,
calendar, tape cast,
sinter
Electrolyte Yttria-stabilized ZrO2 Yttria-
stabilized ZrO2 8 mole% yttria-
stabilized ZrO2
10.5 10#6 cm/cm

#Ca
EVD
EVD, dry press, calendar,
tape cast
Cell interconnect Pt Mn-
doped cobalt 10 mole% Mg-doped
chromite lanthanum chromite

EVD
Dry press, calendar, EVD,
tape cast, plasma spray

Separator
Metal alloy
(planar only)
Support tube Yttria-stabilized ZrO2 Yttria-
stabilized ZrO2 None required (air
(tubular only)
electrode is support for
tubular, none for planar)
a Expansion from room temperature to 1,000#C.
can exhibit protonic conductivity. Compositions such as

Single-phase porous cathodes appear to have an increased
CaTiO (Fe), SrZrO (Fe), LaYO (Sr), and SrUO may be ac-
ionic conductivity with high electronic conductivity. Materials
3 3 3 3
ceptable as alternative materials (39).

that have shown promise include La0.6Sr0.4Mn0.8Co0.2O3-# and
The cathode is generally an Sr-doped lanthanum manga-
La Sr Co Fe O . Two-phase cathodes, containing both
0.6 0.4 0.2 0.8 3
nite (La Sr MnO ) containing about 10 mole% Sr. Research

ionic and protonic conductors, perform even better (42).
1#y x 3
has primarily focused on those materials that do not react

with the ZrO2 electrolyte and have good electronic conductiv-
ity. Substitution of Co for Mn on the B-site has been at-
SUMMARY
tempted, but it appears that only very small amounts of Co
can be introduced (40).
One of the main attractive features of fuel cell systems is
Intermediate Temperature Cell Materials. The yttria-stabi-
their expected high fuel-to-electricity efficiency (40 to 70%
lized zirconia is a poor conductor if it is operated at tempera-
based on lower heating value of the fuel), which is higher
tures below 800#C. Some improvement is shown if thin film
than that of many competing energy conversion systems. In
electrolytes are used. Significant alternative materials are be-
addition, fuel cells operate at a constant temperature, and the
ing examined for operation at lower temperatures. These sys-
heat from the electrochemical reaction is available for cogen-
tems are simpler, have simpler balance-of-plants, and the po-
eration applications. Since fuel cells operate at near constant
tential for lower cost. The most critical component being
efficiency, independent of size, small fuel cells operate nearly
developed is the electrolyte. The materials for the anode
as efficiently as large ones. Thus, fuel cell power plants can
and cathode will need to be compatible with the electrolyte.
be configured in a wide range of electrical output, ranging
Some of the low temperature electrolytes being developed
from watts to megawatts. Fuel cells are quiet and operate
are Ce Gd O , CeO -based materials, (Bi O ) (Y O ) ,
with virtually no gaseous or solid emissions. The two major
0.8 0.2 x 2 2 3 0.75 2 3 0.25
La Bi AlO , Y Ca CrO , (Y Ca ) Cr O , and salts con-

impediments to the widespread use of fuel cells are: (1) high
1-x x 3 1-x x 3 0.6 0.4 1-y 1-y 3
sisting of Li/K/M, Li/Na/M, and Li/Na/K/M, where M is Ca,

initial cost and (2) endurance operation; it is these two as-
Ba, or Sr (41).
pects which are the major focus of technological effort.
Work being conducted on anodes involves alternative ma-
The various fuel cell types (AFC, PEFC, PAFC, MCFC, and
terials as well as Cr-stabilized-Ni. Laboratory-scale cells have
SOFC) have significantly different operating temperatures.
been successfully testing using Ni-Ce0.8Gd0.2O1.9 anodes. Per-
As a result, their materials of construction, fabrication tech-
formance improves with thinner components but, in general,
niques, and system requirements differ. These distinctions re-
is poorer than the Ni-cermet materials (42).
sult in individual advantages and disadvantages which gov-
----------------------- Page 115-----------------------
ern the potential of the various cells to be used for different 22.
D. A. Shores and P. Singh, Proc. Symp. Molten Carbonate Fuel
applications.
Cell Technol., Pennington, NJ: The Electrochemical Society, 1984,
p. 271.
23.
K. Kinoshita, Critical Survey on Electrode Aging in Molten Carbon-
BIBLIOGRAPHY
ate Fuel Cells, ANL 79-55, Argonne National Laboratory, Illi-
nois, 1979.
1. J. A. Hirschenhofer, D. B. Stauffer, and R. R. Engleman (eds.), 24.

Illinois Institute of Technology, Corrosion of Materials in Molten
Fuel Cells A Handbook (Revision 3), U.S. Department of Energy
Carbonate Fuel Cells (MCFC), Morgantown, WV: U.S. Depart-
1–4, 1994. [3. A. J. Appleby, F. R. Foulkes, Fuel Cell Handbook,
ment of Energy, Contract DE-AC21-86MC23265, 1987.
New York: Van Nostrand Reinhold, 1989. 25.
General Electric Co., Develop. of Molten Carbonate Fuel Cell Power
2. S. S. Penner (ed.), Assessment of Res. Needs for Advanced Fuel
Plant, Final Report DOE/ET/17019-20, 1 , Contract DE-AC02-
Cells by the DOE Advanced Fuel Cell Working Group (AFCWG),
80ET17019, 1985, pp. 4–177.
1985, pp. 108–131. 26.

C. D. Icavangelo, Proc. Symp. Molten Carbonate Fuel Cell Tech-
3. K. Kordesch et al., Proc. Symp. Porous Electrodes: Theory and
nol., Pennington, NJ: The Electrochemical Society, 1984, p.
397.
Practice, H. C. Maru, T. Katan and M. G. Klein (eds.). Pen-
nington, NJ: The Electrochemical Society, 1984, p. 163. 27.
D. S. Erickson, E. T. Ong, and R. Donado, Abstracts 1986 Fuel
Cell Seminar, Arizona, 1984, p. 397.

4. H. A. Liebhafsky and E. J. Cairns, Fuel Cells and Fuel Batteries,
New York: 1968, p. 107. 28.
C. D. Icavangelo, J. Electrochem. Soc., 133: 2410, 1986.
5. S. Srinivasan et al., J. Power Sources, 22: 359–375, 1988. 29.

S. H. Lu and J. R. Selman, Proc. Symp. Molten Carbonate Fuel
Cell Technol., Pennington, NJ: The Electrochemical Society, 1984,

6. S. Schwartz and O. Lindstrom, Evaluation and Performance of
p. 372.
an Unconditional AFC-Design, 1996 Fuel Cell Seminar Program
Abstracts, 1996, pp. 694–697. 30.
S. H. Lu and J. R. Selman, J. Electrochem. Soc., 131: 2827, 1984.
7. Y. Kiros, A. Sampathhrajan, and M. Ramanathan, Performance 31.

M. T. Langan et al., Lithium-Ferrate-Based Cathodes for Molten
of Pt-Co Alloys and CoTPP Catalysts for the Reduction of Oxygen
Carbonate Fuel Cells, 1996 Fuel Cell Seminar Program and Ab-
in AFC, 1996 Fuel Cell Seminar Program Abstracts, 1996, pp.
stracts, 1996, pp. 402–405.
698–701. 32.
D. Shores and Y. Qu, Dissolution of Oxides in Molten Carbonates,
8. J. Appleby, Proc. Workshop Electrochem. Carbon, S. Sarangapani,
Proc. 3rd Int. Symp. Carbonate Fuel Cell Technol., 93 (3): Pen-
J. R. Akridge, and B. Schumm (eds.), Pennington, NJ: The Elec-
nington, NJ: The Electrochemical Society, 1993, pp. 356–367.
trochemical Society, 1984, p. 251. 33.

S. T. Kuk et al., Properties of LiCoO2-Coated NiO MCFC Cath-
9. K. R. Kordesch, Survey of Carbon and Its Role in Phosphoric Acid
ode, 1996 Fuel Cell Seminar Program and Abstracts, 1996, pp.
367–370.
Fuel Cells, BNL 51418, prepared for Brookhaven National Labo-
ratory, 1979. 34.
D. C. Fee, S. A. Zwick, and J. P. Ackerman, Proc. Conf. High
Temperature Solid Oxide Electrolytes, Brookhaven National Labo-

10. A. J. Appleby and F. R. Foulkes, Fuel Cell Handbook, New York:
ratory, BNL 51728, 1983, p. 29.

Van Nostrand Reinhold, 1989.
35.
S. Majumdar, T. Claar, and B. Flandermeyer, J. Amer. Ceramic
11. K. Kinoshita, Carbon: Electrochemical and Physicochemical Prop-
Soc., 69: 628, 1986.

erties, New York: Wiley, 1988.
36.
T. Takahaski, H. Iwahara, and I. Ito, Denki Kagaku, 38: 509,
12. J. A. Bett et al., Investigation of Alloy Catalysts and Redox Cata-
1970.
lysts for Phosphoric Acid Electrochemical Systems, FCR-7157F,
prepared by Int. Fuel Cells under Contract No. 9-X13-D6271-1 37.
B. C. H. Steele et al., Solid State Ionics, 38/30: 465, 1988.
for Los Alamos National Laboratory, 1985. 38.

P. H. Middleton et al., Proc. 3rd Int. Symp. Solid Oxide Fuel Cells,
13. P. W. Lu and L. L. France, Extended Abstracts, Fall Meeting of
93 (4): Electrochemical Society, 1993.
the Electrochemical Society, Inc., 84 (2): Abstract 573, Pen- 39.

B. C. Steele, State-of-the-Art SOFC Ceramic Materials, Proc. 1st
nington, NJ: The Electrochemical Society, 837, 1984.
Eur. Solid Oxide Fuel Cell Forum, 1: Switzerland, 1994, pp.
375–393.
14. C. S. Kim et al., Evaluation of the Electrode Performance for
PAFC by Using Acid Absorption Acceleration and AC-Impedance 40.
N. Minh, et al, Monolithic Solid Oxide Fuel Cell Develop.: Recent
Measurement, 1996 Fuel Cell Seminar Program and Abstracts,
Technical Progress, Fuel Cell Seminar Program and Abstracts,
1996, pp. 254–255.
1992 Fuel Cell Seminar, 1992.
15. V. M. Jalan, Proc. Workshop Electrochem. Carbon, S. Saranga- 41.

P. Stonehart (ed.), An EPRI/GRI Fuel Cell Workshop on Fuel
pani, J. R. Akridge and B. Schumm (eds.), Pennington, NJ: The
Cell Technology Res. and Develop., Electric Power Res. Institute,
Electrochemical Society, 1984, p. 554.
California, 1992.
16. A. Pigeau et al., Proc. Symp. Porous Electrodes: Theory and Prac- 42.
N. M. Sammes and Y. Zhang, Effect of Microstructure on the High
tices, Pennington, NJ: The Electrochemical Society, 1984, p. 234.
Temperature Mechanical Properties of (CeO2)0.8(GdO1.5)0.2 Electro-
lytes, 1996 Fuel Cell Seminar Program and Abstracts, 1996, pp.
17. H. C. Maru, L. Paetsch, and A. Pigeaud, Proc. Symp. Porous Elec-
115–118.
trodes: Theory and Practices, Pennington, NJ: The Electrochemi-
cal Society, 1984, p. 20.
D. TRAUB HOOIE
18. R. J. Petri and T. G. Benjamin, Proc. 21st Intersociety Energy
US DOE, Federal Energy
Conversion Eng. Conf., 2, Amer. Chemical Society, Washington
Technology Center
D.C., 1986, p. 1156.
19. R. J. Selman, Energy, 11, 156, 1986.
20. C. Baumgartner, J. Electrochem. Soc., 131: 1850, 1984.

21. W. M. Vogel et al., Proc. Symp. Molten Carbonate Fuel Cell Tech-
ELECTROCHEMICAL SENSORS. See INTELLIGENT BIO-
nol., Pennington, NJ: The Electrochemical Society, 1984, p. 443.
SENSORS.
----------------------- Page 116-----------------------
394 ELECTROENCEPHALOGRAPHY
ELECTRODES, ELECTROCHEMICAL. See ELECTRO-

CHEMICAL ELECTRODES.
----------------------- Page 117-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Electrostatic Precipitators
Standard Article

Phil A. Lawless1 and Ralph F. Altman2
1Research Triangle Institute, Research Triangle Park, NC
2Electric Power Research Institute, Chattanooga, TN

Search
reserved.
DOI: 10.1002/047134608X.W3015
❍ Acronym Finder
Abstract
Detailed Examination of Electrostatic Precipitator Operation
Advanced Topics in Precipitation

%20Conversion/W3015.htm17.06.2008 16:37:10
----------------------- Page 118-----------------------
ELECTROSTATIC PRECIPITATORS edges

opposite a grounded plate. As illustrated in Fig. 1, the
discharge electrode, shown as three vertical wires, has a

Electrostatic precipitation is the removal of solid or liquid small
radius of curvature and the collecting surface, a large
particulate matter from a gas stream with electrical forces. radius
of curvature. This geometry produces a nonuniform
The process consists of particle charging, transport to a col-
electric field (see ELECTRIC FIELD MEASUREMENT) around the
lecting surface, deposition on that surface, and ultimately, re- wire
that is strongest at the surface of the wire. In any gas,
moval from the collection surface. In a modern precipitator, there
are always a few electrons and ions liberated by cosmic
these steps can be separate processes or can take place simul- rays
or radioactivity. In the vicinity of the discharge elec-
taneously, depending on the design and operation of the trode,
these naturally occurring electrons are accelerated by
equipment. the
intense electric field and gain enough energy to ionize
Electrostatic precipitation has a number of distinguishing more
molecules. The ions generated by the corona process are
features that make it suitable for a broad range of applica-
responsible for charging particles. This process is discussed
tions. Most electrostatic precipitators are simple devices with in more
detail later.
few moving parts. This simplicity makes cleaning of very The
charging process is divided into two regimes for dis-
dirty gas streams possible, and, with the use of corrosion re-
cussion: field charging and diffusion charging. Field charging
sistant materials, even corrosive gas streams can be cleaned. results
from the strong polarization of a particle in the elec-
The pressure drop through a precipitator is very low, making tric
field that draws ions toward the particle’s surface. Diffu-
possible the cleaning of large volumes of gas economically. sion
charging results from the random collection of ions that
The electrical power consumed is likewise small for the vol- strike
the surface by chance. Generally speaking, field charg-
ume of gas treated. Finally, it is possible to achieve very high ing in
a precipitator is the dominant mechanism for particles
removal efficiencies in a compact system. The configuration of with
diameters greater than 10 #m, while diffusion charging
the equipment and the effects of particle and gas properties
dominates for particles with diameters below 1 #m. For inter-
on the removal process are discussed in detail in the follow- mediate
particle diameters, both mechanisms are important.
ing sections.
Under typical precipitator conditions, field charging is a
rapid
process, giving 80% of the maximum possible charge
Particle Separation Process within
a fraction of a second. Diffusion charging is slower,
but
continues to increase a particle’s charge throughout its
Particle Properties. It is useful to begin with a discussion of
residence in the precipitator. The charging process is de-
the properties of particles typically collected in electrostatic scribed
in more detail later.
precipitators. Precipitators are well-suited for collecting parti-
cles of the sizes commonly generated by many industrial pro-
Electric Force on Particles

cesses. These particles have effective diameters from less
than 0.1 #m to over 200 #m. Particles larger than 200 #m The
electric fields in the precipitator provide the moving force
settle so rapidly that they will fall out of the gas, unless its that
drives particles to the collecting surface. However, the
velocity is quite high. The term ‘‘effective diameter’’ is used
trajectories of particles are far from simple because of the ef-
because particles are not always spherical in shape. Precipita- fects
of turbulence (1). The flow in precipitators is turbulent
tors readily collect irregular particles, but the theoretical because
the flow rate must be kept high to treat the gas eco-
treatment of the collection process always assumes spherical
nomically—lower flow rates require larger precipitators. At
particles to simplify calculations. typical
precipitator velocities, the flow is considered mildly
Other particle properties that affect the separation process
turbulent with a quiet region near the collecting surfaces. In
are the physical state (solid or liquid), the electrical resistiv- the
turbulent part of the flow, the gas motion dominates the
ity, and the dielectric constant of insulating materials. Solid
particle motion, but near the collecting surfaces, the electric
and liquid particles behave similarly in the gas stream, but field
can drive the particles to the collecting surface at a pre-
quite differently when they reach the collection plate. The re-
dictable velocity. Since the charge on particles depends
sistivity and dielectric constant have some effect on the charg-
strongly on their diameters, the velocities near the surface
ing rate of particles but resistivity plays its major role with vary
considerably with size.
the collected particles. In most precipitators, the collected
particle on the surface form a dust layer through which elec-
Collection. The nature of particle motion leads to the con-
trical current must flow. If the resistivity of this layer is high, cept of
a migration velocity that has been used to characterize
an electrical breakdown in the layer can occur, resulting in the
performance of electrostatic precipitators for many years.
sparking or a phenomenon called back corona. In either case,
Although it is a quantity with the units of velocity, it repre-
the useful electrical power for precipitation is limited. sents
an average collection rate that incorporates many ef-
fects
not related to velocity. Later in this article, the equa-
Corona. An electric corona is the key feature of operation tions
needed to calculate the charge on particles and the
of an electrostatic precipitator because it supplies the ions
electric fields in precipitators are presented. These equations
that charge the particles for subsequent collection. The corona make it
possible to calculate with useful accuracy the electric
discharge (see ELECTROSTATIC DISCHARGE) is created by the ap- forces
on particles as they move through a precipitator and,
plication of high voltage to a wire or electrode with sharp
ultimately, the overall collection performance. Still, the mi-
1
----------------------- Page 119-----------------------
2 ELECTROSTATIC PRECIPITATORS
where eff is the percentage collection efficiency, # is the mi-
gration velocity, A is the plate area of the precipitator, and V

is
the gas volume treated. This is known as the Deutsch equa-
tion, after Walter Deutsch who first derived it (2, p 164). The
High
assumption that collection is proportional to the local particle

voltage
power
concentration is valid if particles are uniformly mixed in the
supply
gas, a reasonable assumption when measured at scales com-
parable to the distance between electrodes. At smaller scales,
uniform mixing may not be maintained. Moreover, the collec-

Gas flow
tion rate parameter (c or #) is expected to vary with particle
size and charge.
The quantity A/V is known as the specific collection area
(SCA); it indicates the size of the precipitator relative to the

Figure 1. Principle of electrostatic precipitator operation: particle-
amount of gas it treats. The ratio of gas volume to plate area
laden gas flows past corona wires at high voltage for charging and col-
has the dimensions of velocity (3, p. 96), and we define it to
lection. be
the specific precipitator velocity, # :
ωp = (3)
A
gration velocity describes some of the performance aspects of
precipitators quite well.
The specific precipitator velocity can range from 0.7–10 cm/s,

as
shown in Table 1. The Deutsch equation can then be writ-
Collection Equations. To derive the classic form of the effi-
ten in terms of the ratio of the migration velocity to the spe-
ciency equation for precipitators, the collection process is as-
cific precipitator velocity, so that it becomes clear that the
sumed to consist of the removal of small numbers of particles
migration velocity needs to be 4 to 5 times as large as the
from a uniformly mixed gas stream along each increment of
specific precipitator velocity for high efficiency collection. Spe-
length in the precipitator. Turbulence is assumed to keep the
cific collector area is more commonly used for describing over-
core of the flow mixed, and electric forces are responsible for
all precipitator performance, while the specific precipitator
removing the particles near the collecting surfaces. Moreover,
velocity is useful in comparisons with velocities of particles of
the number removed is assumed to be proportional to the con- a
specific size.
centration of particles in the channel at any point. These as-
A modified Deutsch equation, called the Matts-Ohnfeldt
sumptions lead to an exponential relationship:
equation (4), was derived to account for the range of particle
sizes, but has also been shown to represent other effects, as
ρ(x ) = ρ(0) exp(−cx) (1)
well:
where #(x) is the particle concentration at distance x into the

k
eff = 100 × [1 − exp (−(ω /ωp ) ] (4)
k
precipitator, and c is a collection rate parameter. This equa-
tion is more commonly written as:
where k is a parameter, typically 0.4 to 0.6. The parameter
#
is much larger than the parameter # for the same precipi-
# # ##
k
ωA
eff = 100 × 1 − exp − (2)
tator. Although the form of the Matts–Ohnfeldt equation may
V
seem strange, it matches the performance of operating precip-
Table 1. Examples of Precipitators
Plate Area Gas Volume Voltage

Current Mass Efficiency Specific Precipitator
Type (m2 3
) (m /s) (kV)
(A) (%) Velocity, # (cm/s)
Home Furnace 0.37 0.03 5

2.4 # 10#4 95 8.1
Industrial Air Cleaner 2.5 0.10 12
1.0 # 10#3 97 4.0
Two-stage Pilot Unit 13 0.025 50–45
4.6 # 10#4 98.3 0.19
Industrial Boiler 2,300 41 30–20

0.03–0.12 91 1.8
Utility Boiler 5,900 117 54–48
0.32–0.68 99.92 2.0
0.60–0.80 98 3.7
Paper Mill 14,000 127 44–53
0.37–1.14 99.94 0.91
0.60–0.81 99.98 1.0
1.4–3.3 99.85 1.6
0.45–0.98 99.85a 0.91
0.60–0.80 99.92 1.5
1.0–2.8 99.76a 0.68
0.55–6.6 99.47a 0.65
a These precipitators experience a degrading condition known as back corona, but

still achieve high efficiencies, primarily at the cost of extra size.
Voltage and current values are given for the inlet and outlet sections. Voltage
usually decreases from inlet to outlet; current usually increases from inlet to
outlet.
----------------------- Page 120-----------------------
ELECTROSTATIC PRECIPITATORS 3
Figure 2. Two wire-tube precipitators.
The gas enters at the bottom (not visible),
flows up through individual small diame-
ter tubes, and out the top sides. Electrical
power to the corona wires is supplied from
the two cylindrical metal enclosures on
top of each unit. Collected liquid particles
flow down through the tubes to drop into
a hopper and drain below each unit. (Pho-
tograph courtesy Research-Cottrell, Inc.,
Somerville, NJ  1997 by Research-
Cottrell. All rights reserved.)
itators in the sense that most large precipitators are not as are
washed with internal sprays, rather than being rapped,
responsive to collector area as the Deutsch equation would for
particle removal.
imply.
Discharge electrodes come in a wide variety. They can be
All these equations are only approximations (5), and mi-
classified as (1) weighted wire, with a weight attached to the
gration parameters are basically fitting parameters. They ig- bottom
of the wire to keep it straight, (2) rigid frame, with the
nore many other processes that occur in precipitators (3, p. wires
or other discharge elements suspended between frame
96), but do show that plate area and gas volume are critical
members, (3) rigid mast, with a central member supporting
in determining the collection efficiency.
discharge elements between cross-members, and (4) rigid dis-
charge
electrodes (single, rigid structures with discharge ele-
Mechanical Configuration ments
attached directly to the central member). Regardless of
design, the electrodes all have elements with points or small

There are two principal geometries in industrial/utility pre-
radii
of curvature to generate the high electric field needed
cipitators: the tube type and the plate type. In the wire-tube for
stable corona.
precipitator, shown in Fig. 2, the wire corona electrode is co-
axial with the tube, and the gas flow is along the same axis.
Electrostatic Precipitation Applications

In the wire-plate precipitator (Fig. 3), the electrodes hang ver-
tically between parallel plates, and the gas flow is horizontal The
earliest application of electrostatic precipitation took
between the plates. In either case, the major components of place
just after the beginning of the twentieth century. These
the precipitator are the collection surfaces and the electrode early
precipitators used high voltage transformers with syn-
system that includes the discharge electrode, an insulated
chronous mechanical rectifiers to produce the dc voltage
support system, and a high voltage power supply. If the parti- needed
for corona. Early uses for these precipitators included
cles are collected dry, there must be a mechanical system to
control of emissions from smelters and cement kilns (see AIR
dislodge the particulate layer from the collection surfaces.
POLLUTION CONTROL). Since that period, the technology has
This is often accomplished with a rapping system that period-
steadily improved in performance and reliability, and precipi-
ically strikes the collections surfaces with a weight or ham- tators
are now used in a wide range of applications, including
mer. The dislodged particles are collected in a hopper below (2,
pp. 8–24; 6, pp. 22–30; 7, pp. 2–6):
the plates, from which they are evacuated for storage or dis-
posal. Most tube precipitators are designed to collect liquid 1.
utility industry for the collection of flyash from coal and
particles, but some plate precipitators (‘‘wet’’ precipitators)
oil combustion
----------------------- Page 121-----------------------
Transformer-Rectifier of
99.5%, must be achieved. Yet, precipitators have long dem-
onstrated the ability to meet such demands. The design of
precipitators is discussed in some detail in later sections since

high
efficiency operation requires careful consideration of
many
conflicting goals.
Experience indicates that to achieve these efficiencies, gas

velocities in the precipitator must be kept low, 1.5 m/s or
lower,
and the precipitator must be constructed with a high
degree
of sectionalization. Sectionalization means that only a
small
number of electrodes are energized by a single power
supply
and that only a small number of plates are rapped at
any one
time. In any precipitator that collects dry particulate
material, some material is inevitably reentrained each time

the
collection surfaces are rapped. Sectionalization deals with
this
problem by placing multiple electrical and mechanical
fields
in the direction of flow, sometimes
as many as ten
Collecting Plate fields.
Most of the reentrained material from an early field is
recaptured in following fields. Typically, the rate of collection

Gas Inlet
in the
last field is very low, and it is rapped infrequently, so
that
the amount of material lost to reentrainment is small
compared to the amount entering the precipitator.
DETAILED EXAMINATION OF ELECTROSTATIC
PRECIPITATOR OPERATION
Hopper The
electrostatic precipitator is best described in terms of its
major
subsystems and their interactions. First of all, it is an
Corona Electrode
electrical machine in which the generation of corona and pro-
Tensioning Weights vision
of strong electric fields are essential for collecting parti-
cles.
Second, it is a large mechanical device whose geometry
Figure 3. Wire-plate precipitator. The gas enters at the left face, has
evolved over the years for providing the best collection of
passes through multiple lanes, and exits at the right face (not visible).
particles in the smallest volume. Third, for particle collection,
The plates shown partially across the left face are perforated with
particles must be charged and moved to the collection sur-
many holes to provide uniform gas flow. Electrical power to three
faces
while large volumes of gas pass through the precipita-
independent sections is supplied at the top. Plate rappers periodically
dislodge the collected particles and allow them to fall into the hoppers tor.
These basic subsystems work together to provide the de-
below. (Illustration courtesy Research-Cottrell, Inc., Somerville, NJ sired
collection, but their operation can be upset by unusual
 1997 by Research-Cottrell. All rights reserved.)
conditions, poorly maintained mechanical equipment, or by
changes
in the properties of the particles being collected.
The
understanding of precipitator operation has been
2. iron and steel industries for particulate control aided
immensely with the use of computer models. The num-
3. cement industry for particulate control in roasting ov- ber of
variables and their interactions that must be consid-
ens (see CEMENT AND BRICK INDUSTRY) ered
create a complex situation. Operations that are easily
4. sulfuric acid manufacturing for the collection of sulfuric
described with a few equations have to be calculated repeat-
acid mists edly
in order to accommodate changing conditions inside the
precipitator, a job well suited for computers. Simple models

5. pulp and paper industry for the collection and recycling
of
precipitator performance can be implemented with a few
of salt cake, crystals of sodium sulfate
spreadsheet equations, but accurate models require the re-

6. nonferrous smelters, for both material recovery and sources
of modern personal computers.
particulate control
Since precipitators consist of multiple parallel gas treat-
7. air cleaning for factories and offices ment
paths, it is common to consider a single lane in the di-
8. carbon black industry to aid in the production of car- rection
of flow as adequate for describing the precipitator op-
bon black
eration. This single-lane approach describes almost all
precipitator phenomena quite well, but even in one lane, the

There are specialty applications, as well, but these eight con- system
is still complex.
stitute the majority of operating precipitators.
In the United States, the largest use of precipitators (size
Electrical Operation of the Electrostatic Precipitator

and number) is in the utility industry for control of particu-
late emissions. Most of the precipitators in Table 1 are exam- The
production of corona as a source of ions is one of the two
ples from utility operations. In this demanding application,
electrical functions of the precipitator; the other is creation of
large volumes of flue gas, from 90 m3 3
/s to 950 m /s , are the
electric field. The characteristics of high voltage coronas
treated, and frequently high collection efficiencies, in excess depend
upon the electrical polarity, corona electrode geome-
----------------------- Page 122-----------------------
try, properties of the gas, and properties of the electrode sur- rona
is a function of wire radius (but not material) and gas
face to some extent. In all cases, the corona is a Townsend
density, as given by
electron avalanche sustained by secondary feedback (8).
An electron avalanche occurs when a free electron moves
6 # # δ #
in a strong electric field. It gains enough energy between colli-

Ec = 3.126 × 10 δ + 0.0301 r
sions with gas molecules to ionize the molecules with which
(7)
it collides. This liberates additional electrons that participate
δ = 298P
in further ionizing collisions. The requisite conditions for the
T
avalanche are (1) an electric field in excess of 3 MV/m (30 kV/
cm) in air, (2) absence of gases that attach electrons strongly, where
Ec is in V/m, r is in meters, T is in kelvin, and P is in
and (3) an initial free electron. The avalanche mechanism can
atmospheres. This equation is valid for negative corona; for
lead to a total electrical breakdown of the gas if it is not lim-
positive corona, the value 0.0301 would be replaced by 0.0266.
ited or stabilized. In precipitators, the stabilization comes The
form of this equation is due to Peek (10).
from a combination of electrode dimensions and gas prop-
For negative polarity, positive ion impact can also liberate
erties. free
electrons from the wire surface (see IMPACT IONIZATION).
Each
electron avalanche away from the wire leaves behind a
cloud
of positive ions that are attracted to the wire. Some
Corona Formation. The coaxial wire-tube precipitator, with
positive ions impact the surface with enough energy to liber-

a wire of radius r and a tube of inner diameter R is not only ate
more free electrons. Once a small area on the wire be-
one of the simplest types of precipitators but also one of the
comes
effective for impact ionization, successive avalanches
easiest to analyze. If the outer tube is grounded and the inner
will
tend to originate at the same spot, creating a ‘‘tuft’’ co-
wire is held at a voltage V in the absence of current, the elec-
rona
(11). This ion impact mechanism is so efficient that the
tric field, E0, at the surface of the wire is related to the voltage
critical field for tuft formation is typically 60 percent of the

by (9, p. 32):
value
given in Eq. (7); under most conditions, glow corona will
eventually collapse into tuft corona.

V
E0 = (5)
For positive corona, positive ions left over from the electron
r ln (R/r)
avalanches move away from the wire to fill the interelectrode
gap.
For negative corona (either glow or tuft), the electrons
The magnitude of the electric field, E(z), for any value of z from
the avalanche are captured by electronegative molecules
between r and R is given by
(typically oxygen) to form negative ions that continue to move
away
from the wire. In industrial flue gases, more acidic mol-
E0r
ecules (sulfur dioxide, sulfuric acid) will rapidly collect elec-
E (z ) = (6)
z trons
or strip them from oxygen ions to form the primary cur-
rent
carrying ions in the gas.
The applied voltage can be made sufficiently high so that the
The corona itself extends into the gap only as far as the
electric field at the surface of the wire and for some distance
electric field is able to support the avalanche. However, co-
into the gap is larger than 3 MV/m. Under such conditions, a rona
serves as an almost unlimited source for ions that con-
free electron (from a passing cosmic ray) would create an ava- tinue
across the gap. Without additional stabilization, the
lanche.
current from a wire would increase without limit. The stabili-
If the wire is positive, the electrons move toward the wire
zation is provided by the accumulation of charged ions be-
and are neutralized on the wire surface. If the wire is nega- tween
the wire and the collection electrode, called the ionic
tive, the electrons move away from the wire into regions space
charge.
where the electric field is too weak to support the avalanche.
The ionic space charge produces an electric field that op-
Each avalanche would produce only a single burst of electrons poses
that applied field. This opposing field stabilizes the co-
and ions, if there were no mechanisms to liberate more free rona
and permits the voltage to be increased well above the
electrons in the high field region.
corona onset. With each increase of voltage, the corona injects
One mechanism for sustaining the corona is photoioniza- more
ions into the electrode gap, but the increased space
tion. When an electron and a positive ion recombine, the atom
charge prevents the number of ions from rising indefinitely.
emits ultraviolet radiation to remove the energy of recombi- The
difference between a corona device and other gas dis-
nation. This radiation can ionize a similar molecule some dis-
charge devices, such as neon lamps or thyratrons, is that the
tance away. Since there are a multitude of photons in the
electrode spacings and gas pressures are very different; ionic
initial avalanche, some electrons are likely to be liberated at space
charge can stabilize the corona, but not the discharge
the proper distance to begin a new avalanche. in
smaller devices.
When photoionization sustains the multiple avalanches of

the corona, a ‘‘glow’’ corona is formed. For positive polarity,
Current Density and Voltage. The voltage–current density
the glow forms a tight sheath around the wire about 1 mm
relationship is important because it establishes the possible
thick, while for negative polarity, the glow appears as rapidly range
of operation. (Current density is used for describing
moving discharges along the length of the wire, forming a
precipitator operation because it is independent of the num-
sheath about 10 mm thick. ber
of corona electrodes and plate area.) For voltages below
The lowest voltage at which corona occurs can be used with the
corona onset value, there is no current, no particle charg-
Eq. (5) to calculate the field at the surface of the wire, the ing,
and no collection of particles. Once the voltage increases
‘‘critical’’ field for corona onset. The critical field for glow co- to
the point of repeated sparking, there can be no further in-
----------------------- Page 123-----------------------
crease. The precipitator must operate between these limits

80
with sufficient current density to provide particle charging
) Space charge = 0
2
and collection.
m
c Ionic
The corona onset voltage, Vc , in the wire-cylinder is calcu-
/ 60 Space charge =
A current
n µ 3
lated by direct integration of the electric field to obtain
( density 6.1 C/m
y
# #
t
R
i
s 40
V = E r ln (8)
n Equivalent
c c
e
r
d current density
n Maximum
For a single wire between plates, a similar relation holds:
e 20
r voltage
r Voltage
u
1 + cos #πr
1/2 C 0 offset
 
2b
#
V = E r ln (9)
 
c c πr
20 30 40 50 60
1 − cos
2b
Voltage (kV)
where b is the wire-plate separation. When multiple wires are

Figure 5. Shifting of the V–j curve with particulate space charge.
2
between plates, there is a mutual coupling effect that raises The
5 kV offset is equivalent to a current density of 20 nA/cm , which
the corona onset voltage for all the wires. When the wire-wire
reduces the amount of real ionic current that can flow at a given
voltage.
separation is greater than twice the wire-plate separation (a
common configuration), the coupling effect is only a few per-
cent of the single wire corona onset voltage.
(12). Eq. (11) and Eq. (12) are inferred from the insights of
Above the corona onset voltage, the relationship between
Sigmond (13) and represent nearly universal corona behavior.

current density and voltage is
That is, the ions are so dense at the wire that mutual repul-
# #a + 1##
sion dominates all other effects (see SPACE CHARGE AND SPACE
V = V + E r a − 1 − ln
c c
2
CHARGE MEASUREMENTS), leading to a voltage dependence that
# j 0x # x #2 (10)
follows Sigmond’s saturation law (with a correction for the
a = 1 +
corona onset voltage) and an angular dependence that ap-
µ# E r
c
proaches a straight line expansion away from the corona wire.
Charged particles in the precipitator complicate the volt-

where j 0 is the current density at the collector, # is the ion
mobility, # is the permittivity of space, and x is the wire-
age–current density relation to the point that there is no ex-
collector separation distance (R for the wire-cylinder or b for act
representation of the V–j curve. A uniformly distributed
the wire-plate). This relationship is exact for the wire-cylin-
cloud of particles moving through a wire-plate precipitator
der (9, p. 34), where the current density is equal over the
will produce an electric field that opposes the formation of
whole collector surface, and nearly so for the wire-plate for
corona, raising the corona onset voltage by an amount:
the central current density, the current density directly under
S b2
the corona wire on the plate.
#V = (13)
A fairly accurate estimate of the current density for a given
# 2
voltage can be obtained with the equation:
where S is the total particulate space charge (C/m3). In addi-

j 0 = µ# (V 2 − V 2 ) (11)
tion, the space charge enhances the electric field toward the
3 c
x
collection plate and so assists in the movement of ions. This
This approximate relation matches Eq. (10) best at the higher

results in a steepening of the V–j curve. A reasonable esti-
values of j 0 . For critical work, Eq. (10) is preferred.
mate of the effects of space charge is obtained by assuming
In the wire-plate geometry, the current density along the
that it is an equivalent current density added to the ionic
plate varies approximately as
current density, as illustrated in Fig. 5. As the space charge
n
increases, the real ionic current available at a given voltage
j (θ ) = j 0 cos (θ ) (12)
decreases. In extreme cases, the ionic current may be sup-

where the angle # is measured from the perpendicular line
pressed or ‘‘quenched’’ altogether.
from wire to plate, shown in Fig. 4, and the value of n is 4–5
The electric field around the wire establishes the charging
conditions and provides the motive force for collecting parti-
cles. The presence of ions and particulate space charge modify

the
Laplacian (zero current) field given by Eq. (6).
Electric Field. The electric field with ionic current and with-
r
out
space charge in the wire-cylinder is given by
θ b
E (z ) = ##E0r #2 + jz #1 − r2 ##1/2 (14)
z µ# z
Figure 4. Wire-plate geometry with wire radius r and wire-plate

Here, j is the current density crossing an imaginary cylinder
spacing b. The electric field and current density at the central loca- at
radius z . The Laplacian portion of the field (j # 0) de-
tion (# # 0) are almost the same as for the wire-tube precipitator.
creases with distance from the central wire, but the ionic
----------------------- Page 124-----------------------
space charge increases the field as the outer cylinder is ap-

Equation (10) and Eq. (11) are approximately valid
proached. At the outer cylinder, with z # R and neglecting
whether or not the voltage is changing with time (at power
the correction term for the wire radius, the field becomes
line frequency) because the ions can cross the electrode gap
in
about a millisecond. Therefore, when the voltage on the
E (R ) = ##E0r #2 + j 0R #1/2 (15)
electrodes rises to its peak value, the current density also
rises to a peak value that may be several times its average

R µ#
value, depending on the wave form. The electric field also fol-
lows the voltage and current density changes. The phase rela-
A similar expression works for the wire-plate geometry at the
tionship between voltage and current may be modified by the
central location:
capacitance of the system in addition to the ion transit time.
Some precipitator phenomena do not respond rapidly

E (b) = ##E0 πr #2 + j 0b #1/2 (16)
enough to be affected by the peak values, but the phenomenon
b µ# of
sparking is directly attributed to the high electric fields
that exist at the peak of the voltage wave form.
These expressions are for single electrodes. In the wire-plate

geometry, adjacent electrodes make a contribution to the
Sparking. Sparking is a complete breakdown of the elec-
Laplacian term, increasing it by up to several percent for
trode gap with enough gas ionization to effectively short cir-
closely spaced wires. In the wire-plate geometry, the electric
cuit the precipitator power supply. Sparking is usually initi-
field along the plate varies approximately as
ated by the formation of positive streamer corona at one
electrode that extends

to the other electrode. The streamers
E (θ ) = E (0) cosn (θ ) (17)
propagate by photoionization into the gap; their repeated pas-
sage over the same volume heats the gas, increases its con-
where the value of n is 2.
ductivity, and leads to the catastrophic spark. The streamers
will form and propagate when the electric field is in excess

Particulate space charge adds a constant amount to the
of
0.5 MV/m (5 kV/cm). The temperature-reduced pressure
field at the plate of Sb/#, and there is no angular dependence
dependence of the formation field is moderate, varying

for this component of the field. As a result, the electric field
3/2
becomes
roughly as (pressure ) (14). This means that altitude, as well
as
the temperature, of the precipitator will affect the field for
Sb
sparking noticeably. The moisture dependence is also fairly
E (θ ) = E (0) cosn (θ ) + (18)
strong, such that the amount of water in a combustion gas
#
stream might raise the field for sparking by 1 to 2 kV/cm.
All control systems for TR sets are designed to turn off the
Strictly speaking, the space charge will increase from charg-
voltage for a few cycles after a spark occurs to protect the

ing as the particles approach the electrode and decrease as
supply and the precipitator electrodes from damage. This

some of the particles are collected near the electrodes. These
does shut down the whole electrical section served by the set
effects are real, but small, for the amount of time particles
for
a few tenths of a second and represents a period of time
spend near any one electrode.
when little collection (except for the space charge field) takes
The interest in the values of field near the plate comes
place. Repeated sparking at the same spot on a wire is one of

from the assumption that the near-plate region is where the
the
primary causes of electrode failure.
collection of particles takes place. The electric field in the in-
terelectrode gap is important in determining the charging
conditions, but some average value, computed from Eq. (14),
Gas Composition Effects. It has been recognized that the
is almost always used.
moisture content of the gas affects the sparkover voltage sig-
nificantly. Other elements in the gas composition can affect

Voltage Wave Forms. Although the precipitator uses high the
electrical operation equally strongly. Ion mobility, defined
voltage dc to operate, the wave forms may vary considerably. as
the ratio of ion velocity to electric field, is characteristic of
Older precipitators were often energized with unfiltered half-
each type of ion. Small, compact ions have high mobilities,
wave rectified dc, probably for economic reasons, since a sin-
while large, complex ions have smaller mobilities.
gle transformer-rectifier (TR) could energize two separate sec-
In air, the negative ions that carry current are typically
tions. Newer precipitators use full-wave rectified dc for all
hydrated oxygen molecules, with a reduced mobility of about
2
# 10#4 2
sections. Although power supply filtering is rarely used, the
m /V # s. In precipitators controlling emissions from
precipitator itself is a large capacitor and smooths the wave
coal-fired boilers, sulfur oxides form the dominant ions be-
form somewhat. Some TR controllers now interrupt the pri-
cause their acidity allows them to capture electrons from the
mary ac voltage for several half-cycles to provide an intermit-
oxygen ions. Sulfur dioxide (SO2) has a mobility of about 1.80
#
10#4 2
tent energization, with the goal of reducing the current con-
m /V # s. at very low concentrations in air (15), but
#4 2
sumed while keeping the peak voltage high.
about 0.41 # 10 m /V # s at high concentrations (9, p. 24).
Generally, the average current density for a given average The
crossover from low- to high-concentration behavior occurs
voltage is not sensitive to the wave form. The current density- at
about 400 #L/L (16). Ion mobility is also a function of gas
voltage characteristic, Eq. (11) for example, is linear enough
density (a combined pressure and temperature effect). Conse-
over the normal operating range that the differences in wave
quently, it is possible for seemingly minor changes in gas
form do not produce different curves. Wave form does dramat-
composition and conditions to affect the operation of the pre-
ically affect the peak voltage and peak current density.
cipitator by a large amount.
----------------------- Page 125-----------------------
Mechanical Design of the Precipitator All

the electrode designs require that the electrodes be iso-
lated
from ground by strong insulators. The insulators must
The mechanical design and construction of precipitators af-
not only
support the gravitational load of the electrodes but
fects their performance in subtle ways. In the wire-cylinder
also
dynamic loads if the electrodes are rapped to remove de-
precipitator (see Fig. 2), the collected particles must be re-
posited
particles.
moved by flowing downward against the upward gas flow.
This limits the use of tube precipitators to collecting liquid
Plate
Design. Tall plates suffer some of the problems that
particles or providing for offline cleaning. In the plate-type
the
electrodes do. Even when suspended from the top, warp-
precipitator (Fig. 3), there are three axes of movement for
age may
occur because of manufacturing flaws, thermal gradi-
particles: along the horizontal gas flow, transverse to the flow
ents,
overheating, or other causes. Many plates are designed
towards the plates, and vertically downward towards the hop-
with
stiffening ridges, at right angles to the plate surface, in
pers. This allows the collected particles to be removed without
them.
The ridges extend into the gas stream somewhat and
serious reentrainment in the gas flow. Most of the following
affect
the gas flow. Sometimes, the stiffeners are also in-
discussion concerns plate-type precipitators.
tended
to shield the deposited particles from the gas flow to
reduce
erosion of the deposits.
Overall Dimensions. The sizing of a precipitator for high
efficiency particle collection for coal flyash aims toward a
Sectionalization. As the need for more efficient collection
treatment time of 4 to 10 s with a gas velocity in the direction grew,
it became apparent that parts of the precipitator
of flow of no more than 1.5 to 1.6 m/s. Higher velocities tend needed
to operate in different regimes. Near the inlet, the
to cause increased rapping losses and continuous erosion of particle
concentrations are high and produce a high space
the collected particles. Other types of particles will modify charge.
This requires a high voltage to drive the corona cur-
these criteria. rent.
Near the outlet, most of the particles have been col-
The overall length is then optimally in the range 6 to 16 lected,
and a high voltage will produce very large corona cur-
m. Invariably, other constraints limit the length: fitting into rents
and sparks. By breaking the precipitator into shorter,
an existing space, costs of land, arrangements of ductwork,
independent sections, each section can operate at its own opti-
and so on. The lane height and overall width need to be ad- mal
conditions. Sectionalization also provides redundancy to
justed to accommodate the volume of gas and the treatment protect
against equipment failures. A four-section precipitator
time. may lose
one section to a failure and still operate fairly well;
The lane height is also constrained by considerations of a single-
section precipitator fails totally.
electrode alignment and the total distance the collected parti- Each
section is much shorter in the direction of flow than
cles must fall to reach the hoppers. Common lane heights it is
tall. This makes collected particles fall a long distance
range from 9 to 12 m. In retrofit situations, it is often easiest toward
the hoppers, during which reentrainment may occur.
to add height to the plates because the footprint of the precip- Plates
in a section may extend the full length of the section
itator does not need to be modified. In those situations, even in the
direction of flow or may be smaller subunits closely
taller plates might be used, but the rapping losses might be- fitted
together. Sometimes, the individual subsections are
come unacceptable. rapped
independently.
The total width has no effect on the operation of the precip-
itator, except for the details of mechanical construction. The Access
Lanes. There are usually access walkways running
ratio of width to height in large precipitators varies from 1 : 1 the
width of the precipitator between sections to allow for
to 8 : 1. Wider precipitators are usually subdivided into sepa-
maintenance and inspection. Particles emerging from one sec-
rate chambers for better mechanical integrity. tion are
thoroughly mixed in the access lanes before passing
Within the optimal treatment times, there is some latitude into the
next section. The dimensions of the access lanes allow
for choosing the plate spacing (lane width) and collector area. large
turbulent eddies and cross-flows from one part of the
Wider plate spacings (30 cm to 40 cm) provide more tolerance
precipitator to another to form. Even if care is taken to assure
for electrode alignment problems and require less plate mate- low
turbulence within the lanes, the access lanes change the
rial but always require higher voltages and may encounter flow
structure and homogenize the particle concentration.
problems with particulate space charge. Narrower plate spac-
ings (22 cm to 27 cm) are more traditional, provide more plate
Sneakage. The requirement for electrical isolation of the
area for collection, and operate at lower voltages but are more high
voltage electrodes means that rather large spaces must
critical to align and maintain. be
allowed at the top and bottom of the lanes to prevent
sparking.
These spaces provide the opportunity for particle-
Electrode Design. Long electrodes are difficult to keep laden gas
to bypass the collection zones, a phenomenon called
aligned between the plates. Wire electrodes are often sneakage.
Internal baffles are often used to suppress the
weighted to keep them vertical and taut; the weights may be sneakage
flow or redirect it toward the active collection zone,
guided at the bottom to suppress swaying. With frame elec- but the
clearances must still be maintained.
trodes, wires are stretched across a rigid tubular frame to To
some extent, the sneakage flows are driven by the pres-
provide alignment; the frame itself must be carefully con- sure
drop along the length of the precipitator. The presence
structed to remain aligned. Rigid electrodes and mast elec- of plate
stiffeners produces a back pressure that is partially
trodes provide a strong tubular spine along which corona elec- relieved
by gas flows out of the top and bottom of the lane. A
trodes are placed at close-spaced intervals; these electrodes simple
model of the pressure drop (17) shows that increased
are often supported on alignment frames at both top and stiffener
protrusion into the lane increases the sneakage flow
bottom. and that
taller plates show proportionally less sneakage than
----------------------- Page 126-----------------------
shorter plates. Sneakage flows represent losses of efficiency

precipitator because each particle must be charged in order to
and may also allow particles to contaminate areas that should be
collected. The ions used for charging the numerous small
be kept clean. particles
are not available to charge the larger ones, and the
strong
particulate space charge limits the ionic current. For-
Particle Collection tunately,
when it is known that a submicron mode is likely to
be
present, the electrode design can be altered to compensate
Particle collection is the primary goal of precipitation, but in
for the
problem.
many aspects, it can be discussed independently of the electri-
cal and mechanical aspects of the precipitator. For much of
the time in the precipitator, the particle cloud is so dilute that
Particle Charging. Particle charging in precipitators has
the charging and collection do not affect the electrical opera- been a
topic of interest for decades because ionic charging cov-
tion of the precipitator. Once the operation of the precipitator ers
nearly the full range of parameters that can be modeled.
is understood for dilute concentrations, then the concentrated For large
particles, the field charging (convection) model gives
regime can be addressed. an
appropriate description. For small particles, diffusion mod-
els give
an appropriate description. In between, composite or
Particle Size Distributions. The particles that the precipita- numerical
models must be used to adequately describe the
tor must collect are rarely all of the same size. The processes charging.
An excellent summary of the classic field charging
that produce particles (combustion, chemical reactions, me- model is
in (9, pp 58–63), and a comparison of all the classic
chanical processes) make them over a range of sizes, called a
precipitator charging models in nondimensional form is made
size distribution (18). The size distribution describes the num- in (19).
When an ion is near the surface of a particle, it is
bers of particles of a specific size that would be found in a attracted
to the particle by image forces, and the concentra-
representative sample. For precipitation, size distributions tion of
ions at the surface is zero. Field and diffusion charging
are considered in terms of number or mass. Mass distribu- models
describe the transport of ions to the vicinity of the
tions are easily measured by aerosol collection equipment, particle
surface.
such as cyclones or impactors, while numerical computations In
fact, the field charging model is only an approximation
are most easily carried out in terms of number distributions. of the
true charging relationship; diffusion effects can be ob-
If the particles are spherical, the conversion between mass served
well into the classical field charging regime. Convec-
and number distributions are straightforward. tion
along the electric field (field charging) delivers ions to the
Figure 6 shows a particle number and mass size distribu-
neighborhood of the particle, but diffusion across the concen-
tion of the type often found in coal-fired precipitator applica- tration
gradient at the surface determines the charging cur-
tions. The mass distribution has a maximum value at about rent
(20).
10 #m diameter, and the shape is roughly lognormal (normal
in the logarithm of the independent variable). Approximately
Particle Charging Theory. The particle charging equations
half the particulate mass is in particles larger than 10 #m, are
simplified if expressed in nondirectional form (19). The
which are easily collected; the remaining mass below 10 #m
nondimensionalized terms are the particle potential, electric
is harder for the precipitator to collect. At the upper end of field, and
charging time. (Other quantities are also needed for
the mass distribution, the large particles are fewer than the
development of the theory (21), but these three are suffi-
might be expected because they have settled in the furnace cient to
understand the results.) The terms are defined as
and ducts. At the lower end of the distribution, a condensa- follows:
tion aerosol may form from material volatilized in the furnace Self
potential of the particle, # (also called the particle
(silica or sulfuric acid, for example). The number distribution charge):
is heavily weighted toward small particle diameters because
the mass of a particle varies as the cube of its diameter.
ν = ne #kT (19)
A number distribution strongly peaked at small particle

4π#a e
sizes (called a submicron mode) can present problems for the
where n
is the number of elementary charge units e, # is the
permittivity of air, a is the particle radius, k is Boltzmann’s

constant,
and T is the absolute temperature. This expression
Number is the
ratio of the electrostatic potential to the quantity
t
n kT/e,
the thermal potential. The thermal kinetic energy of
u
o the ions
in the gas divided by e is about 26 mV at room tem-
m
a Mass perature.
Because ions have this average thermal energy,
e
v
i they
can overcome repulsive forces near the particle and
t
a
l reach
the particle’s surface, even when the classical convec-
e
R tive field
equations forbid such occurrences.
Electric field, w:
0.1 1 10 100
w = aE #kT (20)
Particle diameter ( m)
µ
e
Figure 6. Number and mass particle size distributions typical of

coal-fired precipitator applications. The small mass peak at 0.2 #m is where E
is the uniform applied electric field. The product aE
a condensation aerosol and may not be present in all situations. is the
potential across the radius of the particle. When this
----------------------- Page 127-----------------------
product is smaller than thermal potential, diffusive effects

100,000
will dominate the particle charging.
Time, :
10,000
) Full charging
n
ρµt
( 1,000
τ = (21)
e
#
g
h 100
C
where # is the ion density, # is the ion mobility, and t is the
actual charging time. The quantity #t is often termed the ion
10 Field charging
exposure time.
1
With these definitions in hand, the charging rate for classi-
0.1 1 10
cal field charging is
Particle diameter ( µ m)
#
dν 3w ν 2
Figure 7. Computed particle charge (number of electrons) on parti-
dτ = F (ν, w) ≡ 4 1 − 3w , ν ≤ 3w cles
in a precipitator under normal operating conditions (E # 5 #
(22) 5
#5 2
≡ 0, ν > 3w 10
V/m, j # 1.2 # 10 A/m , time # 1.9 s). With the high electric
field,
sparking occurs in the precipitator but does not affect the
charge. The straight line computed by field charging alone shows the
where 3w is commonly called the unipolar saturation charge;
dependence on the square of the particle diameter and an increasing
it is the upper limit of charge that can be attained by the field
discrepancy from real charges below 3 #m.
charging mechanism. The field charging rate is named F (#,
w) for convenience in referring to it. The value 3 comes from
the assumption of conductive particles. (For most precipita- where
the area fraction, f (w), is described by the equation:
tors, this assumption is reasonable.)
The field charging model has a strong initial charging
1
rate, 3w/4 , that becomes zero at # # 3w. The larger the parti-
f (w) = (w + 0.475)0.575 , w ≥ 0.525 (26)
cle (radius a), or the stronger the field, (E), the greater the
= 1, w < 0.525
charging rate is. When # # 3w, the charge on the particle is
When
the particle charge reaches 3w, the field charging rate
2
ne = 12π#a E (23)
becomes zero, but the diffusive contribution continues to raise
the
particle potential above 3w, at a decreasing rate.
Large particles, therefore, acquire very large charges in the
Figure 7 shows some particle charges computed in model-
precipitator and can be easily collected. ing
a precipitator. Even though the electric field changes in
The classical diffusion charging rate in nondimensional no- time
(because of sparking) and in space (as the particles pass
tation is each
corona wire), the charge on each particle increases con-
tinuously because there is no way for charge (ions) to leave a

dν = Be(ν) ≡ ν (24)
particle once they have been captured.
dτ exp(ν) − 1
Charging at High Particle Concentrations. Once the charging

This function is called the Bernoulli function, designated here
problem for dilute particles is understood, it is possible to ex-
Be(#), because it is a generating function for Bernoulli num- tend
the modeling to more concentrated particle clouds. Par-
bers. The initial diffusion charging rate is 1 and becomes ex-
ticulate space charge is the sum of all the charges on all the
ponentially smaller as the charge on the particle increases,
particles within a volume of space. Eq. (13) shows that the
without ever completely ceasing. space
charge shifts the corona onset voltage upward and re-
The relative importance of the field and diffusion contribu- duces
the current density. Each shift reduces the ions avail-
tions to charging can be estimated by comparing the initial able
for charging and decreases the charging rate. Particles
charging rates. Once the saturation charge is reached, how-
receive only a fraction of the charge that they would under
ever, field charging ceases, but the diffusive component re- more
dilute conditions. Nonetheless, some particle charging
mains. The method of combining the two charging rates de- does
occur, even under these reduced-current density condi-
termines the accuracy of the overall particle charging model.
tions. The charging is slower than under dilute conditions,
As long as the particle charge # is less than 3w, some of and
it is more difficult to calculate. The charging model is
the particle’s surface is at the same potential as the sur- still
correct, but the local ion density and electric field change
rounding space. This means that a portion of the surface re-
rapidly with each position inside the precipitator.
ceives a diffusive current in addition to the field charging. In
addition, that diffusive current is the same as for a particle
Particle Motion. Charged particles experience a force in an
of zero charge in the absence of the field. Taking the frac-
electric field of neE. In still air, a particle accelerates until
tional area receiving the diffusive contribution into consider-
viscous drag in the air exactly opposes the electric force, at
ation, the charging rate becomes which
time the particle moves with a constant drift velocity,
given
by
dν
dτ = F (ν, w) + f (w)Be(0), ν ≤ 3w
neE C
(25)
v(a) = (27)
= f (w)Be(ν − 3w), ν > 3w
6πηa
----------------------- Page 128-----------------------
100
electric fields near the wires and directly under them move
) the
particles toward the collecting plates. When the turbulent
s
/ eddy
velocities are lower than the particle drift velocity, the
m
c
particle can travel to the plate and be captured.
(
y
t
The collection of particles can be modeled as a series of
i
c 10
o
collection zones under the corona wires, followed by mixing
l
e
v
zones between the corona wires. The classical Deutsch rela-
t
f
i tion
for a given particle size can be written as
r
D
1
p (a) = exp(−v(a)/ωp ) (28)
0.1 1 10
Particle diameter ( µ m)
where p is the penetration for particle size a, v is the
drift
5
velocity, and #p is the specific precipitator velocity. Penetra-
Figure 8. Computed particle drift velocities in a field of 5 # 10 V/
m. The minimum in velocity typically occurs in the range 0.3 #m to tion
is 1 # eff / 100, the fraction remaining after collection;
0.5 #m. Although particle velocities greater than 100 cm/s are calcu-
emissions from the precipitator are proportional to the pene-
lated for larger particles, other factors related to the particle’s Reyn-
tration. However, a better collection model has been found
olds number need to be taken into account to compute the velocity to
be
accurately. Even so, such particles are collected very efficiently.
# #
v(a)
p (a) = 1 − (29)
N ω
where C is the Cunningham slip correction factor, and

is
the viscosity of the gas. This is an expression of Stokes’ law
where N is the number of collection zones or wires in the pre-
(18), relating the particle velocity to its viscous drag. Since
cipitator (23). Equation (29) is a two-term series approxima-
the particle charge increases roughly as a2, and the drag in- tion
to Eq. (28), but it actually predicts precipitator collection
creases as a1, the particle velocity increases with the radius
better than the Deutsch equation does, a finding related to
a. The slip correction factor, C, accounts for the fact that as the
relatively low turbulence in precipitators. The Deutsch
particle diameters become comparable to the mean free path
equation does not allow all the particles to be collected in a
of the gas molecules, the particles slip through the gas with
precipitator, but Eq. (29) does. If the drift velocity exceeds
less hindrance (22). As a result, the particle velocity increases N#
the penetration for that particle size is zero.
p
for particles smaller than about 0.5 #m diameter. In turbu-

The total collection performance for a section must be eval-
lent gas flows, this equation indicates the particle velocity
uated by summing over all the particle sizes in the particle
with respect to the local gas velocity. size
distribution. The penetration of each particle size deter-
Particle drift velocities corresponding to the charging con-
mines how many particles of that size appear at the outlet of
ditions in Fig. 7 are shown in Fig. 8. Most of these velocities the
section, and the aggregate total allows the effective mi-
are well above the characteristic precipitator velocities in Ta-
gration velocity, #, to be calculated, if desired. The particle
ble 1, suggesting very effective collection. Near the minimum size
distribution will change from section to section as the
drift velocity, the collection efficiency is the poorest. more
easily collected large particles are removed from the
gas.
Gas Flow in Precipitators. As indicated before, the target

velocity for the gas flow in precipitators is 1.5 m/s or slower.
Rapping. The collected (solid) particles build up a dust
For the lane widths commonly encountered, 0.2 to 0.4 m, com- cake
on the plates that must be removed periodically. It is
puted Reynolds numbers range from 4500 to 15000, well
usually desirable to keep the layer thickness between 2 and
above the threshold for turbulent flow (2000 to 2200), but not 5
mm to obtain good cake cohesion when the material falls
indicative of extremely turbulent conditions. If the gas flow into
the hopper. In the inlet section, the layer thickness may
entering a lane is nonturbulent, turbulence may not fully de-
increase 6 to 8 mm/h, necessitating frequent rapping. In the
velop until most of the lane has been traversed. Plate stiffen-
later sections, longer rapping periods are generally used.
ers, referred to in the section on mechanical design, do induce
Rapping is intended to break the dust cake loose from the
turbulence in the flow, and once established, the turbulence
plate and allow it to fall as a unit until it reaches the hopper
remains constant. or
is trapped under another corona wire. Even if these condi-
Although much research has been devoted to the study of
tions hold, some rapped dust will fall off the trailing edge of
low-turbulence precipitators, it has been found that corona the
section and have to be captured downstream. The desir-
and particulate space charge induce motions in the flow able
dust cake thickness should decrease with each section
stream equivalent to turbulence. Ions and charged particles
downstream of the inlet to reduce these rapping emissions,
drag gas toward the collecting plate, an effect called corona but
the rapping period should still be long enough to allow a
wind. Near the plate, the gas must change direction and move
cohesive dust cake to form.
away from the plate, even though the ions and particles con-
tinue towards the plate. This motion creates a large-scale
Deficiencies in Precipitator Operation
eddy that carries some of the particles back toward the center
There are aspects of precipitator operation that interfere with
of the lane. the
collection and disposal of particles. Some of these occur
with
every precipitator, while others are specific to certain
Local Particle Collection. Turbulence tends to homogenize
types of particles. Improper maintenance of the precipitator
the particle concentration across each lane, while the strong will
also produce problems of operation.
----------------------- Page 129-----------------------
Rapping Losses. The rapping process allows some collected

basis for using a precipitator as an agglomerator in carbon
material to be lost into the gas stream on its way to the hop-
black processing.)
pers. Even though the reentrained material may be caught in
At higher resistivities, when corona current flows through
the later sections, when they are rapped, some of that mate- the
dust cake, strong forces develop to compress the cake, and
rial will be reentrained. In the last section of the precipitator,
erosion is unlikely. In adjacent areas without current flow,
the reentrained material leaves the precipitator and becomes
particles may be collected by the electric forces, but the cake
part of the total emissions. From measurements of precipita- is
much looser and more easily eroded. The patterns of corona
tor emissions with rappers on and off, it has been estimated
current clamping are determined by the electrode design (24).
that the rapping contributes 15 to 65% of the total mass emis-
When continuous erosion is a problem, reductions of velocity,
sions, depending on the precipitator’s configuration.
changes in electrodes, and the use of additives to improve co-
A model (23) of the rapping process gives the fraction reen-
hesion are potential solutions.
trained from each plate as
Back Corona. If the electric field in the dust layer from the
H (0.18vg )2 flow
of corona current becomes large enough, the field can
RR = (30)
L g #x
support corona generation in the interstices of the dust layer.
The
corona in the layer injects ions of opposite normal polar-
where H is the plate height, L is the plate length in the direc- ity
back into the interelectrode gap, hence the name back co-
tion of flow, vg is the average gas velocity, g is the acceleration
rona. The precipitator is intended to be unipolar (charge of
of gravity, and #x is the distance in the direction of flow be- one
sign only). When ions of both polarities are present, they
tween wires. The factor 0.18, obtained from fitting the model
charge particles in opposition, so that the net charge is much
to several measured cases, predicts gas velocity near the plate
reduced below the full unipolar charge.
as a function of the average velocity. The model gives rapping
It has been found that the critical field for back corona
5
reentrainment factors approximating real precipitator opera-
formation in coal flyash is about 5 # 10 V/m. This is much
6
tion. It shows that rapping losses increase dramatically with
smaller than the 3 # 10 V/m required for corona at the
gas velocity and points out the importance of good gas flow
high voltage electrode. Two factors assist the formation of
conditions.
back corona at lower fields in the dust. First, current flow
between touching particles is focused into very small areas

Velocity Maldistribution. Even though the average gas ve- at
the point of contact. This focusing creates high electric
locity through the precipitator may give the proper treatment
fields that make microsparks in the particle-particle gaps.
time, if there are regions of high and low gas velocity, the
Second, the close confines of the pores in the dust cake
precipitator will not work as well as it should. In the low ve-
assure that photoionization and ion impact events will be
locity regions, the precipitator will actually exceed the aver-
very effective at producing secondary electrons to sustain
age collection performance, but in the high velocity regions, the
ionization process.
the performance will be so degraded that the overall perfor-
Back corona has two deleterious effects on precipitator op-
mance will suffer.
eration. First, the injection of opposite ions increases the cur-
Part of the degradation in the high velocity regions is due
rent and lowers the operating voltage; the electric field is re-
to reduced charging and collection time for the particles, but
duced as well. As a result, both particle charging and
a larger part of the degradation comes from the increased rap-
collection suffer in direct proportion to the field reduction.
ping reentrainment where the gas velocity is higher. There
Second, the presence of the positive ions reduces the net
are some subtle effects in the low velocity portions as well.
charge on all particles. The negatively charged particles vig-
Sparking may occur because the particle space charge is so
orously attract the positive ions, partially neutralizing them.
effectively removed. The dust cake will accumulate faster be-
Particle charging models can all account for bipolar ionic
cause of the better collection and may not be rapped at the
charging; all predict zero net charge if the bipolar ionic cur-
appropriate intervals, leading to downstream overload condi-
rents are equal.
tions when the plates are rapped.
Since back corona generation cannot proceed without the
Modern precipitator design calls for the gas velocity at the
normal corona, the back ionic current will always be a frac-
face of the precipitator to be made uniform to within 15 per-
tion of the forward ionic current, and most particles will,
cent rms. Precipitators that operate outside that range may
therefore, carry a net unipolar charge. However, the back co-
have their performance improved with correct flow distribu-
rona current fraction approaches 1 as the resistivity of the
tion. In cases with loosely cohesive dust, the flow conditions
dust cake increases. The degradation of precipitator perfor-
for good performance may be even more stringent.
mance by back corona can be managed if the dust resistivity
can
be controlled.
Continuous Erosion of Dust Cake. In some cases, it has
Resistivity can be controlled by adding chemicals to the
been found that the dust cake experiences continuous erosion gas
that adsorb on the particles and change their characteris-
between rapping intervals. This is a function of the cake
tics. Sulfuric acid, in concentrations of a few #L/L, is effective
cohesivity and gas velocity near the collecting plates. Low for
this purpose. Sodium carbonate is used in some high tem-
resistivity also contributes to the erosion problem because if
perature precipitators where the particle resistivity is af-
the electric field (product of current density and resistivity)
fected by the internal depletion of sodium ions. Ammonia can
in the dust layer is lower than the electric field in the gas,
prevent back corona but appears to affect the gas properties
there is a surface charge that exerts a force pulling the cake
more than the dust cake resistivity.
toward the interelectrode gap. Only the dust layer cohesion
Another way to prevent back corona is to pulse the co-
opposes this force. (This low resistivity effect forms the
rona intermittently. The dust layer breakdown is a function
----------------------- Page 130-----------------------
of the time-averaged current density which can be reduced

Cohesion. The cohesion of the dust cake affects that man-
by as much as a factor of 10 by intermittent energization. ner in
which it falls when rapped. Is it possible to determine
The peak electric field remains unchanged or may increase values
of cohesivity acceptable for rapping? How does the
slightly during pulsing, while the time-averaged electric layer
thickness affect the rapping properties, and are there
field is somewhat reduced. Particle charging responds pri- optimum
values?
marily to the peak electric field, while particle collection is
most affected by the average electric field. Pulsing can,
Size Distribution. In many cases, the rapping emissions
therefore, control back corona onset quite well over about have an
almost universal size distribution with a mean diam-
a decade of resistivity and offer some improvement over a eter of
6 to 10 #m. In others, the rapping distribution is simi-
broader range. lar to
the size distribution of the particles being collected.
What
are the factors that govern the rapping size distribu-
Low-Voltage Sparking. Sparking at lower than expected tion?
Can the size distribution be predicted from particle
voltages reduces the performance of the precipitator in pro-
properties?
portion to the voltage reduction. Back corona does cause low-

voltage sparking over a rather narrow range of resistivities,
Mechanical Design. The rapping reentrainment model sug-
but the correction is to control the back corona. Other more gests
that mechanical dimensions (electrode spacing, length,
likely causes of low voltage sparking are misalignment of and
height of collection plate) affect the amount of material
electrodes and insulator problems.
reentrained. Do these factors play the role that the rapping
Electrode misalignment brings some part of the high volt-
reentrainment model implies? Are there ways to vary the
age system too close to the grounded plate system. As a re-
electrode spacing that would reduce rapping emissions?
sult, the electric field (roughly voltage/distance) becomes high

enough to cause sparking. Such sparking is usually at the
Flow Control. The rapping model suggests that the gas ve-
same spot and may cause noticeable electrode erosion. Until locity
near the collecting plate is critical for controlling rap-
the misalignment is corrected, the voltage can never be raised ping
emissions. Innovative suggestions have been made that
to its normal value. the
normal uniform gas velocity distribution could be made
Insulators perform the unobserved, but necessary function
beneficially nonuniform, with rapping emissions the target of
of supporting the discharge electrodes without conducting the
change. Is this feasible? Plate stiffeners designed to shield
electricity. When insulators become contaminated, repeated the
dust cake from the gas flow have been made. Do they
sparking along an insulator can erode the surface or carbon- perform
as designed?
ize materials on the surface to the point that cleaning is in-

sufficient to restore proper operation. Condensation of mois-
Conditioning Agents
ture or acid on the insulators during a startup is another Gas

conditioning agents have been long used to improve the
source of insulator contamination. Heated air purges are used
operation of precipitators. Water is one of the simplest condi-
to prevent such condensation. tioning
agents. Addition of water droplets to a hot gas cools
and
humidifies the gas. The lower gas volume increases treat-
ment
time in the precipitator, while the humidity reduces the
ADVANCED TOPICS IN PRECIPITATION dust
resistivity for some particles. Sulfuric acid (or SO3) is an
agent
designed to modify the resistivity of the particles for
Electrostatic precipitation is a mature technology, much like back
corona control. Ammonia is an agent that has been used
the automobile. Although we do not expect to find major tech- to
suppress back corona and improve the cohesivity of the
nological advances in precipitators, the variety of applications dust
cake; its back corona action is not certain but may result
and pressure of external regulations will bring about slow, from
the formation of a fine particle mode that increases the
continuous improvements in precipitator operations. Im- space
charge. Sodium conditioning to improve the resistivity
provements are often measured in terms of capital and op- of dust
cakes is another type of conditioning.
erating cost reductions or increased reliability because precip-
Although conditioning agents target specific precipitator
itator technology seems capable of controlling particle
problems, they do not always perform as intended. Some fly-
emissions to any desired level, given a large enough machine. ash
compositions are not responsive to sulfuric acid condition-
Some of the areas that could prove fruitful in these directions ing,
for example, or the precipitator operating temperature
are discussed below. may be
inappropriate. Understanding the methods by which
each
agent improves precipitator operation would guide the
Rapping Loss Reduction best
use of the agent and could lead to development of new
conditioning agents for both the standard problems and new

Reduction of rapping losses has potential for improving the
problem
areas.
collection performance of existing precipitators and reducing
the size of new precipitators. Moreover, the particles in the
Back
Corona Identification and Characterization
emitted dust from rapping have a substantial portion below
10 #m in diameter, a region on which future environmental
Although many cases of back corona are obvious from V–I
regulations may focus. Elimination of rapping emissions curves
and the poor performance of the precipitator, back co-
would improve conformance to such regulations. A better un- rona
may occur marginally in portions of the precipitator
derstanding of rapping emissions would improve chances for without
being detected. Precipitator performance may suffer
control in existing precipitators. The following areas are sug- as a
result. If a better means of detecting back corona were
gested points of investigation.
available, such problems could be identified and corrected. Al-
----------------------- Page 131-----------------------
ternatively, if a conditioning agent is being added to the flue

tor. These particles may be present at the outlet at higher
gas for resistivity control, back corona detection could be used
concentrations than at the inlet. Such growth has been ob-
to optimize the amount of agent introduced. With present
served when, for example, sulfuric acid gas condenses into a
methods of operation, the injection of conditioning agents is
particle phase as the temperature is lowered. The precipitator
usually made to control the resistivity under worst-case con-
is hindered in collecting these particles because they pass
ditions. Small changes in flyash composition may shift the
through most of the machine as a gas. Understanding the con-
resistivity enough that conditioning might not even be re-
centrations of materials and the temperature profiles that
quired.
lead to this phenomenon will be most important in preventing
Ash resistivity can be very sensitive to temperature; tem-
or controlling the emission of such particles, especially since
perature differences across the face of the precipitator have
the most toxic metals often condense preferentially on such
been known to put part of the precipitator into back corona,
small particles.
while the remainder works quite well. Direct detection of back
corona would be useful in finding the causes that contribute
Adaptive Computer Controls
to it.
Despite the obvious degradation that back corona causes,
Large precipitators often run at full power even though the
it has proven difficult to make quantitative predictions of its
boiler may only be at partial output. This practice wastes
severity. Based on the performance of precipitators with mea-
power and may even raise emissions because the temperature
sured dust resistivities, the trends of performance can be esti-
and resistivity of the precipitator and ash are controlled by
mated but not predicted. One technique that has been used is
the amount of gas passing through. The computer models of
to correlate performance with a ‘‘useful’’ current density, the
precipitators are accurate enough to predict many of these
current density available to charge particles before back co-
effects in real time and could be put to use optimizing precipi-
rona sets in. This technique approximates the degradation of
tator operation for changing load conditions (see PREDICTIVE
performance but does not provide insight into the actual
CONTROL). Using real time inputs of boiler and precipitator
mechanisms of degradation.
parameters, such computer control could permit power reduc-
Based on the physics of gas discharges, back corona onset
tions under reduced load conditions, while keeping the emis-
should be affected by the thickness of the dust layer. This
sions low. An approach such as this might be used to offset
means that the removal of the dust by rapping should change
increased precipitator size (for better performance at full
the back corona characteristics. This effect has not been stud-
load) with savings in operating costs.
ied, but it might well alter the rapping strategy in a precipita-

tor with back corona.
BIBLIOGRAPHY
Improved Corona Electrodes

1. S. A. Self and M. Mitchener, Basic Studies to Reduce Electrostatic
Most corona electrodes operate similarly above corona onset;

Precipitator Size and Cost, EPRI Report CS-3226, Palo Alto: Elec-
the nature of the corona almost guarantees it. The design of
tric Power Research Institute, 1983.
a corona electrode can control the corona onset voltage, how-
2. H. J. White, Industrial Electrostatic Precipitation, Reading, MA:
ever, and limit the corona to specific zones along the length
Addison-Wesley, 1963.
of the electrode. These controllable properties may be put to
3. J. H. Turner, P. A. Lawless, T. Yamamoto, D. W. Coy, G. P.
use in cases where high concentrations of particles produce
Greiner, J. D. McKenna, and W. M. Vatavuk, Electrostatic pre-
space charge problems or put heavy loads on the collection
cipitators, In A. J. Buonicore and W. T. Davis (eds.), Air Pollution
Engineering Manual, New York: Van Nostrand Reinhold, 89–

plates. With new electrode designs, it should be possible to
113, 1992.
distribute the charging and collecting of particles over more
of the precipitator sections, reducing the impact of the high
4. S. Matts and P. Ohnfeldt, Efficient gas cleaning with SF electro-
static precipitators, Flakt, 1–12. This was a private company
¨
concentrations.
publication. See White, p. 62, for discussion.
5. P. A. Lawless and L. E. Sparks, A review of mathematical models

Improved Collection of Submicron Particles
for ESPs and comparison of their successes. In S. Masuda (ed.),
Precipitators have a minimum in collection performance be-
Proceedings: Second International Conference on Electrostatic Pre-
low 1 #m diameter, at the point where the particle charge is
cipitation, Kyoto, Pittsburgh: Air Pollution Control Assoc., 513–
522, 1984.
decreasing, and the Cunningham slip factor is still close to
1. This size range, however, contains particles that are most
6. H. E. Rose and A. J. Wood, An Introduction to Electrostatic Precip-
itation in Theory and Practice, London: Constable and Com-

respirable and are most subject to future environmental regu-
pany, 1966.
lation. Improvement of particle collection in this size range
will require improvements to particle charging and mainte-
7. J. Katz, The Art of Electrostatic Precipitation, Munhall, PA: Pre-
cipitator Technology, Inc., 1979.

nance of high electric fields. Losses from sneakage and rap-
ping reentrainment are relatively insignificant for these par-
8. J. B. Loeb, Electrical Coronas—their Basic Physical Mechanisms,
Berkeley and Los Angeles: University of CA Press, 1965.

ticles.
9. S. Oglesby Jr. and G. B. Nichols, Electrostatic Precipitation, New

York: Marcell Dekker, 1978.
Particle Formation in Precipitators and other Control Devices
10. F. W. Peek, Dielectric Phenomena in High-Voltage Engineering,

Experimental measurements in operating precipitators below
New York: McGraw-Hill, 1929.
1 #m diameter have shown that particles condense from the
11. P. A. Lawless, K. J. McLean, L. E. Sparks, and G. H. Ramsey,
gas phase and grow to measurable sizes within the precipita-
Negative corona in wire-plate electrostatic precipitators. Part I:
----------------------- Page 132-----------------------
ELECTROSTATIC PROCESSES 15
characteristics of individual tuft-corona discharges. J. Electrostat-

Others
ics 18, 199–217, 1986.
Industrial Gas Cleaning Institute, Inc. Terminology for Electrostatic

12. K. J. McLean, P. A. Lawless, L. E. Sparks, and G. H. Ramsey,
Precipitators, EP-1, 1984.
Negative corona in wire-plate electrostatic precipitators. Part II:
calculation of electrical characteristics of contaminated discharge
PHIL A. LAWLESS
electrodes, J. Electrostatics, 18: 219–231, 1986.
Research Triangle Institute

13. R. S. Sigmond, Simple approximate treatment of unipolar space-
RALPH F. ALTMAN
charge-dominated coronas: the Warburg law and the saturation
Electric Power Research Institute
current, J. Applied Physics, 53: 891–898, 1982.
14. C. T. Phelps and R. F. Griffiths, Dependence of positive streamer

propagation on air pressure and water vapor content, J. Appl.
Physics, 47: 2929–2934, 1976.
15. P. A. Lawless and L. E. Sparks, Measurement of ion mobilities

in air and sulfur dioxide-air mixtures as a function of tempera-
ture, Atm. Env., 14: 481–483, 1980.
16. P. A. Lawless, Unpublished data, 1993.
17. P. A. Lawless and T. Yamamoto, Improving ESP performance by

reducing losses, In M. Rea (ed.), Proceedings: Third International
Conference on Electrostatic Precipitation, Abano-Padova, 433–
442, 1987.
18. P. C. Reist, Introduction to Aerosol Science, New York: Macmil-

lan, 1984.
19. R. A. Fjeld and A. R. McFarland, Evaluation of select approxima-

tions for calculating particle charging rates in the continuum re-
gime, Aerosol Sci. Tech., 10: 535–549, 1988.
20. J. D. Klett, Ion transport to cloud droplets by diffusion and con-

duction, and the resulting droplet charge distribution, J. Atmos.
Sci., 28: 78–85, 1971.
21. P. A. Lawless, Particle charging bounds, symmetry relations, and

an analytic charging rate model for the continuum regime, J.
Aerosol Sci., 2: 191–215, 1996.
22. D. K. Hutchins, M. H. Harper, and R. L. Felder, Slip correction

measurements for solid spherical particles by modulated dynamic
light scattering, J. Aerosol Sci., 22: 202–218, 1996.
23. P. A. Lawless, Modeling of electrostatic precipitator charging, col-

lection, and rapping reentrainment, Poster Session: 10th particu-
late Control Symposium and 5th International Conference on
Electrostatic Precipitation, Washington, 1983.
24. S. A. Self, G. B. Moslehi, M. Mitchener, and R. Leach, Electro-

mechanics and reentrainment of precipitated ash. In Harry
J.
White (ed.) Proc. Int. Conf. on Electrostatic Precipitation, 399–
440, 1981.
Reading List
The classical reference for precipitators is the out-of-print book by

Harry White (2). This book has been reprinted by the Interna-
tional Society on Electrostatic Precipitation and can be obtained
from: Dr. Robert Crynack, ISESP c/o Wheelabrator APC, 441
Smithfield Street, Pittsburgh, PA 15222-2292.
The next best general reference is the book by Oglesby and Nichols
(9).
A promising new book on the subject is: K. R. Parker, ed., Applied

Electrostatic Precipitation, London: Blackie Academic and Profes-
sional, 1997.
A comprehensive operating/maintenance discussion of precipitators

is the book by Katz, (7).
For a Detailed Look at Corona Processes:
L. B. Loeb, Basic Processes of Gaseous Electronics, Berkeley and Los

Angeles: University of CA Press, 1961, or (8).
----------------------- Page 133-----------------------
} { { } }
●
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Electrothermal Launchers

Standard Article

Mohamed A. Bourham1 and John G. Gilligan 1
1
North Carolina State University, Raleigh, NC

reserved.
Search
DOI: 10.1002/047134608X.W3016
❍ Acronym Finder
Abstract
Electrothermal Plasmas
Electrothermal Devices
Power Supplies
Electrothermal Plasma Modeling
Plasma-Material Interaction
Keywords: arc-driven high-energy plasma; electrical characteristics; plasma

discharge; ablation-controlled arcs; power dissipation;
electrothermal-chemical propulsion; plasma flow modeling; plasma-material
interaction; vapor shield plasma

%20Conversion/W3016.htm17.06.2008 16:37:43
----------------------- Page 134-----------------------
40 ELECTROTHERMAL LAUNCHERS
ELECTROTHERMAL LAUNCHERS
ELECTROTHERMAL PLASMAS
An electrothermal (ET) launcher uses electrically heated gas An

electrothermal plasma is a form of an ionized gas similar
to accelerate projectiles to high velocities. The gas is heated to
that in the sun and stars, except it is characterized by its
to very high temperatures and pressures in a short time to
high density and low plasma temperature. An ET plasma is
provide the accelerating force at, or near, the base of the mov-
generated in a confined volume, for example, a capillary, by
ing projectile. The launcher has a breech region which con- an
ablation mechanism that utilizes electrical energy dissipa-
tains the electrical heating mechanism, and a barrel which
tion from an internal arc. Arc-driven plasmas can be gener-
contains the projectile and into which the high pressure gas
ated over a wide range of pressures, from vacuum conditions
expands. Thus the ET launcher can be thought of as a gun to
atmospheric to high pressure discharges. An arc plasma is
which uses electrical energy instead of chemical propellant to
characterized by its high current density. Because plasmas
create the high gas pressure to accelerate the projectile. The
conduct electric currents, the energy dissipation is similar to
primary advantage of the ET launcher is a higher projectile
that of a simple resistor when a current is passing through
velocity for a given energy input. This is accomplished by con- the
resistor known as ohmic or joule dissipation. Such arc-
trolling the timed profile of electrical power to make the
driven plasma may be used to launch projectiles because of
launch sequence more efficient. A disadvantage of ET launch- the
high pressure gradient developed inside the confined vol-
ers is the large amount of electrical energy storage and
ume. The ET plasma is characterized by its temperature,
switching required for a large gun system. In more advanced
pressure, flow velocity, density, and other important plasma
designs electrothermal sources and chemical propellants are
parameters. An ET plasma has a high-density (1024 to 1027/
3
combined to yield efficient launchers that use a minimum m )

and temperatures of 1 eV to 5 eV (1.0 electron volt, eV, is
amount of electrical energy.
equivalent to 11,600 K) (1–4). Figure 1 shows a schematic
Light atomic weight gases (hydrogen, helium, etc.) are at-
drawing of an arc-driven electrothermal plasma source that
tractive as a projectile propellant because of a low specific
works on the theory of ablation-controlled arcs, where an elec-
heat and higher sound speed for a given gas temperature. Me-
tric arc is initiated between the cathode and the anode when
chanical pistons have been used to generate the high pres- the
switch is closed. The arc extends inside the confined vol-
sures for acceleration in light-gas guns. Ohmic heating of the
ume, which has an ablating wall ablator such that the ablated
gas to high temperatures with or without the pistons has
material (usually a plastic) is vaporized then ionized because
been used to enhance the accelerating pressure since the of
the heat generated from the arc. The plasma travels inside
1950s. In the late 1970s it was found that higher pressure of
the source and is allowed to expand at the source exit. Such
gases could be formed directly with a relatively simple ohmi-
arc-driven plasma source is known as ‘‘ablation-controlled arc
cally heated device called an ablation-controlled arc (1,2).
(ACA) source,’’ with typical dimensions of 4 mm to 50 mm
This device has become the basis for ET launcher work to this
bore diameter and 8 cm to 15 cm channel length. When at-
day. The advantage of an ablation-controlled arc is that no
taching an expansion tube to the source that acts as a
external gas feed is required to form the high temperature
launcher barrel, the plasma continues to travel through the
gas as will be explained later. Temperatures of the arc are
barrel and accelerates the mass payload until it leaves the
barrel.
high enough to not only dissociate molecules into individual
atoms but to excite and ionize the gas atoms themselves to
form a high pressure plasma. The plasma turns out to be in
ELECTROTHERMAL DEVICES
a state of thermodynamic equilibrium (among plasma species)
and ohmically heated by an electrical circuit and hence the
Figure 2 shows a simplified schematic of an ET launcher,
name electrothermal plasma. Unfortunately, ablating plasmas
where the plasma source is attached to an expansion barrel
contain significant quantities of higher atomic weight species
that contains the mass to be accelerated. Examples of the
such as carbon and oxygen.
kind of projectiles accelerated range from very small to very
Anode
Ablating
wall material
Electrothermal
Cathode
plasma jet
Arc
Ablated
mass
Figure 1. Schematic drawing of an arc-

driven electrothermal plasma source
based on the theory of ablation-controlled
arcs. The arc ablates the liner material,
which immediately dissociates, then ion- Power supply
Switch
izes and forms the plasma.

----------------------- Page 135-----------------------
ELECTROTHERMAL LAUNCHERS 41
Electrothermal source
Acceleration barrel
Payload
Figure 2.
Simplified schematic of an
Switch electrothermal
launcher showing an elec-
trothermal
plasma source attached to an
Electric power
expansion
barrel that contains the mass
supply
to be
accelerated.
large. Frozen hydrogen pellets of only a few milligrams can the

electrothermal-chemical launcher concept, where the ba-
be launched with an ET injector to fuel a fusion reactor (3). sic
components are an electrothermal plasma injector, a com-
Electrothermal plasma injected into the breech of an electro- bustion chamber
that contains the propellant, and the barrel
magnetic launcher railgun, forms an armature to accelerate that contains
the projectile (payload).
projectiles of tens of grams.
In another launcher concept, an electrothermal plasma is
POWER SUPPLIES
injected into a chamber that contains a propellant, solid or
liquid, where the plasma ignites and controls the burn of the
The electrothermal plasma
injector is powered via an external
propellant (4,5). Electrochemical-chemical (ETC) launch con-
electric power supply, as
previously shown in Fig. 1. This
cepts provide several advantages among other electrothermal
pulse power system has a
pulse forming network (PFN; see
plasma launchers, especially for large payloads of several ki-
PULSE SHAPING CIRCUITS)
composed of energy storage capaci-
lograms. It provides a considerably lower gas temperature in
tors and coupling inductors. A
simple pulse power system,
the barrel, and thus reduces bore erosion and thermal load-
with pulse forming network is
shown in Fig. 4, where capaci-
ing. Additionally, it requires less electrical energy from the
tor modules are coupled to
each other via inductors to form
external pulse power system because the energy is only used
the desired pulse length of
the discharge current. As shown
to produce the plasma in the electrothermal injector, which
in the figure, a coupling
inductor L is connected between each
then ignites and controls the burn rates of the chemical pro-
two capacitors C, and a
charging resistor R is connected to
pellant. Such an electrothermal-chemical launch concept is al-
the charging high voltage power
supply. Coupling to the load
ready providing the near-term, large scale application of elec-
(the plasma source) is
provided via a high-voltage high-cur-
trothermal launch technology. Figure 3 shows a schematic of
# Discharge current
# # # #
# # and voltage measuring
# sensors
# ## ## #
# # # #
## ## ## #
## ## ## #
# #
# #
## ### ## #
Electrothermal
# #
# # # #
plasma source
# # #
# # # #
# # ##
# # #
Trigger #
# ## # # pulse
# #
# # # #
## ## # #
### #
### ## # #
# #
### #
### ## #
# ## # #
#
Spark-gap #
switch
#
High voltage
# ## # #
#
# ## # #
# #
power supply
# # #
## # # #
Isolation L# L
L R
transformer
## # # #
_
## ## ### #
#
# # #
#
#
### ## ### #
#
# ## # #
#
#
ET source # C C C
C
# ## # #
# #
Plasma jet Propellant Projectile Barrel
#
#
Figure 3. Schematic of the electrothermal-chemical launcher concept Figure 4. A
simple pulse power system with pulse forming network
showing basic process of plasma injection, propellant’s burn and com- (PFN).
Capacitor modules are coupled to each other via inductors to
bustion, and motion of the projectile down the barrel. Basic compo-
form the desired pulse length of the discharge current. Coupling in-
nents are an electrothermal plasma injector, a combustion chamber
ductors are connected between the capacitors, and a charging resistor
that contains a propellant, and a barrel that contains a projectile. is
connected to the charging high voltage power supply.
----------------------- Page 136-----------------------
C = 315 Fµ
will be shown in a following section. The temperature, in real-
(single capacitor, no PFN)
ity, will increase and decrease following closely the time his-
tory of the discharge current. However, maintaining a near
flat-top discharge current would also maintain a near con-
stant plasma temperature.

C = 1933 Fµ
(6 capacitors with PFN)
Plasma initiation in electrothermal sources may be
achieved via exploding a fuse inside of the capillary (required

Discharge at
initial atmospheric conditions). This is a desired opera-
current
tional regime since an electrothermal launcher is not ex-
pected to operate under vacuum. In vacuum, with a back-fill-
ing gas, arc initiation is achievable depending on the
breakdown conditions as is the case in most glow discharges
and vacuum arcs. When operation at atmospheric pressure is

Discharge time
desired, the breakdown voltage would be extremely high; thus
Figure 5. Illustration of how a pulse forming network shapes the an

exploding fuse would be necessary to achieve breakdown
discharge current. Two typical time histories of the discharge current at
considerably low voltages. Once the fuse is exploded, it va-
are shown, one for a single capacitor of 315 #F without PFN, and one
porizes and forms a vapor plasma that is immediately ionized.
for a PFN composed of six capacitors of 1933 #F total capacitance.
Energy dissipation in the form of ohmic heating continues
Scales are arbitrary.
during the current discharge cycle, and the plasma deposits
energy on the inner wall of the ablating surface; thus continu-

ous ablation takes place.
rent switching device. Because the pulse power systems are
Current large caliber test fixture ETC devices utilize fuse
designed to deliver high discharge currents, at high charging
operation of the plasma source. The source may also be de-
voltages, special switching is necessary to close the circuit be-
signed in various configurations. For example thermo-energy
tween the energy storage system and the load (the electro-
cartridges may be used as plasma igniters for the propellants.
thermal plasma injector). Switching may be provided via a
The plasma source for ETC launchers may be a simple car-
simple spark-gap switch that is triggered by a high voltage
tridge attached to the combustion chamber to inject the
pulse, or other electronic switching devices such as ignitrons.
plasma into the chemical propellant as previously shown in
A high voltage trigger pulse initiates the spark-gap switch,
Fig. 3, or be designed in such a way that allows the plasma
and an isolation pulse transformer is used to isolate the trig- to
flow through the propellant via distributed thin channels
ger generator from the electric circuit of the spark-gap and
(known as piccolo configuration). A piccolo configuration
the source. Also shown in the figure are two essential measur-
would allow for a better plasma mixing with the propellant,
ing tools, a potential divider to measure the discharge voltage
better energy deposition into the propellant, and a better uni-
across the source Vd, and a current transformer known as a
form burn (4). The main features of electrothermal plasmas
Rogowskii coil to measure the discharge current flowing into in
ETC launchers are to provide augmentation and controlla-
the source Id . Currents of several thousands of amperes could
bility of burn rates of the propellant. Propellants are typically
be generated with pulse lengths from as short as a few micro-
nitrogen-based compounds not unlike current large gun pro-
seconds (#s) to as long as several milliseconds (ms). Figure 5
pellants. Conventional 127 mm (5 in.) bore guns have been
illustrates how a pulse forming network shapes the discharge
current. Shown in the figure are two typical time histories of
the discharge current, one for a single capacitor of 315 #F
without PFN where the pulse length is short and has a nar-
row peak, and one for a PFN (typical to that of Fig. 4) com-
Ideal interior ballistics Power
posed of six capacitors of 1933 #F total capacitance where the
profiles for an electrothermal
launch device
pulse length is longer and has a wider peak.
It is best to obtain current pulses with a flat top over a
longer period of time to provide a similar plasma pressure
time history into the source. The reason for flat-top pressures
Temperature
Pressure
is to achieve near-ideal electrothermal interior ballistics pro-
files. Ideal electrothermal interior ballistics profiles would
have a flat-top plasma pressure, an increasing velocity during
Velocity
the discharge cycle across the source, a slowly increasing tem-
perature (but kept as low as possible), and an increased elec-
tric power delivery to the source. The idealistic profiles are
shown in Fig. 6, where the pressure is maintained constant
over the discharge cycle, and the electric power is increasing
Discharge time
until the end of the flat-top pressure then decreases rapidly
Figure 6. Ideal interior ballistics profiles for electrothermal launch

at the end of the discharge. Neither the pressure nor the tem-
devices. Ideally, the pressure should be maintained constant during
perature will be ideal because the plasma temperature is a
the discharge cycle, discharge voltage should be near-linearly increas-
function of plasma resistivity, which is governed by collisional
ing, temperature slowly increasing, and the power increasing then
processes (electron–ion and electron-neutral collisions), as
decreasing at the end of the cycle. Scales are arbitrary.
----------------------- Page 137-----------------------
modified to use ETC charges. Electrothermal energies ap-

assumed to be constant across the cross section of the capil-
proaching a megajoule have been shown to significantly in-
lary and the barrel. The ablated material in the source is as-
crease the muzzle velocities of projectiles as compared to pure
sumed to be totally dissociated into the constituent atoms.
chemical propellants.
The heat loss due to conduction inside both the source and the
barrel is assumed to be negligible. Also, the axial radiation
transport is assumed to be negligible inside the source where

ELECTROTHERMAL PLASMA MODELING
the plasma temperature is fairly isothermal. Additional as-
sumptions will be introduced, whenever necessary, through-

Electrothermal plasmas have various applications in launch
out the description of the set of equations.
technology. An electrothermal plasma source may be used as
a launcher by itself, or as a pre-injector to form a plasma
Conservation of Mass
armature in railguns. In plasma-chemical launchers, the
The rate of change in the particle density in each cell is the

source injects the plasma into a propellant to ignite and con-
difference between the rate at which particles are introduced

trol the burn rate and combustion of the propellant. At such
into the cell from ablation of the wall and the rate at which
plasma temperatures and densities, where plasma radiates
particles enter and leave the cell. The equation of continuity

energy like a blackbody, an electrothermal plasma can also
for each cell is given in (14) by:

be used as a high heat flux source for materials evaluation
when surfaces are exposed to thermal shocks, as described by
∂n ∂(vn)
Bourham and Gilligan (6,7). A system of equations describing
= n˙a − (1)
∂t ∂z
the physics of electrothermal plasmas is illustrative. These
equations may be written in a global fashion to calculate the
where n is the number density of plasma particles (atoms/
time and spatial-averaged plasma parameters. A global, time-
m3), v is the plasma velocity (m/s), and is the time rate of
dependent set of equations would help in evaluating the time
change of the number density of ablated material from the
variation of the plasma parameters, as shown by many re-
cell wall (atoms/m3 s) and is given in (15) by:
searchers (see 8–10). However, a one-dimensional, time de-
pendent description yields a more accurate description that
2q##
shows the time and spatial variation of the plasma parame-
n˙a = H A R (2)
sub p
ters inside the plasma generator, and the plasma flow and
acceleration mechanism of the payload inside the launcher’s
where q# is the radiation heat flux incident on the wall sur-

barrel (11,12). The basic equations are the conservation of
face (W/m2), A p is the mass of the atoms that constitute the
mass, momentum, and energy. In a simple description, the
plasma (kg/atom), R is the radius of the cell (meters), and

set of equations describes the balance of masses, momentum
Hsub is the sum of the energy of dissociation of the molecules
and energy from the initiation of the plasma inside of the
to
the constituent atoms (the heat of sublimation and vapor-
plasma source and through its travel into the barrel until
ization). The radiation heat flux incident on the wall surface
leaving the muzzle. These balances must also include the pro-
is
a fraction of the blackbody radiation emitted from the hot
jectile motion and acceleration through the barrel. The one-
plasma core, and is given by:

dimensional, time-dependent description is more appropriate
because of the nature of plasma initiation in the injector and
q## = f σ T4 (3)
s
then its travel along the axial direction into the barrel. A de-
scription of the set of equations for an electrothermal plasma
where f is the energy transmission factor through the vapor
injector that is operating on the principles of ablation-con-
shield, #s is the Stefan–Boltzmann constant (5.670 # 10#8
trolled arcs is given below for a simple capillary discharge
2 4
W/m K ), and T is the plasma temperature (K). The energy

attached to a barrel that contains a payload. When the arc is
transmission factor, f , is a function of the heat of sublimation
initiated inside of the capillary, as previously described in
Hsub, plasma internal energy, plasma pressure and density
Fig. 1, the arc heats the walls of the capillary, ablates materi-
(14,15).
als from the wall, and forms the plasma. One can divide the
source and barrel sections into a specific number of cells and
Conservation of Momentum
look to the plasma as a viscous fluid. Some simplifying as-
sumptions are considered. Each cell is considered to be at lo-
The change in velocity in each cell is due to the pressure
cal thermodynamic equilibrium because the plasma has a
forces, the kinetic energy of particles entering and leaving the
high density and considerably low temperature, and thus it is
cell, and the ablation and viscous drags. The equation for the
highly collisional. The fluid equations are nonlinear due to
time rate of change of the velocity in each cell is given in
the ionization, radiation, and drag effects. The specific inter-
(14,15) by:
nal energy of the plasma is a function of the temperature as-
suming that the ablated material is completely dissociated. A
∂v = − 1 ∂P − 1 ∂v2 − vn˙a − 2τw (4)
mechanism known as the vapor shield will also be considered.

∂t ρ ∂z 2 ∂z n ρR
This mechanism provides a self-protecting nature to the ab-
lating wall because the evolved vapor cloud absorbs a fraction
where 2
w is the viscous drag at the wall (N/m ). In a simplistic

of the arc energy such that the net energy reaching the wall
steady-state way, the momentum equation can be expressed
will be reduced (13). This mechanism will be described in by
the first term on the right-hand side equal only to the first
more detail in the plasma-materials interaction section. Be-
term on the left hand side, means that the rate of change of
cause of the one dimensionality, the plasma parameters are
velocity is equal to the change in velocity due to the axial
----------------------- Page 138-----------------------
pressure gradient, with opposite sign. But due to axial depen-

800
dence, the momentum equation must include additional
terms as appearing on the right hand side of Eq. (4). The sec-
ond term is the change in velocity due to the kinetic energy
) 600
s
gradient. The third term is the velocity loss due to the in-
/
(
crease in the number density from ablated material (abla-
y
t
i
tion drag).
c 400
l
When setting the momentum equation for the barrel, one
e
t
has to include an additional term on the right hand side of
i
x
Eq. (4) to account for losses due to momentum transfer to
E 200
the payload (projectile). This additional term has to be added
Code results
Experiment
negatively to the right-hand side of the equation and is given
in (12) by:
0
0 2 4 6 8
ρproj ∂vproj
Input energy (kJ)
(5)
ρ ∂t Figure
7. A comparison between experimental and computer code
results for the exit velocity of a half gram projectile accelerated in a

15 cm
barrel using an electrothermal launcher.
Conservation of Energy
The rate of change of the internal energy in each cell in the

plasma source section is due to joule heating, radiation, flow charge
state and the ionization potential, and is given in de-
work, changes in density, internal energy entering or leaving tail
in (16). The plasma resistivity, which appears in the joule
the cell due to particle transport, and frictional heating. The
heating term must be a summation of two resistivities due to
time rate of change of internal energy in each cell in the
electron-neutral and electron-ion collisions
#
en #
ei,
source is given in (14,15) by: where
the resistivity due to electron-neutral collisions is
given
in detail by Cambel in (17). The resistivity due to elec-
∂U 2 2q## ∂v 1 2 ∂(nU ) tron-
ion collisions could be determined using the Spitzer re-
n = ηj − − P + ρ˙ v − n˙ U − v (6)
∂t R ∂z 2 a a ∂z
sistivity model, as given by Spitzer and Harm in (18).
Modifications to the Spitzer resistivity for high-density,

where
is the plasma resistivity (-m), and j is the discharge low-
temperature plasmas are better introduced because
current density (A/m2). The first term on the right-hand side such
electrothermal plasmas tend to be weakly nonideal
is the increase in internal energy due to joule heating. The such

that the ideal resistivity model will no longer be valid,
second term is the loss in internal energy due to thermal radi- as
shown by various researchers in (19,20). Also, the
ation and the (2/R ) factor is due to the conversion of surface plasma
viscosity has to be the summation of two viscosity
heat flux to volume radiation. The third term is the change terms,
the viscosity due to the neutral atoms, and that due
in internal energy due to work done by the plasma (flow to
the ions. It is apparent that the given set of equations
work). The fourth term is the increase in internal energy due has
to be solved self-consistently. Although a time-averaged
to friction from ablation. The fifth term is the loss in internal and
spatially-averaged analytical solution may be obtained
energy due to the cold ablated material entering the plasma. with
additional simplifying assumptions, a complete self-
The sixth term is the change in internal energy due to parti-
consistent one-dimensional time-dependent solution has to
cles entering and leaving the cell. When setting the energy be
obtained numerically via appropriate computer codes.
equation for the barrel, one has to remove the joule heating Many
computer codes have been developed for electrother-
term, and include a term to the right hand side of Eq. (6) to mal
plasmas in launch devices to solve the set of governing
account for loss of internal energy which is transferred to the
equations self consistently and to help predict the plasma
energy of the payload (projectile). This additional term has to
parameters for a given discharge configuration and current
be added negatively to the right hand side of Eq. (6) and is
profile. An example of the computer code results for the
given in (12) by: exit
velocity of a half gram projectile accelerated in a 15
cm
barrel is shown in Fig. 7together with experimental
∂vproj
measurements. The velocity reaches 700 m/s for an input
ρprojvproj ∂t (7) energy
of 6 kJ to the electrothermal plasma source.
The energy equation includes several terms that need to be

PLASMA-MATERIAL INTERACTION
defined. The internal energy, for an ideal plasma, is given in
(15) by:
In
most electric launch devices (electrothermal, electrother-
mal-
chemical, and electromagnetic), the heat flux from the
U = 1.5kT (1 + Z) + I + Hsub (8) arc-
formed plasma may exceed 100 GW/m2 for a duration of
0.01
ms to 5 ms. As a result, critical components are damaged
where the first term on the right-hand side is the internal due to
surface erosion and thermal deformations. Surface ero-
energy due to thermal motion, and the second term is the sion
is one of the parameters that has an effect on the per-
internal energy due to ionization. The modified Saha–
formance, durability, efficiency, and lifetime of the launch
Boltzmann equation gives the relation between the effective
device. Minimum deformation and damage of the critical com-
----------------------- Page 139-----------------------
Source fluence (S) nant

mechanism by which energy is transferred to the ma-
Solid surface terial
surface (21,25). The plasma flow in such devices has
a Reynolds
number of 105 7
to 10 , and consequently the

viscous
skin friction generates turbulence.
Experiments on various materials have been conducted us-

ing
varieties of launch devices and simulators to explore the
# performance
and response of such materials to high heat
fluxes
produced from arc-driven plasmas. For example, pure
# copper and
gunsteel have approximately equal erosion, which
# is about
60% less than that of aluminum. Molybdenum is
even
better and has about 75% less erosion than aluminum,
# # while
tungsten has no obvious erosion below 20 GW/m2, but
# surface
coloration and micro cracks may occur. Because re-
# # fractory
materials have less surface erosion, better thermal
Surface heat flux (q)
## resistance,
and better structural strength, launcher compo-
## nents may
be coated with layers of selected refractory materi-
als.
Graphite and carbon materials have no melting tempera-
ture when
direct sublimation takes place.
Time The
energy transmission factor through the vapor shield is
the ratio
between the actual heat flux at the ablating surface
Net erosion to the
incident heat flux from the plasma source, which is
about 10%
for most graphite grades at incident heat fluxes of
60 GW/m2
and greater. Insulators are important materials for
Legend Melt layer

electrothermal and all electrically driven launch devices.
# # Many
polymer materials have
been considered as insulating
Ablative plasma shield components
in electrothermal launch devices, among these is
Lexan,
which is a polycarbonate polymer (C H O ). The va-
#
16 14 3
Energy transmission factor= f = q/S por shield

is more effective for highly ablating materials due
# to the
fast development of the vapor layer. Lexan, boron ni-
Figure 8. A time history illustration of the vapor shield mechanism.
The plasma energy is deposited on the surface and raises the temper- tride, and
aluminum oxide have approximately equal energy
ature. The surface melts then vaporizes. The incoming energy is de-
transmission factors, while silicon carbide composite is ap-
posited in the developed vapor layer and less energy reaches the sur- proximately
a factor of two higher (25). A comparison between
face. The heat flux that reaches the surface is a fraction of the source the energy
transmission factor for three insulators, Lexan
fluence. At the end of the cycle, the melt layer re-solidifies leaving a
polycarbonate, boron nitride, and silicon carbide, is illus-
final net surface erosion. Time scale is arbitrary. trated in
Fig. 9 showing a decreased factor for increased heat
ponents are required in order to achieve efficient operation of

the launcher, especially at high repetition rates of operation. 70
High thermal resistance materials may help in reducing sur-

face erosion of the launcher components. A possible approach
is to use refractory materials or refractory coatings on materi- ) 60
%
als to reduce surface erosion of rails and barrels; also high (
r
tensile insulating materials and specially prepared compos- o
Silicon carbide
t 50
c
ites may also be considered to eliminate the ablation of the a
f
insulators in railguns and electrothermal plasma injection n

o
i 40
sources (21–25). s
s
i
Under such short and intense high heat flux conditions m
s
the material surface suffers melting and vaporization, and n 30
a
r
Lexan
t
a plasma boundary layer will be formed at the ablating y
g
surface. Such vapor plasma absorbs a fraction of the incom- r
e 20
ing energy; thus less heat flux reaches the surface resulting n
Boron nitride
E
in less surface erosion. This is described as the vapor shield
mechanism, and is schematically illustrated in Fig. 8. The 10
energy absorbed in the vapor layer appears as internal

energy that can be transported away from the localized
area due to the large pressure (#1 kbar; 1 kbar is 1000 0
0 10 20 30 40 50
times atmospheric pressure), which is large enough to ex-
2
Source heat flux (GW/m )

pand against an incoming plasma flux (26). The heat flux
is primarily from blackbody spectrum photons, as previously Figure 9.
A comparison between the energy transmission factor for
given by Eq. (3). Once the energy is deposited in the vapor three
insulators, Lexan polycarbonate, boron nitride, and silicon car-
layer then low energy photon transport becomes the domi- bide,
showing a decreased factor for increased heat flux.
----------------------- Page 140-----------------------
Table 1. Erosion Thickness of Various Materials Exposed to High Energy

Electrothermal Plasmas
Input Energy to Electrothermal Plasma 1 2

3 4 5 6 7
Source (kJ) ##
Heat Flux from Electrothermal Plasma 2.8 8.6
15.6 23.6 32.5 43.8 59.4
Source (GW/m2) ##
Material
Erosion Thickness (#m)
Aluminum 10.6 47.3

99.8 168 206
Titanium 7.1 49.6
89.8 114 153 215 260
Gun steel
57.7
Copper 2.6 12.7
24.1 36.4 52.8
Molybdenum (arc cast) 0.1 0.9
9.7 19.2 32.8
Molybdenum (sintered) 0.2 3.5
28.2 15.6 30.9
Glidcop (Cu- 0.15% Al O ) 1.8 9.8
19.9 22.5 28.3
2 3
Molybdenum on copper
18.8
Tantalum carbide on copper
4.3
Tantalum nitride on copper
3.8
Tantalum on copper
0.8
Tungsten #0 #0
#0 #0 #0
Tungsten-rhenium alloy (3% Re) #0 #0
#0 #0 #0 #0 #0
Lexan (polycarbonate, C H O ) 11.1 23.3
30.8 41 43.5
16 14 3
Boron nitride (grade A) 5.2 10.5

15.7 20.6 20.1
Glass-bonded mica 1.2 6.0
8.5
Annealed pyrographite (P-ANN-PG)
12.8
Molded dense electrographite (2020) #0 1.1
1.9 11.4 6.8
High density graphite (6222) #0 #0
#0 3.7 3.3 5.2 9.4
Highly anisotropic pyrographite 2.7 3.6
2.4
flux. This factor is about 10% for most materials at higher

9. S. Cuperman, D. Zoler, and J. Ashkenazy, Analysis of critical flow
heat fluxes. A summary of measured erosion thickness of var-
in a combined discharge capillary-ablative pipe system, Plasma
ious materials exposed to high energy electrothermal plasma
Sources Sci. Technol., 3: 593–601, 1994.
is given in Table 1, for heat fluxes up to 60 GW/m2 over a 100
10. J. D. Powell and A. E. Zielinski, Analysis of the plasma discharge
#s pulse length.
in an electrothermal gun. In A. A. Juhaz (ed.), Technology Efforts
in ET Gun Propulsion, US Army Ballistic Research Laboratory,
Arberdeen Proving Grounds, MD, Vol. II, 1989.
BIBLIOGRAPHY
11. J. D. Hurley, M. A. Bourham, and J. G. Gilligan, Numerical simu-
lation and experiment of plasma flow in the electrothermal
launcher SIRENS, IEEE Trans. Magn., 31: 616–621, 1995.

1. E. Z. Ibrahim, The ablation dominated polymethylmethacrylate
arc, J. Phys. D.: Appl. Phys., 13: 2045–2065, 1980.
12. R. W. Kincaid, M. A. Bourham, and J. G. Gilligan, Plasma gun
pellet acceleration modeling and experiment, Fusion Technol., 30:

2. C. B. Ruchti and L. Niemeyer, Ablation controlled arcs, IEEE
834–839, 1996.
Trans. Plasma Sci., 14: 423–434, 1986.
13. M. A. Bourham et al., Vapor shielding and erosion of surfaces

3. R. W. Kincaid, M. A. Bourham, and J. G. Gilligan, Electrothermal
exposed to high heat load in an electrothermal accelerator, IEEE

plasma gun as a pellet injector, Fusion Technol., 26: 637–641,
Trans. Plasma Sci., 17: 386–391, 1989.

1994.
14. J. D. Powell and A. E. Zielinski, Theory and experiment for an

4. W. F. Oberle, Technology efforts in ETC gun propulsion. Army
ablating-capillary discharge and application to electrothermal-

Research Laboratory, ARL-SR-12, 6, 1-309, 1994; ARL-SR-22, 7,
chemical guns, BRL Tech. Rep. BRL-TR-3355, 1–50, 1992.

1-191, 1995; Proc. JANNAF Combustion Meetings: CPIA 573, I,
165, 1991; CPIA 593, I, 299, 1992; CPIA 606, I, 17-25, 1993; CPIA
15. J. G. Gilligan and R. B. Mohanti, Time dependent numerical sim-
602, I, 17-21, 1994.
ulation of ablation controlled arcs, IEEE Trans. Plasma Sci., 18:
190–197, 1990.
5. D. Hewkin and E. Figura, Fundamental research and numerical
modeling of the internal ballistics of electrothermal chemical
16. Ya. B. Zeldovich and Yu P. Raizer, Physics of Shock Waves and
guns, IEEE Trans. Magn., 29: 561–566, 1993.
High Temperature Hydrodynamic Phenomena, Vol. 1, New York:
Academic Press, 1963.

6. M. Bourham et al., Electrothermal plasma source as a high heat
flux simulator for plasma-facing components and launch technol-
17. A. B. Cambel, Plasma Physics and Magneto-Fluidmechanics, New
ogy studies, Proc. 9th International Conference on High Power
York: McGraw-Hill, 1963.
Particle Beams, Washington, DC, May, 1992, III: 1979–1983,
18. L. Spitzer, Jr. and R. Harm, Transport phenomena in a com-
1992.
pletely ionized gas, Phys. Rev., 89: 977, 1953.
7. J. Gilligan and M. Bourham, The use of an electrothermal plasma

19. R. B. Mohanti and J. G. Gilligan, Electrical conductivity and
gun to simulate the extremely high heat flux conditions of a toka-
thermodynamic functions of weakly nonideal plasmas, J. Appl.
mak disruption, J. Fusion Energy, 12 (3): 311–316, 1993.
Phys., 68: 5044–5051, 1990.
8. E. Jacob, S. Bouquet, and B. Tortel, A global theoretical approach

20. J. Batteh et al., A methodology for computing thermodynamic
for the electrothermal gun: Scaling laws and a 0-D time-depen-
and transport properties of plasma mixtures in ETC injectors,
dent model, IEEE Trans. Magn., 31: 419–424, 1995.
IEEE Trans. Magn., 31: 388–393, 1995.
----------------------- Page 141-----------------------
ELLIPTIC EQUATIONS,
PARALLEL OVER SUCCESSIVE RELAXATION ALGORITHM 47
21. M. A. Bourham, J. G. Gilligan, and O. E. Hankins, Plasma-mate-

rial interaction in electrothermal and electromagnetic launchers,
AIAA 24th Plasmadynamics & Lasers Conference, Orlando, FL,
AIAA 93-3172, 1993.
22. F. D. Witherspoon, R. L. Burton, and S. A. Goldstein, Railgun

experiments with lexan insulators, IEEE Trans. Plasma Sci., 17:
353–359, 1989.
23. R. D. Stevenson, S. N. Rosenwasser, and R. M. Washburn, Devel-

opment of advanced ceramic matrix composite insulators for elec-
tromagnetic railguns, IEEE Trans. Magn., 27: 538, 1991.
24. A. E. Zielinski and C. V. Renaud, Erosion resistance of CuNb

microcomposites in a plasma armature electromagnetic launcher,
BRL Tech. Rep., BRL-TR-3311, 1–30, 1992.
25. M. A. Bourham et al., Review of components erosion in electric

launchers technology, IEEE Trans. Magn., 31: 678–683, 1995.
26. A. Hassanein, Erosion and redeposition of divertor and wall ma-

terials during abnormal events, Fusion Technol., 19: 1789, 1991.
Reading List
P. Aubouin, Electrothermal launcher plasma burner modeling and

comparison to experimental results, Proc. 4th European Sympo-
sium on Electromagnetic Launch Technology, Germany, 1993, pa-
per 1003.
E. Blums, Yu. A. Mikhailov, and R. Ozols, Heat and Mass Transfer in

MHD Flows, Singapore: World Scientific Publishing Co., 1987.
J. R. Greig et al., Investigation of plasma-augmented solid propellant

interior ballistic process, IEEE Trans. Magn., 29: 555–560, 1993.
A. Loeb and Z. Kaplan, A theoretical model for the physical processes
in the confined high pressure Discharges of electrothermal launch-
ers, IEEE Trans. Magn., 25: 342–346, 1989.
B. Schmit and Th. H. G. G. Weise, Performance and results of the

TZN electrothermal gun simulation code, Proc. 4th European Sym-
posium on Electromagnetic Launch Technology, Germany, 1993,
paper 1016.
E. Y. Scholnikov et al., High efficiency electrothermal accelerator,

IEEE Trans. Magn., 31: 447–451, 1995.
D. D. Schuresko et al., Development of a hydrogen electrothermal

accelerator for plasma fueling, J. Vac. Sci. Technol., A5 (4): 2194,
1987.
G. P. Wren et al., Spatial effects of an electrically generated plasma

on the interior ballistics of electrothermal-chemical (ETC) guns,
IEEE Trans. Magn., 31: 457–462, 1995.
MOHAMED A. BOURHAM
JOHN G. GILLIGAN
North Carolina State University
ELECTROVISCOUS FLUIDS. See ELECTRORHEOLOGY.
----------------------- Page 142-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Engineering Workstations
Standard Article

Xing Du 1 2 3
, Xiaodong Zhang , Yong Yan
1University of Virginia, Williamsburg, VA,
2College of William and Mary, Williamsburg, VA,

3College of William and Mary, Williamsburg, VA,
Search
reserved.
❍ Acronym Finder
DOI: 10.1002/047134608X.W1636
Abstract
Architecture
Software Environments
Applications
Current and Future Research and Development Topics

%20Conversion/W1636.htm17.06.2008 16:38:04
----------------------- Page 143-----------------------
114 ENGINEERING WORKSTATIONS
ENGINEERING
WORKSTATIONS
An
engineering workstation is a complete computer system
used for
engineering applications. A typical engineering com-
putation
requires a moderate amount of computing power (es-
pecially
in the form of floating point operations), relatively
high-
quality graphics capabilites, and a large main memory.

----------------------- Page 144-----------------------
ENGINEERING WORKSTATIONS 115
Workstations may be classified at an intermediate level be- a computer

called Sun for the Stanford University Network
tween powerful minicomputers and affordable personal com- in 1981.
Afterwards he formed a company named Sun Micro-
puters in terms of cost and performance. (However, personal systems,
Inc. in 1982, and shipped Sun-1 workstations in the
computers are catching up with workstations in terms of pro- same year.
Since then, workstations have become affordable
cessor power and memory size.) In addition, workstations are computers
for engineers and scientists.
characterized by the following properties:
Currently, the leading workstation manufacturers are Sun
Microsystems, Hewlett-Packard Company, Silicon Graphics

• High Performance. A modern workstation is equipped
Incorporated (SGI), and Digital Electronics Corporation.
with a single or multiple RISC-based or CISC-based pro- (Apollo
Computer was purchased by Hewlett-Packard in
cessors of 100 MHz or more, and a main memory of 64 1989.)
They provide a large variety of workstations in the
megabytes or larger. High-speed and high-resolution market
ranging from single processor desk-top workstations
graphics may be supported by a special processor sepa- to
multiprocessor (up to 64 processors) departmental or even
rately. enterprise
servers.
• Single User. Usually, one workstation is dedicated to one

user and not shared (in most cases) with other users. But
ARCHITECTURE
some powerful workstations are used as departmental
servers as well.
Similar
to most other types of computers, a workstation
• Multitasking. Complex applications require the system
mainly
consists of one or more processors, main memory,
to support multitasking. The user may use the worksta-
mass
storage devices (disks), and other I/O devices such as
tion to do several tasks concurrently, and make full use
mice and
keyboards. Workstations are also equipped with
of the power of the workstation.
network
adaptors and such multimedia devices as CD drives,
• Networking. Workstations are networked for easy access
microphones, and speakers. Initially, most workstations used
to remote computing resources. Workstations can be con- 32-bit
traditional CISC (complex instruction-set computer) (3)
nected to local area networks or even wide area networks processors
such as the Motorola 680x0s. However, at present,
such as the Internet. The commonly used communication 64-bit RISC
(reduced instruction-set computer) (3) based pro-
protocol is transmission control protocol/Internet proto- cessors
running at more than 100 MHz are widely used in
col (TCP/IP) (1). modern
workstations. For example, the RISC processors Ul-
traSPARC
1, MIPS R10000, and PowerPC are used in Sun
In terms of computing power and physical size, worksta- UltraSPARC
stations, SGI O2 machines, and IBM RS/6000
tions can be classified into three types: notebook, desk-top,
workstations, respectively.
and server. Notebook workstations are rare because of their Two
levels of cache (Level 1 and 2) usually help bridge the
low performance/cost rates in comparison with notebook per- gap between
processors and memory. Level 1 cache, which is
sonal computers (PC). From a hardware configuration stand- on the
processor chip, and is often called on-chip cache, is
point, a workstation may be constructed using a single pro- about 32
k bytes. Level 2 cache, sometimes called external
cessor, or using multiple processors ranging from two to up to cache, is
256 kbytes to 1 Mbytes in size. The size of main
a few dozen. A bus-based symmetric architecture is adopted memory
varies from 32 to 128 Mbytes.
for a small number of processors, but when the number of All
components of a workstation are connected via several
processors increases, more complex structures such as hierar- buses as
shown in Fig. 1. The processor and memory are con-
chical buses may be used to enhance the data access band- nected by a
system bus, and all I/O devices share a common
width. I/O bus.
Simplicity and low cost are the major merits of such
The most common operating system for workstations is a
structure, which has been adopted by most workstation
UNIX. It is a time-shared multitasking system supporting
manufacturers.
standard languages such as C. A graphical user interface However,
with the increase of processor computing power
(GUI) is provided to facilitate the use of these workstations. and high
bandwidth requirement of multimedia devices,
Usually such an interface is based on the X Window system buses are
becoming the bottleneck, and their performance im-
(2), and is a WYSIWYG (what you see is what you get) inter- provement
always lags behind the need. Some new architec-
face. Two popular GUIs on workstations are Open Look and tures have
been proposed. For example, Sun developed the
Motif. ultra port
architecture (UPA) (Fig. 2), a crossbar-switched in-
Workstations are used by scientists and engineers to do terconnect
that enables multiple simultaneous data transfers.
computationally intensive or graphics intensive applications In its O2
products, SGI uses the unified memory architecture
such as research, software development, scientific computa- (Fig. 3).
The high-speed multiported main memory is at the
tion, and CAD/CAM. They may also be used for desk-top pub- heart of
the system. All data types required by various sub-
lishing, scientific and medical visualization, image analysis, systems
such as the monitor and video I/O are stored in mem-
and other applications. ory. Each
subsystem accesses the memory through high-speed
The idea of a workstation originated at Xerox during the ports
designed to ensure that no data starvation occurs on
1960s, but it was not until the development of microprocessor any port.
This architecture allows more efficient use of the
technology that workstations became widely available. Mod- entire
system memory and increases performance by reducing
ern workstations began to appear in the market in the 1980s. the buffer
copies among those components.
Apollo Computer, Inc. released its first Apollo/DOMAIN Most
multiprocessor workstations adopt the symmetric
workstations (DN100s) in 1981 with the operating system
multiprocessor architecture (SMP) because of its simplicity
Aegis SR1 installed on them. Andreas Bechtolsheim designed and low
cost. The architecture is very similar to that in Fig.
----------------------- Page 145-----------------------
Processor
Memory
L1 cache
CPU Monitor
Memory
L2 cache
controller
System bus
Graphics
processor
Image and Flat panel
compression display
I/O bus
Main
Video
Memory
PCI I/O
Device Device Device

• • •
Figure 3. SGI’s Unified Memory architecture. It uses a multiported

main
memory shared by all components of the system and allows a
Figure 1. Architecture of a workstation. It consists of one or more more
efficient usage of the entire system memory.
processors associated with one level or multiple levels of caches.
Memory is connected by a system bus. Other devices such as disks
and input/output (I/O) devices are connected by I/O buses.
ers
may provide their own variants of UNIX in their ma-
1 except for the fact that there is more than one pair of pro-
chines. For example, in Sun SPARCStations, the operating
cessor and cache attached to the system bus. Compared with
system in Solaris, which is based on UNIX System V release
single processor machines, multiprocessor ones require a 4;
however, SunOS 4.x is based on BSD UNIX. SGI work-
higher bandwidth of the system bus. The locality property of
stations use Irix. IBM has its own variant, AIX, and HP has
most programs allows each processor to access data often in HP-
UX. There has also been an attempt to standardize UNIX
its own cache, thus reducing the bandwidth requirement.
initiated by the IEEE Standards Board. The standard UNIX
However, when the number of processors increases to more is
called POSIX.
than 8, a hybrid architecture consisting of several buses or a
crossbar-switched interconnect is usually used to balance the
Threads
workload and to reduce the bus bottleneck effect.
The
concept of threads is important in modern operating sys-
SOFTWARE ENVIRONMENTS tems.
It is used in workstations to improve system perfor-
mance
or exploit parallelism when designing and implement-
Operating Systems ing
complex applications. A key concept in traditional
operating systems is the process. A process is basically a pro-

UNIX is the operating system used by most workstations. It gram
in execution. It consists of the executable program run-
is a time-sharing multitasking system. Different manufactur- ning
on the program’s data and stack, which is called the ad-
dress
space of the process. Each process has its own address
space
and a single thread of control. However, in some cases
it
is desirable to have multiple threads of control that share
Processor I/O a
single address space. The major difference between threads
cache subsystem
and
processes is related to the address space. Each process
has
its own address space, and the interaction between pro-
cesses occurs through interprocess communication primitives.
Threads have their own program counters, stacks, and states,

but
share a common address space, through which they com-
UPA
municate with each other. Threads are sometimes called
lightweight processes because of the low overhead for their
creation, deletion, and management in comparison with pro-
cesses.
Threads were invented to allow parallelism to be combined

Memory Graphics with
sequential execution and blocking system calls (system
calls
that may be blocked waiting for some events). In a sin-
gle-
processor machine, they may improve the system utiliza-
Figure 2. Sun Microsystems’ Ultra Port architecture. Rather than tion
and throughput. In a multiprocessor system, threads
using a bus to connect all components, a crossbar is used which allows may
run in parallel to support the exploitation of fine- or
multiple simultaneous data transfer among components.
medium-grain parallelism.
----------------------- Page 146-----------------------
ENGINEERING WORKSTATIONS 117
Communication of
Technology (MIT) in 1984. X Windows is an industry-stan-
dard
software system that allows programmers to develop
The communication protocol often used to exchange data be-
portable
graphical user interfaces. It is a collection of pro-
tween workstations is TCP/IP, which is also the protocol used
grams
and libraries that allow the development of applica-
by the Internet. A new IP protocol called IPv6 is being de-
tions
for graphics workstations in which the output can ap-
signed and standardized to solve problems encountered such
pear in
multiple windows on the screen. It also provides a
as the significant increase in the number of computers on net-
means for
a program to be run on a remote computer, but to
works. Layered protocols such as TCP/IP have high software
display
the results on the local workstation. It is character-
overhead when used to exchange short messages. There are
ized by:
efforts to address this. Usually, workstations are connected
with each other by Ethernet networks. Currently, there are
two bandwidth Ethernets, 10 Mbps (megabit per second) and •
Device Independence. X Windows allows programs to dis-
100 Mbps. Asynchronous transfer mode (ATM) (4) networks play
windows containing text and graphics on any hard-
are more expensive, but provide higher bandwidth ( 100
ware that supports the X Protocol without modifying, re-
Mbps).
compiling, or relinking the application.
•
Client-Server Model. In X Windows, application pro-
Programming Languages
grams are called clients. The clients talk (in X Protocol)
to a
program called the (display) server, which drives the
There are many programming languages available in work-
display and interprets keyboard commands and mouse
stations. The most commonly used ones are C, C##, and
clicks. The client and server may be on different ma-
FORTRAN 77. Designed in 1972 at AT&T, C is a high-level
chines, so sometimes X Windows is called a graphical
programming language used at first to write the UNIX op-
windowing system for computer networks.
erating system, and for system programming. Over the years,
•
Mechanism/Style. X Windows provides mechanisms to
the power and flexibility of C have made it a popular lan-
support many interface styles rather than enforcing any

guage for a variety of applications. C ## is an object-oriented
one
policy.
programming language based on C (superset of C). There are
two standards of C and C ##: ANSI and GNU. Most compilers
comply with either of the two standards. The compiler cc is MIT
has placed X Windows source code in the public do-
an ANSI C compiler, and gcc is the GNU compiler, which main.
Almost all UNIX graphical interfaces, including Open
integrates the C and C ## compilers. Both compilers process Look and
Motif, are based on X Windows.
input files through one or more of four stages: preprocessing, The
Open Look user interface was developed by Sun Micro-
compilation, assembly, and linking. FORTRAN 77 is a lan- systems
in partnership with AT&T. It is the specification of
guage used by many engineers to do engineering computation the user
interface (i.e., how the buttons look, high-level de-
on workstations, and some scientific computation libraries are
scriptions of how the user interacts with the computer).
written in FORTRAN 77. The recent portable Java language Openwin
is a specific implementation of the Open Look User
is already proving popular on workstations as well (5). interface
for Sun computers using X Window. Motif is a look-
and-feel
interface that was developed by the Open System
Foundation (OSF) as an alternative to Open Look. Motif was

User Interfaces
developed
primarily because Open Look at the time did not
UNIX provides two interfaces for users. The basic one is a support
3-D widgets. Motif was designed to be similar to Mi-
command line-based interface called the shell. The advanced crosoft
Windows and IBM’s OS/2.
one is a graphical user interface (GUI) based on the X Win- In
principle, both Open Look and Motif can be regarded as
dow system. toolkits
of X Windows because they extend it with many wid-
There are two common shells available in UNIX: the C gets to
ease the design and implementation of GUIs using X
shell and the Bourne shell. Each is an interactive command Windows.
However, using these widgets, the interface is
interpreter as well as a high-level programming language. As forced to
be in one specific style.
a command interpreter, the shell processes commands the There
are even higher level environments to facilitate the
user enters in response to its prompt. As a programming lan-
development of GUIs. The Tcl and Tk tools are an example.
guage, it processes groups of commands stored in files called The Tcl
scripting language and the Tk toolkit are a program-
shell scripts. System administrators and users often use them ming
environment for creating GUIs under X Window. They
to do file management or other tasks that do not require the are easy
to learn, very powerful, and contain many sophisti-
use of complicated programming languages such as C. The cated
features. They can dramatically reduce development
programming ability of a shell is limited. There is another time for
X programmers.
language called Perl that bridges the gap between shell pro- Tcl
and Tk were created by John Ousterhout (6). Tcl is a
gramming and C programming. It is an interpreted script lan- simple
scripting language for controlling and extending appli-
guage for easily manipulating text, files, and processes. It pro- cations.
It supports generic programming constructs such as
vides a more concise and readable way to do many tasks that
variables, loops, and procedures. Tcl programs can be easily
were formerly accomplished by programming in C or in one
incorporated into applications, and any application can ex-
of the shells. It is available in the public domain on the In- tend Tcl
with new features. One useful extension to Tcl is Tk,
ternet, and widely used on various machines such as work- which is
a toolkit for X Windows. It extends Tcl with com-
stations. mands
for building Motif-like (look and feel) user interfaces
The GUIs for workstations are based on the X Window sys- by
writing Tcl scripts instead of C code, Tk consists of a set
tem, which was first developed at the Massachusetts Institute of
commands for creating user interface widgets and geometry
----------------------- Page 147-----------------------

Command line interface
GUI
Tk
Perl Tcl
Motif OpenLook
Shell
Xlib/widgets
To other X servers
X server
Figure 4. A software environment on a
via networks
workstation. This diagram shows where Unix device drives
each software discussed is in the whole
software environment and the relation-
ship between each.
managers to arrange these widgets on the screen. The Tcl and system and
computation issues to be addressed for effectively
Tk source code is publically and freely available. There are utilizing
networks of workstations for parallel computing. In
many applications developed by academia and industry using practice,
a network of workstations is a heterogeneous and
Tcl/Tk. Figure 4 gives the relationship between all the soft-
nondedicated system. Heterogeneity represents computing
ware components discussed above. power
and architectural differences among workstations.
Nondedication implies possible resource sharing and interac-

APPLICATIONS tions of
both parallel jobs and normal workstation user jobs.
In
addition, it would be too expensive to provide a centralized
Workstations may be used in various fields such as: network
scheduler for parallel jobs. Thus, effective coordina-
tion of
parallel jobs and normal user jobs needs sufficient in-
1. CAD/CAM formation
about system heterogeneity and job interaction. A
practical
scheduler should also utilize the existing worksta-
2. Electronic publishing
tion
operating system, such as a UNIX local scheduler (8).
3. Internet
Another key
enabling technology needed to make high-perfor-
4. Multimedia mance
scientific computing on networks of workstations truly
5. Geographic information system possible
is to build affordable, scalable, and high bandwidth
6. Graphics, imaging, animation, and visualization networks
with low overhead communication protocols (7).
7. EDA (electronic design automation)
Workstation processors are getting much faster, but the
8. Research performance
of I/O is being improved at a much lower pace.
Consequently, applications are becoming more I/O bound

9. Software engineering/tools
than
computation bound. Multimedia devices require high
10. Information management I/O
bandwidth, which also increases the burden for pro-
11. Business-related software cessors.
New architectures such as Sun’s UPA and SGI’s
UMA are
proposed to improve the overall performance of
There are thousands of software products available for
workstations. Some techniques which were used previously
workstations. Some are provided by the manufacturers, and by
minicomputers or mainframes are now used in worksta-
others by third party venders. tions. A
specific I/O processor may be used to manage the I/O
devices,
thus reducing the workload on the main processor.
CURRENT AND FUTURE RESEARCH Redundant
arrays of inexpensive disks (9) has been proposed
AND DEVELOPMENT TOPICS to
increase disk performance and reliability. People are pay-
ing more
attention to the development of fast I/O systems.
With rapid networking software and hardware technology ad- In
addition to traditional applications, the advance in mul-
vances, and with the downsizing of the supercomputing in- timedia
and network techniques promotes the use of net-
dustries, networks of workstations are becoming the major worked
workstations in computer-supported cooperative work
computing infrastructure for science and engineering, from (10)
environments. Applications in this field include telecon-
low-end interactive activities to large-scale sequential and ferencing,
remote education, multi-agent problem solving,
parallel applications (7). However, networks of workstations and
planning.
were not initially designed for parallel computing. Besides A major
difference between PCs and workstations is re-
relatively low network bandwidth, there are other important lated to
processor speed and memory space. A workstation
----------------------- Page 148-----------------------
ENTREPRENEURING 119
traditionally provides more computing power and larger

memory space. With rapid development of the PC and with
more engineering applications running on PCs, the distinc-
tion between PCs and workstations is becoming fuzzier. When
multiprocessor PC systems become available, the PC-based
systems will expand deeper into workstation territory. Appli-
cations and markets of workstations and PCs are quickly
merging.
BIBLIOGRAPHY
1. D. E. Comer, Internetworking with TCP/IP. Volume I: Principles,

Protocols, and Architecture, 3rd ed., Englewood Cliffs, NJ: Pren-
tice-Hall, 1995.
2. D. A. Young, The X Window System: Programming and Applica-

tions with Xt, Englewood Cliffs, NJ: Prentice-Hall, 1990.
3. J. L. Hennessy and D. A. Patterson, Computer Architecture: A

Quantitative Approach, 2nd ed., San Francisco, CA: Morgan
Kaufmann, 1996.
4. D. E. McDysan and D. L. Spohn, ATM Theory and Application,

5. K. Arnold and J. Gosling, The Java Programming Language,

Reading, MA: Addison-Wesley, 1996.
6. J. K. Ousterhout, Tcl and the Tk Toolkit, Reading, MA: Addison-

Wesley, 1994.
7. T. E. Anderson, D. E. Culler, D. A. Patterson, and the NOW

team, A case for networks of workstations: NOW, IEEE Micro,
54–64, Feb., 1995.
8. X. Du and X. Zhang, Coordinating parallel processes on networks

of workstations, J. Par. Distrib. Comp., 125–135, 1997.
9. P. M. Chen et al., RAID: high-performance, reliable secondary

storage, ACM Comp. Surv., 145–185, June 1994.
10. I. Greif, Computer-Supported Cooperative Work: A Book of Read-

ings, Los Altos, CA: Morgan Kaufmann, 1988.
XING DU
University of Virginia
XIAODONG ZHANG
YONG YAN
College of William and Mary
ENHANCEMENT, IMAGE. See IMAGE ENHANCEMENT.

ENTERPRISE RESOURCE PLANNING. See MANUFAC-
TURING RESOURCE PLANNING.
----------------------- Page 149-----------------------
} { { } }
●
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Fuel Cell Power Plants

Standard Article

Mark C. Williams1
1US Department of Energy, Morgantown, WV

reserved.
Search
DOI: 10.1002/047134608X.W3017
❍ Acronym Finder
Abstract
Fuel Cell Technology
Phosphoric Acid Fuel Cells Status

Molten Carbonate Fuel Cells Status
Solid Oxide Fuel Cells Status
Networks
High-Efficiency Fuel Cell Gas Turbine Systems
The World Power Market
The Changing Face of Electricity Generation
Specific Fuel Cell Attributes and Market Applications
International Market Applications
file:///N|/000000/0WILEY%20ENCYCLOPEDIA%20OF%20...ENGINEERING/22.%20Energy
%20Conversion/W3017.htm (1 of 2)17.06.2008 16:38:54
----------------------- Page 150-----------------------
----------------------- Page 151-----------------------
2 FUEL CELL POWER PLANTS
fully
and efficiently leverage funding for the U.S. Stationary
Power
Fuel Cell Program.
Because of investment in the 1980s and early 1990s, first-
generation fuel cells are now crossing the commercial thresh-

old. DOE
and predecessor agencies have funded the develop-
ment of
fuel cell systems since the 1970s. Initially, phosphoric
acid
fuel cells (PAFCs) were the primary focus, and these
units,
operating on natural gas, are now in the initial stage
of
commercialization. In the last few years, focus in the
United
States has shifted to the advanced fuel cell types, in-
cluding
molten carbonate fuel cells (MCFCs) and solid oxide
fuel
cells (SOFCs). These systems offer higher efficiencies and
the
potential for lower capital cost, and because of higher op-
erating
temperatures they are more suitable for cogeneration
than
lower-temperature fuel cells.
FUEL
CELL TECHNOLOGY
Fuel
cells generate electricity and heat using an electrochemi-
cal
process similar to that of a battery. A fuel cell will continu-
ously
produce power as long as a fuel, such as natural gas,
and an
oxidant, air, are supplied to the system. Present early
market
systems are achieving over 40% lower heating value
(LHV)
cycle efficiency. The next-generation systems are ex-
pected
to achieve 55% and eventually 70% LHV cycle effi-
ciencies.
As
shown in Table 1, several different types of fuel cells
are
being developed for stationary power applications. The
electrolyte controls the operating temperature of the cells,

FUEL CELL POWER PLANTS which
in turn determines the materials of construction.
PAFCs
are now becoming commercially available, while
Fuel cell power plants offer the potential for ultrahigh-effi- MCFCs
and SOFCs promise even higher efficiencies for the
ciency energy conversion and the enhancement of the quality future
(1).
of our environment. Because of this, organizations in several A
basic fuel cell (Fig. 1) consists of two electrodes, with the
countries are sponsoring the development of fuel cells for sta- anode
and cathode separated by an electrolyte. Fuel cell types
tionary power generation market applications. are
characterized by their electrolyte. For example, PAFCs
Concerns for the global environment are driving future utilize
a phosphoric acid electrolyte in a matrix between
power generation systems toward technologies that produce anode
and cathode electrodes. To produce a useable quantity
extremely low environmental emissions. Because of their high of
electric power, individual cells are assembled into a vertical
efficiencies, fuel cell power plants will help in reducing carbon
‘‘stack’’ of repeating components which are electrically inter-
dioxide emissions. Since combustion is not utilized in the pro-
connected. A fuel cell power plant (Fig. 2) consists of the stack
cess, fuel cells generate very low amounts of nitrogen oxide or power
section integrated with a fuel processor and a power
(NOx ). Fuel cell power plants have been exempt from air per-
conditioner to convert the power from direct current to alter-
mitting requirements in northern and southern California nating
current.
and in Massachusetts. Relying on electrochemistry instead of
combustion, the fuel cell is attractive for both heavily polluted
urban areas and remote applications. Not only will it emit
none of the smog-causing pollutants associated with conven- Table 1.
Types of Fuel Cells
tional powerplants, it is ideal as a distributed power source;

Characteristic PAFC MCFC SOFC
that is, it can be sited at or near the electricity user—for ex-
Electrolyte Phosphoric acid Lithium Stabilized

ample, at electrical substations, at shopping centers or apart-
carbonate/po- zirconia
ment complexes, or in remote villages—minimizing long-dis-
tassium car-
tance transmission lines.
bonate
The U.S. Stationary Power Fuel Cell Program is a market-
Operating
driven program which has over 40% cost-sharing from the pri-
temperature 200#C 650#C 1000#C
vate sector. The U.S. program is being implemented by the
Electrical
U.S. DOE Federal Energy Technology Center (FETC). The
conversion
stationary power fuel cell developers enjoy the support of user
efficiency
(LHV)
45–50% 50–65% 50–60%
groups with over 75 utility and other end-user members. In
Materials Carbon Nickel stainless Ceramic

addition, DOE cooperates with the Gas Research Institute
platinum steel
(GRI) and the Electric Power Research Institute (EPRI) to

----------------------- Page 152-----------------------
FUEL CELL POWER PLANTS 3
Oxidant (O or CO )
The DOE/FETC-sponsored PAFC development work at
2 2
IFC
was completed in 1992. ONSI
Corporation, located in
Cathode
South Windsor, Connecticut, has been actively involved in the
development and marketing of on-site PAFC systems and has
Electrolyte a 40
MW/year manufacturing facility. In their PAFC commer-
Anode
cialization, the ONSI Corporation, a subsidiary of IFC, is of-
fering acomplete packaged phosphoric acid fuel cell power

Fuel H2 e–
plant for $3000/kW. Named PC25, over one hundred 200 kW
units are in
operation in the United States and around the
• Electromagnetic process
world. An additional 22 units were added in the United States
• Direct conversion to electricity
through the Climate Change Fuel Cell Program. Operating
H + 1/2 O H O + Electricity
2 2 2
experience has been excellent, with availabilties of over 90%.
Figure 1. Basic fuel cell. The

PAFC is so reliable that it is being considered for unin-
terrupted power supply (UPS) applications.
Although PAFC technology is the most mature of the fuel

The fuel cell is inherently modular. Constructed as an as- cell
types being developed and cell-and-stack performance ex-
sembly of individual cells, stacks ranging from 100 to 250 kW
hibited by all designs is close to acceptable for early commer-
form a modular building block. Depending on the generating cial
operation, cost remains as an issue. Power-plant costs
capacity required, 10 to 20 stacks can be grouped with a fuel must
be reduced to be competitive with other advanced tech-
processor and a power conditioner to create a 1 to 2 MW
nologies. A current goal is to reduce these costs to less than
power plant. Larger plants will use a larger number of stacks.
$1500 to $2000/kW. An operating life of 40,000 h is desired
In high-growth areas or remote sites, modular power plants and
may not be an issue. In fact, 70,000 h of life is now
located near the demand can offset the cost of right-of-way
thought to be attainable. IFC is currently developing 1 MW
access and transmission lines.
class units based on a five-stack design and developing the
PC25C, which is lower in size and cost. The major improve-

ment
represented by the PC25C was the smaller, lighter-
PHOSPHORIC ACID FUEL CELLS STATUS
weight invertor whose smaller size helped lower the PC25’s
weight by 20,000 pounds (9091 kg).

In the PAFC technology, the electrochemical reactions occur
ONSI claims several things about its large 1.2 MW PAFC
on highly dispersed electrocatalyst particles supported on car-
plants it is offering: The PAFC is not a strandable asset since
bon black. Platinum or platinum alloys are used as the cata- it
is movable; repackaging of the PAFC into 1.2 MW plants
lyst for both electrodes. The platinum is supported on carbon will
lower cost and footprint; availability is increased by using
black for both electrodes. The electrodes also use a polymeric
multiple, high-reliability units, making it a natural for UPS
binder and a carbon paper substrate. The separator plates are
applications; and the PAFC can provide high power availabil-
all graphite (1). ity
with low reserve margins.
DOE and GRI, beginning in the late 1970s, supported an
In its premium power application, IFC uses a static switch
on-site PAFC effort that included an R&D program and a to
switch to grid only when the fuel cell, which is baseloaded,
manufacturing and field test program with international fuel is
to be maintained. The grid is the UPS. This is quite unlike
cells (IFCs). The program resulted in the production and test- the
applications where the reciprocating engine is used as a
ing of over fifty 40 kW, on-site cogeneration power plants dis- UPS.
The engine is not baseload run and is used only in an
tributed to sites throughout the United States and Japan.
emergency, being tested daily.
This program was successfully completed in 1986 and formed
the technology base for the current 200 kW, on-site work. The
PAFC is a proton-conducting fuel cell, which has routinely
MOLTEN CARBONATE FUEL CELLS STATUS
reached an operating performance level of 200 W/ft2 (2150
W/m2) at ambient pressure.
Overall system efficiencies of 50 to 60% are forecast for natu-
ral
gas and coal gasification MCFC power plants. The MCFC
operates at 650#C. The MCFC, like other fuel cells and unlike
turbines and diesels, offers high efficiency at small size and

Air (O )
2 at
part load. Furthermore, an MCFC power plant can operate
Natural on
coal or natural, refinery, or processed gas. MCFC stack
gas or
designs incorporate either internal or external fuel and oxi-
derived H2 dant
manifolding and either internal or external reforming.
coal gas Fuel Power Dc Power All
MCFC designs include flat cell components in the cell
processor section conditioner Ac
package (i.e., anode, matrix to hold carbonate, cathode, cur-
Steam power rent
collector, and separator plate).
The main components of an individual cell are the anode,

the
cathode, and the molten carbonate electrolyte. Electrode
materials are usually porous nickel alloys for reducing atmo-

By-product
spheres (anode) and nickel oxide for oxidizing atmospheres
heat
(cathode). The electrolyte, typically a combination of molten,

Figure 2. Fundamentals of a power plant.
alkali (lithium, potassium, sodium) carbonates, is contained
----------------------- Page 153-----------------------
2 2
within a porous ceramic matrix, commonly made of lithium
A/ft (160–270 mA/m ), at 0.60–0.80 V, with 50–85% fuel uti-
aluminate. An individual cell is approximately 6 mm thick.
lization (2–9).
The electrolyte is about 1 mm thick.
The goal of the U.S. MCFC program is to develop and com-
Figure 3 illustrates the structure of an MCFC stack. Con-
mercialize low-cost, packaged, simple, and modular fuel cell
ductive, bipolar separator plates connect the individual cells
systems. DOE is accelerating the drive for private sector com-
in a stack, both structurally and electrically. The bipolar sep-
mercialization of multifuel, MCFC power plants.
arator plate is made of stainless steel, and each plate physi-
The two MCFC developers have collected impressive stack
cally separates the fuel gas stream of one cell from the oxi-
test performance data under the 1990 program R&D an-
dant gas stream of the adjacent cell. One side of each
nouncement (PRDA). ERC is developing an externally mani-
separator plate channels a fuel stream so that it flows over a
folded, externally reforming MCFC and has constructed a 2
porous anode, while the flip side channels an oxidant stream to
17 MW/year MCFC manufacturing plant. ERC has con-
over a porous cathode. Each bipolar separator plate also col-
structed a 100 kW test facility in Danbury, Connecticut, and
lects current, thus, electrically connecting adjacent cells of a
has scaled up to a 6 ft2 2
(0.56 m ) area stack.

stack in series. Electrons are conducted from the anode
MCP is developing an internally manifolded, externally re-
through the bipolar separator plate and into the cathode of
forming MCFC and has constructed a 4 to 12 MW/year MCFC
the adjacent cell. There they react with the oxidant gas
manufacturing plant. MCP has constructed a 250 kW accep-
stream and form carbonate ions. The carbonate ions diffuse
tance test facility in Burr Ridge, Illinois, and has scaled up to
an
11.4 ft2 2
through the electrolyte and into the anode, where they react
(1.06 m ) full-area stack.
with the fuel gas stream, releasing electrons into the anode.
DOE, in conjunction with EPRI, GRI, and the Department
Electrons are conducted in this manner through all the cells, of
Defense (DOD), is also funding product development tests
thus establishing direct current through the stack. An exter-
(PDTs) concurrently with system development at ERC and
nal circuit connects a load between the two end plates of the
MCP. A successful demonstration track record will enhance
stack, completing the circuit.
support for MCFC technology from utilities and other end us-
Figure 4 depicts an MCFC with the fuel and oxidant
ers in the distributed, repowering industrial and commercial
streams flowing perpendicular to each other within the cell.
markets.
This internal flow geometry is known as cross-flow. Other pos-
The initial MCFC PDTs was in California in 1996–1997.
sible internal flow geometries are cocurrent and countercur-
ERC is conducting a 2 MW PDT in Santa Clara, California,
rent. In general, a countercurrent internal flow geometry pro-
funded by the Santa Clara Demonstration Group, EPRI, and
duces the most uniform current densities and temperature
DOE. MCP will conduct a 250 kW PDT in San Diego, Califor-
distributions within a cell. A cocurrent internal flow geome-
nia, funded by DOE, GRI, and San Diego Gas and Electric at
try, in general, results in the least uniform current densities
the Miramar Naval Air Station.
and temperature distributions.
DOE/FETC recently competed a Product Design and Im-
At least two MCFC developers, Energy Research Corpora-
provement (PDI) PRDA to resolve technology, system, and
tion (ERC) and M-C Power (MCP), have conceptual designs of
network issues. There remain major issues in MCFC opera-
efficient integrated MCFC power plants. Operating conditions
tion, such as cathode corrosion (3,4,10). Major network and
for these MCFCs are projected to be in the range of 150–250
system issues are cost, heat loss management, footprint,
Internal manifold stack
(no external manifolds)
Fuel out
External manifold stack
Air out
Manifold Fuel cell

Air in
seal stack
Fuel
out
Fuel in
Oxidant
out
Oxidant
inlet
manifold
Fuel inlet
manifold
Fuel
Figure 3. MCFC stack structural de-
Oxidant
in
signs.
in
----------------------- Page 154-----------------------
Fuel
Fuel
feed
exhaust
H2 CO2 H O
2
2 –
=
e
H2 + CO 3 CO + H O + 2
2 2 e
Anode
=
Molten
CO
Load
carbonate 3
electrolyte
Cathode 1 =
CO + O + 2 CO
2 2 2 e 3
2 –
Oxidant CO2 O2
Oxidant
feed
exhaust
Figure 4. Operation of an MCFC.
packaging and integration, parasitic power losses, pressuriza-

The high-temperature (1000#C) SOFC can provide greater
tion, and reforming. The objective of this work is to aim cur- fuel
flexibility than lower temperature fuel cells, since the re-
rent MCFC stack development toward the development of a
forming reaction is favored at higher temperatures. Re-
packaged, commercializable MCFC product. The PRDA will
forming heat requirements with low-temperature fuel cells
bring a multifueled, integrated, simple, low-cost, modular, can
actually lower overall system efficiency for some fuel
market-responsive MCFC power plant to the marketplace.
cells. Reforming is an important system consideration which
The development program will be based on a commercializa- will
remain important in the absence of a low-cost hydrogen
tion plan to manufacture and package, demonstrate, and
supply. In addition, a higher-quality heat produced by the
aggressively market MCFC power plants. The PDI PRDA
high-temperature SOFCs results in better bottoming cycle
will culminate in the manufacture and construction of high-
performance in some system configurations.
performance, low-cost, 500 to 2000 kW MCFC power-plant
Westinghouse Electric is the acknowledged world leader in
modules.
tubular SOFC technology. The Westinghouse Electric tubular
configuration is shown in Fig. 5. Several completely packaged

SOLID OXIDE FUEL CELLS STATUS and
self-contained generators, up to nominal 25 kW size, have
been
manufactured and tested by Westinghouse Electric. A
Some general characteristics appear to be shared by many of pre-
pilot manufacturing facility currently produces the cells
the SOFC technologies being developed. While there is vari-
(tubes), bundles, and generators. The length of the tubes has
ability in materials being used for various components, the been
scaled up to a nominal 2 m in length. The porous air
SOFC is an oxygen ion-conducting, solid-state device com-
support tube has recently been eliminated. The cell is now
posed of a nickel–zirconia cermet anode, an yttria-stabilized
supported by the air electrode. The Westinghouse Electric
zirconia electrolyte, a strontium-doped lanthanum manganite
technology has been validated to a far greater extent than
cathode, and a doped lanthanum chromite interconnect (1). any
other SOFC technology. Multiple tube tests have been
The solid-state electrolyte of yttria-stabilized zirconia oxide is
successfully conducted for more than 65,000 h, with less than
characterized by ionic conduction. The solid-state character of 1%
per 1000 h degradation. Pressurized operation of the tubu-
the SOFC electrolyte means there are few constraints on de- lar
SOFC has recently been demonstrated at Ontario-Hydro.
sign. There is no problem of electrolyte containment, hence
the flexibility and the wide variety of designs or forms being
pursued.
The flexible SOFC may be operated over a wide range of
Interconnection
temperatures. The theoretical thermodynamic efficiency (73%
based on the hydrogen oxidation reaction at 927#C) is slightly
Electrolyte
lower for the SOFC than for the MCFC and the PAFC. How-
ever, the overall efficiencies of SOFC systems are more than
Fuel
those of the PAFC and certainly rival those of MCFC system
Air flow
configurations.
electrode
Power densities for SOFCs are promising. Power densities

of 2.0 W/cm2 on hydrogen at 1000#C have been reported for
SOFCs. The high-power density with thin-layered compo-
Air
Flow
nents could make the SOFC an attractive power-plant alter-
Fuel electrode
native. However, packaging and cost reduction will be re-
quired to make the SOFC promise a reality.
Figure 5. Westinghouse SOFC design.
----------------------- Page 155-----------------------
A 100 kW generator test, in the Netherlands, is also planned

temperature fuel cell with a potentially low manufacturing
for the 1996–1997 timeframe (12–16). cost
(24,25).
Several planar designs are also under development. Orga-
It is often difficult to determine if the SOFC materials and
nizations developing planar designs include the Institute of their
electrochemical and physical properties, per se, or if the
Gas Technology (IGT), Ceramatec, Ztek, Technology Manage-
individual SOFC designs contribute more to performance, as
ment Incorporated, and Allied Signal Aerospace Corporation.
measured by power density, efficiency, longevity (or durabil-
These developers hold strong patent positions on cell designs, ity),
cost, packagability, and system integrability. A variety of
which is essential for low-cost manufacturing. both
material and design-related issues are being addressed.
IGT is developing an 800#C, intermediate-temperature, in-
ternally manifolded planar design. This trilayer IGT design,
NETWORKS
shown in Fig. 6, has, according to IGT, the advantages of
more effective gas flow patterns, more compact design and
In
conventional fuel cell systems, multiple stacks have been
cell stacking, more efficient current and voltage transfer from
arranged in parallel with regard to the flow of reactant

cell to cell, and more cost-effective manufacture (17). The IGT
streams. Networking (26–30) improves upon conventional

design is an internally manifolded fuel cell design using
MCFC
system designs in which multiple stacks are typically
pressed metallic plates called IMHEX#. Because the
arranged in parallel with regard to the flow of reactant

IMHEX# design has no external gaskets and seals, only com-
streams. As illustrated in Fig. 7(a), the initial oxidant and

pression seals are necessary to obtain good sealing. The ce-
fuel
feeds are divided into equal streams which flow in paral-
ramic bipolar separator plates in the SOFCs currently under
lel
through the fuel cell stacks.
development are the single most expensive component. These
In an improved design, called an MCFC network, reactant

make up more than 80% of the total materials and fabrication
streams are ducted such that they are fed and recycled among
costs of the cell components (18). IGT replaces the ceramic
multiple MCFC stacks in series. Figure 7(b) illustrates how

separator plates with nickel-based metallic separator plates,
the
reactant streams in a fuel cell network flow in series from
thus lowering cost significantly. Since at 800#C the zirconia
stack
to stack. By networking fuel cell stacks, increased effi-
electrolyte will have high-internal-resistance losses, IGT is
ciency, improved thermal balance, and higher total reactant

using the provskite gadolinium-doped barium cerium oxide.
utilizations can be achieved. Networking also allows reactant

IGT may utilize the Argonne National Laboratory glass/ce-
streams to be conditioned at different stages of utilization.

ramic composite seals, which could sidestep most of the prob-
Between stacks, heat can be removed, streams can be mixed,

lems associated with glass-only or cement-only manifold
and
additional streams can be injected.
seals.
MCFC stack networks produce more power than conven-

The Ceramatec design, CPn#, consists of stacks and a fuel
tional configurations because they more closely approximate

processor, and it places some cells in a series rather than in
a
reversible process. The Nernst potential is the voltage
parallel to obtain greater efficiency. Ceramatec has attained
which
drives reversible electrode reactions. This reversible
a power density of 0.18 W/cm2 2
(167 W/ft ) and a current den-
2 2
voltage, generated by the overall cell reaction, is a function of
sity of 250 mA/cm (230 A/ft ). Ceramatec has tested a 1.4 kW
the
local temperature, pressure, and reactant concentrations.
module and has a limited partnership with Babcock and Wil-
As
reactants are utilized, their concentrations change. Since
cox (19,20) for the commercialization of the technology. Ztek
Nernst potential is dependent upon the concentrations of re-

uses a radial design stacked into two-stack modules which
actants, it varies with the degree of utilization.

are then combined into arrays. Ztek, along with EPRI and
In a conventional power plant, the fuel is utilized in a sin-

Tennessee Valley Authority, has completed testing a 1 kW
gle
stack, and all the current is generated at a single voltage.
stack (21,22). Technology Management Incorporated uses an
In
networks, stacks in series each utilize only part of the fuel.
Interscience Radial Flow design in which each cell is made up
The
network can produce more power because most of the
of four layers, with sealing being achieved through the use of
total
charge is transferred at increased voltages. When the
rings which also form the internal fuel and air manifolds.
total
fuel utilization of each system is optimized for maximum
Small stack testing from one to ten cell stacks has been per-
formed. Power densities around 0.08 W/cm2 2
efficiency, the efficiency of the fuel cell stacks networked in
(75 W/ft ) have
series can be nearly 10% greater than that of the stacks ar-
been attained (23). Allied Signal Aerospace Corporation is de-
ranged in parallel.
veloping the monolithic and flat planar designs and is now
Arranging fuel cell stacks in series offers several other ad-

using tape-calendaring to produce a thin-electrolyte, reduced-
vantages over conventional fuel cell powerplates. Placing
stacks in series also allows reactant streams to be conditioned

at
different stages of utilization. Between stacks, heat can be
consumed or removed (methane injection, heat exchange),

Porous cathode which
improves the thermal balance of the system. The com-
(screen printed)
#########
position of streams can be adjusted between stacks by mixing
Porous anode (tape cast)
exhaust streams or by injecting reactant streams.
HIGH-
EFFICIENCY FUEL CELL GAS TURBINE SYSTEMS
Manifolding Dense
hole electrolyte
One
of the most promising developments in fuel cell power
(thin film)
plants is the
conceptual development of very high efficiency
Figure 6. IGT SOFC. fuel
cell gas turbine power plants (31–40). Studies have indi-
----------------------- Page 156-----------------------
IRMCFC Exhaust
Stack
Fuel
and IRMCFC Exhaust
oxidant
Stack
flow
IRMCFC Exhaust
Stack
(a)
Fuel
and IRMCFC IRMCFC IRMCFC
Exhaust
oxidant
Stack Stack Stack
flow
Figure 7. (a) Parallel flow of reactant
streams through stacks. (b) Series flow of

(b)
reactant streams through stacks.
cated that this combination has the potential to increase the in

the 1–5 MW size range. Table 2 summarizes some of these
overall efficiency for the conversion of natural gas into elec- fuel
cell gas turbine power plants (40).
tricity to over 70%.
The combination of the fuel cell and turbine has the poten-
Because of the synergistic effects leading to the higher ef- tial
for enormous synergies, in that it offers a solution to two
ficiencies and lower emissions achieved by combining a fuel
important problems: (1) the low efficiency and relatively high
cell and a gas turbine into a power generation system, many NOx
emissions of small gas turbines and (2) the high cost of
potential system configurations have been developed (39). small
fuel cell power plants.
These include the natural gas, indirect-fired, carbonate fuel
Small gas turbines, with capacities of less than 10 MW,
cell bottomed, combined cycle and the topping natural gas/
typically have efficiencies in the 25 to 30% (LHV) range.
solid oxide fuel cell combined cycle for distributed power and Small
high-temperature solid oxide and molten carbonate fuel
on-site markets in the 20 to 200 MW size range. Most of these cell
power plants are predicted to cost $1000 to $1500/kW,
large fuel cell/gas turbine systems utilize a steam cycle to when
commercially available in the years after 2000. By com-
achieve high thermal efficiency. The latter is shown in Fig. 8.
bining the two systems, and in effect allowing the fuel cell to
In addition, smaller systems not incorporating a steam tur- serve
as the combustor for the gas turbine and the gas turbine
bine are ideal for the distributed power and on-site markets to
serve as the BOP for the fuel cells, the combined efficiency is
raised to the 58 to 63% range even at sizes of less than 3 to 10

MW,
and NOx emissions are essentially eliminated. The capital
cost
of the combined system is markedly reduced relative to the
CH4 cost
of a stand-alone fuel cell power plant of that size and is
equal
to or less than that of a gas turbine power plant of that
A Combination size.
The combined efficiency is much higher than either stand-
Fuel alone
plant of either technology.
cell
If the early efforts are successful in commercializing these
Air B
combination cycle products, the foundation will be laid for
scaling up the technology to large-scale powerplants. This is

Stack
important, in that the combination at the scale of 200 MW or
more
can achieve efficiencies of 75% or more. This is signifi-
C C T
cantly higher relative to other

technologies for generating
Gas turbines
electricity from natural gas, and as a result, has the potential
to
significantly reduce carbon dioxide emissions. In compari-
Intercooler son,
the best currently available, large-scale, gas-fired, com-
bined
cycle power plants have an efficiency of about 58%.
Recuperator That
level will likely increase to 60 to 62% over the next de-
T
cade.
The highest efficiencies currently projected for several
fuel
cell technologies, which are now under development, are
in
the range of 55 to 65% for stand-alone fuel cell power
Figure 8. Gas turbine/fuel cell.
plants.
----------------------- Page 157-----------------------
Table 2. Potential Power Plants
Product Efficiency
Mature Price
Vendor Size (MW) (LHV) Fuel Cell
Gas Turbine Availability Target ($/kW)
Westinghouse 3 61 Pressurized
Heron 1999 Prototype 1200–800
5 69 Tubular SOFC
Heron
10 60 One 1.8 MW unit
Allison
Two 1.8 MW units
Multiple 1.8 MW units
Solar 1–2 58–63 Pressurized 4 :
1 PR 1999 Prototype #650
Planar SOFC
Turbo-expander
ERC 3.3 65 Direct MCFC
Steam bottoming gas 1250
3.8 68
turbine topping and
steam bottoming
Ztek 0.2 Pressurized 50–
100 kW 1998 Prototype 1000–1500
50 Planar SOFC
Allison 10–25 59–62 Unspecific
Allison 501 KB/KM or ATS engine in 425–450 for
ATS 1998–2002 engine only
THE WORLD POWER MARKET tentially

more cost-effective than the current technologies in
the
baseload market segment.
Fuel cell technology is expected to play a role in the world Some
utilities consider that the success of fuel cells and
power market. By the year 2010, it is estimated that approxi- some
other technology hinges on the emergence of dispersed
mately 130 gigawatts (GW) of new generating capacity will power
generation. Dispersed power generation is one of the
be installed in the United States, while in world markets and phenomena
accompanying the deregulation or disruption of
within a much closer timeframe, nearly 550 GW of generating the
electric power industry. Hence, fuel cells are viewed by
capacity will be added (41). Fuel cell commercialization oppor- some as a
disruptive technology since it is helping ‘‘introduce
tunities in the U.S. market are focused in several areas: re- customer
choice’’ and offers a set of attributes suitable for dis-
powering, central power plants, industrial generators, and persed
power generation.
commercial/residential generators.
Deregulation of the electric industry is about capturing
The worldwide market for additional electric generation system
economies and efficiencies down to a point where the
capacity dwarfs the domestic market. Nearly 550,000 MW of
payout/return is not worth the investment/trouble. Self-dis-
new capacity will be added by 2002. Estimates of plant repow- patching
of fuel cells in the deregulated industry would be
ering installations between 1999 and 2010 range from 15% to done to
minimize cost or maximize profit—that is, make the
approximately 65% of the installed generating capacity. Most most
money or save the most money. However, economics
repowering will occur in central power plants: Fuel cell instal- cannot
control decisions such as frequency control, voltage
lations of 100 MW or more are targeted to this market, pow- control,
and spinning reserve since decision-making takes too
ered initially by natural gas and later by coal gas. long.
These control decisions will probably not be economic
New generating capacity of approximately 100 GW will be ones.
required in the central powering market by 2010. Coal gas- Fuel
cells should be able to capture economies in a deregu-
powered fuel cell power plants are targeted to this market, lated
industry. The more aggressive, nonpassive decision-
with plants sized at 100 MW or more. making
which will accompany deregulation will lead to oppor-
The market for additional industrial capacity by 2010 is tunities
for fuel cells. However, utilities need help in de-
estimated at 3 GW, and the market for additional com-
termining
where fuel cells would benefit them; passive deci-
mercial/residential capacity is estimated at 6 GW. These
sion-
making by utilities, not looking at other economic
markets are targeted for early entry and will be a proving
alternatives, just going ahead and doing the standard substa-

ground for natural-gas fuel cell power plants sized from 500
tion
upgrade—trashing power quality and raising costs for all
kW to 20 MW.
customers
—hurts fuel cells and other new technologies.
Technologies for the distributed power and cogeneration

THE CHANGING FACE OF ELECTRICITY GENERATION market
segment include gas turbines, diesel engines, hydro-
electric
plants, solar and wind generation, and already com-
Fuel cell power plants should provide a significant share of
mercialized PAFC. In this market, MCFC and SOFC plants
our electrical power in this decade and well into the next cen- also hold
distinct advantages: The smaller applications favor
tury (42–45). They are set to play a major role in a deregu- fuel
cells for their high-efficiency, low-emission, and load-fol-
lated power industry. Large-scale plants will compete in the lowing
capabilities. In addition, the attractiveness of economi-
baseload power generation market, while smaller plants will cal and
reliable on-site power generation may significantly ex-
penetrate the distributed power and cogeneration markets. pand the
market for small-scale commercial and industrial
Baseload generation currently relies on coal-fired, nuclear, power
plants. The Clean Air Act mandates significantly re-
or natural-gas-fired technologies. The natural-gas-fired fuel duced
emissions of sulfur and nitrogen compounds from ex-
cell is more efficient, more environmentally friendly, and po- isting
power plants and sets strict limits on emissions from
----------------------- Page 158-----------------------
new sources. In the short term, these restrictions may encour- about
exposure to electromagnetic fields from high-voltage
age the use of underutilized fuels, particularly natural gas, lines.
by electric power producers.
Smaller-scale distributed configuration power plants are
perfect
for commercial buildings, prisons, factories, hospitals,
telephone switching facilities, hotels, schools, and other facili-

SPECIFIC FUEL CELL ATTRIBUTES ties. In
these applications, consumers get the best of all
AND MARKET APPLICATIONS worlds:
high-quality power that is economical and reliable.
On-site
power conditioning eliminates the voltage spikes and
Fuel cells have many attributes which make them suitable harmonic
distortion typical of utility grid power, making fuel
for distributed generation applications (42). These include low cell
power plants suitable even for sensitive electronic loads
emissions, high efficiency, production of high-grade waste like
computers and hospital equipment; and in many cases,
heat, modularity, reliability, unmanned operation, and fuel utility
grid backup reduces the need for expensive UPS
flexibility, to name a few. systems.
Increasing power generation without increasing emissions Fuel

cells promise to be one of the most reliable, if not
is the challenge facing power producers today, and fuel cells the most
reliable, power generation technology. They are now
are a key approach to balancing our energy needs with our being
used by hospitals, hotels, and telephone companies as
desire for a cleaner, healthier environment. Fuel cell power part of
critical UPS systems.
plants produce dramatically fewer emissions; and their by-
Unmanned fuel cell operation may mean big savings in
products, primarily water and carbon dioxide, are so environ- some
applications. This is especially true for dangerous and
mentally friendly that natural-gas fuel cell power plants have
metropolitan areas. Fuel cell designs with small footprints
a blanket exemption from regulations in California’s South and easy
installation are a must in cities. The footprint of the
Coast Air Quality Management District—possibly the strict- fuel
cell is currently higher than that of turbines.
est in the nation. Fuel
cells need hydrogen, which can be generated inter-
Fuel cells convert a remarkably high proportion of the nally
from natural gas, coal gas, methanol landfill gas, or
chemical energy in fuel to electricity. Even without cogenera- other
fuels containing hydrocarbons. Although most market-
tion, fuel cell power plants promise to be nearly twice as effi- entry
fuel cell plants are fueled by natural gas, fuel flexibility
cient as conventional power plants, and efficiency is not a means
that power generation can be assured even when the
primary
fuel source is unavailable.
function of plant size or load: Small-scale fuel cell plants are
Potential customers have also identified premium power,

just as efficient as large ones, and operation at partial load is
grid
support, voltage control, reliability improvement, VAR
as efficient as at full load. Higher efficiencies mean fuel sav-
control,
frequency control (fuel cell is a smart transformer),
ings for the producer and cost savings for the consumer.
spinning
reserve, incremental (modular) load growth (small
High-grade waste heat from fuel cell systems is perfect for
incremental cost), emission offset, transmission and distribu-

use in commercial, industrial, and residential applications,
tion
(T&D) deferral, and customer retention as uses for the
including cogeneration, heating, and air-conditioning. When
fuel
cell. Fuel cell’s value is dependent on ‘‘what it does,
by-product heat is used, the total energy efficiency of fuel cell
where it
does it, and when it does it.’’
systems approaches 85%.
The
ideal fuel cell application would be for a new prison,
The fuel cell stack is the basic component of a fuel cell
hospital, or orphanage, that is, something with bed and re-

power plant. Stacks are combined into modules, and plant ca-
quiring
heat, electricity, and a UPS in an area with no
pacity is determined by the number of modules. Individual
T&D
infrastructure, so credit could be given for deferment of
modules can go from idle to full load in minutes. Modular a
substation upgrade; the fuel cell could be owned and oper-
plants can help planners overcome many difficult expansion ated by
a ‘‘utility entity’’ having a distributorship taking cred-
problems. Mass-assembly construction techniques and its for
environmental benefits. The point is that the right ap-
shorter lead times for installation reduce the capital risk in
plication needs to be a high-value application which can take
adding generating capacity. Capacity can be better matched credit
for many quantifiable benefits.
to load, and the high costs of large new plants with underuti- The
building application is potentially important for fuel
lized capacity can be avoided. cells.
It is obvious that fuel cells can compete only on a life-
Modularity also produces a flat economy of scale: The cost cycle
cost basis. The longer the life, the better. Building oper-
per kilowatt is about the same in small plants as in large ators do
not want to get into the power business. One oppor-
ones; and because electrical efficiency is determined by indi- tunity
is to own and lease power plants for the operators. New
vidual cell performance, the number of modules in the power
ownership modes need to be explored. The retrofit market is
plant has little or no effect on overall efficiency. As a result,
important for buildings. Developers must be patient. It will
fuel cell power plants offer the same advantages at 25 kW as take
many years for fuel cells to penetrate any market.
they do at 50 MW.
The modular nature of fuel cells allows power capacity to
be added wherever it is needed. In the typical central power
INTERNATIONAL MARKET APPLICATIONS
configuration, additional capacity is sited at the central plant
or at substations. In a distributed power configuration, capac- Some
countries, such as Japan and the European community,
ity is placed close to the demand. In high-growth or remote are
firmly committed to the entire gamut of fuel cell develop-
areas, distributed placement offsets the high costs of acquir- ment, as
is evidenced by their funding commitment from both
ing rights-of-way and installing transmission and distribution
government and private industry, extensive research facili-
lines. A distributed configuration also eases public concerns ties,
and commitment of personnel (46).
----------------------- Page 159-----------------------
Large companies, such as Fuji, Mitsubishi Electric Com-

14. E. Ray and S. Veyo, Tubular solid oxide fuel cell demonstration
pany, Mitsubishi Heavy Industries, Toshiba, and Hitachi are
activities, Proc. ’94 Contractor’s Rev. Meet., DOE/METC-94-1010,
involved in fuel cell development in Japan. This contrasts
1994, pp. 9–14.
with the situation in the United States, where several of the
15. E. Ray and S. Veyo, Tubular solid oxide fuel cell demonstration
fuel cell developers are smaller, entrepreneurial companies.
activities, Proc. ’94 Contractor’s Rev. Meet., DOE/METC 94/1010,
The capital investment required for establishing manufactur-
1994, pp. 9–14.
ing and distribution infrastructure should be easier to obtain
16. S. Veyo et al., A solid oxide fuel cell power system—1992/93 field
for these large Japanese corporations than it will be for the
operation, 56th Annu. Meet. Amer. Power Conf., Chicago, 1994.
small U.S. corporations.
17. L. G. Marianowski et al., Fully internal manifolded fuel cell
stack, U.S. Patent 5,227,256, 1993.

Japan’s electric industry has very high capital and op-
erating costs and might be an excellent place to introduce
18. K. Krist and J. D. Wright, in S. C. Singhal and H. Iwahara (eds.),
Foundation methods for reduced temperature solid oxide cells,
new, higher-cost technologies like the PAFC. This high cost
Proceedings of the Third International Symposium on SOFC’s, The

may be attributable, in part, to both high fuel costs and inef-
Elec. Chem. Soc., 1993, pp. 782–791.

ficiencies in the Japanese goods and services distribution sys-
19. E. A. Barringer, A. Khandar, and A. Yagiela, Development and

tem. Some Japanese companies favor the deregulation of the
Demonstration of a 50-kW Planar SOFC Power Generation System,
electric industry, which is occurring in the United States and
a white paper for EPRI, GRI, and Utilities, Pennington, NJ, 1994.
promises to lower electricity costs.
20. R. Privette et al., Status of SOFCo SOFC technology develop-
There is a PAFC unit in Japan at a Nippon Telegraph and
ment, Proc. Fuel Cell Semin., Orlando, FL, 1996, pp. 40–43.
Telephone switchboard in Yokohama, Japan. The waste heat
21. M. Hsu, D. Nathanson, and E. Hoag, Ztek’s radiant thermal inte-
is also used to cool the equipment, using absorption chillers.
gration program for efficient and cost-effective electric utility
This is a large market in Japan, where 0.5% of electricity is
powerplants, EPRI/GRI Fuel Cell Workshop Fuel Cell Technol.
consumed by telephone switchboards. Also, Tokyo Gas will be
Res. Develop., 1994, Vol. 20.
selling heat-recovery-type absorption chillers as a future
22. M. Hsu et al., Ztek’s ultra-high efficiency fuel cell/gas turbine
product.
system for distributed generation, Proc. Fuel Cell Semin., Or-
Due to low capital costs, Tokyo Gas considers the natural
lando, FL, 1996, pp. 183–186.
gas engine to be the primary competition of the PAFC, espe-
23. M. Petrick, T. Cable, and R. Ruhl, Stack development status of
cially in peak shaving and UPS applications.
the Interscience Radial Flow (IRF) SOFC, EPRI/GRI Fuel Cell
Workshop Fuel Cell Technol. Res. Develop., 1994, Vol. 19.
24. N. Minh, W. Wentzel, and R. Gibson, Tape-calendaring mono-

BIBLIOGRAPHY
lithic and flat plate SOFC’s, in S. C. Singhal and H. Iwahara
(eds.), Proceedings of the Third International Symposium on
SOFC’s, The Elec. Chem. Soc., 1993, pp. 801–809.

1. Fuel Cells—A Handbook (Revision 3), Report Number DOE/
25. N. Minh et al., Allied signal SOFC technology, Proc. Fuel Cell
METC-94/1006, US Department of Energy, 1994.
Semin., Orlando, FL, 1996, pp. 40–43.

2. M. C. Williams and T. J. George, The developmental status of
26. J. Wimer and M. C. Williams, MCFC Networks—Principles, Anal-

coal-fueled molten carbonate fuel cell powerplants, Proc. 25th
ysis and Performance, DOE/METC-93/4112, U.S. DOE/METC,

IECEC, 3: 185–192, 1990.
1993, pp. 1–44.
3. M. C. Williams and T. J. George, Cathode corrosion in MCFC’s,
27. J. Wimer, M. C. Williams, and D. Archer, Networking of EMS
Proc. 26th IECEC, 3: 40–51, 1991.
devices, AIChE Annu. Meet., Miami, FL, 1992, Abstr. Ext. Abstr.,
4. H. Maru, ERC Commercialization activities, Proc. ’95 Contractor’s

pp. 59–60; Pap. Prepr., 1st Sep. Div. Top. Conf. Sep. Technol.: New
Rev. Meet., DOE/METC-95/1020, 1995, pp. 151–155.
Develop. Opportunities, October 1992, pp. 976–983.
5. A. Leo, Status of the Santa Clara MCFC PDT, Proc. ’95 Contrac-
28. J. Wimer, M. C. Williams, and D. Archer, EMS device net-
tor’s Rev. Meet., DOE/METC 95/1020, 1995, pp. 165–171.
working, in Gas Purif. Sep., 22: 121–124, 1993.
6. H. Maru et al., ERC product improvement activities, Proc. ’95
29. M. C. Williams and J. Wimer, Mathematical modeling of MCFC
Contractor’s Rev. Meet., DOE/METC 95/1020, 1995, pp. 178–188.
cells/stacks and networks, Prepr. Pap., 206th Amer. Chem. Soc.
Natl. Meet., Chicago, 38 (4): 1435–1440, 1993.

7. J. Scroppo et al., IMHEX fuel cells progress toward commercial-
ization, Proc. ’95 Contractor’s Rev. Meet., DOE/METC 95/1020,
30. M. C. Williams et al., System and method for networking electro-
1995, pp. 156–164.
chemical devices, U.S. Patent 5,413,878, 1995.
31. P. Micheli, M. Williams, and E. Parsons, Indirect-fired MCFC-

8. J. Scroppo et al., MCFC PDT at SDG&E, Proc. ’95 Contractor’s
bottomed cycle, U.S. Patent 5,449,568, 1995.

Rev. Meet., DOE/METC 95/1020, 1995, pp. 172–177.
32. M. Williams and C. Zeh, Proc. Workshop Very High Effic. FC/GT
9. J. Scroppo et al., M-C Power’s PDI, Proc. ’95 Contractor’s Rev.
Power Cycles, 1995.

Meet., DOE/METC 95/1020, 1995, pp. 189–196.
33. D. Archer, M. Williams, and J. Wimer, FC/GT systems—power

10. Energy Research Corporation, Molten Carbonate Fuel Cell Devel-
generation with heat and fuel recovery, Proc. Workshop Very High
opment, Final Report to U.S. DOE under Contract No. DE-AC21-
Effic. FC/GT Power Cycles, 1995, pp. 33–40.
76ET11304, Danbury, CT, 1987.
34. M. C. Williams, E. L. Parsons, and P. Micheli, Engineering a 70%

11. R. Selman, Y. Yazici, and Y. Izaki, NiO cathode dissolution in the
efficient IFCFC, Proc. 9th Fuel Cell Contractor’s Meet., 1995,
MCFC: A review. Prepr. Pap., 206th Amer. Chem. Soc. Natl. Meet.,
Poster 10, pp. 121–129.
Chicago, 38: 1429–1434, 1993.
35. P. Micheli, M. Williams, and F. Sudhoff, Dual fuel cell tandem

12. E. Ray, Tubular solid oxide fuel cell development program, Proc.
cycles, U.S. Patent 5,541,014, 1994.
’94 Contractor’s Rev. Meet., DOE/METC-94/1010, 1994, pp. 3–8.
36. D. Archer, J. Wimer, and M. Williams, Power generation by com-
13. E. Ray, Tubular solid oxide fuel cell development program, Proc.
bined fuel cell and gas turbine systems, Proc. 10th Fuel Cell Con-
’95 Contractor’s Rev. Meet., DOE/METC 95/1020, 1995, pp. 9–14.
tractor’s Meet., 1996.
----------------------- Page 160-----------------------
FUNCTIONAL AND SMART MATERIALS 11
37. M. C. Williams, E. L. Parsons, and P. Micheli, Configuration and

performance of the IFCFC, J. Eng. Gas Turbines Power, 13, 1996.
38. D. Archer, J. Wimer, and M. C. Williams, Power generation by
combined fuel cell and gas turbine systems, IECEC Proc., 2:
1117–1122, 1996.
39. L. Rath, P. Lee, and F. Sudhoff, Configuration and performance
of fuel cell combined cycle options, Proc. Workshop Very High Ef-
fic. FC/GT Power Cycles, 1995, pp. 21–32.
40. M. Williams and C. Zeh (eds.), Proceedings of the Second Work-
shop on Very High Efficiency FC/GT Power Cycles, US Depart-
ment of Energy, 1996.
41. EIA, 1996 Energy Outlook, Washington DC, 1996.
42. M. C. Williams, Fuel cell market applications, Proc. 11th U.S./
Korean Workshop, 1995, pp. 259–276.
43. M. C. Williams, Distributed generation, GRI Workshop Distrib-
uted Generation, San Diego, CA, 1995.
44. M. C. Williams, Status of and marketing opportunities for fuel
cells, Fuel Cell Technol. Forum, Market Opportunity Panel, Santa
Clara, CA, 1995.
45. M. C. Williams, Status of and marketing opportunities for fuel
cells, Fuel Cell Technol. Forum, Market Opportunity Panel, Pasa-
dena, CA, 1995.
46. M. C. Williams, Stationary power FC commercialization status
worldwide, Proc. Fuel Cell Seminar, Orlando, FL, 1996, Abstr.,
pp. 1–3
MARK C. WILLIAMS
US Department of Energy
----------------------- Page 161-----------------------
} { { } }
HOME
●
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Fusion Reactors

Standard Article

Larry R. Grisham1
1Princeton University, Princeton, NJ

reserved.
Search
DOI: 10.1002/047134608X.W5208
❍ Acronym Finder
Abstract
Nuclear Fusion
Fusion Reactor Fuels
Possible Types of Confinement for Fusion Reactors
Magnetic Confinement Fusion
Major Components of a Tokamak Fusion Reactor
Inertial Confinement
Progress in Fusion Energy

%20Conversion/W5208.htm17.06.2008 16:39:11
----------------------- Page 162-----------------------
FUSION REACTORS gives rise

to a potential barrier, referred to as a coulomb
barrier,
which the approaching nuclei must overcome in or-
der to
merge. Due to quantum mechanical tunneling, some
NUCLEAR REACTORS, FUSION pairs of
nuclei can fuse even when their combined kinetic
energy is
less than that required to exceed the coulomb
barrier.
However, the likelihood of this tunneling occurring
REACTORS, FUSION declines
very rapidly as the kinetic energy of the reactants
falls
farther below the Coulomb barrier height. For most
nuclear
reactants, the reaction rate drops to inconsequen-
tial
levels if the approaching nuclei do not have kinetic
NUCLEAR FUSION energies of
at least 70% to 80% of the barrier height. For
heavier
nuclei, the height of the Coulomb barrier becomes
The fission plants which now provide about 20% of the several to
many millions of electron volts (MeV ), where an
world’s electricity accomplish this feat through neutron- electron
volt is a unit of energy corresponding to the energy
driven chain reactions in which heavy atomic nuclei, such one
electron gains when it falls through a potential differ-
as U235 , split apart into lighter nuclei, releasing large ence of 1
V. A temperature of 1000 eV (1 keV) corresponds,
amounts of energy (on the order of 180 million eV) in the in more
conventional units, to about 10,600,000◦ C.
process. Beginning about 1951, when one of the first fusion Thus, in
order to fuse most of the elements in the peri-
programs was begun at Princeton University by Lyman odic table,
one would need to cause them to approach each
Spitzer, Jr., many of the world’s developed nations, as well other at
energies of several to many MeV. This can be ac-
as some developing ones, have pursued research to even- complished
with particle accelerators for nuclear physics
tually produce a fundamentally different type of nuclear research
purposes, but is not practical for producing net
reactor. Since 1958, when research on peaceful uses of nu- power with
significant quantities of thermal reactants. As
clear energy was declassified, these many countries have a
consequence, only the lightest elements in the periodic
shared their research and openly collaborated, even during table have
sufficiently large probabilities (also called reac-
decades when some of the principal contributing nations tion cross
sections) of undergoing nuclear fusion for them
were political adversaries. This new type of reactor is a fu- to be
considered as fuels for fusion reactors.
sion reactor, and the very great progress which has been Fusion
reactions drive the core of the sun at a temper-
made in the decades since 1951 stands as a landmark to ature
thought to be roughly 10 million to 11 million ◦ C.
what may be the most universal and long-lived collabora- Fusion also
powers all the other stars, and thus supplies
tive effort in human history. most of the
universe’s observable energy. Our sun, in com-
Fusion reactors, like fission reactors, will use exother- mon with
other main sequence stars, obtains its energy
mic nuclear reactions to release energy. The fusion reac- through a
number of nuclear fusion reactions, beginning
tions themselves, however, are very different in character with the
fusion of 2 protons into deuterium. The deuteron
than are those in fission reactors, and they require entirely is
accompanied by a positron and a neutrino, and together
different conditions in order to proceed at an acceptably they carry
1.44 MeV of energy. The reaction probabilities
high rate for power plant applications. Due to these differ- (or cross
sections) for this and the succeeding solar nuclear
ences, fusion reactor designs will look very different from reactions
are much too small to be of any use to a commer-
those of fission reactors. Moreover, many of the problems cial fusion
reactor on the earth’s surface. The sun is able to
that have had to be surmounted in the pursuit of fusion produce
enormous amounts of energy with these reactions
power are of a fundamentally different nature than those only
because it is very large compared to the earth. The
encountered in making fission reactors practical. sun has a
diameter of about 897,000 mi and, because of its
Nuclear fusion is a reaction in which two atomic nu- size and
its gravity, has excellent energy confinement. The
clei merge to form a heavier element. If the reaction is time for
energy produced in the sun’s fusion driven core to
an exothermic one, then the fusion process will result in reach the
sun’s surface is estimated to be of the order of
the release of energy. This energy is carried as kinetic en- thousands
of years or more.
ergy by the reaction products, consisting of the product nu- Because
a practical fusion reactor for electricity produc-
cleus (which is positively charged) and another particle, tion needs
to be much smaller than the sun, it will have
such as a neutron (which is electrically neutral) or a pro- to rely
upon different nuclear fusion reactions with larger
ton (which is positively charged). Because all nuclei carry cross
sections, and it will also need to operate at temper-
positive electric charge, they repel each other, with the re- atures that
are at least 10 to 20 times those of the sun’s
sult that fusion reactions can proceed at significant rates core.
Temperatures in this range were achieved within a
only at very high temperatures which give the nuclei suf- mirror
confinement device in 1979 and within the Toka-
ficient energy to overcome their repulsion and approach mak Fusion
Test Reactor (TFTR) at Princeton University
each other close enough to merge. Alternatively, this same in the mid-
1980s, and have now also been obtained on the
end may be achievable by using somewhat lower temper- Joint
European Torus (JET) tokamak in England, and the
atures, but at high densities achieved by compressing the JT-60U
tokamak of Japan. Ion temperatures as high as
fusion fuel with high symmetrically applied pressures. 300 million
to 500 million ◦ C (roughly 30 to 50 times the
The electrostatic repulsive force between two nuclei is temperature

of the sun’s core) were routinely produced in
proportional to the product of the positive charges (and Princeton
University’s TFTR over a period of years.
thus to the atomic numbers) of the two reactant nuclei. This

Copyright © 2007 John Wiley & Sons, Inc.
----------------------- Page 163-----------------------
2 Reactors, Fusion
FUSION REACTOR FUELS jor

factors: the fusion reactivity is much higher at tem-
peratures which can be reached by some of today’s fusion

There are a number of nuclear fusion reactions which have devices,
and the radiation loss rates from bremsstrahlung
cross sections sufficiently large to be potential candidates and
synchrotron radiation are lower for likely reactor con-
as fuels for commercial fusion reactors. These are listed ditions.
Synchrotron radiation would be less important for
below, where D represents a deuteron, the nucleus of 2 H, T
alternate plasma confinement schemes which do not use
stands for a triton, the nucleus of 3 H, and the energies in strong
magnetic fields, but these approaches have not been
parentheses are the amounts of kinetic energy each of the

developed as far as the strong field path to fusion.
reaction products carries away from a fusion event. An n
stands for a neutron, and a p for a proton, the two types of
Deuterium Fuel
baryons which occur in atomic nuclei.
Deuterium is a stable, naturally occurring isotope of hydro-

gen. On
the earth, one out of every 6700 atoms of hydrogen
is
deuterium. Thus, enormous supplies of deuterium are
available from the earth’s water. If later generations of re-

actors
operate with the D + D reaction, then there is enough
fuel to
supply the world’s energy needs for hundreds of
millions
to billions of years, depending upon assumptions
about
the future growth of energy usage. The deuterium
Of these reactions, the D + T reaction has by far the
can be
concentrated and extracted from any water by uti-
largest cross section at energies of tens of keV, which is
lizing
such enrichment techniques as diffusion through a
the region which should be obtainable in the first gener-
series
of filters or by electrolysis.
ation of commercial fusion reactors. For a fusion reactor
The
earth’s water contains about 4 × 1016 kg of deu-
with a thermal reacting population, the parameter which
terium.
If this were used to fuel D + T fusion reactors with
describes the fusion reaction rate is #σv#, where σ is the nu-
an
overall operating efficiency of 33%, then this would al-
clear reaction cross section, and v is the relative speed of
22
low the
production of about 10 GJ of electricity, which
the reacting nuclei. The angle brackets mean that the prod-
11
is close
to 3 × 10 the present annual electricity produc-
uct of σv is averaged over the Maxwellian velocity distri-
tion of
the entire world. As we will see later in this section,
butions of the reactants. This fusion reactivity parameter
the
real limit on the amount of energy potentially avail-
is also much larger for the D + T reaction than for the other
able
from D + T fusion reactors is the supply of feedstock
possible fuel mixtures. Due to the importance of the reac-
to
produce the tritium used in the reaction.
tions arising from the high energy tail of the Maxwellian
If
later reactors used D + D as their primary reaction,
velocity distributions, the optimum range for ignition of a
then,
because this reaction is less exothermic that the D +
D + T fuel mix lies across a range of about 20 keV to 40
T
reaction, and because two deuterium atoms, instead of
keV, well within the operating range of some of the largest
one,
would be required for each reaction, then the estimate
of existing fusion devices. These experimental devices have
of
total electricity production if all the deuterium in the
not actually reached ignition because, although their tem-
21
waters
of the world were used would drop to about 10
peratures were adequate, the product of the density and
GJ,
which is still a very large number, and is equivalent to
the energy confinement time was not yet sufficiently large. about 3
× 1010 times the world’s current annual production
The energy confinement time is a measure of how long is
of
electricity. The cost of deuterium is of the rough order of
required for energy to leak from the plasma. In the 1950s
one
dollar per gram (1), with a gram of deuterium being
a typical experimental device had an energy confinement
sufficient to produce 300 GJ of electricity if fusioned with

time of a few milliseconds for the overall plasma; by the
tritium.
The cost of the deuterium fuel for a D + T reactor
mid-1990s this reached values of as much as 1.4 s.
is thus
about $0.003 per gigajoule of electricity. If the deu-
In addition to the magnitude of the reactivity as a func-
terium
were instead used in a D + D reactor, the cost would
tion of temperature, there are also other factors which bear
rise to
about $0.02 per gigajoule of electricity, which is still
upon the ease with which different fusion fuels can be used
small
compared to the cost of bulk electricity, which runs
in a commercially viable reactor. There are two types of ra-
in the
vicinity of $20 per gigajoule. The fuel costs would
diation loss which take energy from the confined fuel and
be a
negligible portion of the cost of electricity from a fu-
transport it directly to the outside, and both of these tend
sion
plant, which would be dominated by capital costs and
to be more serious for higher temperatures. The first of
maintenance. It is likely that the price of electricity from

these is bremsstrahlung, which scales in intensity propor-
a fusion
reactor would be similar to or perhaps somewhat
tional to the square of the density times the square root of
more
expensive than electricity from a fission reactor, at
the temperature. For D + T fueled reactors, this would be
least
in the near term, while there is still adequate fuel
a relatively tolerable portion of the overall power balance.
for the
simple once-through fuel cycles used in most fission
However, for the other fuels it would play a larger role. For
power
plants.
approaches to fusion reactors which employ strong mag-
netic fields, the electron synchrotron radiation becomes a
Comparison to Fossil Fuel Energy Densities
major factor at the higher temperatures (60 keV to 120
keV) required by reactions involving D + D or D + 3 He.
Producing a gigawatt (109 W) of electricity for a year in a
Thus, the appeal of a D + T fuel mix in at least the early fusion

reactor would require roughly 1000 kg of deuterium.
electricity-producing fusion reactors arises from two ma-
Producing the same amount of electrical energy from a
----------------------- Page 164-----------------------
Reactors, Fusion 3
power plant burning coal would require about 2 × 109 kg of or more)

are taken into the body, it is not very likely to pro-
carbon. The actual weight of coal required would of course duce ill
effects. This is the result of two factors. The first is
be greater than this, since coal contains other elements be- that the
ionizing radiation released into the body by each
sides carbon. The fact that the mass of fuel which has to tritium
decay is much less than the decay energies of fission
be carried from the fuel concentration source to a fusion products
(which are typically at least hundreds of keV) or
power plant is more than a factor of a million smaller than the
energies in the decay chains of heavy elements, which
for competing fossil fuel plants is a significant advantage. can run to
over 10 MeV). The second factor is that most of
It means that moving the fuel for an entire fusion econ- any tritium
absorbed by the body would enter as water, and
omy would impose no requirements upon the transporta- water is
continuously excreted, with a biological half-life in
tion infrastructure, since the masses being moved would human
bodies of about 12 to 13 days. The absorption rate
be thousands of kilograms instead of billions of kilograms. for tritium
breathed into the lungs as gaseous molecular
The fact that fusion reactors will be able to produce a hydrogen is
very low, roughly 20,000 to 25,000 times less
gigawatt-year of electricity while using fuel which weighs than for
tritium in water molecules.
more than a million times less than that required for a Thus,
while tritium should always be treated with due
fossil fuel burning plant reflects the large difference be- care, its
possible health effects if mishandled do not appear
tween nuclear binding energies and electron binding ener- to be
significantly worse than those of many other chemi-
gies. When a reaction occurs between two parent particles, cals
routinely handled by an industrialized society. Indeed,
the extra kinetic energy carried by the daughter particles tritium is
already used in conjunction with phosphors to
comes from the change in the overall binding energy. This provide
light without the need of electricity in several ap-
kinetic energy is distributed among the daughters in ac- plications
such as school exit lights, some airfield land-
cordance with the laws of energy and momentum conser- ing
lights, and some modern illuminated watches. These
vation, with the result that the lighter daughter carries
applications incorporate significant amounts of tritium.
most of the kinetic energy. Tritium-
powered school exit lights typically use about 15 to
Burning a fossil fuel is a chemical reaction, which in- 25 curies
of tritium, and emergency runway landing lights
volves exchanges and rearrangements of the electrons out- use much
more. This compares with an amount on the or-
side the nuclei. The outer electrons of atoms typically have der of 100
to 150 curies injected into the Tokamak Fusion
binding energies of several to a few tens of electron volts. Test
Reactor at Princeton University for a high-power fu-
Thus, the changes in the net binding energy that occur in sion shot.
This device has operated on a routine daily basis
chemical reactions, such as burning fossil fuels, are typi- using
deuterium and tritium for experiments over a period
cally only a few electron volts. Nuclear binding energies, spanning
three and a half years without any significant
on the other hand, are typically many millions of electron incidents
involving tritium contamination of personnel or
volts. This has the consequence that the changes in binding the
environment. This demonstrates that tritium can be
energy involved in nuclear reactions are also much greater handled on
a large fusion system without unduly imped-
than for chemical reactions. For instance, in the D + T re- ing the
manner in which it is operated, although the total
action, 17.58 MeV of energy is released as kinetic energy quantities
of tritium being handled were orders of magni-
in inverse proportion to the masses of the daughters. Thus, tude lower
than would be required in a power plant. The
because a neutron weighs about one fourth as much as does neutron-
induced activation and damage of materials would
an 4 He nucleus, the neutron carries four times as much of be much
more challenging in a fusion power plant.
the kinetic energy. While

the beta decay of tritium does not result in either
The disparity in binding energies between nuclear and undue
hazard potential or in excessive constraints upon
chemical reactions arises in turn from the fact that the the
operation of fusion reactors, it does have an inconve-
strong nuclear force is much more powerful across nu- nient
consequence. The half-life (the time for half of the
clear dimensions than is the coulomb force (which binds nuclei in
any assemblage to undergo beta decay) is only
the negatively charged electrons to the positively charged 12.3 years.
Thus, any primordial tritium that fell into the
nucleus) across dimensions typical of atoms. Consequently, earth
during its formation decayed away billions of years
the huge difference in the magnitudes of potential energy ago. Some
tritium is continuously produced in the upper
available per unit of mass available from nuclear, as op- atmosphere
through nuclear reactions initiated by cosmic
posed to chemical reactions, arises from a major difference rays.
However, due to the short half-life of tritium it does
in the strengths of two of the fundamental forces of nature. not build
up, so the equilibrium concentration of tritium in
Thus, there is no possibility that any chemical reaction in- air is very
low, and far too small to economically extract as
volving changes in electron configurations could ever begin fuel for a
fusion fuel.
to approach the energy releases available from nuclear re-
Consequently, tritium fuel for fusion must be manu-
actions. factured
through nuclear reactions. The tritium used for
present-day
fusion experiments is made in heavy water
nuclear
fission reactors. However, the amount produced by
Tritium Fuel
these
reactors would be insufficient to fuel a fusion econ-
Tritium, the heaviest isotope of hydrogen, is unstable. It omy.
Equally important, the long-term goal of fusion re-
decays by emitting a beta particle (an electron) with an av- search is
to produce fusion reactors which supplant fission
erage energy of 5.7 keV. This beta particle is easily stopped reactors as
we know them, so fusion reactors must produce
by even a piece of paper, so it does not pose a hazard unless their own
tritium. Fortunately, this is feasible using reac-
it is ingested. Even then, unless large amounts (millicuries
----------------------- Page 165-----------------------
4 Reactors, Fusion
tions such as: fact, if

it appeared advisable to use an isotopically enriched
blanket,
it would be possible to do so with only a modest
effect
upon the price of electricity. The energetics are such
that the
electricity produced with a Li6 blanket might be
slightly
cheaper than with a natural Li blanket, because
the extra
4.8 MeV from the n + Li6 reaction is much larger
Tritium Production than the
amount of energy required to enrich the lithium.
The incident neutrons that induce these reactions arise The
United States has large reserves of easily ex-
from the D + T reactions, and perhaps also from other re- tractable
lithium, mostly in dry salt lakes and saline lakes,
actions initiated by the primary neutrons. The reaction in- which
could be produced at prices not greatly different from
volving 6 Li has the advantage that it is exothermic, adding the $20
per kg of today. This reserve is estimated to be
another 4.8 MeV of kinetic energy to the 17.58 MeV of ki- about 5
× 109 kg (1. With an electricity production effi-
netic energy released by the D + T reaction that produced ciency of
33%, this would yield about 3 × 1014 GJ of elec-
the neutron. Moreover, because it is exothermic, there is tricity,

an amount that is of the order of 800–1000 times
no threshold energy for initiation of the reaction, mean- the
primary energy consumption of the world in the early
ing that even low-energy neutrons which have undergone 1990s.
Inasmuch as the United States comprises only 6% of
the
earth’s land area, it is reasonable to assume that the to-
many elastic and inelastic collisions can still produce tri-
tium in this way. tal world
reserves of cheaply extractable lithium might be a
The reaction involving 7Li is less advantageous in one few times
greater than these United States reserves. Thus,
sense, in that it is endothermic, consuming 2.5 MeV of ki- the
reserves of easily obtainable lithium are sufficient to
netic energy in order to occur. This also means that only run an
economy powered by D + T fusion reactors for a pe-
energetic neutrons above about 3 MeV (after allowing for riod of
several to many centuries. There is a much greater
amount
(of order 103 ) of lithium dissolved in the world’s
center of mass effects among the reactants) can initiate
this reaction. On the other hand, the n + 7Li has the ad- oceans
than in land deposits (1). This could be extracted at
vantage that it does produce an additional neutron, which somewhat
greater cost than from the saline lakes and dry
6 salt beds.
However, this cost premium might be relatively
may initiate an n + Li reaction to produce still another
modest.
Dikes could be built across tidal flats to isolate
triton.
seawater
which could then be treated the same as saline
Thus, for a fusion reactor using deuterium and tri-
tium fuel, the raw material for the production of tritium lakes and
salt flats after it had undergone sufficient solar
is lithium. The natural abundances of these lithium iso-
evaporation.
topes on earth are 7.4% for 6 Li and 92.6% for 7Li. In or-
der to produce sufficient tritium to at least continuously Reduced

Activation Fuels
replenish its fuel supply, a deuterium tritium fusion re-
actor would be surrounded by a lithium blanket which The high-
energy neutrons produced by fusion reactors us-
would produce tritium and capture the kinetic energy of ing
deuterium or deuterium + tritium as fuels undergo nu-
the fusion neutrons and neutron-induced reaction prod- clear
reactions with the materials forming the structure
ucts. As these particles slow down by collisions with the of the
reactor. Some of these reactions result in the pro-
lithium blanket, their kinetic energy will be converted to duction
of radioactive nuclei with a variety of half-lives.
heat. This heat will, in turn, be used to produce steam to This
activation of structural components, and particularly
drive electricity-producing turbines. This tritium-breeding of those
components close to the fusion core, makes main-
blanket may also include other materials such as beryl- tenance
more difficult, and also will require at least short-
lium. Terrestrial beryllium is 100% 9 Be, which can act as term
storage of removed components, or possibly longer
a neutron multiplier primarily through the reaction: term,
depending (as will be discussed in a later section)
upon the
materials used to build the reactor. In addition,
the
energetic neutrons can introduce lattice defects as they
scatter,
which in turn reduce the lifetimes of components,
This reaction is modestly endothermic, requiring an input requiring
more frequent replacement.
of 1.67 MeV in order to occur, meaning that the thresh- In
order to reduce these deleterious effects of the neu-
old energy in the laboratory frame for incident neutrons is trons upon
fusion reactor structures, studies have been car-
around 2 MeV. ried out
to evaluate the feasibility of alternate fuels which
If, for instance, 7Li is used to make tritium (which is a would
produce fewer neutrons per unit of fusion energy
fairly good approximation to using natural lithium), then released.
These reactions include the D + 3 He → 4 He +
with a system efficiency of 33% for net electricity produc- p
reaction and the p + 11B → 3 4 He. As discussed ear-
tion, 1 kg of 7Li would be sufficient to produce 7 × 104 lier, the
first of these reactions requires higher tempera-
GJ of electricity from a D + T fusion power plant. At the tures

than are presently obtainable, and is more subject
present cost of roughly $20 per kg (1) for natural lithium, to
bremsstrahlung radiation losses, and is much more vul-
the lithium in a tritium breeding blanket would contribute nerable
to synchrotron radiation losses in systems with
only about $0.001 per GJ of electricity. This is negligible magnetic
fields than is a D + T fuel mixture. Due to lower
compared to the bulk price of electricity of about $20 per fusion
cross sections and the higher atomic number of 11B,
GJ. Thus, the price of breeding lithium could rise many-fold these
problems are more difficult for a reactor using the p
before it had a noticeable impact upon electricity costs. In + 11B
reaction.
----------------------- Page 166-----------------------
Reactors, Fusion 5
A reactor using the D + 3 He → 4 He + p reaction would on

earth due to the effects of weather, water, plate tectonic
have the advantage over a D + T reactor that both com-

movements of the land, and volcanoes. On the moon, how-
ponents of the fuel would be stable. It would still produce ever,
which lost these processes in the distant past, many
some 3.51 MeV neutrons from D + D reactions, and it would large
craters still stand from the later periods of the bom-
produce some 14.1 MeV neutrons from the burnup of tri-
bardment.
tium produced in one of the two branches of the D + D re- As
a result of these differences between the histories
action. However, almost all of the tritium produced would of
the lunar and terrestrial surfaces, much of the lunar
be consumed in D + T fusion reactions, and thus would not
surface has remained relatively undisturbed for perhaps
leave the reactor. If the temperature and fueling profiles of
billions of years. During all this time the solar wind has
such a reactor could be optimized appropriately, the neu-
continued to deposit 3 He into the lunar surface, so that it
tron production rate could be reduced by a factor of 100 now

exists there at a concentration of about 10 parts per
compared to a D + T reactor (2).
billion. A study of the feasibility and economics of collect-
Unlike all the other potential fusion fuels, which oc- ing
and concentrating the 3 He from the dust which covers
cur in abundant commercially attractive concentrations on much

of the lunar surface found that, even after includ-
or near the earth’s surface, 3 He is extremely rare in the ing
the high transportation costs of carrying equipment
earth’s crust. The small supplies that are available arise to the
moon and sending the 3 He back, this would be eco-
from the beta decay of tritium. This tritium, in turn, has to
nomically feasible if a commercially viable 3 He + D reactor
be bred through nuclear reactions of one sort or another. could

be developed (3). As discussed earlier, however, the
At the present time, all substantial production of tritium
parameters required for a 3 He + D reactor to be viable are
is carried out through neutron capture reactions on deu-

significantly more daunting than for a D + T reactor.
terium in heavy water moderated fission reactors. In the
Thus, although there are a number of light element
future, tritium might be bred through other nuclear reac-
isotopes which might someday be suitable as fuels for
tions either by fusion reactors or high-energy particle ac-
advanced fusion reactors, the remainder of this article
celerators. However, if one must first produce tritium in will
concentrate on concepts for fusion reactors fueled
order to obtain the 3 He fuel for a 3 He + D fusion reactor, by
deuterium concentrated from water, and by tritium
then much of the rationale for using 3 He disappears. which
would be produced by fusion-produced neutrons
What is needed in order that 3 He + D fusion reactors from
lithium in blankets surrounding fusion reactors.
have the possibility of commercial viability is a naturally

occurring source of 3 He in concentrations and quantities
sufficient for economic exploitation. As it happens, while

POSSIBLE TYPES OF CONFINEMENT FOR FUSION
3
no He deposits occur on earth, they do occur in abundance
REACTORS
on the surface of the moon.
Because temperatures much higher than those to which

Lunar 3 He we
are accustomed are required for any fusion reactor
with a
useful reaction rate, it is important that there be
Because the moon, unlike the earth, lacks both an atmo- some
method for keeping the reacting fusion fuel out of
sphere and a magnetic field, its surface is raked by the so- direct
contact with material objects which would quickly
lar wind, a flux of energetic particles driven outward from cool
the fuel to temperatures below which fusion reactions
the sun. Among other constituents, this wind carries 3 He, were
negligible. The principal mechanism responsible for
which embeds itself in the lunar surface when it strikes. this

rapid cooling would not be conduction into the in-
The moon, unlike the earth, does not have enough internal
truding material, but rather radiation losses from the fuel
heat to drive plate tectonic motions of its surface now, nor due to
enhanced bremsstrahlung and line radiation from
for a very long time into its past. For similar reasons, much atomic
transitions due to heavier impurities entering the
of its surface has not experienced fresh outpourings of lava fuel.
Thus, a reactor requires some sort of restraining force
or ash since the interior cooled. The early moon probably which
balances the outward pressure of the fusioning fuel,
possessed an atmosphere, but this long ago escaped into which,
since it is absorbing part of the energy it releases, is
space because of the moon’s weak gravity. Consequently, also
producing pressure. There are a number of restraining
there has been no weather and no water to erode and re- forces
one might imagine, not all of which are practical in
arrange the lunar surface in billions of years. When our a
power plant.
solar system was young, the moon, in common with the
earth, was subjected to an intense bombardment by large
and small fragments of matter left over from the formation
of the planets. The ejecta arising from this bombardment
Gravitational Fields
deposited new layers of material over much of the moon’s
surface, burying pre-existing surface layers. However, most
Gravitational fields produce suitably high restraining
of the solar system debris was swept out of the planetary forces
only for very large assemblages of mass, because the
orbits long ago through the direct interception of fragments
gravitational force is much weaker than the other known
by the planets and moons, and through deflections of the
fundamental forces of nature. Thus, while all the successful
orbits of fragments arising from close encounters with the fusion
reactors we can see, namely stars, use gravitational
gravitational fields of the much larger planets and moons. fields
for confinement, this is impractical for a commercial
The evidence of this early bombardment has eroded away
reactor by an enormous factor.
----------------------- Page 167-----------------------
6 Reactors, Fusion
Dc Electric Fields
approximately given by the time required for the hot ions,
with a mean
velocity of about 108 cm/s, to traverse the ra-
If the reacting fusion fuel possessed a strong net electri-
dial
dimension of the fuel, which is much less than 1 cm.
cal charge, then it might be possible to confine it with dc
Thus, the
confinement time is in the sub-nanosecond range.
electric fields. However, in the types of fuel assemblages so
The fusion
output is proportional to the product of the den-
far used in fusion research, any net imbalance in charge
sity, the
ion temperature, and the confinement time. This
which developed in the fuel was much too weak to permit
is more
conveniently referred to as nτT, where n is density,
confinement solely through the forces that could be applied
τ is
confinement time, and T is the ion temperature. For re-
through dc electric fields. Significant electric fields do de-
alistic ion
temperatures of a few tens of keV, the very short
velop under some conditions inside the fuel using the mag-
confinement
time of the inertial approach requires com-
netic field confinement approach discussed later, and they
pression of
the fuel to very high densities, 40 to 400 times
may play a role in altering the quality of the confinement
the normal
density of the solid deuterium and tritium fuel
when they do occur.
(4), in
order to produce fusion energy releases relevant to
a reactor.
Radio Frequency Electromagnetic Fields
At the high temperatures required for a practical fusion Confinement

by Magnetic Fields
reactor, matter exists not in any of the three states with An
alternate strategy for maximizing the nτT product of
which we have everyday experience: solid, liquid, or gas, density,
confinement, and ion temperature is to use a much
but rather in a fourth state known as a plasma. In this lower
density, but a correspondingly longer confinement
state, the negatively charged electrons have been stripped time. The
technique which most naturally fits this ap-
away from positively charged nuclei which they otherwise proach is
to place the plasma in a magnetic field. Since
normally encircle to form electrically neutral atoms. The the plasma
is composed of electrically charged particles,
plasma consists of unbound electrons and these nuclei, they are
constrained to move in helical paths along the
called ions. This has the result that both the ions and lines of
magnetic force, with the negatively charged elec-
the electrons are highly mobile, and can rapidly rearrange trons
spiraling in one direction, and the positively charged
themselves to counter external electric fields applied to ions in
the other direction. In the simplest instance of a
the plasma. Because electrons are much lighter than ions, uniform
magnetic field, and in the absence of collisions, a
they have far higher velocities if they are at a tempera- charged
particle remains tied to its line of force, although it
ture roughly similar to that of the ions. Thus, the higher is free to
move along it. Consequently, the overall effect of a
electron mobility normally accounts for most of the charge uniform
magnetic field is to restrict the outward motion of
rearrangement which takes place in a plasma to shield out particles
across magnetic lines of force, while leaving mo-
externally applied electric fields. Plasma will be discussed tion
parallel to the magnetic field unimpeded. For a plasma
in more detail later in this article. with a
pressure perpendicular to the magnetic field lines of
An approach which has been considered for confining a 100 kPa (1
atm) (an overall pressure that is of the general
fusioning plasma is to counter the outward pressure of the order
required for a fusion reactor), a countering perpen-
plasma with the inward radiation pressure of a radiofre- dicular
magnetic force of 100 kPa (1 atm) can be produced
quency electromagnetic field. However, for plasmas that with a field
strength of about 5000 Oersteds, which is quite
would produce fusion power densities appropriate to this readily
achievable.
type of fusion reactor, the overall outward plasma pres-
sure would be one atmosphere or more. In order to pro- Principal
Confinement Approaches of Fusion Research
duce a countering inward radiation pressure of about an
atmosphere, the electric field strength in the applied wave Research
toward the goal of a fusion reactor began in a
would need to be about 1 million V/cm. This is difficult to significant
way in 1951 at Princeton University, and has
achieve without inducing electrical breakdowns, and eddy since
spread to many different laboratories in a large num-
current losses due to image charges in nearby walls might ber of
nations. The overwhelming majority of this research
be large. has
followed either the low-density, high-confinement time
approach
using magnetic fields, or the high-density, low-
confinement
time path of inertial confinement. Accordingly,
Inertial Confinement
the
remainder of this article will cover only magnetic con-
Inertial confinement is the method used in fusion bombs, finement
and inertial confinement, with an emphasis on
more commonly called hydrogen bombs. It works well un- magnetic
confinement because this has profited from the
der those conditions when the radiation from the explosion most
research, and is presently closer to practicality.
of a fission bomb, usually called an atomic bomb, is used
to produce soft X-rays, which isoentropically compress a
MAGNETIC
CONFINEMENT FUSION
fuel mixture of deuterium and tritium. The goal of inertial
confinement research for fusion reactor applications is to
Physics of
Plasmas
reproduce this effect on a much smaller amount of fuel, and
with a far less energetic driver than an atomic bomb. This Debye
Length. Because a plasma is made up of posi-
approach is referred to as inertial confinement because it tively
charged ions and negatively charged unbound elec-
is simply the inertia of the assemblage which confines the trons,
these two populations of particles interact strongly
fuel while it is fusioning. The fuel is confined for a time with each
other through the coulomb force. Any substan-
----------------------- Page 168-----------------------
Reactors, Fusion 7
tial displacement of one species relative to the other leads first

criterion is that the spatial extent of the plasma should
to a large electrostatic potential, which would require some be
much greater than a Debye length. The second crite-
countering force to maintain. In the absence of any exter- rion
is that there should be many charged particles within
nally applied countering force, there is always an intrin- a
Debye sphere (with a radius λD ) so that the statistical
sic force available from random thermal fluctuations in a
treatment underlying the definition of the Debye length is
plasma. The electron thermal energy density per degree of
valid. The Debye length is usually small for the plasmas
freedom is 0.5neTe in a plasma with an electron density of used
for laboratory or industrial applications. For a 3 eV
ne and an electron temperature of T . This energy is avail- arc
discharge with an electron density of 1019 m−3 , the De-
e
able to drive charge separations between the positive ions bye
length is about 3 × 10−6 m, with about 103 charged
and the negative electrons. The restoring force is provided,

particles within a Debye sphere, sufficient to validate the
in turn, by the electrostatic energy density associated with
underlying statistical treatment (5).
the electric field established by the charge separation. For A
plasma more characteristic of those produced in fu-
a charge separation length of d, the electrostatic energy sion
research devices might have a density of 3 × 1019 m−3 ,
density E scales approximately as: T =

10 keV, T ≈ 0.5T , and Z ≈ 1.2, with the fact that
e
e i
Z is
not unity arising from light contaminants in the deu-
terium and tritium fuel. For these conditions, the Debye
length is about λD ≈ 8 × 10−5 m, where almost all of the

where ε0 is the permittivity of free space, and e is the fun-
increase relative to the arc discharge is due to the much

damental charge of an electron. If this electrostatic energy
higher temperature of the fusion plasma. Since the char-

density is compared to the thermal energy density, then it
acteristic dimensions of fusion plasmas are usually of the

is apparent that substantial charge separations within the
order of a meter or meters in magnetic confinement de-

plasma can only take place over distances not significantly
vices, it is always the case that they are much greater than
greater than d ≈ λD , where the value of λD is given by:
a
Debye length.
The plasmas found in electric arcs, lightning, or fluo-
rescent lights are usually only weakly ionized, with per-

This is called the Debye length, because it was first calcu- haps
one to a few percent of the gas molecules ionized.
lated theoretically for electrolytes by Debye and Huckel in On
the other hand, plasmas with the much higher temper-
1923. This equation is the one most commonly used for λD ,
atures needed for magnetic confinement fusion research
although a more precise definition would include a term (1 are
highly ionized, with thermal neutrals only penetrating
+ ZTe/Ti ) in the denominator to account for the ion effects. into
the outer few centimeters of the plasma.
Here Ti is the ion temperature and Z is the ionization state
of the plasma ions. It is of most importance for plasmas
Cyclotron Motion in Magnetized Plasmas. If a magnetic
with high charge state ions, which is not the case for the field
is applied or arises within a plasma, then the motion of
fuel components of a deuterium and tritium fusion reactor, the
constituent electrons and ions is significantly altered,
although there is usually some admixture of light impuri-
becoming more ordered along a preferred axis. The equa-
ties such as carbon and perhaps small amounts of oxygen tion
of motion of a particle with charge Z , vector velocity
in the plasma of today’s experimental fusion devices. This v
and mass m in a magnetic field of vector strength B is
ion term can also be of importance if Te is much greater
given by:
than Ti . However, in plasmas of interest to fusion work, Te
rarely exceeds Ti , and is often lower than it.
Since both the ions and electrons in a plasma can move
If
we choose the z axis to be along the magnetic field, then,
freely, they will tend to move so as to neutralize the elec-
in
cartesian coordinates, the components of the particle mo-
tric fields arising from charge imbalances, with electrons tion
are as follows:
positioning themselves to shield the electrostatic field from
positive ions, and the ions moving to cancel the field from
the negative electrons. The electrons, being much lighter
and therefore much more mobile than the ions for all plas-
mas of interest to fusion, account for most of this charge
shielding. with
ωc = (eZB)/m, and with B the scalar magnitude of the
The Debye length is one of the fundamental parameters
magnetic field. The fact that the particle velocity is con-
of any plasma. Charge imbalances can occur over distance
stant along the magnetic field arises from the fact that, as
of the order of the Debye length, but over distances much
shown in the vector cross product above, magnetic fields
greater than the Debye length the plasma will not sustain a have
no effect on a velocity component parallel to the field.
net electric field unless there is some additional countering The x
andy components are both perpendicular to the mag-
force to support it. Similarly, over distances much greater
netic field. The time derivatives of a particle in these di-
than a Debye length, a plasma will shield out externally
mensions correspond to circular motion with a frequency
applied electrostatic fields. Thus, the electric potential is of
ωc , which is called the cyclotron frequency. The combina-
normally the same throughout a plasma, unless some other tion
of this circular motion perpendicular to the magnetic
force due, for instance, to rotation, alters the balance. field
and the uniform velocity parallel to the field (for uni-
The two criteria for defining an assemblage of ions and form
magnetic fields) give rise to a net helical particle orbit
electrons as a plasma both involve the Debye length. The
spiraling along the magnetic field, as shown in Fig. 1.
----------------------- Page 169-----------------------
8 Reactors, Fusion
Figure 1. Charged particles gyrat-
ing along a magnetic field line. For
electrons and protons with the same
energy, the Larmor radius for the
electrons would be 1/43 of the pro-
ton’s. The average motion can be
described by the guiding center, so
long as no parameters change signif-
icantly across a Larmor orbit. Drawn
by T. Stevenson.
In a plasma with an isotropic Maxwellian velocity dis-

In
these equations the electron and ion temperatures are
tribution, the temperature perpendicular to the magnetic
expressed in keV. For comparison, room temperature is

field is equal to the temperature parallel to it. However, be-
about 0.025 eV. The magnetic field is given in units of tesla,
cause the perpendicular motion encompasses two degrees
where a tesla is equal to 10,000 gauss, or roughly 25,000

of freedom while the parallel motion has only one, for these
times the average strength of the earth’s surface magnetic

conditions the perpendicular motion of the plasma particle
field.
distribution carries twice the energy content of the parallel
In a 5 T magnetic field, which corresponds to a typi-

motion. It is not always the case that the plasma tempera-
cal
field strength on the plasma axis of the Tokamak Fu-
ture is isotropic with respect to the magnetic field direction.
sion
Test Reactor at Princeton University, a deuteron would
Some heating mechanisms transfer energy preferentially
8
have
a cyclotron frequency of ωi = 2.4 × 10 per second. A
into either the parallel or the perpendicular motion of the
plasma electron in the same field would have a cyclotron

particles with respect to the magnetic field direction. Simi-
11
frequency of ωe = 8.79 × 10 per second. The much higher

larly, some particle loss processes involving collisions pref-
frequency of the electron cyclotron motion is a consequence

erentially deplete either the perpendicular or the parallel
of
the much smaller mass of the electron relative to a
energy, depending upon whether the process is stronger for
deuteron. For the same reason, the Larmor radius of an ion

large perpendicular energies or large parallel energies.
is
much larger than that of an electron. For the case of a 5
T
magnetic field confining a plasma with central tempera-
Larmor Radius in a Magnetized Plasma. The radius of the
tures of 40 keV for the ions and 12 keV for the electrons,
perpendicular motion around the field line is called the
which is reasonably representative of higher performance
Larmor radius, denoted as ρ. It is given by the ratio of
plasma parameters on Princeton’s Tokamak Fusion Test
the perpendicular velocity to the cyclotron frequency, and
Reactor, the Larmor radius of an electron with the ther-
scales for a particle moving at the thermal velocity as: mal
energy would be ρe = 0.0073 cm, and for an ion with
the
thermal energy it would be 0.58 cm. The full range
of
Larmor radii would include values ranging from a bit
smaller than this up to a few centimeters (on the tail of
where vt is the thermal velocity, and the absolute value of the
Maxwellian distribution) that would be a factor of 2 or
the charge is taken to ensure a positive length. The factor more
larger. These Larmor radii are small compared to the
of square root of two arises from the two degrees of free-
dimensions of the fusion devices in which they presently
dom in the perpendicular motion. For any plasma, whether
occur. The position of a charged particle averaged over its
Maxwellian or not, there will clearly be a range of values
cyclotron motion is called the guiding center. In the ab-
of the Larmor radius corresponding to the perpendicular
sence of perturbations such as collisions, the guiding center
velocity distribution. Inserting the values of electron and of
an electron or ion moves along a line of magnetic force
baryon masses leads to the following more specific formu-
(with the electron guiding centers moving in one direction,
las for the cycloton frequencies and Larmor radii of thermal and
the positive ion guiding centers in the other). So long
particles in plasmas: For an electron: as
the gradient scale lengths for plasma properties such
as
density, temperature, and magnetic field strength are
much
larger (usually a factor of a few is sufficient) then
it
is a good approximation to model most types of particle
behavior by following the guiding centers rather than the
detailed gyromotion, which is more complicated computa-

For an ion of charge Z and a mass of A atomic mass units:
tionally, and requires more computer time.
Plasma Diamagnetism. The cyclotron orbits of particles

in a
plasma each enclose small volumes of magnetic flux.
----------------------- Page 170-----------------------
Reactors, Fusion 9
As a consequence of Lenz’s Law, the directions of rotation

of both the electrons and the ions are such that the tiny
solenoidal currents they represent produce magnetic fields
in the opposite direction from the field they are enclos-
ing. Thus, the gyromotion at the cyclotron frequency re-
duces the total strength of the field inside a particle orbit.
This is referred to as plasma diamagnetism. Its importance
increases as the plasma pressure increases or the exter-
nally applied magnetic field decreases. For high ratios of
the plasma pressure to the applied magnetic field strength,
this diamagnetism can hollow out the magnetic field and
produce a region in the central plasma with a magnetic
well, that is, a region which is everywhere surrounded by
increasing magnetic field strength. This approach has been
Figure 2. The simplest form of mirror confinement. The trapped
suggested as a technique for confining plasma, but its prac-
portion of the particle velocity distribution reflects from the
stronger magnetic fields at each end of the cell, forming a confined

ticality is not yet clear.
plasma in the region between the mirrors.
Magnetic Moment. A particle gyrating around a mag-

netic field line constitutes an electric current I = |Zq |ωc/2π, as:
which encompasses an area of A = πr2L , where Z is the
ionization state, q is the fundamental charge, ωc is the cy-

clotron frequency, and rL is the Larmor radius. The product
of the current and the enclosed area is called the magnet
Clearly there is one class of particle orbits for which mag-
moment of the particle orbit, and is generally denoted as
netic mirroring would have no effect at any magnetic field
µ. Thus,
strength, no matter how strong. Particles which are every-
where moving entirely parallel to the local magnetic field
have no gyro-orbit, and thus µ = 0. As a consequence, their
parallel velocity is unaffected by changes in the magnetic

Here vp is the velocity of the gyrating particle in the plane
field strength, and these particles are not reflected.
perpendicular to the magnetic field line it is following, and
Wp is the kinetic energy associated with this perpendic-
Magnetic Mirror Confinement. More generally, for finite
ular velocity. It can be shown (5) that, in the absence of
ratios of magnetic field strengths in a plasma confinement
collisions or electric fields, the magnetic moment µ of a
device, a considerably broader range of particle orbits is not
particle orbit is an invariant. This has far-reaching con-
mirror confined. If we define the minimum magnetic field
sequences for plasma confinement schemes utilizing mag-
strength along a line of force to be Bm and the maximum
netic fields. The invariance of a particle orbit’s magnetic to
be BM , then the constancy of µ and Wt lead to the condi-
moment means that as the particle gyrates along a line of
tion that all particle orbits with µ > Wt/BM are trapped by
magnetic force into a region of stronger magnetic field, the the
magnetic mirror field. If this were not so, then parti-
perpendicular velocity increases so that the energy of rota-
cles could reach the point of maximum field strength with
tion perpendicular to the field line increases by the same
their perpendicular kinetic energy greater than their total
factor as the magnetic field strength. Inasmuch as the par-
kinetic energy, which is not possible. Applying this princi-
ticle’s total kinetic energy is also constant in the absence of ple
to determine the boundary between mirror trapped and
collisions or electric fields, the increase of the perpendicu-
untrapped particles, one finds the conditions for marginally
lar energy as a particle gyrates into an increasing magnetic
trapped particles, where Wp (min B ) is defined as the energy
field implies that the kinetic energy of the particle paral- of
perpendicular rotation when the particle is in the region
lel to the field line decreases, and thus that the velocity of
minimum magnetic field strength, and Wz (min B ) is de-
along the field line also decreases. Thus, for a particle orbit
fined as the parallel energy of the particle at the minimum
which has a finite energy of perpendicular rotation along
magnetic field:
any part of its path, there can exist some magnetic field
strength at which its velocity parallel to the field line goes
to zero. When this happens, the particle is reflected back
into the region of weaker magnetic field. This is referred to
as magnetic mirroring, and plays a significant role in every
These conditions can also be written in terms of the ratios of
form of magnetic confinement. If a region of weaker mag- the
perpendicular and parallel velocities evaluated at the
netic field is bounded by a stronger magnetic field at each
minimum magnetic field with respect to the total velocity,
end, then particles can be reflected back and forth between v,
of the particle, giving:
the regions of stronger magnetic field, producing a trap,
as shown in Fig. 2. This works for both positively charged
particles (most ions) and negatively charged particles (for
example, electrons). For a particle with parallel velocity vz
and total kinetic energy W , the parallel velocity will vary
t
----------------------- Page 171-----------------------
10 Reactors, Fusion
will
rapidly escape the confinement device unless their ra-
tio
of perpendicular to parallel velocity is scattered back
into
the trapped region by a subsequent collision or colli-
sions. Accordingly, although the loss cone is the same for
different types of particles in a plasma, the particle species
which collides most frequently, and thus will most rapidly

Figure 3. Schematic sample of loss cones in the perpendicu-
lar versus parallel velocity space of a magnetic mirror-confined
scatter into the loss cone, is the one which is preferentially
lost.
plasma. Particles which undergo collisions such that their vp/vz
drops into the loss cone quickly escape along the field lines at the
For a plasma in which the temperatures of the electrons
magnetic mirror throat. and
the ions are of the same order, the electrons are the
more
collisional species, so they initially escape through
the
mirror throats more rapidly than the ions. This im-
These equations define the boundary between mirror
mediately gives rise to a net negative charge leaving the
trapped and untrapped particles in the space of paral-
mirror throats, and a net positive charge remaining in the
lel versus perpendicular velocities. Thus, charged particles
central cell of the mirror confinement device. This charge
with a velocity ratio vz (min B )/v which is sufficiently low
imbalance produces an electric potential which is positive
will be trapped. On the other hand, charged particles will in
the central cell, and thus an electric field which retards
promptly escape along the mirror field if they satisfy the the
loss of electrons along magnetic field lines through the
criterion:
mirror throats. This electric field causes lower energy elec-
trons to be confined which would otherwise be lost, and,
correspondingly, it slightly increases the parallel veloci-

ties
of positive ions as they approach the mirror throats.
This can be rewritten in terms of the ratio of the paral-
Thus, the electric field strength will rapidly build up only

lel and perpendicular velocities evaluated at the minimum
to
the strength at which it impedes the loss of lower energy
magnetic field strength:
electrons sufficiently so that the net electron loss just bal-
ances the loss of positive ions. In a mirror-confined plasma

it
will be the more energetic electrons which are able to
This condition defines two regions in velocity space, in
escape over the electrostatic potential hill formed at the
which particles with higher ratios of v (min B )/v (min B )
mirror throats, so the energy balance of mirrors is dom-
p z
are trapped, and ones with lower values of this ratio are
inated by these electron thermal losses. The equilibrium
lost. Since the perpendicular velocity corresponds to ro- rate
at which electrons and positive ions escape from the
tation in two dimensions, the boundary between trapped
mirror-confined plasma is governed by the rate at which
orbits and loss orbits in a three-dimensional velocity space the
less-collisional ions scatter into the loss cone, so ion
forms a cone. Thus, particles with orbits that have ratios
collisions set the rate of particle loss.
of parallel velocity to perpendicular velocity which are too
During the early decades of fusion research, devices
great for trapping are said to fall into the loss cone, depicted
incorporating magnetic mirrors as their primary confine-
in Fig. 3. ment
mechanism were extensively investigated. In their
A mirror-trapped plasma can never be isotropic in veloc-
simplest form these machines consisted of a central cell
ity space. This is because the transit time along a magnetic
permeated by a uniform solenoidal magnetic field, bounded
field line is very fast, so particles with velocity ratios that by a
stronger magnetic mirror field at each end. These mir-
fall within the loss cone will rapidly escape through the ror
machines were examples of “open” confinement devices
mirror throat (region of maximum B ). Thus, the average
because their magnetic field lines left the region of plasma
perpendicular energy in a mirror-confined plasma is larger
confinement through the throats of the mirrors. This ge-
than the parallel energy by more than the factor of 2 one
ometry had a practical appeal because it is mechanically
would expect just from the relative number of degrees of
simpler to build a linear device than one with some more
freedom involved.
complex shape. Moreover, mirrors are devices which do not
Note that neither the charge nor the mass of the par-
require a time varying field, and thus are intrinsically ca-
ticles enters directly into the criteria for trapping. In an
pable of continuous operation if the engineering systems
entirely collisionless plasma, the particles within the loss are
designed accordingly. In a practical mirror fusion re-
cone would be entirely lost within a transit time (the time
actor the magnetic field coils would be superconducting in
required for the slowest particles to travel the length of the
order to avoid ohmic dissipation.
region enclosed by the magnetic mirrors). After this very
Simple mirror confinement machines encountered dif-
rapid loss, the remainder of the plasma would remain con-
ficulties due to macroscopic instabilities which rippled
fined for as long as the magnetic field configuration was the
plasma surface and greatly augmented the loss of
maintained.
plasma particles and energy. These problems were ad-
In all systems of interest to fusion research, this static
dressed by confinement schemes which added additional
condition never arises because coulomb collisions between
current-carrying conductors to alter the magnetic field in
mirror-confined particles alter their ratio of perpendicular the
central cell so that it increased toward the edge of
to parallel velocity, creating a dynamic loss due to particles the
plasma. The intrinsically anisotropic velocity distri-
which scatter from trapped velocity ratios to untrapped ra-
bution residing within a mirror drives another smaller
tios. Once in the loss cone, these newly scattered particles
scale length microinstability, called the loss-cone instabil-
----------------------- Page 172-----------------------
Reactors, Fusion 11
ity, which increases particle and energy loss. Finally, of Closed

Magnetic Field Line Confinement Devices. Exper-
course, there is the overriding consideration that even if all imental
confinement machines with closed magnetic fields
losses across the magnetic field were eliminated, the loss of have in
most cases used a topology which is basically
plasma particles and energy along the magnetic field lines toroidal,
or toroids elongated with straight sections. Such
through the mirror throats would remain. devices
were pioneered in western countries by Professor
Even in the absence of macroscopic or microscopic insta- Lyman
Spitzer at Princeton University in 1951, and at
bilities, the theoretical limit on the energy gain of a simple about the
same time by researchers at the I. V. Kurchatov
mirror with a single confinement cell is relatively modest. Institute
in Moscow. Although similar in many respects,
We can define Q as the ratio of the nuclear power (ther- the
approaches which these laboratories pursued in the
mal) produced by the plasma to the amount of power sup- 1950s and
early 1960s differed in some important ways.
plied via external means to the plasma in order to heat
it to the high temperatures required for fusion reactions.
Rotational Transform. All closed field confinement sys-
Assuming a conservative efficiency for converting thermal tems are
defined by the fact that their magnetic field lines
energy to electricity of about 33%, and keeping in mind wrap around
to reconnect with themselves after some an-
that a practical reactor must produce considerably more gular
translation along the major circumference of the de-
electricity than it uses, an energy multiplication factor of vice. The
simplest topology for a closed system, and the
Q = 15 to 20 or more is often considered to be sufficient for one which
has been most often used for large fusion ex-
an economic fusion reactor using a fusion-heated steam cy- periments,
is a torus. In this approach, current-carrying
cle to drive the electricity-producing turbines. coils are
arranged around a vacuum vessel so that they
Due to the end losses along the magnetic field lines, it produce a
toroidal, or donut-shaped, magnetic field along
would be difficult even in principle to achieve an econom- which the
confined plasma ions and electrons can gyrate.
ically acceptable Q with a simple mirror machine. For a This
geometry is the topological equivalent of deforming a
simple mirror, free of any instabilities, and using as fuel linear
magnetic solenoid into a torus by bending it into a
deuterium and tritium, the energy multiplication would be circle such
that the ends touch. In a toroidal geometry, in
at best Q = 2 log(BM/Bm ). The logarithm in this expression
contradistinction to the case for a linear system, each of
can never be greater than an order of unity (that is, with these
toroidal field coils has an inner leg and an outer leg
a value of one to at the very most a few). This is because with
respect to the vacuum vessel and the plasma it con-
the mirror ratio, the ratio of the maximum to minimum tains. It
is clear that the inner legs of the toroidal field coils
magnetic field in the device, cannot be made arbitrarily carry the
same current as the outer legs, but the outer legs
large. Such practical considerations as mechanical forces on a
torus are much farther apart from each other than
and the maximum current densities superconductors can are the
inner legs, which are often sized so that they abut
carry limit the maximum field strength that can be pro- or nearly
abut each other. This gives rise to an asymmetry
duced in a mirror of useful size, while the minimum field between the
average current density flowing along the in-
strength cannot be reduced too much without allowing inner
circumference of the coil system where the coils nestle
stabilities and diffusion across the magnetic field in the near each
other compared to the outer coil circumference
central cell to increase. where the
coils are widely separated. This assymetry in
In order to circumvent these limitations, researchers turn
produces a gradient in the strength of the magnetic
contrived more sophisticated mirror confinement schemes field
produced by the coils; the field is stronger on the inside
such as tandem mirrors. In this approach, the central con- of the
torus, and weaker on the outside.
finement cell is enclosed by an additional mirror cell at A
toroidal geometry is most naturally described by two
each end which accepts particles escaping through the ve- dimensions,
the major and minor radii of the plasma toroid,
locity loss cone of the mirror field of the central cell. The where the
major radius, R0 , is the distance from the cen-
plasma in these end cells, and particularly the electron tral axis
of the coil arrangement to the axis of the confined
component, is heated to very high temperatures to increase plasma. The
minor radius, r0 , is the distance between the
the magnitude of the electrostatic field plugging the mirror plasma axis
and the outside of the plasma cross section.
throat from the central cell. Increasing this negative poten- For
circular coil systems and plasmas where the ratio of
tial in the throat means that more energetic electrons are the
perpendicular plasma pressure to the inward magnetic
confined in the central cell, thus reducing the parallel heat pressure is
fairly low, the effective major and minor radii
loss, and permitting the theoretically obtainable Q to be are quite
similar to what one would get from simply looking
greater than for a simple mirror. at the
plasma as a donut. For elongated coils and plasmas,
Maintaining the proper velocity distributions of parti- and for
high-pressure plasmas which rearrange flux sur-
cles in tandem mirrors without driving instabilities proved faces,
these effective radii are somewhat modified.
quite daunting so that, in the mid-1980s the experiments If the
only magnetic field threading a plasma torus is
with the mirror approach to fusion were largely aban- the
toroidal field, then the plasma will be only momentarily
doned. Since that time almost all magnetic confinement confined, no
matter how strong the toroidal field is made.
fusion has gone into the other general class of devices called This arises
as a result of the fact that there is a gradient in
closed field systems, in which the magnetic field lines per- the
toroidal magnetic field strength running from a high
vading the central confinement region do not leave the level on
the inside to a lower level on the outside. This mag-
plasma, but instead circle around to eventually reconnect netic
gradient causes the negatively charged electrons to
with themselves. drift
transverse to the gradient in one direction, and the
positively
charged ions to do the same thing in the other
----------------------- Page 173-----------------------
12 Reactors, Fusion
direction. These oppositely directed vertical drifts produce to return

to its initial poloidal location, then q = 2.7. For
an imbalance in the net neutrality of the plasma above the values of q
> 1 in a circular cross section toroid with the
midplane, compared to the plasma under the midplane. major
radius significantly larger than the minor radius, a
One side will have excess positive charge, while the other particle
gyrating along a field line experiences a net mag-
will have excess negative charge. These charge accumu- netic well
when averaged along its orbit, which tends to
lations will establish a vertical electric field which is per- abet
confinement.
pendicular to the toroidal magnetic field. In turn, the com- Within a
confined plasma of a device of the tokamak type
bination of this electric field and the magnetic field drive to be
discussed shortly, there are arbitrarily many local val-
a type of particle drift with a drift velocity of v = (E × ues of q
associated with flux surfaces. However, the class of
B )/B2 . For toroids as described here, the direction of this surfaces
with rational values of q often play a special role
drift is radially outward along the direction of the major either in

the large-scale internal stability or the gross sta-
radius. This mechanism drives both the negative electrons bility of
the plasma. Rational magnetic surfaces are ones
and the positive ions out together, so no electric field is es- for which
the safety factor can be represented as q = m/n,
tablished which would limit this drift. As the plasma drifts where m
and n are integers, and a field line comes back
out to larger major radii it quickly runs into the material to its
original position after m toroidal and n poloidal ro-
boundary of the containment device, and is extinguished tations
around the torus. Rational magnetic surfaces have
as impurity influx leads to large line transition radiation proven
particularly susceptible to magnetohydrodynamic
losses. Reference 5 contains excellent descriptions of the (MHD )
instabilities, which are large-scale fluid-like pertur-
two types of particle drifts mentioned here, as well as other bations of
the plasma and the field lines within it. If they
sorts of drifts which occur in magnetized plasmas. occur on
rational flux surfaces well inside the plasma, they
The fact that a simple toroidal magnetic field cannot can
increase the radial loss of energy and particles; if they
confine a plasma is a consequence of the field gradient that occur near
the outside of the plasma, they can cause the
necessarily arises from the geometry of any toroid. Thus, plasma to
disrupt.
there is nothing which can be done to prevent the oppo- The
magnetic surfaces with low values of rational q tend
sitely directed drifts of the electrons and ions perpendicu- to be the
most pathological, since they are most suscepti-
lar to the magnetic gradient. ble to
current-driven instabilities. Early researchers found
What can be done instead is to short out the charge im- that it was
usually easier to maintain gross plasma stabil-
balance which would otherwise develop by adding a helical ity if the
surfaces with low rational values of q were buried
twist to the magnetic field lines so that particles gyrating well inside
the plasma, which meant that the q value of the
along these lines of magnetic force will spend half of their outside was
large (as much as 8 to 10 or so). Thus, the di-
time above the tokamak midplane and half below the mid- mensionless
quantity q came to be called the safety factor,
plane. Since the vertical drift of each species is independent since it
was easier to maintain the overall stability of most
of whether the particle is above or below the midplane, this sorts of
toroidal plasmas if the edge q was larger.
means that the unidirectional drift of each particle is out-
ward half of the time, and inward the other half, so there Ways to
Introduce Rotational Transform. The pioneers of
is no net change of position. fusion
research in 1951 realized from the beginning that a
The helical twist in the confining magnetic field lines is rotational
transform would be needed in the field lines of
described by the rotational transform, which is the amount a toroidal
or semitoroidal confinement device. Where they
a field line moves in poloidal angle (the angle around a differed
in the mechanisms they used to produce the ro-
cross section of the plasma donut) as it traces itself around tational
transform. The group at Princeton University im-
the plasma in toroidal angle. The inverse of this quantity, posed the
rotational transform externally by applying mag-
called the safety factor, and denoted as q, is commonly used netic fields
from helical coils spiraling around the vacuum
in describing toroidal magnetic confinement systems. vessel
inside the much more powerful toroidal field coils.
The local value of q will in general be different for each
Alternatively, in some early experiments researchers ob-
magnetic flux surface within a toroidal plasma, but it will tained the
rotational transform without the helical coils,
be uniform on any given flux surface. For flux surfaces with but instead
with an elongated racetrack type of design dis-
a minor radius significantly smaller than the major radius, torted so
that the two straight sections crossed over each
which is the case for all of the large devices so far built, the other.
These closed field devices in which the rotational
value of the dimensionless quantity q can be well approxi- transform
arose from the external configuration or special
mated as: coils were
named stellarators because of the astrophysics
background
of their inventor, Lyman Spitzer, Jr., of the
Princeton
University astrophysics department.
Stellarators had two particularly appealing character-

Here r is the minor radius of the flux surface being de- istics. One
was that they did not require any net current in
scribed, B # is the toroidal component of the magnetic field, the plasma
to maintain the rotational transform, making
R0 is the plasma major radius, and B is the poloidal com- them
less susceptible to damaging disruptions and obviat-
ponent of the magnetic field (the component which imparts ing the
need to find a way to drive the current. Since the
the helical twist to field lines on the flux surface). If a field required
fields were static, a stellarator was intrinsically
line returns to its starting position after exactly one circuit capable of
steady-state operation if the coils were designed
of the torus, then q = 1 on that flux surface. If, for example, for this.
The other major desirable feature was that, with
2.7 transits around the torus are required for a field line the
magnetic fields all imposed by external coils, position
----------------------- Page 174-----------------------
Reactors, Fusion 13
control was relatively straightforward, so the plasma could

limitations upon the current densities which conductors
be kept out of contact with material surfaces. This meant can
carry, inductively driven tokamaks are intrinsically
that keeping the plasma fairly free of impurities, which pulsed
devices.
would dilute the fuel and increase energy losses through In
the present generation of tokamaks, these pulse
radiation, should be feasible. lengths
range from a few seconds to a minute. Inductively
A disadvantage of stellarators was that the helical coils driven
fusion power reactors of the future might have much
producing the rotational transform crossed the toroidal longer
pulses, perhaps of many hours, but they would still
field lines produced by the encircling toroidal array of coils. have to
pause at some regular interval to reset the current
This produced large mechanical stresses which were dif- in the
primary transformer winding. The resulting period
ficult to accommodate in the 1950s. Partly as a result of during
which the fusion plasma was turned off would be
this engineering difficulty, no stellarators with large mi- short
compared to the time when it was making power. De-
nor radii were built during that period. These early small- pending
upon the thermal inertia of the heat blanket sur-
cross-section stellarators (typically with minor radii of or- rounding
the tokamak, where heat exchange and tritium
der 10 cm or even less) achieved disappointing results, breeding
take place, the transformer recharge interval
which may have been in some part because they were might
produce either relatively little or significant fluctu-
so small and relatively cool (compared to a stellar core) ation in
the electricity output. Perhaps more significantly,
that low-energy neutral particles could penetrate through- pulsed
thermal and magnetic field loads are expected to in-
out the plasma, undergoing charge exchange with the hot- crease
the stress on components near the plasma, thereby
ter confined particles which, once neutralized, would es- reducing
their lifetimes. As will be discussed later, much
cape the confinement. In the late 1960s the stellarator ap- progress
has been made in finding noninductive ways to
proach was largely abandoned in response to more favor- drive
the plasma current, allowing steady-state operation
able results obtained at the I. V. Kurchatov Institute in of
future power reactors.
Moscow using the tokamak approach, to be discussed next.
In later years, as the understanding of both the physics
Plasma Limiters in Tokamaks. Another possible disadvan-
and the engineering of fusion devices improved, some re- tage of
tokamaks in their simplest form is that, since the
searchers have returned to the stellarator approach and plasma
is a current carrying loop, it tries to expand in
variations thereon. Better performance was obtained with major
radius. This tendency can be countered, however,
larger plasma cross-section stellarators built in Germany by
adding coils which produce a vertical magnetic field.
by the Max Planck Institute, and a much larger stellarator A more
serious concern of early researchers was that sim-
variant is being built outside Nagoya, Japan, by the Na- ple
circular-cross-section tokamak plasmas required con-
tional Institute of Fusion Studies. These later stellarators tact
with a material limiter in order to stabilize their po-
have mostly employed magnetic field coils with complex sition.
This could act as a source of impurities, and would
shapes which fulfill the role of both the toroidal field coil require
large amounts of cooling in a power reactor. Most
and the helical coils of the early stellarators. of the
more recent tokamak experiments such as DIII-D
at
General Atomics in San Diego have replaced the limiter
Tokamaks. Soviet scientists chose another way to intro- with
divertors, to be discussed later, which appear to be
duce the required rotational transform which employed the much
more suitable for reactors.
toroidal field coils of the stellarator concept, but no heli-
cal coils to introduce the poloidal magnetic field compo-
MAJOR
COMPONENTS OF A TOKAMAK FUSION
nent. They instead drove an electric current which flowed
REACTOR
toroidally through the plasma. This current produced an
encircling poloidal field which vector-added to the toroidal
Plasma
Confinement System
field from the coils to produce a spiraling field with dif-
ferent values of the rotational transform, and thus of q, The
quality of energy confinement that can be achieved
on each flux surface. The value of the outer surface q, and within a
tokamak design is the single most important fac-
consequently of the inner surface q’s could be easily al- tor in
determining the most feasible design for a fusion
tered by driving different amounts of plasma current for a power
plant. If the rate at which energy leaks out of the
given value of the toroidal field. This device was named the plasma
is too high, then prohibitively large amounts of
tokamak, which means something like “magnetic bottle” in power
will be required to heat and maintain the plasma
Russian. Figure 4 shows the basic schematic of a tokamak. at the
temperatures required to produce useful amounts
of
fusion energy. The confinement time is characterized
Inductively Driven Plasma Current in Tokamaks. In all the by a
quantity, τ, which gives the time required for the
early tokamaks, and in many later ones, all or nearly all of energy
content of the plasma to decline by an e-folding.
the plasma current was driven inductively. This was accom- At the
dawn of fusion research, energy confinement times
plished by adding a transformer solenoid to the tokamak, were
typically much less than a millisecond. Today they
with the plasma acting as a one-turn secondary winding. As are
commonly hundreds of milliseconds, and in some high-
with any transformer, current would continue to be driven
confinement plasmas the confinement time is more than a
in the secondary, that is, the plasma, only so long as the second.
flux produced by the primary winding was changing. Since The
basic components of the confinement system are the
there is always a practical limitation on the volt-seconds various
current-carrying coils which produce those mag-
available to drive such a flux swing, and since there also netic
fields that are externally applied. In addition to the
----------------------- Page 175-----------------------
14 Reactors, Fusion
Figure 4. Basic components of a tokamak magnetic confine-

ment
configuration. The coils surrounding the plasma produce
a
toroidal magnetic field, while the current in the plasma cre-
ates
a weaker poloidal magnetic field. The combination of the
two
makes field lines which spiral as they are followed around
the
torus.
toroidal field coils and vertical field coils, there are coils ar- ing
from toroidicity. The excess energy leakage was called
rayed around the plasma to alter its cross-sectional shape,
anomolous diffusion, and was presumed to be driven pri-
and in some cases additional coils might be added to con- marily
by collective instabilities in which great numbers of
trol gross instabilities. More recently, the systems which ions
and electrons moved together to form waves and large-
deposit energy, momentum, and additional current into the scale
plasma deformations that transported energy much
plasma have also become important in improving confine- more
rapidly out of the plasma than could two-particle col-
ment.
lisions. The early years of fusion research were plagued by
anomalous energy losses so large that the confined plas-

Collisions among Confined Particles. Getting to the high mas
could not be successfully heated to and maintained at
levels of energy confinement that can be obtained in to- the
very high temperatures necessary for practical fusion
day’s tokamaks has been a long journey. The tokamak con- power.
finement configuration is characterized by a magnetic field

topology which guides the confined ions and electrons along
Enhanced Confinement. During the decades stretching
helical field lines on toroidal flux surfaces. A single ion or from
the 1970s through the 1990s, progress occurred as
electron could travel along such field lines forever if there
successive generations of tokamaks led to a better under-
were no imperfections. In practical applications, there are
standing of how the anomalous energy leakage could be
always some magnetic field imperfections, and, much more reduced
by varying such factors as the dimensions of the
importantly, there are many ions and electrons instead plasma
toroid and the magnitude of the current flowing
of one. The ions and electrons each collide among them- within
it. At the same time, modes of tokamak operation
selves and with each other. Sometimes when the ions col-
were
discovered with improved confinement: that is to say,
lide they fuse, releasing the energy which is the purpose of
significantly reduced anomalous energy leakage. These en-
the tokamak, but much more often they scatter from each hanced
confinement regimes were characterized by differ-
other. The scattering within and among the different con- ent
profile shapes for the density and temperature than
fined species gives rise to transport of energy and particles were
found in the usual tokamak plasmas, and also by re-
across the confining magnetic field lines, and thus eventu-
ductions in the rate at which neutral gas was recycling
ally out of the plasma. If this scattering by Coulomb colli- from
the walls into the outer part of the plasma. Some of
sions between pairs of particles were the only mechanism these
enhanced modes also involved changes in the current
by which energy leaked from the plasma, then achieving density
profile, and in the relative speed at which different
the quality of confinement necessary for a fusion power
portions of the plasma rotated.
plant would have been a daunting, but nonetheless con- The
improved modes were usually called by phenomeno-
ceptually straightforward, undertaking. logical
names because they were mostly discovered experi-
mentally rather than first being predicted theoretically: fu-

Anomalous Diffusion. The reality has been that the sion
research makes progress by building new tokamaks.
early plasma confinement experiments encountered en- Two of
the most famous confinement modes were the H-
ergy leakage rates much greater than would be expected if Mode,
which was discovered in Germany, with the H stand-
the energy loss were primarily due to diffusion arising from ing for
high confinement to distinguish it from low confine-
two-particle collisions. This type of two-body collisional dif- ment,
and the Supershot Mode, discovered at Princeton
fusion was somewhat understood, and could be calculated in 1986
in the Tokamak Fusion Test Reactor. A supershot
with neoclassical theory, which took account of effects aris- plasma
was distinguished by having a region in its core
----------------------- Page 176-----------------------
Reactors, Fusion 15
with very good energy confinement, and by temperatures to 1 s

because the required plasma current profile is pro-
much higher than had previously been achieved. Super- duced by
rapidly increasing the total plasma current in an
shot plasmas were also the first ones in a tokamak to ex- already hot
plasma. Because the plasma is hot, its electri-
hibit the phenomenon of bootstrap current. The bootstrap cal
conductivity is high, so the extra current takes hun-
current arises from a dynamo effect within the plasma dreds of
milliseconds or more to diffuse into the central
as the charged particles press across magnetic field lines. plasma.
During this period, the current density in the outer
Finding the bootstrap current enhanced the economic vi- plasma is
elevated. It is expected that in the future it will
ability of future fusion power plants because its existence be possible
to maintain the required reversed shear pro-
means that most of the electric current needed to sustain file with
active current profile control using injected beams
the poloidal component of the magnetic field can be sup- of
energetic particles or injected waves. This confinement
plied at little or no extra cost by the plasma itself. Similar mode may
lead to fusion power plants which are physically
sorts of plasmas were subsequently produced in Japan’s smaller and
cheaper than had been previously envisioned,
JT-60U tokamak and in Britain at JET, the principal toka- but much
work remains to be done in understanding how
mak of the European Union. It was the Supershot mode to control
the mode, and in how to fuel and remove helium
which was selected as the best route to follow in the Toka- ash from
it.
mak Fusion Test Reactor to produce fusion thermal power
levels of up to 10.7 MW.
Plasma
Heating Methods
Enhanced Reversed Shear Mode. In the mid-1990s, an Alongside

these improvements in energy confinement,
even better operating mode was found at Princeton’s TFTR equally
significant progress occurred in technologies for
as well as at the European Union’s JET, and subsequently heating the
confined plasma. There are four general ways
at General Atomics in San Diego, and Japan’s JT-60U. This of heating
a magnetically confined plasma. One is with the
discovery grew out of a study of reversed shear plasmas fusion
reactions themselves. When a deuterium nucleus
which had been pursued on several tokamaks. Shear is the and a
tritium nucleus fuse, 80% of the released energy is
rate of change of the rotational transform of the magnetic immediately
carried out of the plasma by a neutron which,
field lines in passing from one nested flux surface to the being
electrically neutral, is not confined by the magnetic
next one. It is governed by the shape of the current within field that
holds the plasma. In an actual fusion power plant,
the plasma, which produces the poloidal field component the neutron
will be captured in a special blanket where it
that determines the transform. Most tokamak discharges produces
heat for electricity production and new tritium
have current profiles that yield shear which is always in the for fuel.
The remaining 20% of the energy is borne by the
same direction in passing through successive flux surfaces. nucleus of
a helium atom, called for historical reasons an
Techniques were found in recent years which modified the alpha
particle. Since this nucleus does not have any neg-
plasma current profile in such a way (by putting more cur- ative
electrons bound to it, it carries a net positive elec-
rent in the outer portion of the plasma) that the magnetic trical
charge, and is confined by the magnetic fields in the
shear was reversed in the plasma interior. In the central plasma.
plasma the rotational transform changed oppositely from
Consequently, the alpha particles produced by fusion
what it did in the outer plasma in going through successive reactions
remain in the plasma, giving up energy to the
flux surfaces. plasma
particles through collisions until the alphas are
In many cases this reversed shear magnetic field con-
thermalized, that is, until they cool down to the tempera-
figuration did not produce dramatic improvements in con- ture of the
plasma. The energy transferred from the alpha
finement. However, in some cases, the confinement in the particles,
which are initially much hotter than the plasma,
central region of the plasma improved markedly in terms heats it.
In an ignited fusion reactor, such as will one day
of ion energy leakage, and particle confinement of both the be used
to produce electricity, this will be the principal
ions and electrons. Named the enhanced reversed shear source of
plasma heating; in fact the definition of ignition
mode by researchers at Princeton University’s TFTR, who is that the
alphas supply sufficient energy to maintain the
were the first to observe the abrupt transition to this mode plasma
temperature. However, some other technique will
with high levels of energetic particle heating power, it was be needed
to heat the plasma of even a future electricity-
quickly produced on other tokamaks. In retrospect, it was producing
reactor to the ignition conditions under which
found to be very similar to a type of confinement mode alpha
particle heating can take over, and, in any event, the
which had been studied earlier in the European Union’s
experimental tokamaks in operation today are of too mod-
JET tokamak with a different heating technique. est a size
and capability to reach ignition conditions.
Enhanced reversed shear plasmas have nearly perfect The
other three types of plasma heating technologies
particle confinement in their cores, and the rate of energy are called
ohmic heating, wave heating, and neutral beam
loss through ion–ion collisions is for the first time down heating.
Of these, ohmic heating is the most readily im-
to the theoretical minimum due to two-body collisions. In plemented,
since it arises automatically from the electric
fact the conduction is so low that it required a revision current
that flows through the plasma to maintain the
of the neoclassical theory, which had previously not taken poloidal
component of the magnetic field which confines
account of orbit effects in regions with very steep gradients the plasma.
In much the same way that an electric current
in plasma parameters. flowing
through a copper wire heats it as the electrons car-
At present this enhanced reversed shear regime can rying the
current scatter as they move along it, the current
only be achieved transiently for periods of roughly 0.2 s moving
through the tokamak plasma also heats it.
----------------------- Page 177-----------------------
16 Reactors, Fusion
There is, however, an important difference in how an portion of
the ions each pick up an electron from the gas
ordinary electrical conductor and a plasma behave when molecules.
After picking up an electron, a beam ion becomes
heated. In an ordinary conductor, such as a copper wire,
electrically neutral, and thus once again an ordinary atom.
the electrical resistance rises as the wire’s temperature in- Unlike an
ordinary atom, which at room temperature has
creases, which is to say that the scattering of the electrons an energy
of about one fortieth of a volt, these atoms in
increases. Thus, in an ordinary conductor, if the current what is now
a neutral beam have energies of 120,000 V. The
passing through it is kept constant by raising the driving remaining
ions in the beams are bent out of it with a mag-
voltage as the conductor’s temperature increases, then the net, and
then the purely neutral beam is able to pass unim-
current will become progressively more effective in heat- peded
across the tokamak’s magnetic fields to enter the
ing it until the wire eventually melts. In the sorts of plas- plasma.
Once inside the plasma, the neutral atoms of the
mas of relevance to fusion, the behavior is just the oppo- beam are
again ionized through collisions with the plasma
site. The frequency of scattering decreases as the tempera- particles
which detach the electrons from the beam atoms.
ture increases, so the electrical resistance drops. This has The
resulting 120,000 V ions, being electrically charged,
the consequence that, as the plasma gets hotter, more and are confined
by the magnetic field of the tokamak. They cir-
more current must be run through it to achieve smaller culate
along the magnetic field, colliding with the plasma
and smaller additional increases in the temperature. It particles.
Since the beam ions are much more energetic
would be extremely difficult to heat a plasma to thermonu- than the
plasma electrons and ions, they transfer energy
clear ignition temperatures in this way, and even if one to them,
and thereby heat the plasma. Eventually the beam
could handle the large current required, the poloidal field ions slow
down and become part of the bulk plasma. At this
it would produce might be greater than the optimum for point they
are said to have been thermalized because their
maintaining plasma stability and confinement. In light of energy is
similar to that of the bulk, or thermal ions.
these limitations, ohmic heating serves only as the initial In 1973,
when neutral beams were selected as the prin-
heating mechanism in tokamaks, typically raising the tem- cipal
heating technique for the Tokamak Fusion Test Reac-
perature to 10 million to 20 million ◦ C, modest by fusion tor, this
technology had reached the point of development
standards, during the startup phase of a tokamak pulse. where it

could inject a few tens of thousands of watts of
One or both of the other two heating techniques—waves power into
a tokamak plasma. In the experiments which
or neutral beams—must be used to further raise the tem- took place
at Princeton University in the 1990s, the neutral
perature to the point where significant numbers of fusion beams
injected a maximum of forty million watts of power
reactions can occur. Wave heating works in much the same into the
Tokamak Fusion Test Reactor, an increase of about
way that a microwave oven does, except that instead of a thousand-
fold in the power capability of the technology
heating food by causing molecules to vibrate, the waves in- from the
time when the decision was made to use it.
crease the energy of the ions or electrons gyrating along the The fact
that the beam ions do become part of the con-
magnetic field within the tokamak plasma. Various forms fined
thermal plasma means that it is important not only
of wave heating have been used in many tokamak experi- that they
carry in energy to heat the plasma, but also that
ments, and it is expected that wave heating will be impor- they be the
right hydrogen isotopes to fuel it as well. Thus,
tant for future reactors. However, so far the highest tem- the TFTR
plasma heating systems inject high-energy neu-
perature and fusion power results in tokamaks have been tral beams
of both tritium and deuterium in order to main-
brought about with the other technology: neutral beams. tain the
correct fuel mix in the reacting core of the plasma.
As the
plasmas in successive generations of tokamaks
Neutral Beam Injection. Most neutral beams in use to- become
larger and denser, there is a corresponding increase
day are born in ion sources where the negatively charged in the beam
energy required to ensure that most of the en-
electrons are stripped from the positively charged atomic ergy and
fuel are deposited in the central plasma. However,
nuclei to produce ions. These ions are then accelerated by the
efficiency with which positive ions can be converted
passing them between grids with different electric volt- back to
neutral atoms is a strong function of the ion ve-
ages applied to them. This forms a beam of energetic ions locity. For
velocities corresponding to a beam energy above
moving toward the tokamak, in much the same way that 120 keV for
deuterium or 180 keV for tritium, the neutral-
the electron source in the back of a television picture tube ization
efficiency is steeply declining into unacceptably low
forms a beam of energetic electrons moving toward the values.
phosphor screen to form the image. The ions in a beam at In
response to this limitation, a new technology is being
Princeton University’s TFTR tokamak, however, have more developed
based upon ion sources which produce negative
energy than the electrons in the beam of a picture tube— ions, or
ions which have one more electron than the neu-
120,000 V instead of a few tens of thousands of volts. In tral atom
would have. Production of negative ions of deu-
addition, the electric current in the ion beam is thousands terium or
tritium is much more difficult than is produc-
of times greater than the current in a picture tube. ing their
positive ions, but the neutralization efficiency of
The electrically charged ions that emerge from the accel- high-energy
negative ions is nearly independent of energy
erator grids would not be able to enter the tokamak plasma across a
range of several million electron volts, with neu-
in their charged state. This is because the very magnetic tralization
efficiencies of 58% to 59% being quite feasible.
field which confines the plasma would bend the ion beam The first
generation of a negative ion beam system began
out of its path and prevent it from entering the plasma. To operating
on the JT-60U tokamak at Naka, Japan, in 1996.
circumvent this difficulty, the ion beam is passed through When it
reaches full power it will inject 10 megawatts of
a neutralizer cell filled with low pressure gas, where a 500 keV
neutrals. Another large negative ion system will
----------------------- Page 178-----------------------
Reactors, Fusion 17
go into operation in the late 1990s on the Large Helical chines.

These particles are referred to as trapped particles,
Device, a type of stellarator, outside Nagoya. in
distinction from the passing particles which circulate
freely
along field lines around the tokamak. Their orbits,
Current Drive. All of the tokamaks which have been op- when
projected onto a plasma cross section, resemble ba-
erated through the mid-1990s normally drove most or all nanas, with
the tips corresponding to mirror reflection from
of their plasma current inductively. However, as discussed higher
magnetic field strengths. In the Soviet Union, these
earlier, this technique necessarily requires that the toka- were
sometimes referred to as sickle orbits. In lower tem-
mak plasmas operate in pulses. In order to run future toka- perature
collisional plasmas, the trapped particles often
maks steady state or in very long pulses, other techniques scatter out
of trapped velocity space before they complete
must be used to drive the current. Over the years, a number a mirror
bounce, and thus have relatively little effect on
of techniques have been demonstrated to drive substantial the plasma
behavior. At the high temperatures typical of to-
amounts of plasma current, or even all of it in some spe- day’s
tokamaks, the particle collisionality is much reduced,
cial cases. The methods have used either the high-energy and
trapped particles may complete many bounces. It is
neutral beams which also heat the plasma, or waves which these
trapped particles which are particularly important
either transfer momentum preferentially to one of the con- in the
production of bootstrap current. Trapped particles
fined species in one direction along the magnetic field lines in a
tokamak are an effect which arises from the toroidic-
or, alternatively, waves which preferentially heat the elec- ity, since
this is what creates the higher magnetic field at
trons or ions in the direction perpendicular to the mag- smaller
major radius. Thus, the fraction of the total parti-
netic field, thus changing their collisionality. The different cles which
are trapped is proportional to r/R, ratio of the
current drive methods tend to drive current in different minor
radius to the major radius, which is the inverse of the
regions of the plasma, meaning that appropriate combina- aspect
ratio. Thus, in order to maximize bootstrap current,
tions of them can modify the shape of the current profile to which is in
some sense free, reactor designs are sometimes
achieve better stability and less energy leakage. driven
toward lower aspect ratios, in the general vicinity
of about 3
or less. However, this is somewhat offset by the
Bootstrap Current. All of the continuous direct current fact that
the fraction of the current that can be driven by
drive techniques described above have in common that bootstrap
effects is in part determined by (R/r )0.5 . In ad-
their efficiency for driving current is significantly less than dition,

there are also other constraints, such as the scaling
the efficiency of neutral beams or some sorts of waves in of
confinement in some plasma regimes, which may drive
heating the plasma. Thus, it would be more appealing if the optimum
aspect ratio to larger values (6, 7).
much of the plasma current could be driven by the ther-
Comprehensive design studies of a number of variants
mal energy of the plasma, rather than being driven by a of tokamak
reactors for power plants have been carried out
less efficient externally applied drive. A thermal energy at the
University of California at Los Angeles, resulting in
drive would be fed primarily by the efficient neutral beam a number of
designs dubbed with the nameAries (8). These
or wave systems heating a sub-ignited plasma, or by the for the
most part anticipate providing much of the plasma
energy of the fusion-produced alpha particles in an ignited current
with the bootstrap effect, and the balance with one
plasma. or more of
the direct-current drive techniques.
A current drive mechanism arising from the bulk ther-
mal plasma was theoretically predicted in the 1970s, and Plasma
Exhaust System. A long pulse or steady-state
was first identified in the early 1980s on the octupole (a tokamak
reactor requires something better than a sim-
type of magnetic confinement device) at the University of ple
limiter to handle the power coming out the edge of
Wisconsin. It was named the bootstrap current because it the
plasma, and to dispose of the helium ash which re-
was generated by the plasma itself through a sort of dy- sults from
fusion reactions. This is done by altering the
namo effect as particles press against magnetic field lines. closed field
line topology of the normal tokamak configura-
The strength of this effect depended upon the plasma pres- tion. By
placing an additional shaping coil in the vicinity
sure gradient. Thus, it was not until 1986 that bootstrap of the
plasma top, bottom, or in both locations, it is possible
current was found experimentally in a tokamak when high to redirect
the outer field lines so that a thin layer of the
confinement plasmas with steep gradients were produced plasma on
the edge of the plasma cross section is diverted
on the TFTR tokamak at Princeton. The existence of the out of the
main plasma chamber into a divertor chamber
bootstrap current was subsequently verified on the other where the
plasma, and the helium ash it entrains, are con-
major tokamaks of the world, and it is now expected to verted to
neutrals through a combination of encountering
supply much of the current for future tokamaks. One or a region of
higher neutral particle density and of directly
more of the other direct-current drive techniques may also striking
plates of carbon or some other material. The di-
be needed to provide lesser amounts of localized current vertor
region is baffled from the main chamber to reduce
drive to produce the optimum profile for confinement and reflux, and
much of the neutralized plasma outflow can be
stability. pumped away
as gas. Figure 5 displays one possible diver-
Because the strength of the total magnetic field in a tor
configuration.
tokamak increases sharply as one progresses inwards in Ensuring
that the divertor plates have an economi-
major radius, particles with sufficiently small ratios of par- cally
attractive lifetime is a significant engineering prob-
allel velocity to perpendicular velocity with respect to the lem which
is not yet fully solved for fusion power reactors;
field will become trapped in local magnetic mirrors, much the power
density in the scrapeoff plasma flowing toward
as happens to the main plasma in mirror confinement ma- the
divertor will be high, causing thermal problems, and
----------------------- Page 179-----------------------
18 Reactors, Fusion
or
helium are examples of possible choices. This extracted
heat
will then be used to produce electricity, probably by
generating steam to drive a conventional turbine.

The
neutrons, which carry the other 80% of the fusion
power,
will be slowed down and captured in a blanket sur-
rounding
the tokamak. In the process, their kinetic energy
will be
converted to heat, which will be further augmented
by
exothermic nuclear reactions with 6 Li. This heat can
then be
removed from the blanket to produce electricity, ei-
ther
through steam turbines, or possibly by more efficient
techniques taking advantage of the fact that the blanket

may
contain a liquid metal.
This
blanket must also breed enough tritium to at least
replace
the amount consumed in nuclear reactions by the
plasma,
and it is preferable that it produce somewhat
more in
order to start up additional fusion power plants.
The
reactions of neutrons with lithium isotopes which pro-
duce
tritium were discussed in an earlier section. Proposed
blanket
designs have of necessity included either natural
lithium
or 6 Li, but they have differed in what neutron mul-
tipliers
were added to the mix. The most commonly as-
sumed
multiplier is beryllium, which undergoes an (n,2n)
reaction, but heavy multipliers such as lead have also been
considered. The chief difficulty with some heavy multipli-

ers is
that they tend to produce radioactive daughter nuclei
with
undesirable half-lives. Beryllium does not pose this
difficulty.
Figure 5. Example of tokamak plasma with top and bottom di-
vertors. The divertor coils produce open flux surfaces after the
Maintenance and Materials. Developing techniques to
separatrix field line, allowing the outer plasm to leave the main maintain
and repair fusion power plants is an area which
chamber to strike target plates. This reduces the influx of impu- still
requires substantial engineering development. It may
rities to the main plasma, allows the power density striking the be
somewhat simplified by the time commercial plants
target plates to be reduced through the expansion of flux tubes,
begin
operating some decades from now by advances in
and may allow further power density reductions through radia-
robotics. Present tokamaks are commonly made of stain-

tion in the divertor region.
less
steel alloys or inconel, primarily because they are rel-
atively
inexpensive, and can be machined and formed by
the energetic particle flux can cause high sputtering rates drawing
upon a large body of fabrication experience. These
on the divertor plates. Some of the approaches being pur-
materials, however, would stay significantly radioactive for
sued to ameliorate these problems include reducing the many
years after removal from a fusion plant, meaning
incident power density by reducing the angle of incidence, they
would require storage. They would not be volatile or
and by expanding the flux envelope of the diverted plasma, very
reactive, so there would be little threat of contaminat-
and reducing the sputtering rate by introducing strongly ing
water or air.
radiating heavier impurties such as noble gases into the
However, one of the advantages of fusion relative to fis-
divertor chamber plasma to lower the temperature of the sion is
that one has some degree of freedom to choose mate-
ions so they will produce less sputtering. This deliberately rials
which reduce activation. This is not possible with fis-
enhanced radiation can also be used to disperse energy at sion,
since the nuclear waste products are produced by the
reduced intensity over a larger expanse of material.
fissioning of the fuel itself. Accordingly, fusion power plants
probably
will use different materials which will have lower
Tritium Breeding and Thermal Conversion Blanket. About initial
levels of radioactivity per gigawatt year of exposure,
80% of the energy produced in a deuterium-tritium fusion and
shorter half-lives which decrease the storage time re-
reactor immediately leaves the plasma in the form of 14.1 quired
for removed components. Two of the low activa-
MeV neutrons, since plasmas with dimensions and densi- tion
materials which have been considered by studies of
ties appropriate to reactors are almost perfectly transpar- the
Aries group (8) include a vanadium alloy with small
ent to high-energy neutrons. The remaining 20% of the en- amounts
of chromium and titanium, and silicon carbide. In
ergy carried by the fusion alpha particles heats the plasma, their
reactor models, they found that the total radioactiv-
and eventually either leaks from the plasma as electromag- ity per
watt of reactor thermal power remaining in a fusion
netic radiation that will be absorbed and converted to heat power
plant one year after shutdown would be a factor of
by the first wall, or it flows in the scrapeoff plasma to the 103
lower than for a comparably rated fission breeder reac-
divertor, where it heats surfaces. Thus, all of the alpha par- tor, if
the fusion plant was built of HT-9, a ferritic steel. If
ticle energy is ultimately collected on surfaces. These sur- the
plant were instead constructed of the vanadium alloy,
faces will be cooled with some fluid or gas, of which lithium the one-
year radioactivity was depressed to a level about
----------------------- Page 180-----------------------
Reactors, Fusion 19
105 less than for a fission breeder, and if silicon carbide was the sphere
has proven very difficult. In addition, with some
used for the fusion reactor, then the one-year radioactivity types of
drivers there may be preheating of the fuel due to
of the fusion reactor declined further to a value about 108 penetration
by driver ions, and laser-driven instabilities
3 may
degrade the symmetry of the compression. Ways of
less than the fission breeder. If D- He reactors ever proved
feasible through lunar mining, then the silicon carbide de-
ameliorating these problems are being studied.
sign would give a one-year radioactivity that would be more The
other implosion method is indirect drive. In this ap-
than 1010 less than for a fission breeder. proach the
fuel capsule is not struck by the primary driver
These and other low-activation materials require engi- beams (4).

Instead, the capsule is placed inside a Hohlraum
neering validation through tests exposing them to large which is of
considerably larger dimensions than the cap-
neutron fluences. In addition, some materials, especially sule, as
depicted in Fig. 7. This Hohlraum is made of a
silicon carbide, require progress in fabrication technology. high atomic
number material, and has openings through
The structures normally made of silicon carbide are signif- which the
driver beams can enter so as to directly strike
icantly smaller than would be required for a reactor. the inner
wall of the Hohlraum, but not the fuel capsule. A
portion of
the driver energy striking the Hohlraum is con-
verted to
soft X rays, and they in turn drive the implosion of
INERTIAL CONFINEMENT the fuel
capsule. An advantage of this method compared to
direct
drive is that quite uniform irradiation of the capsule
Parameter Regime with the
soft X rays can be achieved even with anisotropic
A practical fusion power plant fusing deuterium–tritium primary

driver beams, enabling fewer beams to be used.
fuel requires temperatures of over 10 keV, with a fuel den- In
addition, there is a large body of experience available
sity, n, and an energy confinement time, τ, such that their with this
sort of indirect drive from the nuclear weapons
product lies in the range of nτ = 1014 to 1015 3 program.
The majority of research being conducted on iner-
s/cm . A prac-
tical fusion reactor using the magnetic confinement ap- tial
confinement in the US today follows the indirect drive
proach discussed earlier will likely operate in a regime with approach.
a density of about 1014 fuel nuclei/cm3 , and a confinement
time of about a second or so. Inertial confinement operates

Inertial
Confinement Drivers
at the other extreme of the parameter range. Because the
small nuclear explosions in an inertial confinement reac- Laser
Drivers. Over recent decades, a number of classes
tor would occur with high-velocity fuel particles travers- of drivers
have been considered, and in some cases tested,
ing very small distances, the confinement time would be for
inertial confinement applications. Experiments have
of the order of 10−10 s, requiring that the target fuel be been driven
by neodymium-glass lasers, and krypton flu-
compressed to about 1025 3
fuel particles/cm . oride gas
lasers have been studied as possible primary
This compression must be carried out by a strong im- drives.
Most of the laser drivers suffer from low efficien-
plosion. A typical inertial confinement target capsule is a cies for
conversion of electricity into laser light. In the case
small sphere formed of an ablator material which is lined of KrF
this efficiency is about 6% to 8% (9). This would
with solid deuterium–tritium fuel. The central spherical require
that the target capsule produce 140 to 160 times
cavity thus formed, which composes most of the target cap- more
energy than was in the driver in order to make a
sule’s volume, is filled with deuterium–tritium gas. The im- power
plant feasible. Such high gains are vastly beyond
plosion occurs when the ablation shell is illuminated with current
achievements, but may, in principle, be achievable.
a brief burst of extremely high-energy density, on the order Laser
diodes have much higher efficiencies of as much as
of 10 TW/cm2 . This causes the outer portions of the abla- 60%. If the
prices of these can be reduced by a large factor,
tion shell to leave it with high momentum which, on the and if

techniques can be found to interface them with the
average, will be directed radially outward from the shell.
thermonuclear environment of a reactor cell, then these
Conservation of momentum requires that an equal inward- may become
attractive.
directed momentum be imparted to the remaining target,
driving the collapse. For a reactor to have energy gains of
economic interest, it is also necessary that this implosion Ion
Drivers. The other classes of drivers are light ions
proceed isoentropically, with as little preheating ahead of and heavy
ions. In the light ion approach, extremely high
the compression wave as possible. This, in turn, requires currents of
lithium ions are produced by an array of diode
that the compression be uniform to a level of about 1% over sources and
accelerated to energies of hundreds of keV to
the entire sphere. Figure 6 shows one possible design for a a few MeV.
Lithium diode drivers have been studied at the
target capsule. Sandia
National Laboratory. The other method would use
lower
currents of ions from the high mass end of the peri-
odic table
such as xenon or cesium, which would be acceler-
Direct and Indirect Drive
5
ated to
energies of 2 GeV to 10 GeV. About 10 A of beam at
The most obvious way to compress a target sphere is to these
energies would be required to achieve the necessary
shine the driver energy directly onto the ablation sphere.
illumination intensity. Although the required pulse dura-
The driver might be an array of lasers, light ion acceleration is
very brief, such currents are well beyond the capabil-
tors, or heavy ion accelerators. One of the principal chal- ities of
existing high-energy accelerators. High-current ac-
lenges in implementing this approach is that obtaining the celerator
concepts such as inductive linacs are being stud-
required high degree of uniformity in the driver power at ied.
----------------------- Page 181-----------------------
20 Reactors, Fusion
Figure 6. Target capsules for inertial confinement fusion are
compressed by ablation from the outer shell. After ignition in the
center, the fusion burn region will need to spread outward to pro-
duce significant energy gain.

Inertial Confinement Plant Systems
An
inertial confinement fusion power plant would consist
of a
target chamber, the driver system of ion accelerators or
lasers, and a target fabrication system introducing several
capsules (and Hohlraums if needed) per second into the
chamber. A lithium layer would breed tritium and would
convert the fusion energy to heat, which could then be con-
verted to steam to drive turbines. Concepts for this lithium
layer have included liquid metal flowing over the inner sur-
face
of the target chamber, as well as jets of liquid metal
around the target.

Figure 7. Target placement for indirect drive inertial confine-
ment fusion. High-power laser beams strike the inner surfaces of
the Hohlraum producing X rays which compress the target cap-
PROGRESS IN FUSION ENERGY
sule. A variant of this would use heavy ion beams striking ab-
sorbers at the ends of the Hohlraum. Indirect drive requires much
Over
the past four decades, large advances have been made
less perfect symmetry in the drivers than does direct drive.
in
understanding the physics of plasmas suitable for a mag-
netically confined reactor, and the technologies for heating

the
plasmas have similarly progressed. In this period the
power released through fusion reactions has increased a
factor of 108 , with over 10 MW produced by the TFTR toka-
Both of the ion drive techniques have the major advan- mak
funded by the US Department of Energy at Princeton
tage, relative to most of the laser approaches, that their
University. Similarly the energy leakage of the ions has
efficiency for conversion of electricity to ion energy can
dropped from being many times higher than that theoreti-
be quite high, 25% to 30%. With the heavy ion approach
cally predicted to values which in some cases are at or near
a target energy gain of a factor of 40 might be sufficient the
theoretically best confinement that could ever occur.
for a feasible plant, which is roughly a factor of 4 below
Temperatures in tokamaks have climbed from a few hun-
that required for a KrF laser driver. On the other hand, dred
electron volts to as much as 40 keV to 45 keV in TFTR
for either technology the final path of approach of the ion and
in JT-60U in Naka, Japan. Energy confinement times
beams toward the Hohlraum will be after the final focus- have
grown from a few milliseconds to a second or more on
ing elements of the ion optical system. Maintaining the
Japan’s JT-60U and Europe’s JET tokamaks. The power
necessary illumination intensity and uniformity will be dif- of
neutral beam systems heating the plasma has climbed
ficult as these high space charge density beams converge from
the tens of kilowatts prevalent a four decades ago to
toward even higher concentrations of space charge. Space 40
MW on TFTR and JT-60U. In recent years, neutral beam
charge effects, which arise because of the coulomb repul-
heating systems based upon negative hydrogen and deu-
sion between like charges, are strongly defocussing. In re-
terium ions have been deployed in Japan on the JT-60U
cent years, the vapor pressure in the most common target
tokamak in Naka and the Large Helical Device at Tokai.
chamber concepts is much higher than in earlier scenarios.
Negative ion based beam systems can maintain good neu-
This may result in sufficient plasma being produced along
tralization efficiency at the much higher megavolt beam
the beam trajectory to partially or totally cancel the ef-
energies which will be needed for magnetically confined fu-
fect of the beam space charge forces, and thus simplify the sion
reactors. The JT-60U negative ion based neutral beam
problem of obtaining small beam diameters at the target. has
operated at over 400 kilovolts.
----------------------- Page 182-----------------------
Reactors, Fusion 21
New superconducting tokamaks are nearing completion

in China, India, and South Korea, while a compact stellara-
tor is being built at Princeton in the U.S. to study a con-
figuration combining some aspects of both stellarators and
tokamaks. A coalition consisting of the European Union,
Japan, the United States, Russia, South Korea, China, and
India is building the International Tokamak Experimen-
tal Reactor (ITER) in Caderache, in the south of France.
This will be the first fusion device to demonstrate useful
amounts of fusion power, and will test many of the tech-
nologies necessary for a real fusion reactor. It is expected
to begin operations in the latter part of the second decade
of this century.
Inertial confinement fusion has made advances in un-
derstanding the physics of matter at the very high pres-
sures and densities required for this approach, and under-
standing has been gained about the ways in which energy
and momentum are transferred to the target capsule. A
major new laser-driven test bed, named the National Igni-
tion Facility, is nearing completion at Lawrence Livermore
National Laboratory in California. It is expected to produce
target ignition early in the second decade of this century.
Meanwhile, investigations continue on more efficient and
compact heavy ion driver beams for inertial confinement
fusion reactors.
Producing an economically attractive fusion power
plant will require more work for either the magnetic or
inertial confinement approaches. Based upon the present
state of these fields, it appears that the magnetic confine-
ment approach enjoys a clearer path forward to a reac-
tor, but this could change in the future depending upon
progress in inertial confinement.
BIBLIOGRAPHY
1. J. Wesson Tokamaks, Oxford: Clarendon Press, 1987.

2. J. G. Cordey R. J. Goldston R. R. Parker Progress toward a
Tokamak fusion reactor, Phys. Today , 22–30, 1992.
3. L. J. Wittenberg J. F. Santarius G. L. Kulcinski Lunar source
of 3He for commercial fusion power, Fusion Technol. , 10: 167,
1986.
4. J. Lindl Development of the indirect-drive approach to inertial

confinement fusion and the target physics basis for ignition and
gain, Phys. Plasmas , 2 (11): 3933–4024, 1995.
5. R. J. Goldston P. H. Rutherford Introduction to Plasma Physics ,

Bristol and Philadelpia: Institute of Physics, 1995.
6. L. R. Grisham et al. The scaling of confinement with major ra-
dius in TFTR, Phys. Rev. Lett. , 67: 66–69, 1991.
7. L. R. Grisham et al. Scaling of ohmic energy confinement with
major radius in the Tokamak fusion test reactor, Phys. Plasmas ,
1: 3996–4001, 1994.
8. R. W. Conn et al. Economic, safety, and environmental aspects
of fusion reactors, Nucl. Fusion , 30: 1919–1934, 1990.
9. W. J. Hogan R. Bangerter G. L. Kulcinski Energy from inertial

fusion, Phys. Today , September, pp. 42–50, 1992.
LARRY R. GRISHAM
Princeton University, Princeton,
NJ
----------------------- Page 183-----------------------
} { { } }
HOME
ABOUT US
●
CONTACT US
HELP

Your Librarian
Profile
Gas Turbine Power Stations

Standard Article

Aristide F. Massardo 1
1Università di Genova, Genova, Italia

reserved.
Search
DOI: 10.1002/047134608X.W3019
❍ Acronym Finder
Abstract
Thermodynamic Analysis of The Gas Turbine Power Cycle
Existing Gas Turbines
Gas Turbine Components and Technological Innovations
Advanced Cycles
Summary

%20Conversion/W3019.htm17.06.2008 16:39:44
----------------------- Page 184-----------------------
258 GAS TURBINE POWER STATIONS
GAS
TURBINE POWER STATIONS
Dramatic increases in unit size and performance, coupled

with
significant price reductions and very low emissions, have
made
the large gas turbine the dominant form of electric
power
generation.
In
the period between 1970 and 1993 worldwide orders for
pure
steam (Rankine cycle) power plants decreased contin-
ually
while those for combined cycle and single cycle gas tur-
bine
(Brayton cycle) power plants increased. These strong
sales
trends have carried over into the past few years, and
show
every sign of continuing into the next century.
Heavy frame industrial gas turbines, designed specifically

for
central and distributed power plants, and aeroderivative
gas
turbines—modified jet engines—are in fact playing an in-
creasingly important role in the generation of electricity.

Key
factors in the success of gas turbine electricity genera-
tion
have been very low emissions coupled with low values of
the
unit cost of fuel brought about by burning natural gas.
This
value, coupled with high gas turbine efficiencies, tends
to
minimize the unit cost of the electricity produced, creating
the
most significant boost for gas turbine sales.
To
summarize the main positive factors of gas turbine
power
plants:
1. Low specific cost ($/kWe)
3
2.
High specific power (kWe/kg/s; kWe/m )
3.
Flexibility, availability, and reliability
4.
Short build-up time (preassembled module)
5.
Short start-up time
6.
Remote control (i.e., pipelines)
7.
Water cooling not necessary
8.
Wide power range commercially available (100 kWe to
300 MWe)
9.
Very high efficiency in the cogeneration configuration
10.
Waste gas temperature well suited for combined cycle
(Rankine bottoming cycle solution)

11.
Very low CO, unburned hydrocarbon (UHC), and NO
emissions
while,
as a negative factor, we can consider the need for clean
fuel
(kerosene, distillate oil, natural gas).

----------------------- Page 185-----------------------
GAS TURBINE POWER STATIONS 259
In this article, after a simplified thermodynamic analysis

1800
of the Brayton gas turbine cycle, the data (efficiency, specific )
1700 Advanced G.T regime with cooled
C
°
1600 single crystal metallic,ceramic,or
work, turbine outlet temperature, pressure ratio, costs) of ex- (
e
1500 coated carbon/carbon composite
isting gas turbines are presented and discussed. r
u
1400 turbine blades
t
A detailed analysis of the main components of gas turbines a
r
1300
e
Approximate upper limit
and the technological innovations, particularly regarding ma- p
1200 Open-cycle gas turbine
m
for metallic heat
terials, blade cooling, turbomachinery design, and combustion e
1100 trend (OBC)
t
t
source exchanger
chamber emissions is also presented. e
1000
l
n
In conclusion, the cycle innovations under investigation i
900
e
n
800
(regenerative cycle, intercooling, reheating, steam injection, i
b
Closed-cycle gas
r
700
humid air gas turbine, etc.) are presented, including a com- u
turbine trend (CBC)
T
600
parison with actual and short term combined cycle perfor-
500
mance.
1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 2030
Year
THERMODYNAMIC ANALYSIS OF
Figure 2. Open and closed cycle turbine inlet temperature trend.

THE GAS TURBINE POWER CYCLE
This parameter is one of the major factors in OBC efficiency advance-
ment. It has been continuously increased in the last 40 years.

The gas turbine cycle refers to the Brayton (or Juole) cycle
(1,2), composed of an adiabatic compression, a constant pres-
sure heating, an adiabatic expansion, a constant pressure CBC
systems. Indeed, in the CBC constraints are applied to
cooling. This is a description of a closed Brayton cycle (CBC)
maximum cycle temperature, due to the use of a heat ex-
(3), however, gas turbines for power generation operate with
changer and heating system—see Fig. 2—(CBC maximum
an open Brayton cycle (OBC), see Fig. 1. In the OBC the
temperature 850–900#C) (4). While in the OBC, higher maxi-
working fluid is air, while in the CBC various working fluids mum
temperatures (1400–1450#C) are possible thanks to the
can be used (helium, nitrogen, CO2, air, etc.); the OBC com- use
in the hot section (combustor and turbine blades) of ad-
pressor inlet conditions are coincident with ambient condi-
vanced cooling techniques. CBC systems can also utilize dirty
tions (ambient pressure and temperature), while CBC com-
fuel thanks to the external combustion solution, while for the
pressor inlet pressure can be higher than ambient pressure OBC
the fuel must be clean because the combustion gases are
(pressurized CBC system). In the OBC, the heating phase is
utilized inside the expander thus increasing the high temper-
realized through the combustion of fuel inside the system
ature corrosion/erosion effects.
(combustion chamber), and the gases are directly utilized as
For electricity generation an OBC is usually utilized, while
working fluid in the expander. In the CBC, the expander a
CBC is used for special applications: space and underwater
working fluid is clean because the heating is realized with a
power generation; gas-cooled nuclear reactor power systems;
heat exchanger (external combustion cycle). Finally, in the
cryogenic fluid (LNG, LH2) applications (5,6,7,8).
OBC, the heat is directly wasted to the environment without
As an example, Fig. 3 shows the performance of an ideal
the need for a gas cooler as in the CBC. OBC
utilizing the perfect gas hypothesis or considering the
Therefore, the OBC does not need heat exchangers and the
specific heat cp as a function of the temperature. Efficiency
cost, weight, and dimensions are greatly reduced compared to and
specific work (the ratio between the power generated and
inlet air mass flow rate) are represented versus cycle pressure
ratio #, while maximum cycle temperature T3, is considered

Open Brayton Cycle (OBC) as
a parameter. The efficiency is always an increasing func-
tion of # and the

influence of T3 is negligible; while specific
2 3
work curves always present a maximum value, and the effect
CC of
T3 is quite evident. When the gas perfect hypothesis is uti-
C T
lized, the specific work is lower than for cp # f (T) conditions,

Wt – Wc = Wu
while regarding efficiency, the difference is quite reduced.
1 4
The next step to improve the thermodynamic analysis is
m cooling the
introduction of the irreversibilities (nonideal cycle). The
irreversibilities can be associated to (see Fig. 4):
1. Nonisentropic compression and expansion: compressor

2 3
and expander efficiency lower than unit
H
C T
2. Pressure losses in the inlet section-filter, inlet pipes, in
W – W
the combustion chamber, in the connecting pipes be-
GC t c
tween the compressor, combustion chamber and ex-
1 4
pander, in the silencer, in the waste heat recovering
system, if present
Closed Brayton Cycle (CBC)
3. Thermal losses in the gas turbine hot section
Figure 1. Gas turbine cycle lay-out. C: compressor; T: turbine; CC:
4. Chemical losses correlated to incomplete fuel combus-
tion
combustion chamber; H: heater; GC: gas cooler; W : compressor work;
c
W : turbine work; W : useful work.

5. Compressed air leakages
t u
----------------------- Page 186-----------------------
0.7
700
0.6
600
T3 = 1300 C° )
0.5
500 g
Perfect gas k
y η
(
c
T = 1000 C° k
n
3 r
e 0.4
400 o
i
w
c
i
f
c
f
i
f
E
i
e
0.3
300 p
T3 = 1300 C°
0.2
Specific work 200
Figure 3. Ideal open Brayton cycle

0.1
100
(OBC) performance: perfect gas hypothe-
2 6 10 14
18 22 26 30 34 38 42 46 50
sis and maximum cycle temperature in-
β
fluence is evident.
Pressure ratio
6. Hot section-combustion chamber, nozzle, and rotor

When cooling is utilized, the expansion is no longer adia-
blades-cooling batic,
and taking into account the mixing between the main
7. Mechanical losses (ventilation, bearings, auxiliary sys- flow
and the cooling air, the specific entropy can be reduced,
tems, etc.) while
the mixing is highly irreversible, and this produces an
8. Electrical losses
efficiency reduction.
The influence of nonunitary turbomachinery efficiency is

Regarding the irreversibility associated with blade cooling, one
of the most important differences between the ideal and
it is necessary to point out that it has always been the prac-
nonideal cycle; if the network of the cycle is considered:
tice to pass a quantity of cooling air over the turbine disks
w
and blade roots, and a substantial quantity of coolant to the
wu = wt − wc = wis ,t · ηis ,t − is ,c
nozzles and the rotor blades. Indeed, to obtain a very high
ηis ,c
gas temperature coupled with a satisfactory blade material
temperature, cooling must be utilized (combustor gas exit and
taking into account that compressor and turbine adia-
temperature 1100–1300#C, material temperature 800– batic
work are quite similar, the effect of turbomachinery in-
950#C).
efficiency is very clear. If the efficiency versus pressure ratio
Combustor pressure loss
p3 p4
Cooling
mass flow
Nonisentropic
Heat loss 4 Outlet pressure
compression
loss
Mass loss
4is
Nonisentropic
2
expansion
2is
T
p2
Actual expansion line
p0 = pamb.
Exhaust gas
p1
1
Figure 4. Entropy diagram (T, s) of a real Inlet pressure loss

open Brayton cycle (OBC): cycle irreversibilit-
s
ies are shown.
----------------------- Page 187-----------------------
0.7
T = 1000 °C β = 40
0.6 T = 1300 °C η p = 1
β = 40
30
22
η = 0.9
0.5 p β = 40
16
12
y 0.4 8
c
n
e
i β = 40
c
i
f
f 0.3
E
1612 8
22
0.2
β = 4 β = 4
η = 0.8
0.1 β = 30 p
Figure 5. Efficiency vs. specific work of

0
OBC: compressor and expander efficiency
0 100 200 300 400 500
600 700
and maximum cycle temperature influence

Specific work (kJ/kg)
is evident.
is analyzed in the nonideal OBC, a different behavior can be only

by engine manufacturers, who can resort to the informa-
found compared to the results illustrated in Fig. 2. In the non- tion
provided by exhaustive measurements to calibrate their
ideal cycle, the efficiency is not a monotonic function with #,
computer programs and verify the accuracy of their design
but it presents a maximum. The pressure ratio value where
methods. While for general thermodynamic analyses it is use-
# is maximum depends on maximum temperature value T . ful
to use much simpler computing procedures, concentrating
3
Figure 5 shows the nonideal OBC performance-efficiency the

previously mentioned loss sources in few parameters.
versus specific work where the pressure ratio values, the
Another useful strategy is the use of performance data
maximum temperature, and the turbomachinery efficiencies about
gas turbines currently produced around the world,
are those normally utilized in OBC systems. When ideal tur- which
can be found either in handbooks and catalogues issued
bomachinery are considered (#p # 1), the influence of T3 is by
specialized publishers, or in technical literature, or in com-
present only for specific work data, while when #p # 0.8, the pany
publications. As an example, Fig. 6, based on the Gas
OBC efficiency is greatly reduced and optimum # value is
Turbine World Handbook (9), shows the efficiency values ver-
around 12 # 13, while the T3 influence on # is quite evident. sus
power for most of the world’s existing gas turbines. The
If #p # 0.9, typical value for large size industrial gas turbines,
output power data shown are for base load at ISO conditions
a large difference between the pressure ratio value for opti-
(15#C, 60% relative humidity, sea level, burning natural gas).
mum efficiency (# # 30 depending on T3 value) and pressure The
efficiency increase versus power is quite evident—for
ratio value for optimum specific work (16 # # # 22) exists. In power
under 3 MW the efficiency is under 25%, while for aer-
this way a design parameter as pressure ratio shows a very
oderivative turbines, in the range between 10 to 40 MW, the
wide range, and this is very important for the gas tubine de-
efficiency is in the range 35–40%. Large industrial (heavy-
signer. Indeed, maximum efficiency (high #) corresponds to duty)
turbines show efficiency values in the range 35–37%. It
minimum fuel consumption (aeroengine solutions), while the is
important to point out that aeroderivative gas turbines are
maximum specific work condition (low #) minimizes the sys- based
on aeroengine gas turbines, while heavy-duty gas tur-
tem costs. More detail about this aspect will be provided in bines
are directly designed for electricity generation purposes.
the next section. The
difference between these two different designs is shown
more
clearly in Fig. 7, where the existing gas turbine pressure
EXISTING GAS TURBINES ratio
is plotted versus power. Aeroderivative systems, taking
also
into account the very high T3 values utilized in aeroen-
The calculation of the thermodynamic cycle of a real OBC gas
gines, show very high pressure ratio values-# # 20 (maximum
turbine, as already stated, is a very complicated matter. Just
efficiency design—see Fig. 5—minimum fuel consumption),
to mention some of the phenomena to be accounted for, it while
lower values for the pressure ratio are utilized for in-
must be considered that: air flows bled from various points of
dustrial gas turbines (maximum specific work-minimum
the compressor are reinjected in the main stream after cooling plant
cost; high outlet turbine temperature T4 value-combined
turbine nozzles, rotor blades, and disks; the compression and cycle
application), considering also the reduced T3 values uti-
expansion process in turbomachinery is neither adiabatic nor lized
in heavy-duty gas turbines.
at constant polytropic efficiency; leakages, pressure, and heat
For small gas turbines (P # 10 MW) # is between 5 and
losses occur in various parts of the machine. Accurate evalua- 15,
due to the impossibility, at commercial costs, of realizing
tion of such complex phenomena can be completely performed
small, high pressure and high efficiency turbomachinery. Fi-
----------------------- Page 188-----------------------
0.45
Aeroderivative Siemens Rolls-Royce
GE Westinghouse
Allison
0.40
Hitachi
Zosen
0.35
Mashproekt
Solar Turbines
ABB
y 0.30
c
n
e
Heavy-duty
i Pratt & Whitney
c
Mitsubishi HI
i
f
European GT
f 0.25
E
Turbomeca
0.20
Small size
0.15
Kawasaki
Figure 6. Efficiency vs. power of existing
gas turbines (ISO conditions). Three dif- 0.10
0.1 1
10 100 1000
ferent groups are evident: small size, aer-
oderivative, and heavy-duty gas turbines.
Power (MW)
nally, it is worth noting that only two large gas turbines pres- heavy-
duty values are higher than aeroderivatives, but this
ent a very high pressure ratio (# # 30): this value is corre-
difference is not so evident as in the case of the previous data
lated to the nonsimple OBC configuration (reheating cycle),
analyzed (Figs. 7 and 8). Large industrial gas turbines (P #
analyzed in the final part of this paper. 50 MW)
show TOT values higher than 500–520#C, which are
Figure 8 shows a simultaneous increase in specific work
particularly well-suited for a combined cycle solution.
and efficiency of the existing gas turbines. At fixed specific To
complete the analysis of existing gas turbines Fig. 10
work, aeroderivative gas turbines show higher efficiency, shows
the average prices listed only for equipment (not turn-
while at fixed efficiency industrial gas turbines show higher key).
Prices per electrical kilowatt generated shown include a
specific work values (see also Fig. 5). basic
gas turbine generating package with a single-fuel gas
Another thermodynamic feature of existing gas turbines is
turbine, an air cooled electric generator, standard starting
reported in Fig. 9, where turbine outlet temperature (TOT) is system,
skid enclosure, standard inlet and exhaust ducts, ex-
shown. This parameter is particularly important when a gas haust
silencer, and standard controls. Not included are any
turbine is utilized as a topping system in combined cycle con-
financial or debt service costs, contingency and insurance
figurations (the bottoming system is a Rankine cycle heated costs,
legal and environmental permitting costs, and so on.
with the heat recovered from the gas turbine exhaust gas). The
price per kilowatt drops as the unit size (output) of the
All the TOT’s data are in the range 400–600#C, and the gas
turbine increases as a function of the economy of scale.
40
Westinghouse EL Rolls-Royce
Aeroderivative
35
Reheating
turbine
Orenda Aerospace
30
ABB
GE
o
i
t 25
a
r
Small size
e
Allision
r β
u 20 Hitachi
zosen
s
s
e
r
P
15 Kawasaki
Pratt &
Mitsubishi H.l.
10 Whitney
Heavy-duty
5
10 Mashproekt
10 Siemens
Nuovo pignone
Turbomeca
Solar Turbines European GT
Figure 7. Pressure ratio vs. power of existing 0
gas turbines. Three different groups are evi- 0.1 1
100 1000
dent: small size, aeroderivative, and heavy-
duty gas turbines.
Power (MW)
----------------------- Page 189-----------------------
0.45
Westinghouse Rolls-Royce Siemens
Allison
Mitsubishi HI
0.40
GE
Solar Turbines
European GT
0.35 Mashproekt
ABB
Orenda
y Aerospace Nuovo Pignone
c 0.30
n
e
i
c Turbomeca
i
f
f 0.25
E
0.20 Pratt & Whitney

0.15 Kawasaki
Figure 8. Efficiency vs. specific work of ex-
isting gas turbines. In this case the three dif-

0.10
ferent groups (small size, aeroderivative,

100 150 200 250 300 350 400
450 500
and heavy-duty gas turbines) are not ev-

Specific work (kJ/kg)
ident.
Obviously prices can vary significantly depending upon the •

Combustion and pollution control
purpose of the plant equipment, geographical area, special •
Compressor and turbine aerodynamic
site requirements, and competitive market conditions. •
Materials technology and blade cooling techniques
•
Fabrication methods
GAS TURBINE COMPONENTS AND •
Instrumentation and diagnostic systems
TECHNOLOGICAL INNOVATIONS •
Testing approaches
The main components of OBC gas turbines are (see Fig. 1): As
previously shown one of the major factors in OBC effi-
an axial flow compressor, combustion chamber, and axial flow ciency
advancement has been the continuously increasing
turbine (expander). Only for small systems are the compres- level of
turbine inlet temperature (Fig. 2), made possible by
sor and/or expander of the radial type (centrifugal compres- materials
technology and turbine blade cooling techniques.
sor; centripetal turbine) instead of the axial type. There
are several approaches to the problem associated
To design more effective components, research and devel- with high
gas temperature; in general they can be categorized
opment activities have been carried out mainly in the follow- as
developing suitable materials and cooling systems.
ing areas:
Materials
• Analytical and design methodologies (computational The

components that suffer most from a combination of high
fluid dynamic, optimization techniques)
temperatures, high stress, and chemical attack are those of
650
Pratt & Whitney Allison Nuovo Pignone
ABB
Siemens
600 Hitachi Zosen
Kawasaki HI
550
Mitsubishi HI
) 500
C
°
(
p
m 450
e
Westinghouse
t
t Turbomecca
s
u 400
a Solar Turbines Rolls-
Royce
h
x Orenda Aerospace
E 350 GE
300
European GT Mashproekt
250
Figure 9. Turbine outlet temperature of ex-

200
isting gas turbines. In this case the three dif-
0.1 1 10 100
1000 ferent groups (small size, aeroderivative, and
Power (MW)
heavy-duty gas turbines) are not evident.
----------------------- Page 190-----------------------
1000
900
800
700
)
W
k
/ 600
$
(
t
s
o 500
c
c
i
f
i
c 400
e
p
S
300
200
Figure 10. Specific cost vs. power of 100
some existing gas turbines. Not included

are any financial or debt service costs, 0
100 1,000
10,000 100,000 1,000,000
contingency and insurance costs, legal
and environmental permitting costs, etc.
Power (kW)
the turbine first stage fixed blades (nozzles) and moving surface as
well as the purity and temperature of the hot gas.
blades. Heat resistent materials and precision casting are two For
surface temperatures up to around 800#C, chromium
recent advances, largely attributable to developments in air- based
diffusion coatings have proved to be quite effective and
craft engines (Fig. 11). Cobalt based alloys have been used are still
being used today in the rear blade turbine stages
for the first stage fixed blades. These alloys are now being that run
at lower temperatures. Blades operating at higher
supplemented by vacuum-cast nickel-based alloys that are
temperatures require coatings that are thicker and more com-
strengthened through solution and precipitation hardened plex in
composition. Here MCrAlY coatings (M comprises Co
heat treatment. For the moving blades, cobalt based alloys and/or Ni)
are used worldwide for this purpose. The coatings
with high cromium content are now used (4). Ceramic materi- are
applied using the vacuum plasma-spray process. All the
als are also being developed, especially for combustor and tur- coatings
mentioned are consumed while the turbine is in use.
bine inlet fixed blades. A problem here is inherent brittleness, At
present, the service life is still much shorter than actual
which causes fabrication problems and raises uncertainties blade
material and, consequently, the coatings need refur-
about the mechanical properties of ceramic materials (10). bishing
after some 25,000 h of operation.
To ensure the integrity of the blades against chemical at- A
different type of coating is the thermal barrier coating
tack and increasingly higher operating temperatures, protec- (TBC)
(11,12); this has been used for several years in high
tive coatings are utilized (11). Indeed, the intensity of chemi-
performance aero-engines. The importance of such coatings
cal attack is dependent on the temperature of the blade’s will
increase significantly over the coming years because their
1200
Aircraft 3rd Gen SC
engines
PWA 1484
1100
PWA 1480
)
C
MM200 = Hf
°
(
Trend 5 °C/Year
e
MM 247
r 1000
u IN
100
t
MM 247
a 713 – U
r
e
R 80
p U – 500
SC ALLOYS
m
B1900
e 900
t
e X – 40
d
Waspaloy GTD 111
a
DS GTD 111
l
RENE 77
B Nimonic 80 A
U
500
800
IN 738
INCO X-750
General Electric
Figure 11. Aeroengine and heavy-duty N 80 A
heavy-duty
S 816 M
252
gas turbine blade temperature trend. The
700
blade temperature increase is largely at-
1940 1950
1960 1970 1980 1990 2000
tributable to development in aircraft en-
gines.
Year
----------------------- Page 191-----------------------
use will allow the operating temperature to be increased by 2.

The internal flow (cooling passage) distribution must be
approximately 100 K. A TBC comprises at least two layers:
designed to optimize the internal heat transfer coeffi-
an outer ceramic layer that limits the flow of heat into the
cient distribution.
blade, and a metallic bonding layer that is usually a member
of the MCrAlY family. Zirconia (ZrO2) is used as the ceramic Both
aspects are very complex and a lot of work has been
layer because it has a thermal expansion coefficient similar carried
out utilizing theoretical and experimental analysis. In
to that of the materials. The ZrO2 is stabilized with yttria to the case
of internal flow detailed data have been obtained uti-
avoid detrimental phase transformation. For TBC two domi- lizing
the naphtalene sublimation technique (13) and, more
nant failure mechanisms must be taken into account: flaking recently,
the thermocromic liquid crystals technique (14).
(incompatible transient stresses between the inner and outer Both are
useful to study the very complex internal passage of
surfaces of the ceramic materials), and peeling (oxide growth cooled
blades, including pins, fins, and others.
on the bonding layer as a result of oxygen diffusion through As
shown in Fig. 2, by improving both materials and cool-
ZrO2). A satisfactory service life for a TBC is, nevertheless, ing the
turbine inlet, temperature has been steadily increased
achieved by: in the
last few decades. However, to further improve the cool-
ing
efficiency, closed loop steam cooling in the heavy indus-
trial
gas turbines designed for combined cycle operation has
• Limiting the thickness of the ceramic layer
been
under development (15), and it can be expected to enter
• Increasing the strain tolerance of the ceramic material to service
in the late 1990s. The higher heat capacity and heat
improve resistance to expansion and thermal shock transfer
capability of water permit lower metal temperatures
• Selecting a bounding layer with maximum oxidation re- (for the

same gas temperature) and hence reduced hot corro-
sistance sion and
deposition from contaminated fuels. Water cooling
also
eliminates the need for air passage through the blades
as in film
cooling, which would be subject to plugging combus-
There are two principal methods for applying these coat-
tion
gases.
ings: thermal spraying and physical vapour deposition. The
This
approach is suitable for use with a combined cycle,
second method is much more costly than the first, but yields
where
steam is readily available, but requires the use of so-
better results.
phisticated sealing technology to prevent loss of steam. How-

ever,
the losses due to bleeding high pressure air from the
Cooling
compressor for use in an air cooled turbine are eliminated.
The contribution of maximum cycle temperature to the in-
Turbomachinery
crease in the cycle thermodynamic efficiency of gas turbines
was described earlier. However, to obtain very high tempera-
Significant improvements in existing gas turbines have also
ture coupled with satisfactory blade material temperature, been
obtained through improving the performance of turbo-
cooling must be utilized. The most widespread cooling tech-
machinery. For example, the general trends seen in axial
nique is air cooling, which includes: convection, film, or tran-
compressor design over the past 30 years are: higher speeds,
spirational cooling (1). It is believed that combined convection higher
spool pressure ratio, higher pressure ratio per stage,
and film cooling offers the most promise for air-cooling ap- higher
aerodynamic loading, lower blade aspect ratios, higher
proaches. At current levels of turbine inlet temperature, three blade
row solidities, and improved configurations and blade
or four stages of a turbine rotor may be cooled, and air would shapes
(16). The trend toward higher tip speeds, higher pres-
be bled from earlier stages of the compressor to cool the later sure
rise per stage, and lower aspect ratios is seen for the
stages of the turbine. Bleeding air from earlier stages reduces first
three compressors in Table 1, each of which has a pres-
the work input required to pressurize the cooling air, with sure
ratio between 12 : 1 to 13 : 1. A time of about 20 years
beneficial effects on the net output. is
covered by these three machines while the Ge-NASA E3
As already stated, cooling is absolutely necessary to oper-

compressor is a recent design that continues the trend illus-
ate gas turbines at very high gas temperature, however, it trated
by the first three compressors. In this case the ratio
also has some negative influences on cycle efficiency. The 23 : 1 is
developed in 10 stages!
sources of losses are as follows: loss of turbine work due to Core
compressor pressure ratio values, for a single spool,
the reduction in mass flow; nonadiabatic expansion with a have been
moving steadily upward as a result of the efficient
negative reheat effect in multistage turbines; mixing of spent use of
improved materials and advanced mechanical design
cooling air with the main stream; pumping work done by the technique
(Fig. 12). Pressure ratios per stage have been in-
blades on the cooling air. creasing
partially because higher speeds are being used, but
To minimize these negative aspects, the cooling mass flow also
because the nondimensional loading has been increased.
rate must be minimized at fixed T3 and blade temperature.
To obtain this result improvements must be carried out in the
Table 1
following areas:
Compressor Pressure Corrected Tip Speed

Name
Ratio (m/s) Stages
1. Knowledge of the external (main flow) flow distribution
(velocity, pressure, turbulence) and of the heat transfer CJ805
12.5 291 17
CF6-50
13 359 14
coefficient distribution on the blade surface. Particular
CFM56
12 396 9
attention must be devoted to the transition point (or E3
23 455 10
zone) position and tip clearance/endwall effects.
----------------------- Page 192-----------------------
1.8
1.6
y
t
i
d
i
l
o
s 1.4
e
g
C
a
r
e
v
a 1.2
e
g
a
t
S
1
0.8
0.5
q 0.45
/
p
g
n
i
d
a 0.4
o
l
e
g
a
r
e
B
v
A
0.35
0.3
25
6
Spool
pressure ratio
5
20
o
o
i
i
t
4 t
a
a
r
r
e 15
t
c
r
e
u
Average aspect ratio p
s
s
3 s
e
a
r
e
p
g
l 10
o
a
r
o
A 2 e
p
v
S
A
1
Figure 12. Axial flow compressor design param-
eters trend: higher speed, higher spool pressure
0
0
ratio, higher aerodynamic load, lower blade as-
1950 1955 1960
1965 1970 1975 1980 1985 1990 1995
pect ratios, higher blade row solidities (see also
Table 1).
Year
Lower aspect ratios, plus higher solidity and higher stagger pactness
and ruggedness, blading arrangements, and higher
blading, are the major design advancements that make this stage
loading with good performance.
possible. The trend toward lower blading aspect ratios in Similar
considerations can be carried out for the gas tur-
multistage compressors over the past 30 years is clearly bine
expander, particularly for blade stages, transonic and su-
shown in the figures. Average aspect ratios of about 1.3 to 1.4 personic
design, tip clearance, and endwall effects.
are now common. Similar considerations can be carried out
for solidity values.
Combustion
Chamber
In conclusion, the application of advanced design methods
and research and development results has led to the compres- Figure 13
shows a simplified representation of a gas turbine
sor system having: improved efficiency and stall-margin, com- combustion
chamber: the diffuser, the swirler, the liner, the
----------------------- Page 193-----------------------
Primary Intermediate Diluition

The purpose of water (or steam) injection is to provide a
Fuel
zone zone zone
substantial decrease in flame temperature; SCR is a system
injector
for exhaust clean-up, where a catalyst is used together with
injection of controlled amounts of ammonia resulting in the
conversion of NO to N and H O. SCR has been used in situa-
x 2 2
tions where extremely low (#10 ppm) limits of NO have been
x
Air
specified. For dry low NOx systems three different designs
exist, and they depend on the gas turbine type: heavy-duty,
aero-derivative, and aeroengine.

Diffuser
Air swirler Liner
Cooling slot Air casing
ADVANCED CYCLES
Figure 13. Simplified lay-out of a gas turbine combustion chamber.
As
already stated, the simple OBC gas turbine performances
are good, due mainly to the high efficiency and reliability of

primary, secondary, and diluition zones, and the fuel nozzle
the components. However, several different thermodynamic
are evident. More complex configurations can be found in the
cycles can be utilized to improve simple OBC performance.
literature (17). Usually combustors are classified according to
Unfortunately, they usually lose the compactness and other
their geometry as: single can, multi-can, annular, can annu-
characteristics of simplicity of simple cycle gas turbines.
lar. Different types of classification are based on fuel systems,
As an example, three modifications of the simple cycle are
air direction inside the chamber, and others.
presented here: regeneration, intercooling, and reheating (see
The main factor of importance in assessing combustion
Fig. 15).
chamber performance are: pressure loss, combustion effi-

ciency, outlet temperature distribution, combustion stability
Regeneration
limits, and combustion intensity. As permissible turbine inlet It
consists in the heat recovery of exhaust gas to heat the
temperatures have increased a problem which has assumed
compressor outlet air before the combustion chamber (in the
greater importance is that of cooling the flame tube. A recent
CBC regeneration it is absolutely necessary to obtain high
technique is the use of transpiration cooling, allowing cooling
efficiency). In this way, a heat exchanger between the com-
air to enter a network of passages within the flame tube wall
pressor and combustion chamber must be utilized. A complete
before exiting to form an insulating film of air. Another im-
analysis of the regenerative cycle is presented in Ref. 2; here
portant aspect is correlated to the higher turbine inlet tem- in
Fig. 16 the efficiency versus specific work of such a cycle is
peratures: they imply the use of lower air fuel ratios, with
compared to simple cycle data. The efficiency of the regenera-
consequently less air available for film cooling. Furthermore,
tive cycle is less dependent on compressor pressure ratio, par-
a higher cycle temperature is usually accompanied by the use
ticularly for low # values; for simple cycle # has the opposite
of a higher cycle pressure ratio (see Fig. 5). Thus the tempera-
influence; the efficiency improvement is not very high for high
ture of the air leaving the compressor is increased and its #
values, but is evident for low # values; efficiency is around
cooling potential is reduced.
43–44%. This value, for industrial power generation, does not
Gas turbine combustion is essentially a clean and efficient
seem to be decisive for the use of the regenerative cycle in-
process and for many years there was no concern about emis-
stead of the simple cycle. As for specific work it decreases a lot
sions, with the exception of the need to eliminate smoke from
due mainly to the regenerator pressure losses. Regenerative
the exhaust. Recently, however, control of emissions has be-
solutions are utilized only for terrestrial and marine propul-
come the most important factor in the design of industrial
sion systems (18) and for small gas turbines employed to gen-
gas turbines.
erate electricity; for large industrial gas turbines regenerative
A modern well-developed combustor will produce negligible
cycles are not utilized due to the increment of cost and size of
carbon monoxide (CO) and hydrocarbons (UHC), while sulfur
the plant, while there is not great improvement in efficiency.
emissions are practically absent (virtually no sulfur is present
in natural gas). The principal pollutants emitted are nitrogen
oxides NO . The most important factor affecting the formation
Intercooling
x
of NOx is the flame temperature (Fig. 14). The rate of forma- In

this case, the compression phase of the OBC is obtained in
tion of NOx varies exponentially with the flame temperature,
more than one part (max 3) and between every compression
so the key to reduce NOx is reduction of flame temperature;
phase the air is cooled (usually until ambient temperature)
NOx is also slightly dependent on the residence time of the
utilizing a water cooling heat exchanger. This allows the com-
fluid in the combustor. Unfortunately, the reduction of flame
pression work to be minimized (approximation of the isother-
temperature corresponds to an increase in CO and UHC emis-
mal compression). Net OBC work increases, due to the reduc-
sion, and the residence time has a similar effect.
tion of compression work, but at the same time the cycle
There are basically three major methods of minimizing
efficiency decreases due to the higher fuel rate necessary to
NOx emissions:
heat the air until turbine inlet temperature. Indeed, the out-
let compressor temperature with intercooling is lower than
the corresponding value in simple cycle configuration. Never-

• Water or steam injection into the combustor
theless, the effect of the colder air in the last stages of the
• Selective catalytic reduction (SCR)
compressor, which is very useful for turbine cooling purposes,
• Dry low NOx design at

the same # values as for simple cycle must be considered.
----------------------- Page 194-----------------------
120
30
Temperature range
) 100
for low emissions 25 )
v
NO v
m
x m
p
p
p
p
(
(
80
20
e
n
d
e
i
CO g
x
o
o
r
60
15 t
n
i
o
n
m
f
o
n
o 40
10 s
e
b
d
r
a
i
x
C 20
5 O
Figure 14. Carbon monoxide and oxides of nitrogen vs. com-

0
0
bustion chamber primary zone temperature. The rate of ox-
1500
1600 1700 1800 1900 2000 2100
ides of nitrogen varies exponentially with the flame temper-
ature.
Primary zone temperature (K)
Figure 16 also shows intercooled cycle efficiency versus spe- heat

the working fluid (recombustion is possible due to the
cific work. At fixed # values, efficiency is practically constant, high
excess of air in the first combustor). In this way, the
while specific work increase is quite evident (corresponding to
expansion work is greatly increased; however, the negative
a high increase in power generated). It is possible to note that
thermodynamic aspect of this solution is the increase in tur-
with the available technology, the intercooled cycle pressure bine
outlet temperature. Nevertheless, the influence of this
ratio can arrive at # # 80 instead of 30 as for the simple defect
is greatly reduced, or it vanishes, if the waste heat is
cycle. Intercooling is very useful to improve the performance
recovered in a Rankine bottoming cycle (combined cycle).
of small OBC gas turbines, while for large systems it is less
Figure 16 shows the performances of the reheating cycle,
so. While specific work increases (positive economic aspect), which
are greatly dependent on the expansion ratio in the
the cost of the system increases due to the heat exchangers, first
(high pressure) and second (low pressure) turbine. The
connecting pipes, and compressor modifications. Cooling wa- increase
in specific work is evident in both the cases pre-
ter must be available, while for the simple OBC this con- sented
here, while the efficiency is similar to simple OBC data
straint is not present (see for example gas turbine installa- only
when the expansion ratio of the high pressure turbine is
tions in deserts). equal to
2.
One
of the most important gas turbine companies launched
Reheating
a large
industrial reheating cycle gas turbine (19), a revolu-
In this case the expansion phase is divided into two parts and tionary
proposal in a market where the simple cycle gas tur-
after the first, a second combustion chamber is introduced to bine is
absolutely dominant. The efficiency of this machine is
B
A
C
2
1′ 2′′ 3 3′ 3′′ 4
2′
LPC
HPC HPT LPT
Wu
m cooling
3′
3′
2"
4
Figure 15. Three modifications of OBC simple cycle T
Q reg.
are presented: regeneration, intercooling, and re-

2′
5
heating (A: intercooler, B: regenerator, C: combustion
2
chamber, LPC and HPC low and high pressure com-
1′ 1
pressor, LPT and HPT low and high pressure
turbine.
S
----------------------- Page 195-----------------------
0.6
Near-term combined
Combined cycles
#####
0.55 cycles β = 30
24 18 15 12 3 lev. + RH.
2 lev.
Reheating cycles, #####
β hpt = 2
1 lev.
0.5
β =
30 27 24 21 18
15
Regenerative cycles G /G =
0.127 0.152 0.167 0.186
w a
0.212
y 0.45 10 12
0.249
c 8
65
n β = 80
e 15
50
i β = 6
c
i
40
f
f 0.4 18 β = 36
30 STIG cycles
E 30
β = 30 24
24
24
18 18
0.35 18
15 15
Intercooled cycles
OBC 12
(Simple cycles) 12 β = 30
0.3 24
Reheating cycles,
10
8 18
β hpt = SQRT( )β
15
6
0.25
250 300 350 400 450
500 550 600 650 700
Specific work
(kJ/kg)
Figure 16. Efficiency vs. specific work for several

advanced gas turbine-based cycles and com-
bined cycles (simple OBC cycle, regenerative cycles,
intercooled cycles, re-heating cycles, steam
injection cycles, existing combined cycles, near-term
combined cycles).
similar to the simple cycle gas turbine, but not so high as to tive
systems. Another interesting aspect of the steam injec-
justify the technological difficulties associated with the re- tion
cycle is the very high performance when electricity is
combustion process. However, the associated combined cycle
generated together with heat—cogeneration power plant—
performance is very high, about 58%, thanks to the high tur- and
also when water recovery is utilized (24). However, the
bine outlet temperature values (#600#C) and the economic re- steam
injection solution requires very clean water (demineral-
sults also seem to be very interesting. ized
water; careful control of the compressor surge margin
Obviously, several combinations of nonsimple solutions limit;
new compressor-expander matchline; mechanical stress
can be utilized, with intriguing performances. An interesting
limitation due to the high power increase (see Fig. 16). An
aspect of the regeneration and intercooled or reheated cycle
offshoot of steam injection in OBC systems is the coupling
is that regeneration avoids the negative thermodynamic as- with
the intercooling steam injection gas turbine cycle (IS-
pect of intercooling or reheating previously discussed (more TIG);
in this case the efficiency can be higher than 50%.
heat entering in the cycle).
New cycles are also under investigation all around the
Other systems based on OBC technology are steam injec- world,
and, at this moment, the most promising is the HAT
tion turbines (STIG (20), CHENG (21)), humid air turbines (humid
air turbine) or CHAT (cascaded humid air turbine).
(HAT (22)), cascaded humid air turbine (CHAT (23)) or more The
HAT cycle is a regenerative cycle with a complex lay-out:
complex solutions. the
high temperature heat at the expander outlet is trans-
The most utilized is the steam injection cycle, where the ferred
to the hot and water saturated air at the compressor
steam is generated by recovering the waste heat of the ex-
outlet. Some preliminary performance evaluations for the
haust gas at the expander exit, through the use of a heat HAT
show a cycle efficiency value around 55% and very high
recovery steam generator. Normally the steam pressure is
specific work values (600–650 kW/kg/s), and higher for the
1.25–1.30 times the maximum OBC pressure to allow steam CHAT
solution.
injection in the combustion chamber and in the power turbine A
last very promising solution is the possible coupling be-
stages. The high cp value of steam or water increases the out- tween
the gas turbine simple-cycle with solid oxide fuel cells
put of the gas turbine for fixed temperature drop (the ex- (SOFC)
(25). In this case an efficiency value higher than 70%
pander working fluid is a mixture of combustion gases and seems
to be possible.
steam) and decreases the combustion flame allowing a reduc-
tion of No emissions.
x
Figure 16 shows a comparison between the simple OBC

SUMMARY
system and steam injection cycle performance: utilizing the
available technology, an interesting improvement in efficiency The
success of gas turbine power plants for generating elec-
and specific work is quite evident. In this way, it is possible to
tricity is dependent on the technological development of the
understand the potential cost reduction of the steam injection
components of the plant (axial flow compressor, combustion
solution, particularly for small gas turbines and aeroderiva-
chamber, expander, materials, cooling, control systems) that
----------------------- Page 196-----------------------
270 GATE AND TUNNEL DIELECTRICS, MANUFACTURING ASPECTS
has taken place mainly in the last twenty years, and corre-
18. T. Korakianitis and K. Beier, Investigation of the part-load per-
lates closely to the research and development activities in the
formance of two 1.12 MW regenerative marine gas turbines,
aeroengine field.
ASME Paper 92-GT-86, ASME Trans. J. Eng. Gas Turbines
However, at this moment, very important reasons for the
Power, 1994.
gas turbine power plant’s success are also the low cost of en-
19. W. Jury and D. Searles, Process optimization of an integrated
ergy generation utilizing natural gas, the reduced environ-
combined cycle. The impact & benefit of sequential combustion,
mental impact, the wide availability and reliability, and fi-
ASME Paper 97-GT-490, 1997.
nally the possibility to be utilized in the combined cycle
20. D. H. Brown and A. Cohn, An evaluation of steam injected com-
configuration (the heat of the high temperature exhaust
bustion turbine systems, ASME Trans., J. Eng. Power, 103: Janu-
ary 1981.
gas—see Fig. 9—is recovered in a bottoming steam Rankine
cycle). The combined cycle solution shows efficiency greater
21. M. A. Saad and D. Y. Cheng, ‘‘CHENG CYCLE II’’ recent ad-
than 55%, and in the short term period close to 60% (see Fig.
vancements in gas turbine steam injection, Flowers ’92, Proc., I:
83–96, 1992.
16), while an increase in the specific work of about 60% is
evident compared to the OBC simple cycle.
22. A. D. Rao, United States Patent 4,829,763: Process for Producing
Power, May 16, 1989.
23. M. Nakhamkin et al., CHAT technology: an alternative approach

BIBLIOGRAPHY
to achieve advanced turbine systems efficiencies with preset com-
bustion turbine technology, ASME Paper 97-GT-142, 1997.
24. T. Korakianitis et al., Parametric performance of combined-co-

1. H. Cohen, G. F. C. Rogers, and H. I. H. Saravanamuttoo, Gas
generation power plants with various power and efficiency en-

Turbine Theory, 4th ed., Harlow, UK: Longman, 1996.
hancements, ASME Paper 97-GT-285, 1997.

2. D. G. Wilson and T. Korakianitis, The Design of High Efficiency
25. D. Stephenson and I. Ritchey, Parametric study of fuel cell and

Turbomachinery and Gas Turbines, 2nd ed., Englewood Cliffs, NJ:
gas turbine combined cycle performance, ASME Paper 97-GT-

Prentice Hall, 1997.
340, 1997.
3. C. F. McDonald, The future of the closed-cycle gas turbine for
terrestrial applications, Proc. 1987 ASME Cogen Turbo Symp.
ARISTIDE F. MASSARDO
IGTI, 1: 238–246.
`
Universita di Genova
4. A. F. Massardo and C. F. McDonald, Closed cycle gas turbine
perspective: enabling technologies from combustion turbines the
key factor for future deployment, VIII Advanced Energy Systems
Conf., Bologna, Italy, 1995.
5. A. F. Massardo, Design and performance evaluation of a CBC

solar space power system, ASME Paper 93-GT-180.
6. C. F. McDonald, Nuclear gas turbine plant (GT-MHR) perfor-

mance potential, ASME Paper 94-GT-416.
7. A. Agazzani, A. F. Massardo, and T. Korakianitis, An assessment

of the performance of closed cycles with and without heat rejec-
tion at cryogenic temperatures, ASME Trans. J. Eng. Gas Tur-
bines Power, 1997.
8. T. Korakianitis, N. E. Vlachopoulos, and D. Zou, Models for the

prediction of transients in closed regenerative gas-turbine cycles
with centrifugal impellers, ASME Paper 94-GT-342, ASME
Trans., J. Gas Turbines Power, 1996.
9. Anonymous, Gas Turbine World Handbook, Pequot Publication,

1997.
10. M. Tatsuzawa et al., Development of 300kW-class ceramic gas

turbine (CGT301) engine systems, ASME Paper 95-GT-201, 1995.
11. Anonymous, Sermatech Review, n# 51, Limerick, PA: Sermatech

Spring 1995.
12. Anonymous, Sermatech Review, n# 56, Limerick, PA: Sermatech

Fall 1996.
13. M. K. Chyu, Regional heat transfer in two pass and three-pass
passages with 180-deg sharp turns, ASME Trans., J. Heat Trans-
fer, 113(1): 63–70, 1991.
14. A. F. Massardo, Flow visualization and surface heat transfer

from three pass trapezoidal blade cooling passage, 2nd Euro.
Conf. Turbomachinery, Antwerpen, Belgium, 1997.
15. D. L. Chase et al., GE Combined-Cycle, Product Line and Perfor-

mance, GER-35741-1997.
16. D. C. Wisler, Advanced compressor and fan systems, GE Aircraft

Engine Business Group, Cincinnati, OH, 1986.
17. A. H. Lefebre, Gas Turbine Combustion, McGraw Hill, 1983.
----------------------- Page 197-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Geothermal Power

Standard Article

John E. Mock1
1Massachusetts Institute of Technology, McLean, VA

reserved.
Search
DOI: 10.1002/047134608X.W3021
❍ Acronym Finder
Abstract
The Nature of Geothermal Resources
Applications of Geothermal Energy
Geothermal Heat Pumps
Geothermal Power Development
Future Developments

%20Conversion/W3021.htm17.06.2008 16:40:02
----------------------- Page 198-----------------------
374 GEOTHERMAL POWER
GEOTHERMAL POWER
Electric
power was first produced from geothermal energy in
1904 when
a 560 W generator was placed in operation at the
Larderello steam field in the Tuscany region of Italy. It was

not
until 1960 that geothermal power was first generated in
the
United States (at The Geysers, California) (1). Since then,
geothermal power capacity has expanded worldwide to over

6925 MW
(of which 2850 MW is installed in the United
States).
Although geothermal energy represents an extremely
large
worldwide resource, its exploitation has been slow. Its
commercial viability depends upon a number of factors: tech-

nical,
economic, legal, and environmental. There is a special
need for
technological improvement: (1) in the geosciences to
help
locate and characterize geothermal resources, (2) in
hard-rock
drilling to reduce geothermal drilling costs, which
are
currently up to three times oil and gas drilling costs, and
(3) in
energy conversion systems to decrease the cost of pro-
ducing
electricity by reducing power plant capital costs as
well as
reducing the costs of operations and maintenance.
THE
NATURE OF GEOTHERMAL RESOURCES
Geothermal energy is the generic term for the heat energy of

the
earth, which is contained in magma, underground rocks,
and
fluids. Volcanoes, geysers, fumaroles, and mineral
springs
represent surface manifestations of this ubiquitous
source of
energy, whose origin lies in (1) the decay of natural
radioactive elements in the crust of the earth, (2) the residual

----------------------- Page 199-----------------------
GEOTHERMAL POWER 375
heat of planetary formation (from the kinetic and potential

Table 1. Estimated US Geothermal Resources
energy of material accreted by the early earth), and (3) the
Accessible Resource Resource
heat generated by the friction of tectonic plates grinding
Type Base (1018 J) (1018 J)
against one another.
Hydrothermal (T # 90#C) 9,600 2,400

At certain locations, the earth’s crust thins or cracks, and
Geopressured
magma rises close to the surface, occasionally penetrating to
Thermal energy 107,000 270–2800

create hot molten fissures or volcanoes. In other cases, it
Methane energy 63,000 158–1640
comes close to the surface, transferring heat by conduction to
Total 170,000 430–4440
rocks or underground bodies of water. The geysers and hot Hot
dry rock 450,000 (Uncertain)
springs observed at Yellowstone National Park are graphic
Magma 500,000 (Uncertain)
examples of manifestations of near-surface hot spots. It is
Source: US Geological Survey Circular 790.

such hot spots that represent attractive targets for the com-
mercial generation of electricity.
Geothermal resources can be classified based upon their
intrinsic properties as hydrothermal, geopressured, hot dry
base at a production cost competitive with other forms of en-
rock, or magma. Hydrothermal resources consist of naturally
ergy at a foreseeable time and under reasonable assumptions
occurring steam or hot water carried upward by convective of
continuing technological improvement (4). Estimates of US
circulation. The porosity of the underlying reservoir rocks de-
geothermal resources are shown in Table 1. For comparison,
termines the total amount of fluid available; their permeabil- the
total consumption of energy in the United States in 1995
ity determines the rate at which fluid can be produced. These was
approximately 85 # 1018 J.
resources are found from several hundred to 4300 m beneath

Geothermal resources, as shown in Fig. 1, are not distrib-
the earth’s surface with temperatures up to 400#C. Hydro-
uted uniformly. In the United States, hydrothermal reser-
thermal resources are the easiest geothermal resource to ac-
voirs are located primarily in the West, where relatively re-
cess and the only resource currently exploited commercially.
cent geologic activity has occurred (creating shallow and
The other three geothermal resources will require ad-
accessible high-temperature sites). Hydrothermal electricity
vanced technology before becoming commercial. Geopressured
production currently is based in California, Nevada, Utah,
resources consist of hot pressurized brines containing dis- and
Hawaii. A large geopressured resource exists along the
solved natural gas, lying at depths ranging from 3600 m to
Texas/Louisiana Gulf Coast. Although advanced technology
over 6000 m, and characterized by temperatures of 50# to
using hot dry rock could extend development of geothermal
260#C, pressures 50 MPa to 140 MPa, salinities of 20,000 to
resources across the entire United States, early developments
300,000 parts per million, and gas content of 0.7 m3 to 3 m3
will most naturally occur in the tectonically active West,
of methane per barrel of brine (2). The unique advantage of

Alaska, and Hawaii. The geographic distribution of potential
geopressured resources is that they contain three forms of en-
magma resources is purely speculative at this time, but the
ergy—thermal, chemical, and hydraulic—which can be con-
best prospects lie in the western part of the United States.
verted individually or in combination to generate electricity.
Hot dry rock resources consist of hot, relatively water-free
APPLICATIONS OF GEOTHERMAL ENERGY

rocks at depths of 2400 m to 9000 m, with temperatures up
to 350#C. These hot rocks have few pore spaces or fractures;
Production of Geothermal Power

hence, they contain little water and little or no interconnected
permeability. Heat can be extracted from the rocks by creat- Two
separate steps are required in the development of geo-
ing artificial fractures connecting two wells, injecting water
thermal power. The first step, developing the geothermal field
through one well and recovering the hot fluid through the sec- to
provide thermal energy and fluids to the power plant, can
ond well, extracting the heat therefrom for the generation of be
lengthy and expensive. It consists of exploration to locate
electricity (3). Magma resources are molten or partially mol- a
suitable reservoir, testing to determine its size and quality
ten rock within the upper 10,000 m of the earth’s crust with
(temperature, pressure, enthalpy, salinity), and flow tests to
temperatures as high as 1300#C. Magma comes close to the
determine its impedance (resistance to flow) and to optimize
earth’s surface primarily at the edges of the major tectonic the
location of production wells. The second step—after the
plates that float on the molten underlying mantle. As the tec-
adequacy of the reservoir has been established—is the con-
tonic plates move apart from one another (producing rift
struction of the power plant suitably connected through dis-
zones or spreading centers) or subduct one under the other,
tribution lines to both production and injection wells, as
magma rises close to the surface. Occasionally, as in the cre-
shown in Fig. 2. Depending on the state of the geothermal
ation of the Hawaiian Island chain, a geological ‘‘hot spot’’
resource (liquid or steam) and on its temperature and pres-
continuously extrudes magma as a tectonic plate slowly
sure, one of three conversion technologies is generally used.
moves over it. In
the case of dry steam, as at The Geysers and Larderello,
Individually, or together, these geothermal resources rep- the
steam is treated to remove any entrained particulate mat-
resent a major source of energy for the world. The US Geologi-
ter, and then passed directly into a conventional steam tur-
cal Survey defines geothermal resources as (1) the ‘‘accessible
bine. After the steam is condensed, the condensate provides
resource base,’’ which includes all geothermal resources shal-
makeup water for the cooling tower or is reinjected into the
low enough to be reached by production drilling in the fore-
ground.
seeable future regardless of near-term economic viability, and
For high-temperature liquids (above 200#C), flash steam
(2) the ‘‘resource,’’ which includes only those geothermal re-
technology is generally utilized (5). In flash systems, the liq-
sources that can be extracted from the accessible resource
uid’s pressure is dropped as it reaches the surface, allowing a
----------------------- Page 200-----------------------

Temperature above 194°F
Temperature below 194°F
Geopressured resources
Figure 1. US high-temperature geothermal resources are

located primarily in the West, Alaska,
and Hawaii. (From US Geological Survey.)
portion of the fluid to flash into steam, which is used to drive

Commercial electricity was first produced in the United
a conventional steam turbine, as shown in Fig. 3(a). For mod-
States in 1960 from superheated steam at The Geysers. By
erate-temperature liquids (150# to 200#C), binary technology is
1975 the installed capacity at The Geysers reached 500 MW,
more efficient (6). In binary systems, as shown in Fig. 3(b),
and by the late 1980s it peaked at 2000 MW. Flash plants
the heat from the geothermal fluid is used to vaporize a sec-
and binary plants were first installed in the early 1980s in
ondary working fluid (such as isobutane or isopentane), which
several reservoirs in the Imperial Valley of California. Cur-
is then used to drive a vapor turbine analogous to a steam
rently the US geothermal industry has over 2800 MW of in-
9
turbine but smaller in size for the same output power.
stalled capacity, and produces some 17 # 10 kWh of electric-
Conversion
Geoscience
Discovery
Effluent
abatement
Geoscience
modeling
Numerical
modeling
Well drilling
Injection
and
completion
wells
Production
wells
Figure 2. Geothermal power projects are devel-

oped in two phases: (1) the discovery, validation, Sensors
and development of a geothermal field and (2) the
construction of a power plant designed to convert
efficiently the geothermal heat into electricity.
----------------------- Page 201-----------------------
Turbine
Flash
generator Cooling
tower
separator
From
production
wells
Gas
removal
To
injection
To injection
wells
wells
Condenser
(a)
Turbine
generator Cooling tower
Heat exchanger
From
production To
Figure 3. (a) In a single-flash system, the high-pressure

wells
injection
wells
fluid passes
into a separator where reduced pressure pro-
To injection Condenser
duces steam which rotates a conventional turbogenerator to
wells Makeup water
produce electricity. (b) In a binary system, the geothermal
fluid transfers heat to a secondary fluid which vaporizes, and

(b)
expands through a turbogenerator to produce electricity.
ity annually (7). Geothermal energy is the second largest grid-
where projects under construction or planned will bring 2400
connected renewable electricity source in America—exceeded MW
of new power on line by the year 2000 (10). New projects
only by hydropower. Table 2 compares estimated costs of are
also under consideration in Central America, South
baseload electricity in the United States for fossil, nuclear,
America, and East Africa. These developments reflect the
and geothermal resources.
steady growth of geothermal power worldwide which has in-
Geothermal power plants have a number of desirable prop-
creased at a robust rate of 8.5% per year since the early
erties making them attractive to developers, especially in rap-
1920s (11).
idly growing nations lacking hard currencies for fossil fuels.
They are modular and can be installed in increments from
less than one to over 50 MW. They can be designed to provide
Geothermal Heat Pumps
either baseload or peaking power and offer short construction
times: as short as 6 months for plants in the range of 0.5 MW The
earth maintains a relatively constant temperature at
to 10 MW, and less than 2 years for clusters of plants totaling
shallow depths below 1.5 m, warmer on average than the air
250 MW or more (8). Many developing countries are located in
above it in winter, cooler in summer. The term geothermal
areas of active geologic processes—areas generally containing
heat pumps is generic for all heat pumps which utilize the
high-grade geothermal resources. It has been estimated that
earth’s thermal capacity as an energy source (for heating) or
as much as 78,000 MW of geothermal electrical power from
energy sink (for cooling). The earth’s thermal capacity can be
hydrothermal resources are available for development in
utilized either directly or indirectly (for example, by using
known resource areas in some 50 developing countries (9).
groundwater as an intermediary heat transfer agent).
Currently total worldwide installed capacity is 6925 MW
Geothermal heat pumps are more energy efficient than air-
as shown in Table 3. The most rapid growth of geothermal
source heat pumps, central air-conditioners, and gas furnaces.
power is taking place in the Philippines and in Indonesia, The
heat pump itself operates on the same principle as the
----------------------- Page 202-----------------------
Table 2. Estimated Costs for New Baseload Capacity
Installed and
Breakeven
Annual Plant Cost
O&M Costs Fuel Costsa Busbar Price
Resource Type ($/kW) (¢/kWh)

(¢/kWh) (¢/kWh) (¢/kWh)
Oil 800 2.0

0.3 1.7–3.4 4.0–5.7
Gas 600 1.5
0.3 0.8–2.3 2.6–4.1
Coal 1200 3.0
0.3 0.6–4.0 3.9–7.3
Nuclear 3200 7.8
0.4 1.0 9.2
Hydrothermal
(high-grade) 1000–1500 2.4–3.6
0.3 2–3 4.7–6.9
Hydrothermal
(low-grade) 2000–2500 5.1–6.3
0.4 4–10 9.5–16.7
Hot dry rock
(high-grade) 1000–1500 2.4–3.6
0.3 3–4 5.7–7.9
Hot dry rock
(low-grade) 2000–2500 5.1–6.3
0.4 20 25.5–26.7
a Geothermal ‘‘fuel’’ costs represent the cost of drilling

additional wells when needed. The cost ranges for fossil energy are
dependent on market fuel prices: oil ($10–$20/bbl), gas ($1–

$3/MBTU), coal ($15–$100/ton).
Source: J. E. Mock, J. W. Tester, and P. W. Wright,
Geothermal energy from the earth, Annual Reviews of Energy and the
Environment, V. 22, Palo Alto, CA: Annual Reviews, Inc., 1997.
home refrigerator, which is actually a one-way heat pump.

used in which a loop of plastic pipe is placed below the frost
Because electricity is used only to transfer heat, not to gener-
zone in a horizontal trench and backfilled with soil (12).
ate it, the geothermal heat pump delivers three to four times
In addition to providing the home or building owner with
more energy than it consumes. In a typical installation a loop
lower heating and cooling costs, several significant advan-
of plastic pipe is placed in a vertical drill hole up to 120 m
tages also accrue to the local utilities. GHPs are ideal demand
deep, and the hole is backfilled with clay. A water–antifreeze
management tools, reducing summer cooling peak loads by
solution is circulated through the loop, then through the heat 1
kW to 2 kW and winter heating peak loads by 4 kW to 8 kW
pump. There is no consumption of groundwater, nor is there
for the typical residence (13). These impressive load-leveling
any contact between the solution in the plastic pipe and the
capabilities and energy savings permit utilities to manage
earth or groundwater. An alternative installation is often
their operations more efficiently (both diurnally and season-
ally) and postpone the construction of new generating ca-
pacity.
Table 3. Geothermal Electrical Plants: Country and World
Direct Thermal Use

Total Installed Capacity (MW)
Nation Existing Capacity Year-2000 Capacity

There are many energy uses which do not require high-grade
energy sources such as electricity, but can be satisfied with
Argentina 0.7 0.7
low- to moderate-temperature sources of heat. Low-tempera-

Australia 0.2 0.2
China 28.8 81.0
ture geothermal resources have found significant use in a
Costa Rica 57.0 170.0
wide variety of commercial applications ranging from 10#C for
El Salvador 105.0 165.0
soil warming (for agriculture) and ice melting, to 200#C for
France 4.2 4.2
cement drying. Historically, geothermal heat was first used
Guatemala 0 240.0 in
the United States by small resorts and district (or home)
Iceland 49.4 49.4
heating systems. By the mid-1990s, geothermal heat was
Indonesia 309.8 1957.0
used in a wide variety of applications, providing over 14 #
Italy 631.7 856.0
1015 J annually (14). Industrial applications now include: pulp
Japan 50.4 600.0
and paper processing (200#C), dehydration of vegetables
Kenya 45.0 45.0
(130#C), heap leach mining operations to extract precious
Mexico 753.0 960.0
metals (110#C), enhanced oil recovery (90#C), and mushroom

New Zealand 286.0 440.0
growing (60#C). Geothermal fluids are also finding increasing

Nicaragua 35.0 35.0
use in aquaculture (to raise catfish, tilapia, sturgeon, lobster,

Philippines 1227.0 1976.0
Portugal 6.4 6.4
shrimp, and tropical fish) and greenhouse operations (to raise
Russia 11.0 110.0
many commercial crops such as flowers, house plants, vegeta-
Thailand 0.3 0.3
bles, and tree seedlings). Geothermal energy serves as the
Turkey 20.6 125.0
heat source for 23 district heating systems in the United
United States 2849.9 3395.0
States including the nation’s oldest in Boise, Idaho, and the
World total 6925.0 11216.2
nation’s largest in San Bernardino, California (15).
Source: Geothermal Energy Association, International Geothermal Electric
Environmental Considerations
Power Plants (Davis, CA, 1991); L. McLarty, DynCorp EENSP, Inc. (Alexan-
dria, VA, August 1995); G. W. Huttrer, International Geothermal Association,
The exploitation of geothermal energy has a net positive im-
Proceedings of the World Geothermal Congress (Florence, Italy, 1995).
pact on the environment. Modern geothermal power plants

----------------------- Page 203-----------------------
have extremely low levels of SO , NO , CO , and particulate the

field by using a conductivity meter and a pH meter. The
x x 2
emissions. Sulfur oxides and nitrogen oxides average only a

amount and nature of dissolved chemical species are func-
few percent of those from fossil fuel alternatives. Geothermal
tions of temperature and the local geology. Many of the high-
energy use also reduces markedly the emissions of green-
temperature resources in the western United States contain
house gases. The current generation of geothermal power
6,000 mg/L to 10,000 mg/L TDS, whereas some resources in
plants emits only 0.14 kg of carbon (in the form of CO2) per the
Imperial Valley of California are saturated with salts at
megawatt-hour of electricity generated, compared to 128 kg/
300,000 mg/L. The pH of geothermal resources ranges from
MWh for natural gas plants, 190 kg/MWh for a plant op-
moderately alkaline (8.5) to moderately acidic (5.5) (17).
erating on No. 6 fuel oil, and 226 kg/MWh for a plant using
Geophysical exploration makes use of physical measure-
bituminous coal (16). Air-cooled, closed-loop geothermal
ments: (1) to detect a resource directly, (2) to provide indirect
power plants, which employ 100% injection of all geothermal
evidence of its existence and location, and/or (3) to determine
fluids and gases, have essentially zero air emissions. The di- and
map its physical and chemical characteristics. Such phys-
rect use of geothermal heat in many cases displaces electrical ical
parameters as the distribution of temperature over the
heat, reducing demand for electricity with its associated pol-
surface of the earth and at depth, the electrical, magnetic, or
lutants. In the same vein, geothermal heat pumps (which pro-
density properties of the ground, and the manner in which
duce no pollution) reduce the demand for electricity.
seismic waves are propagated in the earth all respond in
characteristic ways to the presence of a geothermal resource.
Geophysical surveys are valuable to help locate resources that

GEOTHERMAL POWER DEVELOPMENT
have
no evident surface expression, to site production and in-
jection wells, and to monitor production from and injection

Hydrothermal resources share with nonconcentrated solar
into
a reservoir.
and ocean thermal systems the disadvantage of low resource
temperatures, which limits the efficiency of electric conver-
Selecting a Geothermal Power Plant
sion processes. Whereas fossil energy and nuclear plants op-
erate at efficiencies of 35% to 50%, geothermal plants perform
Geothermal power plants operate on the same Rankine cycle
typically at efficiencies as low as 10% to 20%. Improved tech- used
by fossil and nuclear plants; however, hydrothermal con-
nologies are needed to decrease the overall cost of conversion
version systems are constrained to a relatively small op-
of geothermal energy to electrical power, and to reduce the
erating range of temperatures. Most commercial hydrother-
substantial costs associated with geothermal exploration and mal
systems operate with fluid temperatures of 250#C or
field development. less
—with heat rejection at ambient temperatures around
35#C, leaving a temperature differential of only 215#C for op-

Locating a Geothermal Field
erating the power cycle. Consequently, a high premium is
placed on designing all parts of the geothermal system to op-

The first step in geothermal power development is explora-
erate at peak efficiency.
tion, which includes (1) locating suitable reservoirs and (2)
siting wells for the production of geothermal fluids. Even
Thermodynamic Considerations: Cycle and Utilization Effici-
within well-explored fields such as The Geysers, the drilling
encies. The second law of thermodynamics imposes an upper
success rate is only 80%, whereas for wildcat drilling in rela-
limit on the production of electricity from low-temperature re-
tively unknown areas, the success rate is as low as 10% to
sources and provides a basis for defining utilization efficiency,
20% (17). The key problem is not in finding a source of heat, # :
u
but in finding both adequate permeability and fluids that are
recoverable in amounts sufficient to supply a commercial-size
Wnet Wnet
power plant. In any geothermal exploration program, an ade-
ηu = #B = m · #B
quate understanding must be developed of the regional and

local geology. Geologic mapping is the important first step,
where Wnet represents the net useful work from an actual sys-
conducted by field geologists who (1) identify and locate the tem;
#B (thermodynamic availability) represents the maxi-
various rock units in the area (sedimentary, plutonic, volca- mum
amount of work which theoretically could be extracted
nic); (2) map the structural elements of the geology (faults, in a
reversible process in which a condensed geofluid is cooled
fractures, folds); (3) search for evidence of geothermal activity from
its well-head temperature, Tw , to ambient temperature,
from such obvious indicators as thermal springs, geysers, and To ;
m is the mass fluid flow rate; and #B is availability per
fumaroles, to such subtle indicators as hydrothermal alter- unit
mass (18). #B can be calculated from
ation of rocks, or ancient spring deposits of sinter or traver-
#B = (#H − T #S) |Tw

tine; (4) collect samples of rocks and minerals for microscopic
0 T0
examination, radioactive age dating, and geochemical analy-
sis; and (5) collect samples of fluids from wells and springs
where #H is the enthalpy difference and #S the entropy dif-
for geochemical studies (17). Based on these results, promis-
ference between the two states. Thus, #u is a direct measure
ing areas are identified for more detailed geochemical and of
the effectiveness of resource utilization; for a fixed Tw ,
geophysical investigations.
higher values of #u correspond to lower required flow rates
In geothermal geochemistry, the chemistry of the geother- (m)
for a given power output (Wnet). This efficiency concept is
mal fluids is investigated as well as the chemistry of the rocks
especially useful in comparing flash- and binary-cycle perfor-
in which the geothermal resource exists. The simplest chemi-
mance for the same resource.
cal parameters used to characterize geothermal fluids are to-
If the utilization efficiency is low, then the resource is be-
tal dissolved solids (TDS) and pH, which can be measured in ing
utilized wastefully, and an unduly large investment in
----------------------- Page 204-----------------------
wells is required. On the other hand, as we approach utiliza-

individual field owners, power plant improvements to utilize
tion of the full potential of the geothermal resource, total well low
pressure steam more efficiently, and fluid injection, for
costs decrease, but the required investment in highly efficient
example, by the Southeast Geysers Effluent Pipeline Proj-
power conversion equipment is high. The economic optimum ect—
the world’s first wastewater-to-electricity system. A 29
occurs when #u takes on some intermediate value; for exam- mi.,
20 in. diameter pipeline has been designed and con-
ple, at The Geysers, #u # 0.55 is typical with To # 26.7#C (19).
structed to carry 7.8 million gallons a day of Lake County, CA
An alternative approach is achieved by defining a cycle ef-
wastewater for injection to depths of approximately 2430 m
ficiency,#cycle, which represents the ratio of the net work, Wnet ,
(8000 ft.) at The Geysers to produce a total of 70 MW of power
to the amount of heat actually transferred from the geother- from
six existing geothermal power plants (21).
mal fluid, QH . As the cycle efficiency decreases, the amount of
Flash Steam Plants. In the western United States many geo-
heat rejected to the environment increases. For ambient tem-
thermal reservoirs are found that produce hot water at tem-
peratures of 25#C with a geothermal heat source of 100#C,
peratures above 170#C and pressures above 10 atm, making
cycle efficiency is less than 10%. As the source temperature them
economically attractive for flash-steam plants. As
increases to 150#C, #cycle # 12.5%; at 200#C, # 17.5%; and at
shown in Fig. 3(a), flash systems consist basically of one or
250#C, # 20%. Because power conversion efficiencies are low, two
large tanks, wherein part of the geothermal fluid vapor-
the amount of heat transferred may be 5 to 15 times greater izes
(flashes) into steam at pressures less than reservoir pres-
than the power produced, requiring large heat exchangers at
sure. The steam, typically 18% to 25% of the fluid from the
significant cost. For example, a 50 MW geothermal plant with
reservoir, is sent to the turbogenerator. The remaining water
2 2
a 12% cycle efficiency requires about 30,000 m (325,000 ft ) (75%
to 82% by weight of the initial fluid) is disposed of in
of heat-exchanger surface area.
injection wells (8). For some fields in the Salton Sea area (but
In general, to obtain efficient utilization of a geothermal not
at most US flash plants) the high-temperature brines con-
power plant it is necessary that (1) most of the heat be ex- tain
substantial amounts of dissolved silica, which, if not
tracted from the geothermal fluid before disposal, (2) temper-
treated, precipitates upon equipment walls in the form of
ature differences across heat transfer surfaces be minimized, hard
scale. Ameliorating measures available include: (1) in-
(3) turbines and pumps be designed for optimum perfor-
creasing the brine exit temperature above that optimal for
mance, and (4) heat be rejected at the lowest possible ambient
power production, (2) using a ‘‘crystallizer–clarifier’’ system
temperature, To (19). For example, for a 200#C geothermal rein
conjunction with the first flash tank to precipitate and re-
source, a decrease in condensing temperature from 50# to move
silica crystals, or (3) using a ‘‘pH-modification’’ system
25#C increases the potentially available work by more than
which injects small quantities of an acid (H2SO4 or HCl) up-
40%.
stream of the first flash tank to help keep the silica in solu-
tion
(22).
Design of Geothermal Power Plants. Commercial geothermal
Flash-stream plants can be designed using either condens-
power plants range in size from 0.5 MW to 180 MW (8). The ing
or noncondensing cycles. Single-flash, noncondensing
specific design of each plant depends primarily on the physi-
plants with steam exhausted to the atmosphere through a dif-
cal and chemical state of the geothermal fluid, and to a lesser
fuser–silencer do not optimize the use of the resource, but
extent on the local ambient temperatures. Seasonal and diur- are
simple to operate and can be installed at minimum cost.
nal variations of dry bulb temperatures can also affect cycle
Geothermal resources having very high noncondensable gas
performance. Power conversion technologies in current com-
content may make condensing cycles impractical or uneco-
mercial operation include dry-steam, flash-steam, binary
nomical and thus favor the use of such noncondensing sys-
plants, and steam/binary combined cycle plants. tems
(23). The addition of a condenser can double the output
Dry Steam Plants. Dry steam resources are very rare; only of a
flash plant, at the expense of increasing its cost and com-
two dry-steam fields are being utilized commercially—one at
plexity. With low-temperature resources, up to half of the
Larderello, Italy, the other at The Geysers, CA. For a typical
power developed by the turbine comes from the expansion of
50 MW plant at The Geysers, 10 to 20 production wells are the
steam below atmospheric pressure (23).
drilled about 1000 m apart to provide sufficient steam for the A
dual-flash cycle represents a simple extension of the sin-
turbogenerators. Gathering lines are constructed to deliver gle-
flash cycle, making use of the energy remaining in the
steam from the wells to cyclone separators which remove en-
separated brine from the first flash tank. By flashing this
trained particles and water droplets. The steam then passes
brine in a low-pressure separator, additional steam is gener-
through the turbines to the condensers and to steam ejectors, ated
which can increase total power by as much as 50%.
which remove noncondensable gases. The condensate from
Binary Plants. For geothermal resources with temperatures
the condensers is used to replace water evaporated in the
below 170#C, the most efficient and economical plant is one
cooling tower; any excess condensate is pumped to injection
employing a secondary working fluid in a binary cycle. Tem-
wells, which helps to maintain reservoir pressure, replace lost
peratures as low as 100#C and as high as 200#C are suited to
fluid, prevent land subsidence, and dispose of wastes. Gas-
binary operation, depending on the availability of cooling wa-
eous emanations from the condensers, primarily CO2, may re- ter,
range of ambient temperatures, and cost of wells (5). In
quire chemical treatment to remove contaminants such as hy- this
system, shown in Fig. 3(b), the geothermal brine flows
drogen sulfide and traces of methane, arsenic, and boron (5).
through the tubes of a shell-and-tube heat exchanger, va-
By the late 1980s, The Geysers had become a mature
porizing the binary fluid, which expands through a turbogen-
steam field, characterized by a general pressure decline and
erator, generating electricity. The binary fluid is then cooled
a gradual decrease (7% to 8% per year) in steam production. in
a water-cooled condenser and sent to a storage tank. The
Measures were taken to reduce the decline in production in-
heated water from the condenser is pumped to a conventional
cluding (20): cooperative steam field management among the
cooling tower. In spite of its greater complexity and capital
----------------------- Page 205-----------------------
cost, the binary system may be preferred in some cases to the
Advanced
flash system—even for high-temperature resources—because 100
conventional
of its higher efficiency and environmental acceptability.
technology
A geothermal combined cycle power plant, commercialized
Base case:
by ORMAT, efficiently extracts the energy contained in the
Today’s technology
typical mixture of steam and brine flowing from geothermal
s
wells. In this system the geothermal fluid flows directly into r
a
l
l
Problem
a steam separator with the separated high-pressure steam o 10
burdened
used to drive a back-pressure turbine. Low-pressure steam, S
U
which exits the back-pressure turbine, flows into the vapor- 4
izer of an organic cycle binary system wherein its heat of con- 9
9
1
densation is added to the thermal energy of the separated f
o
brine to vaporize an organic fluid, which is used to drive a s
Linear drilling
n
technology
binary turbine. o 1
i
l
l
System Application. Geothermal power plants are generally i
M
baseload systems, but may sometimes be used in a load-fol-
Oil and gas
lowing mode. Current contractual capacity factors for most
average
geothermal plants are on the order of 80%. However, actual
capacity factors for some operating plants approach 100%
0.1
5000 10000 15000 20000 25000 30000
(24). System capacity factor is defined based on nameplate
(ft)
rating:
0 2 4 6 8 10
kWh output per year
Well depth (km)
Capacity factor =
(Nameplate kW)(8760 hours per year) HDR
actual Hydrothermal actual JAS ultra-deep SPE oil and gas
System availability factors (the percentage of a year in which

HDR predicted Hydrothermal predicted JAS correlation
the system is capable of delivering its rated power) are also
very high, typically 95% or better (8). The capacity and avail- Figure
4. Drilling costs for oil/gas/geothermal wells increase expo-
ability factors of geothermal power systems tend to be higher
nentially with depth. Geothermal wells can cost up to three times
than other baseload systems primarily because of the intrin- more
than oil/gas wells drilled to the same depth.
sic simplicity of geothermal plants.
wells
from the same drill pad, rather than drilling at several
Advances in Geothermal Technology
different locations, minimizes pipeline, site, and access costs.

Drilling Technology. Drilling to depths of several thousand During
exploration, there are significant advantages to dril-
feet is required at all stages of geothermal development: ex- ling
small wells—so-called slimholes. Geothermal exploration
ploration, production, and reinjection. Geothermal wells are has
traditionally entailed the drilling of large-diameter (30
difficult and expensive to drill since geothermal reservoirs are cm)
wellbores for production testing to prove a new resource.
typically found in hard, abrasive, high-temperature, fractured A
newer cost-effective method is to drill small-diameter (10
rock formations. Unique problems arise in drilling through cm)
wells to obtain the required reservoir data. Slimhole dril-
fractured formations, such as the loss of drilling fluids, lead- ling
is up to 40% cheaper because the drilling rigs, crews,
ing to wellbore instability, stuck drill pipe, inadequate casing and
drilling fluid requirements are smaller and because site
cementing, and increased costs—accounting for 10% to 20%
preparation and road construction in remote areas is signifi-
of the costs of a typical geothermal well. Figure 4 shows that cantly
reduced (for example, by using helicopter-portable drill
drilling costs increase exponentially with depth and that geo- rigs)
(28).
thermal wells cost, on average, two to three times more than As
part of component developments, Sandia National Lab-
oil and gas wells at similar depths (25). Costs per typical geo-
oratory has collaborated with General Electric and drill-bit
thermal wells range from $1 to $3 million. Drilling and well
manufacturers in applying human-made diamonds to drilling
completion costs generally represent 35% to 50% of the total bits.
Field tests in the shales and sandstones of geothermal
cost of a geothermal power project; and being accrued early wells
in the Imperial Valley of California demonstrated bit
in the life of the project, their financial impact is particularly lives
and penetration rates two to ten times those achievable
significant (26). Three mutually supporting approaches ap- with
conventional roller cone bits; however, tests in hard,
pear promising to reduce drilling costs: (1) well emplacement
abrasive, highly fractured rock formations were less success-
optimization, (2) drilling components development, and (3) ful,
leading to further advances in polycrystalline diamond
smart systems development. Under well emplacement optimi-
compact (PDC) bits, impregnated diamond bits, and ther-
zation, one approach is to maximize well production—that is, mally
stable polycrystalline (TSP) bits (29). Other significant
to aim for large wells capable of producing 20 MW each, in-
technical advances include the development of: high-tempera-
stead of the more typical 5 MW per well. [The most productive ture
drilling muds, high-temperature elastomers for downhole
of the Unocal Salton Sea wells supplies enough brine to pro-
drilling motors, high-temperature cements, CO2-resistant ce-
duce 45 MW, a world record (27).] Multileg completion and ments,
high-temperature logging instruments, lost-circulation
side tracking wells are also methods of improving well pro-
materials, and acoustic technology to transmit downhole data
ductivity. The successful directional drilling of four or five to the
surface in real time (30).
----------------------- Page 206-----------------------
The National Research Council concluded in a 1994 study vapor

tends to become drier (more superheated) as it expands.
that revolutionary advances (cost reductions up to 50%) are In a
test conducted on a 3.5 MW binary unit at Mammoth,
now within reach through the development of smart drilling CA,
supersaturated turbine expansion showed an improve-
systems—that is systems capable of sensing and adapting to ment in
power output of up to 35% (36).
conditions around and ahead of the drill bit in real time with
Geothermal turbines are conventional in concept; however,
minimal operator intervention (31). Rapid innovation in mi- a number
of special-purpose power generation devices have
croelectronics and other fields of computer science and minia- been
investigated. For example, ‘‘total flow turbines’’ such as
turization technology holds great promise for significant im- the
Biphase turbine have been designed to extract efficiently
provements. The National Advanced Drilling and Excavation both
hydraulic and thermal energy from the two-phase flow
Technologies (NADET) Program was established by the US (of steam
and water) from wet geothermal wells. In the Bi-
Department of Energy in collaboration with industry with the phase
machine, pressurized brine (or a water/steam mixture)
goal of reducing drilling costs for deep geothermal wells by at impinges
tangentially on a rotary-separator wheel which is
least 50% within the next two decades (32). set
spinning by frictional drag. Impulse steam blades,
attached
to the rotary wheel, extract additional kinetic energy
Energy Conversion Technology. Although geothermal energy from the
high-velocity steam. Tests of an experimental Bi-
conversion technologies are mature—with dry steam plants phase
turbine at Roosevelt Hot Springs, UT, reported an in-
in operation for over 30 years, flash steam plants since the crease in
power (up to 20% depending upon flow conditions)
early 1980s, and binary plants in commercial operation since compared
to a single-flash steam turbine (37). An advanced
the mid 1980s—substantial room for improvement still exists. system,
diagrammed in Fig. 5, is designed to increase produc-
During the period 1986 to 1992, Ormat Inc. reduced binary tion at a
plant in Cerro Prieto, Mexico from 7,410 kW to
system costs by approximately 30%, largely through equip- 10,760 kW,
a potential gain of 45% (37). The Biphase turbine
ment design improvements that decreased manufacturing can also
operate as a bottoming unit using the hot water from
costs (33). Flash system costs were reduced in the same time steam
separators or can be used as a stand-alone wellhead
period by 20%; the most cost-effective improvement was made generator
to serve remote communities.
by Unocal in their Salton Sea flash plants by replacing the Flash
plants can be made more cost-effective by using
older crystallizer–clarifier technology (at about $17 million more
efficiently the lower-temperature fluid flowing from the
per 40 MW plant) with newer pH-modification technology for first flash
tank. One of the more promising cycles is the Ka-
silicate scaling control (at only a few million dollars per plant) line
cycle, invented and developed by Exergy, Inc., which re-
(8,20). These successes result from the continuing research places
hydrocarbon working fluids with an ammonia–water
efforts of the geothermal industry to improve geothermal mixture,
and uses a number of high-temperature and low-
power conversion systems.
temperature heat exchangers, as shown in Fig. 6, to improve
Geothermal power plant costs are projected to continue to
thermodynamic efficiency. Exergy, Inc. speculates that their
decline (1) as the number of operating personnel, instru- Kalina
system can reduce capital costs by 35% to 40%, in-
ments, controls, and safety systems are reduced as experience crease
brine utilization by 20% to 30%, and reduce the overall
is gained and (2) as improved conversion cycle designs are cost of
power by 30% to 35% (27).
utilized which produce more electricity per pound of geother- Many
of these cycle improvements produce a synergistic
mal fluid through the addition of (a) topping cycles (with the effect.
Although they add components and cost to the system,
ORMAT Biphase rotary separator or Rotoflow turbines) that the
ultimate result is a lower cost per kilowatt-hour, since the
extract extra power from high-temperature fluids, (b) hybrid increased
efficiency requires less geothermal fluid per kilo-
cycles that extract extra power from moderate temperature watt-hour,
which in turn reduces (1) the size (and cost) of
fluids (e.g., by using the proposed Kalina cycle or the ORMAT those
parts of the plant through which the fluids flow and (2)
combined cycle), (c) bottoming cycles that extract extra power the number
of wells needed to be drilled and maintained.
from low-temperature fluids (e.g., by using the vacuum-flash
cycle), and (d) cycles combining combustion turbines with bi- Small
Geothermal Power Plants
nary systems to extract power from the lowest temperature
geothermal reservoirs (34). Small
geothermal power plants, a few megawatts or smaller
Synchronous speed turbines offer significant advantages in size,
can enhance the reliability and backup aspects of off-
for geothermal binary systems. Commercial binary turbines grid or
end-of-grid powering at remote locations—such as on
are high-speed, radial inflow turbines, which require a speed the many
isolated Indonesian islands or in remote villages in
reduction gear box between the turbine and the generator. the Rift
Valley of East Africa. They are also valuable as ‘‘ice-
Synchronous speed turbines rotate at the same speed as the breaker’’
plants installed during the early development of new
generator (being coupled directly to it), thereby avoiding the fields,
providing both (1) the power needed for field develop-
energy losses and cost of the gear box. Synchronous turbines ment
activities and (2) an early source of revenue to help off-
reduce capital costs by 17% while increasing brine utilization set front-
end costs.
by 3% (34). Small
geothermal plants are readily transportable: For 100
Other technological improvements include the use of mixed kW to 300
kW systems, the entire plant including the cooling
working fluids for binary plants (generally mixtures of bu- system
can be built on a single skid fitting into a standard
tanes, pentanes, and hexanes), and the use of metastable, su-
transoceanic shipping container. These small plants are de-
persaturated turbine expansion cycles, capable of producing signed to
be self-starting, with only semiskilled labor needed
up to 10% more power (35). Isobutane, commonly used in bi- to
monitor plant operation on a part-time basis. Completely
nary systems, has a retrograde dew point curve on a tempera- unattended
operation is possible with plant performance
ture–entropy diagram; thus, in contrast to steam, isobutane monitored
and controlled remotely through a satellite link
----------------------- Page 207-----------------------
Existing
flash
tank
Steam
Central steam
turbine plant
Existing low pressure two-phase flow ###
############
###
############
Brine
Steam
back
############
Flash into
orifice Separated existing
steam
############
system
############
############
Biphase Back Generator
############
High turbine pressure

pressure steam
############
two flow turbine
############
Geothermal
wells
Figure 5. Schematic of a Biphase turbine system

designed to increase power up to 45% at a
geothermal well at Cerro Prieto, Mexico.
(38). These systems are environmentally friendly, releasing discussed

above is the improvement of today’s hydrothermal
no greenhouse gases or other pollutants to the atmosphere. technology—
especially drilling and conversion technol-
Power plants such as these have been installed by ORMAT ogy—to reduce
costs, thus making geothermal more competi-
in Thailand. tive with
conventional forms of energy. The second path is
The demand for electric capacity per person at off-grid advanced
research on geopressured, hot dry rock, and magma
sites ranges from 0.2 kW in less-developed areas to greater resources,
whose successful exploitation will greatly expand
than 1.0 kW in developed areas. Thus, a 100 kW geothermal the geographic
availability of geothermal energy and, in view
plant can serve communities of 100 to 500 persons; a 1000 of the large
size of these resources, provide a virtually inex-
kW plant, 1000 to 5000 persons. The estimated cost of power haustible
supply of energy.
for a 300 kW system on a 120#C reservoir is approximately
10.5 ¢/kWh, and it drops to 4.7 ¢/kWh for a 1000 kW plant
on a 140#C reservoir (38). These costs compare quite favorably Advanced
Technology
with alternatives such as diesel generators (46 ¢/kWh to 103

Geopressured. Geopressured resources are not yet com-
¢/kWh) and solar photovoltaic systems with adequate battery mercially
viable, primarily because of today’s low price of nat-
storage (75 ¢/kWh to 100 ¢/kWh) (38). ural gas.
However, as conventional sources of natural gas are
depleted and
prices rise and as production costs of geopres-
FUTURE DEVELOPMENTS sured
resources are reduced, these resources will become com-
petitive.
Geopressured resources represent one of the largest
Research efforts aimed at the increased use of geothermal en- US sources
of unconventional natural gas—with estimates
ergy are proceeding along two broad paths. The first path as of more than
63,000 # 1018 J (63,000 trillion cubic feet) just
Recuperative Ammonia–water
heat exchanger working fluid
Turbine
Air-cooled
Recuperative condenser
preheater
Brine Brine
heat heat
exchanger exchanger
Pump
Figure
6. Schematic of Kalina cycle system, which utilizes
Hot brine To reinjection well
ammonia–water working fluids and cascaded recuperative
from geothermal well heat
exchangers to increase power conversion efficiency.
----------------------- Page 208-----------------------

in the coastal region of Texas and Louisiana (39). Other po- jection
and production well system in order to provide low
tential geopressured basins of the United States identified by impedance
across sufficiently large rock volumes with accept-
the USGS include: Mississippi salt basin, Appalachian basin, able
water losses. Economic analyses show that the perfor-
Anadarko-Ardmore basin, north Louisiana salt basin, Dela- mance of
HDR systems can be improved markedly by having
ware basin, Unita basin, Santa Barbara Channel/Los more
than one production well—preferably two or three.
Angeles/Ventura/Tanner Banks basin, and Gray’s Harbor to Other
concepts have been advanced; for example, a patent
Hoh Head basin area in Oregon and Washington. It has been (42) was
issued to Shulman (Geothermal Power Co., Inc.) on
estimated that an additional 46,000 # 1018 J of thermal en- a
completely enclosed system (eliminating the need for a frac-
ergy exists in the upper 10 km in these basins, with a similar tured

reservoir) with the working fluid passing through con-
amount of energy contained in dissolved natural gas (39). tinuous
metallic pipe installed from the surface, through the
Research carried out under the sponsorship of the US De- hot rock
zone, and back to the surface where the heated fluid
partment of Energy has demonstrated that geopressured is
processed to recover the thermal energy. Economic analyses
wells can be flowed at rates of 40,000 barrels per day, and the indicate
the advantage of several injection wells connected to
brine reinjected underground at depths of 1500 m to 2500 m a manifold
at the surface permitting rotation of the descend-
without causing subsidence or associated seismic activity. ing
liquid among the pipe loops (as individual loops cool) for
Two large sandstone aquifers—at Pleasant Bayou in Texas continuous
operations. Because of the low thermal conductiv-
and at Gladys McCall in Louisiana—each estimated to con- ity of
hard rock, this system will have lower heat exchange
tain in excess of five billion barrels of geopressured brines— rates
than the more conventional system using a large frac-
were tested and characterized (2). A 1 MW hybrid power sys- tured
reservoir. However, such closed systems may be neces-
tem was constructed in 1990 at the Pleasant Bayou site in sary in
regions of highly fractured hot rock where large water
which gas was burned in an engine to generate electricity di- losses
would be unacceptable, and they may also prove to be
rectly. The exhaust heat from the engine was then combined
advantageous for small power plants and/or for direct use.
with heat from the brine to generate additional electricity us- Cost of
energy from HDR resources are highly speculative
ing a binary cycle. Heat from the gas engine, available at a and
highly dependent on the characteristics of the source—
high temperature, improved markedly the efficiency of the bi- especially
its thermal gradient, that is, the rate at which the
nary part of the hybrid cycle. This demonstration showed that
temperature increases with depth, measured in degrees centi-
hybrid systems can yield 30% more power than stand-alone grade per
kilometer of depth. The worldwide average thermal
geothermal and fossil fuel power plants operating on the
gradient
is about 25#C/km. Approximately 16% of the land
same resources (40).
area in
the Untied States can be categorized as a thermal
area—that
is, an area with a significant fraction containing
Hot Dry Rock. Estimates of the useful US hot dry rock re-
regions
with gradients of 60#C/km to 80#C/km. Gradients are
source exceed 500,000 quads (4). In 1970, scientists at the Los
important
for HDR economics because the higher the gradi-
Alamos National Laboratory conceived the idea of extracting
ent, the
shallower a well to reach a given temperature,
heat from this large resource by creating an artificial reser-
greatly
reducing well costs. An economic analysis has esti-
voir through hydraulic fracturing of competent hot rock hav-
mated
that for thermal gradients of 80#C/km with today’s
ing low porosity and low permeability. This reservoir would
technology, busbar electricity costs would be on the order of 5

be interconnected to a heat exchanger at the surface through
¢/kWh to 6
¢/kWh; for gradients of 50#C/km, 8 ¢/kWh to 9 ¢/
a pair of wells (a production well and an injection well) form-
kWh; and
for gradients of 30#C/km, 16 ¢/kWh to 18 ¢/kWh
ing a closed convective circulation loop. The basic HDR con-
(25).
cept was subsequently expanded to encompass the heat min-
ing of all geothermal resources requiring artificial measures HDR
research programs have also been established in Ja-
beyond current technology to achieve commercial heat expan (at
Hijiori), in Great Britain (at Rosemanowes), and in
traction. France
(at Soultz) under the auspices of Germany and
The world’s first HDR system was created in 1977 at Fen- France,
in union with the European Community, subse-
ton Hill, NM. The system was constructed by drilling a well to quently
joined by Great Britain and the United States. Under
3000 m into granitic rock at 185#C; hydraulic fractures were this
program, a consortium of European industrial firms has
produced at 2600 m depth; and after redrilling the production undertaken
a $300 M effort aimed specifically at the develop-
well to intersect the fractures, hydraulic communication was ment of
HDR in areas of low thermal gradient (41).
achieved. Pressurized water was circulated through the frac- The
US HDR research program is aimed at developing
tures bringing heat to the surface at temperatures up to technology
to enable industrial HDR projects to generate
140#C, with a thermal energy output of 5 MW, some of which power at
less than 9.5 ¢/kWh early in the twenty-first century
was used to operate a 60 kW binary-cycle, electrical generator (43).
Specific research areas include: (1) drilling—to develop
(41). Based on the successful operation of this system, a better
technology for creating fractures and for completing
larger, deeper (4390 m) system was constructed in the early and
logging wells, to develop means of locating accurately the
1980s at Fenton Hill with maximum rock temperatures of
intersection of fractures with the wellbores, and to reduce the
327#C and a thermal output of 10 MW. cost of
drilling deep wells in hard, hot rock; (2) reservoir defi-
Although HDR technology has tremendous potential, its nition—to
improve instrumentation to locate, measure, and
commercialization will depend on resolving several technical control
fracture propagation in HDR reservoirs; (3) reservoir
uncertainties, such as reservoir productivity and lifetime, wa- evaluation
—to develop technology to monitor changes in res-
ter loss rates, flow impedance, and corrosion and scaling se- ervoir
volume and temperature and to study reservoir draw-
verity. The most critical technical obstacle is centered on the down
characteristics; (4) system optimization—to evaluate
formation and connection of the fractured network to the in- and model
the performance of HDR reservoirs in order to de-
----------------------- Page 209-----------------------
velop improved cost estimates for electricity production and fluid

circulated down the annulus and up the inner pipe, ex-
to evaluate the efficiency of various power plant designs.
tracting heat without the working fluid ever contacting the
HDR systems can both generate baseload electricity and
magma, and (2) an open heat exchanger system also con-
be used in load-following modes. An experiment at Fenton
sisting of two concentric pipe strings down to the magma
Hill in 1995 demonstrated that a HDR reservoir is capable of
chamber but with only the inner pipe penetrating the magma.
a significant, rapid increase in thermal power output on de- Fluid
is circulated down the inner string and returned
mand. It is estimated that the thermal output could be in-
through the fractured crust (formed around the inner pipe by
creased up to 65% for four hours each day without requiring the
rapid freezing of the adjacent magma) and then up the
additional wells or a larger reservoir (43). The price premium
annulus to the surface. Engineering analyses indicate that
for peaking power paid by utilities would more than cover the the
amount of energy which can be extracted from a single
additional capital expense required to increase the power magma
well is 3 MW to 5 MW for the closed system and 25
plant capacity, ultimately improving overall economics by ap- MW to
45 MW for the open system—showing the clear superi-
proximately 10%. ority
of the latter (47). Economic analyses conducted at San-
dia
estimate that for a magma reservoir at 5500 m depth the
Magma. The idea of extracting energy directly from cost
of power using the open system would be in the neighbor-
magma emerged as an energy alternative during the early hood
of 8 ¢/kWh to 10 ¢/kWh (49). A second independent cost
1970s when it was realized that molten magma reservoirs
analysis developed for the California Energy Commission es-
within 10 km of the earth’s surface in the continental United
timates that a 50 MW power plant could produce magma
States contain up to half a million quads of energy (0.5 # power
for 5.6 ¢/kWh. Such estimates are clearly speculative;
24 J).
10
however, they do indicate the likelihood that energy can ulti-
Large magma bodies insulated within the earth’s crust
mately be extracted economically from the world’s abundant
have a very slow cooling rate, retaining significant amounts magma
resources (3).
of heat for hundreds of thousands of years. Geophysical data
indicate that large magma chambers exist in various parts of
the world, including Kamchatka, Iceland, Sicily, Japan, the
Global Climate Improvements: CO2 Reduction
Azores, Alaska, and the western United States (44). Several

Various geothermal options, including electricity production,
calderas in the United States are known to be large enough
direct heat application, and heat pump utilization, can be im-
and young enough to retain significant residual magma: the
portant components in a global strategy to transition to re-
Yellowstone caldera in northwestern Wyoming (formed about duced
fossil energy dependency, if and when needed. The pro-
600,000 years ago), the Valles caldera in north-central New
duction of carbon dioxide from the burning of fossil fuels is
Mexico (formed about 1,100,000 years ago), and the Long Val-
currently perceived as a serious threat to climate stability.
ley caldera in east-central California (formed about 730,000 The
United Nations Conference on Environment and Devel-
years ago) (45). The size of these magma bodies may be as
opment was held in Rio de Janeiro in 1992 to ‘‘achieve . . .
large as 1000 km3 with temperatures as high as 1300#C. It is
stabilization of greenhouse gas concentrations in the atmo-
estimated that 2 km3 of magma could provide the energy re-
sphere at a level that would prevent dangerous anthropogenic
quired to operate a 1000 MW power plant for 30 years—an

interference with the climate system’’ (50). Follow-up confer-
energy output of approximately 1018 J, equivalent to 172 mil- ences
were held in Berlin in 1995, Geneva in 1996, and New
lion barrels of oil (46). York

in 1997 to reaffirm this commitment.
The US government initiated a research program in 1974
which successfully demonstrated the scientific feasibility of
this novel concept. The program was then extended to investi-
Reduction of Demand for Fossil Fuels. The most significant
gate engineering feasibility. Several significant findings
greenhouse gas, in terms of both quantity and growth poten-
emerged from this research: (1) Drilling into ultrahigh-tem- tial,
is carbon dioxide, and its principal source originates in
perature lava—using high-velocity water jets in advance of
fossil fuel emissions from energy production, residential and
the drill bits to freeze the magma and maintain a stable bore-
commercial energy use, manufacturing, and transportation.
hole—was successfully demonstrated at the Kilauea Iki lava In
response to the Rio Conference, the United States promul-
lake in Hawaii. (2) Energy extraction through the production gated
a Climate Change Action Plan (CCAP) in 1993 aimed
of dry steam was also demonstrated in the Hawaiian experi- at
returning US greenhouse gas emissions by the year 2000
ment. (3) Engineering materials needed for drilling into to
1990 levels (50). These emissions were projected to grow by
magma and suitable for long-term energy extraction were 7%
by 2000 without the CCAP—from 1462 million metric
evaluated in reconstituted magma environments at 850#C, tons
of carbon equivalent (MMTCE) to 1568 MMTCE, an in-
showing that nickel-based superalloys have excellent chemi-
crease of 106 MMTCE. Under the CCAP, the private and
cal resistance and strength in this hostile environment. (4) It
public sectors established collaborative efforts to accelerate
was also shown that many of the problems associated with
market acceptance of renewable energy technologies. A con-
ultrahigh temperatures (including accelerated drill bit wear,
sortium of geothermal developers and utilities was created to
drilling fluid degradation, drillstring corrosion, and wellbore cost-
share exploration and drilling activities to expand known
instability) can be eliminated or mitigated by using insulated
hydrothermal reserves. The substitution of geothermal en-
drillpipe in conjunction with surface mud coolers to keep dril- ergy
for fossil fuels can markedly reduce CO2 emissions. The
ling fluids cool (47).
annual pollutants per 1000 MW effective electric plant capac-
Two proposed methods of extracting energy from magma ity
are 2.609 MMTCE for a typical coal-fired plant, compared
were analyzed (48): (1) a closed heat exchanger system con- to
0.001 MMTCE for a geothermal flash steam plant. Thus,
sisting of two concentric pipes inserted into the magma, with the
substitution of each 1000 MW of geothermal power for an
----------------------- Page 210-----------------------
equivalent coal-fired plant reduces the greenhouse gas burden to

less than 1% of the hydrogen market, because of the high
by 2.6 MMTCE (51).
cost of electricity. However, the cost of electrolytic hydrogen
The use of geothermal heat pumps (GHPs) reduces the can
be reduced using high-temperature electrolysis with effi-
amount of electricity needed to satisfy the nation’s heating
ciencies greater than 80%. Jonsson (55) explored the feasibil-
and cooling requirements, thus reducing the nation’s CO2 bur- ity
of using geothermal steam as a heat source for a high-
den. GHPs can be categorized as ground-coupled, groundwa-
temperature electrolyzer and found that geothermal-heated
ter, and hybrid. In the first type, a closed loop of pipe is buried
steam at 200#C can reduce the specific electricity require-
horizontally beneath the frost zone (or vertically 30 m to 120
ments by 30% compared to conventional electrolytic pro-
m deep). The second type, used in the United States since the
cesses. Furthermore, the capacity of geothermal fields, most
1930s and until recently the most popular, delivers ground-
efficiently used for base-load electric power, can be increased
water to a heat exchanger installed in the heat pump loop, to
above peak-load demand and the integrated excess capacity
then discharges the groundwater on the surface or into an
used in the production of hydrogen. The lower incremental
injection well. The third type, the hybrid system, is used pri-
cost of the excess power along with appropriate credits for
marily in commercial buildings. Due to the high cost of meet- air
pollution abatement, reduction of greenhouse gases, and
ing peak cooling loads, hybrids typically incorporate a cooling
mitigation of other market externalities can help lower the
tower allowing the engineer-designer to (1) size the ground
cost of hydrogen produced from geothermal energy to competi-
loops for heating loads and (2) use the tower to help meet the
tive levels (56).
larger peak cooling loads.
Over two-thirds of the United State’s electricity is used in
Geothermal’s Growing Global Role

buildings. Space heating and cooling, along with water heat-
ing, account for over 40% of the electric power used by resi-
Geothermal resources are large and widely distributed, espe-
dential and commercial buildings. GHPs have the potential to
cially in many of the rapidly developing countries of the
reduce electric energy consumption and related emissions by
world, including the Philippines, Indonesia, and nations in
20% to 40%. A Geothermal Heat Pump Consortium (GHPC)
East Africa, Central and South America. Geothermal power
was formed in 1994 to accelerate the development and rapid
globally has grown steadily since the early 1920s at a rate of
commercialization of GHPs by promoting research to reduce
more than 8.5% per year, reaching 7000 MW installed capac-
drilling costs for the emplacement of subsurface heat-ex- ity
in 1997, and projected to exceed 11,000 MW by the year
change loops and by developing training programs for engi-
2000. Geothermal power already makes a significant contri-
neers and installers. The GHPC goal is to increase the annual
bution on a regional basis; for example, over 7% of Califor-
installation of geothermal heat pumps from 40,000 to 400,000
nia’s electricity is produced from geothermal energy.
by the year 2000, reducing greenhouse gas emissions by 1.5
Geothermal power plants offer numerous advantages: they
MMTCE annually (52,53). are
simple, safe, and modular (1 MW to 50 MW); have short
construction periods (one year for a 50 MW plant); and are

Displacement of Fossil Fuels: Hydrogen Production. Automo-
capable of providing base, load-following, or peaking capacity.
biles account for approximately one-half of the oil consumed
Moreover, geothermal plants provide significant societal ben-
in the United States while producing more than half of urban
efits: they reduce the demand for imported oil along with the
pollution and one-quarter of that nation’s greenhouse gases.
concomitant national defense and balance-of-payments prob-
Many metropolitan areas in the United States, such as Los
lems, and offer benign environmental attributes (negligible
Angeles, fail to meet the Environmental Protection Agency’s
emissions of CO , NO , SO , and particulates, and modest land
2 x x
air-quality standards. Elsewhere in the world, in cities such and

water use). These features are fully compatible with the
as Mexico City, Tokyo, Jakarta, and Sao Paulo, air pollution
sustainable growth of global energy supplies, making geother-
is even more severe. Many of these cities have initiated com- mal
energy an attractive option.
prehensive studies to understand better the nature of the
The robust growth in geothermal power has been based
problem and to develop technically viable, politically accept-
almost exclusively on the use of high-temperature hydrother-
able, cost-effective solutions. Generally, government solutions mal
resources. If geothermal power is to become more univer-
embrace both regulatory controls (such as limiting the num-
sally available with a significant impact on global energy sup-
ber of vehicles, rationing fuel, and enacting driverless days)
plies, then low-temperature resources (and advanced concepts
and technical advances for pollution abatement (such as im-
including hot dry rock, geopressured, and magma) must be
proved fuels, catalytic converters, and electric vehicles). A
pursued vigorously to make them economically competitive.
major US research program, the Partnership for a New Gen-
This will require an aggressive research program to reduce
eration of Vehicles, was initiated in 1993 by the automobile
field development and energy conversion costs.
industry and the government to create by 2004 cars which
Low-temperature resources provide an economical source
would meet stringent clean-air standards and have markedly of
energy for GHPs and for direct use in domestic, industrial,
improved energy efficiencies (equivalent to 80 mi. per gallon)
agriculture, aquaculture, and district heating applications.
(54). One of the promising approaches is the development of The
installation of GHPs in the United States has been grow-
electric-drive vehicles, which include not only cars powered ing
rapidly at a rate exceeding 15% per year. GHPs offer us-
by batteries, but also vehicles that generate electricity on- ers
an inexpensive means of space heating and cooling, along
board by the use of fuel cells. The ideal fuel for these cells,
with domestic hot water, while offering utilities the benefits
from both a technical and environmental perspective, is hy- of
reduced peak demands for power, and the deferred need for
drogen, which, when burned, produces only water vapor. new
plant capacity.
The electrolytic production of hydrogen—the most environ-
Research programs designed to increase understanding
mentally benign process for generating it—currently amounts and
improve technology for heat mining in the severe environ-
----------------------- Page 211-----------------------
ment of hot dry rock and magma resources are underway in 20.
Anonymous, Managing maturity, in United States Geothermal
Europe, Japan, and the United States. If successful, these ef-
Operations, Los Angeles: Unocal Corp., 1995, p. 2.
forts will make abundant geothermal energy universally 21.
M. Dellinger, Turning community wastes into sustainable geo-
available to humankind.
thermal energy—the S.E. Geysers effluent pipeline project,
Trans. Geotherm. Resour. Coun. 20: 205–208, 1996.

22.
D. L. Gallup et al., Brine heat exchanger treatment method, U.S.
BIBLIOGRAPHY
Patent No. 5,190,664, 1993.
23.
G. Shulman, Low temperature flashed steam power generation,
1. P. D. Blair, T. A. V. Cassel, and R. H. Edelstein, Geothermal En-
in Proceedings of the World Geothermal Congress, 1995, Taupo,
ergy: Investment Decisions and Commercial Development, New
N.Z.: International Geothermal Association, 1995, Vol. 3, pp.
York: Wiley, 1982.
2059–2061.
2. D. B. Lombard and F. L. Goldsberry, Geopressured brine well 24.

U.S. Department of Energy, Availability of geothermal power
tests, Geotherm. Sci. Technol., 225–251, 1988.
plants, in Geothermal Progress Monitor, Washington, DC: DOE,
1987.
3. D. Duchane, Geothermal energy, in Kirk-Othmer Encyclopedia of
Chemical Technology, New York: Wiley, 1994, 4th Ed., Vol. 12, 25.
J. W. Tester and H. J. Herzog, Economic Predictions for Heat
pp. 512–539.
Mining: A Review and Analysis of Hot Dry Rock (HDR) Geothermal
Energy Technology, MIT-EL-90-001, Cambridge, MA: MIT En-

4. L. J. P. Muffler and M. Gaffanti (eds.), Assessment of geothermal
ergy Lab., 1990.

resources in the United States, Geol. Surv. Circ. (U.S.) 790: 1978.
26.
G. V. Beeland, Geothermal Drilling—Access to a Useful Resource
5. J. E. Mock and R. S. H. Toms, Power generation, geothermal, in
and Benign Disposal of Waste Fluids, MCC-331-ETC-94, Alexan-

J. A. Wilkes and R. T. Packard (eds.), Encyclopedia of Architec-
dria, VA: DynCorp-Meridian Corp., 1994.

ture: Design, Engineering and Construction, New York: Wiley,
1989. 27.
D. J. Entingh, Technology Evolution Rationale for Technology
Characterizations of U.S. Geothermal Hydrothermal Electric Sys-

6. U.S. Energy Information Administration, Renewable Resources in
tems, NA-006 B, Falls Church, VA: NOVA Analytics, 1993.

the US Electricity Supply, Washington, DC: DOE/EIA, 1993.
28.
Sandia National Laboratories, Slimhole drilling technology, in FY
7. G. W. Huttrer, The status of world geothermal power production
97 Geothermal R&D Program Plan, Albuquerque, NM: Sandia,

1990–1994, in Proceedings of the World Geothermal Congress,
1996, pp. 121–122.

1995, Taupo, N. Z.: International Geothermal Association, 1995,
Vol. 1, pp. 3–13. 29.
D. A. Glowka and J. L. Wise, Development of advanced synthetic-
diamond drill bits for hard-rock drilling, in Federal Geothermal
8. D. J. Entingh, Geothermal Hydrothermal Electricity, Rockville,
Research Program Update 1996, Rockville, MD: Princeton Eco-

MD: Princeton Economic Research, 1997.
nomic Research, 1997, pp. 3.1–3.5.

9. Geothermal Energy Association, Geothermal Energy in Developing
30.
D. S. Drumheller, Wireless data telemetry, in Federal Geothermal
Countries—Opportunities for Export of U.S. Goods and Services,
Research Program Update 1996, Rockville, MD: Princeton Eco-

Davis, CA: GEA, 1993.
nomic Research, 1997, pp. 3.31–3.34.

10. US Energy Information Administration, Renewable Energy An- 31.
National Research Council, Drilling and Excavation Technologies
nual 1995, Washington, DC: DOE/EIA, 1995.
for the Future, Washington, DC: National Academy Press, 1994.
11. J. E. Mock, J. W. Tester, and P. W. Wright, Geothermal energy 32.

C. R. Peterson, National advanced drilling and excavation tech-
from the earth: Its potential impact as an environmentally sus-
nologies program and institute, in Proceedings of Geothermal Pro-
tainable resource, in Annual Reviews of Energy and the Environ-
gram Review XIV, Washington, DC: U.S. Department of Energy,
ment, Palo Alto, CA: Annual Reviews, 1997, Vol. 22.
1996, pp. 239–245.
12. K. Rafferty, A capital cost comparison of commercial ground- 33.

A. Elovic, Advances in binary organic Rankine cycle technology,
source heat pump systems, in Proceedings of Geothermal Program
Trans. Geotherm. Resour. Coun., p. 511, 1994.
Review XIV, Washington, DC: U.S. Department of Energy, 1996.
34.
J. Brugman et al., Next Generation Geothermal Power Plants,
13. L. W. Pratsch, Geothermal: A household word by the year 2000,
EPRI RP 3657-01, Palo Alto, CA: Electric Power Research Insti-
in International Energy and Environmental Congress, Richmond,
tute, 1995.
VA: Association of Energy Engineers, 1995.
35.
G. L. Mines, Operation of Mammoth-Pacific’s MP1-100 turbine
14. P. J. Lienau, J. W. Lund, and G. G. Culver, Geothermal direct use
with metastable-supersaturated expansions, in Proceedings of
in the United States: 1990–1994, GHC Q. Bull., 16 (2): 1–6, 1995.
Geothermal Program Review XIV, Washington, DC: U.S. Depart-
15. P. J. Lienau and B. C. Lunis (eds.), Geothermal Direct Use Engi-

ment of Energy, 1996, pp. 115–122.
neering and Design Guidebook, Klamath Falls: Geo-Heat Center, 36.
G. L. Mines, Improvements in binary cycle performance: super-
Oregon Institute of Technology, 1991.
saturated turbine expansions, in Federal Geothermal Research
16. UURI, Geothermal Energy: Clean Sustainable Energy, Salt Lake
Program Update 1996, Rockville, MD: PERI, 1997, pp. 5.17–5.21.
City: Earth Science Laboratory, University of Utah Research In- 37.
W. R. Studhalter and J. L. Hahn, Demonstration of economic ben-
stitute, 1993.
efits of a Biphase turbine in a geothermal application, in Federal
17. P. M. Wright, Exploration for direct heat resources, in P. J. Lie-
Geothermal Research Program Update 1996, Rockville, MD: PERI,
nau and B. C. Lunis (eds.), Geothermal Direct Use Engineering
1997, pp. 5.65–5.78.
and Design Guidebook, Klamath Falls: Geo-Heat Center, Oregon 38.
D. J. Entingh, E. Easwaran, and L. McLarty, Small Geothermal
Institute of Technology, 1991, p. 55.
Electric Systems for Remote Powering, Rockville, MD: PERI, 1994.
18. J. W. Tester, Energy conversion and economic issues for geother- 39.
R. H. Wallace et al., Assessment of geopressured–geothermal re-
mal energy, in Handbook of Geothermal Energy, Houston, TX:
sources in the northern Gulf of Mexico basin, Geol. Surv. Circ.
Gulf Publishing, 1982, pp. 471–588.
(U.S.) 790: 1978.
19. J. W. Tester, Geothermal energy, in Kirk-Othmer Encyclopedia of 40.

R. G. Campbell and M. M. Hattar, Design and Operation of a
Chemical Technology, New York: Wiley, 1980, 3rd Ed., Vol. 11,
Geopressured–Geothermal Hybrid Cycle Power Plant, Holt Rep.
pp. 746–790.
30008, Pasadena, CA: Ben Holt Co., 1991.
----------------------- Page 212-----------------------
388 GETTERING IN SILICON
41. USGS, Potential of hot dry rock geothermal energy in the eastern
United States, Geol. Surv. Open-File Rep. (U.S.) 93-377: 1993.
42. G. Shulman, Geothermal heat mining and utilization, U.S. Pa-

tent No. 5,515,679, 1996.
43. L. McLarty and D. Entingh, Technology Characterization; Geo-

thermal Hot Dry Rock Electricity, Rockville, MD: PERI, 1997.
44. H. C. H. Armstead and J. W. Tester, Heat Mining, New York:

E. & F. N. Spon, 1987.
45. S. Bender-Lamb, Magma energy exploratory well Long Valley

caldera, Calif. Geol., April, pp. 85–92, 1991.
46. J. C. Eichelberger and J. C. Dunn, Magma energy: What is the

potential? Geotherm. Resour. Coun. Bull., February, pp. 53–55,
1990.
47. T. Y. Chu et al., The magma energy program, Geotherm. Resour.

Coun. Bull., February, pp. 42–52, 1990.
48. C. C. Carson and J. Haraden, The economics of magma power

generation, Trans. Geotherm. Resour. Coun., pp. 171–175, 1985.
49. J. C. Dunn, Status of the magma energy project, Geotherm. Re-

sour. Coun. Bull., pp. 3–9, July, 1988.
50. U.S. Department of Energy, The Climate Change Action Plan,

Washington, DC: DOE, 1993.
51. National Geothermal Association, Geothermal Energy for a

Cleaner Environment, Davis, CA: NGA, 1992.
52. L. W. Pratsch, Geothermal, a household word? Trans. Geotherm.

Resour. Coun., 18: 381–385, 1994.
53. Edison Electric Institute, National Earth Comfort Program: A Pro-

posal for an Industry—Government Collaborative, Washington,
DC: EEI, 1994.
54. E. Marshall, Slower road for clean-car program, Science 276:

194, 1997.
55. V. K. Jonsson, Feasibility of using geothermal energy in hydrogen

production, Geothermics, 21 (5/6): 673–681, 1992.
56. P. Kruger, Geothermal Hydrogen Program (Annu. Rep. No. 2),

Stanford, CA: Civil Engineering Department, Stanford Univer-
sity, 1996.
Reading List
J. E. Mock, J. W. Tester, and P. M. Wright, Geothermal energy from

the good earth: An abundant and environmentally sustainable re-
source, in Annual Reviews of Energy and the Environment, Palo
Alto, CA: Annual Reviews, 1997, Vol. 22.
Proceedings of the World Geothermal Congress 1995, Taupo, N.Z.: In-

ternational Geothermal Association, 1995.
National Research Council, Drilling and Excavation Technologies for

the Future, Washington, DC: National Academy Press, 1994.
J. C. Bresee (ed.), Geothermal Energy in Europe, Philadelphia: Gordon

and Breach, 1992.
P. J. Lienau and B. C. Lunis (eds.), Geothermal Direct Use Engi-

neering and Design Guidebook, Klamath Falls: Geo-Heat Center,
Oregon Institute of Technology, 1991.
H. C. H. Armstead and J. W. Tester, Heat Mining, New York: E. &

F. N. Spon, 1987.
P. D. Blair, T. A. V. Cassel, and R. H. Edelstein, Geothermal Energy:

Investment Decisions & Commercial Development, New York: Wi-
ley, 1982.
S. L. Milora and J. W. Tester, Geothermal Energy as a Source of Elec-

tric Power, Cambridge, MA: MIT Press, 1976.
P. Kruger and C. Otte, Geothermal Energy Resources, Production,

Stimulation, Stanford, CA: Standford University Press, 1972.
JOHN E. MOCK
Technology Transfer International
----------------------- Page 213-----------------------

} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Homopolar and Inductor Machines
Standard Article

Manfred H. Stiebler1 and Kezhong Guo2
1Technical University of Berlin, Berlin, Germany
2Shanghai Jiaotong University, Shanghai, PR China

Search
reserved.
DOI: 10.1002/047134608X.W3004
❍ Acronym Finder
Abstract
Machines with Homopolar Excitation
Homopolar DC Machines
Inductor Machines
Variable-Speed Inductor Machine Drive
Hybrid Stepping Motors
Appendix

%20Conversion/W3004.htm17.06.2008 16:40:21
----------------------- Page 214-----------------------

c
HOMOPOLAR AND INDUCTOR MACHINES
Electric machines are energy converters in which force and torque are normally
produced by a load current
in the presence of a magnetic field. This article is devoted to a
special class of machines in which the load
winding is induced by a homopolar flux that generates dc voltage or by a flux
modulation due to the variable
reluctance of one member that produces ac voltage. These machines differ in
construction from conventional
dc and ac machines, and they are designed mostly for special applications. The
following paragraphs describe
the concepts of heteropolar and homopolar excitation and the derived machine
concepts within the scope of
this article.
Heteropolar Excitation
Rotating electrical machines are mostly heteropolar, where magnetic north and
south poles are placed
with alternating polarity along the circumference. In synchronous structures,
the magnetic flux is supplied
by one member, which is generally the rotor. In conventional dc machines which
belong to the synchronous
structures, the flux supplying member is the stator. Asynchronous structures, where
the field generated by ac
currents in the primary member is also heteropolar, are not considered here.
In machines with alternating pole polarity, the airgap field curve does not
contain a homopolar component.
The voltage induced in the armature winding on the other member of the machine
during rotation is determined
by the law of induction. Its basic integral form, valid for a conductor loop of
area A with boundary line s moving
with velocity v in a field with magnetic flux density B , is given here:
Note that B , v, A , and s are vectors. Applied to a coil in a machine, the first
term on the right side of Eq. (1)
represents the transformer voltage component due to time-variable flux, and the
second term, due to the speed
of the coil moving in the field, denotes the rotational voltage component.
Heteropolar structures have a pole
pitch defined by the field exciting member. In a dc machine where axial conductors of
active length l move with
constant tangential velocity v in a radial field B , the voltage induced in a coil
having N c turns of 180◦ pitch is
a well-known special case of Eq. (1):
where
Hence for constant speed, the ac voltage waveform reflects the flux
density waveform. For sinusoidal field
distribution and constant speed, the generated voltage is also a sine
wave. In dc machines the alternating
----------------------- Page 215-----------------------
2 HOMOPOLAR AND INDUCTOR MACHINES
voltage is rectified by a mechanical commutator. In synchronous machines the no-load

coil voltage follows Eq.
(2) directly.
Homopolar Excitation
In contrast to the heteropolar magnetic circuit, with homopolar excitation
the flux passes through the
magnetic circuit in only one direction from the field-supplying member
which is generally the stator. This
homopolar field excitation is characteristic for two different machine
types, differing in armature winding
design and the voltage-induction mechanism.
In an acyclic dc machine the armature winding is fixed to the rotating member.
The winding can either be
a cylinder of a conducting material or consist of conductors arranged as a cage.
The dc load current has to be
transmitted to and from the rotor via sliding contacts. This type is called a
homopolar machine in a narrower
sense.
The cyclic ac machine has its armature winding placed on the
stationary member together with the
excitation winding. A variable reluctance rotor, which itself carries no winding,
modulates the flux linked with
the load winding in such a way that ac voltage induction occurs. The inductor
machines are of this type. In
smaller machines the homopolar flux is usually produced by permanent magnets,
preferably fixed to the rotor.
Such machines, known as hybrid stepping motors, are also considered in this
article.
Machine Concepts
A systematic overview of the different machine concepts with homopolar and
heteropolar excitation is
given in Fig. 1. The following features can be assigned to the different types of
machines:
• Machines with homopolar excitation

Type (1): acyclic dc machines
field supplied by a ring winding on the stator

load (armature) winding in the form of a cylinder or bar-winding on the
rotor, slip-ring brush contacts
for current transmission
Type (2): cyclic machines, homopolar inductor machines, and hybrid motors
field supplied by a ring winding on the stator or by a

permanent magnet, mostly on the rotor,
producing an axial field
ac armature winding on the stator
• Machines with heteropolar excitation

Type (3): heteropolar inductor machines including medium-frequency generators
field supplied on the stator by coils

armature winding on the stator
Type (4): claw-pole machines, also transversal-flux machines featuring
field supplied on the rotor by permanent magnets with alternating

polarity
an ac ring winding with claw-pole arrangement on the stator
or alternatively
field supplied by a ring winding together with claw-poles on the rotor

ac coil winding on the stator
The solutions indicated in Fig. 1 are of synchronous structure, but they do

not cover conventional syn-
chronous machines of the turbo or salient-pole type.
----------------------- Page 216-----------------------
HOMOPOLAR AND
INDUCTOR MACHINES 3
Fig. 1. Schemes of machines with homopolar and heteropolar

excitation.
Although the medium-frequency generators, well known as Lorenz and Guy type
machines, have lost their
market to static power electronic equipment, claw-pole machines and especially
transversal-flux machines are
under consideration for direct-driven, low-speed generators or vehicle motors, see
Ref. 17. Within the scope
of this article, of the heteropolar structures only the inductor machine is
considered because its behavior is
similar to the homopolar inductor machine.
Machines with Homopolar Excitation
Acyclic Induction. The mechanism of voltage induction in a homopolar

acyclic machine is not easily
understood. It was shown that an exact theory requires a relativistic
approach. However, for a practical
understanding, the following description may be sufficient (see Fig. 2). A conductor
cylinder 1 of length l is
connected to slip rings 2 and 2# with brushes 3 and 3# gliding on them. They
supply the load 4. The induced
voltage u = −d#/dt in this circuit is given by the rate of change of the flux linked
with area A . The boundary
of A represents one turn, and the flux linkage is equal to the flux, # = # =B ·
A . Because the flux density is
constant, u = −B · dA/dt. There is only one portion of A which changes due to
rotation. When the rotor surface
speed is v, the winding area decreases so that dA/dt = −lv, and we end up with:
This equation is formally the same as for one armature conductor of a conventional
dc machine; see also Eq.
(2). Here with constant flux density and speed the equation describes a dc voltage.
The inherent absence of
voltage harmonics is the reason that homopolar acyclic machines are recommended as
tacho generators.
Cyclic Induction. For the cyclic types of machine, consider Fig.
3. Both excitation 1 and armature
winding 2 are placed on the same stationary member. The rotor has a salient
structure. The homopolar flux
passes from stator to rotor in the left part I and from rotor to stator in the
right part II. The pole pitch of the
ac winding is related to the number of rotor teeth. Depending on the rotor
position, the flux linkage of a coil
varies between a maximum and a minimum value. Suppose the variable reluctance has
only a fundamental
----------------------- Page 217-----------------------
Fig. 2. Illustration of homopolar acyclic

induction.
Fig. 3. Principle construction of a homopolar ac

machine.
component. Then we can describe the flux (Fig. 4) by
#0 is the mean value, and #a the resulting flux component linked with the ac
winding coil. When a machine
that has zr rotor projections rotates with constant angular speed #, the induced
voltage in a coil of N c turns is
given by
These machines utilize only up to half of the power per volume compared with
heteropolar machines because
twice the amplitude of the ac flux density component cannot exceed the homopolar flux
density.
----------------------- Page 218-----------------------
HOMOPOLAR AND
INDUCTOR MACHINES 5
Fig. 4. Flux waveform of a homopolar ac machine.
Fig. 5. Principle construction of a homopolar dc

machine.
Homopolar DC Machines
Designs and Construction. For the principle of the classical

homopolar dc machine, see Fig. 5.
Generally the machine is designed as a double system to obtain symmetrical
conditions for flux and armature
reaction. It consists of a stationary member with an excitation winding 1 and a
rotating member carrying the
armature 2, here, in two parts. In its simplest form the winding consists of a
conducting cylinder. Slip rings 3
at both ends of the conductors on which brushes are gliding transmit the load
current.
This machine is typical for very low voltages of only a few volts. A series
connection of conductors cannot be
made in a simple way. On the other hand, the ability to supply very large currents
drew attention to homopolar
generators for special purposes. However, static equipment with power electronic
circuits has become a serious
competition, so homopolar generators have had no breakthrough for electrolytic
plants.
However, homopolar dc machines received new consideration some years ago for
supplying impulse loads,
such as in electromagnetic launch facilities (rail guns) or nuclear fusion
(tokomacs). The concept is to couple
the homopolar generator with rotating energy storage (flywheel). Designs for rated
values of up to 12.5 MA, 50
V impulse loads with 1 GJ energy have been reported. See also references 1,2,3.
The construction can have an iron or a non-iron core. Especially
for the latter, construction with a
superconducting excitation winding offers advantages. In the 1970s a homopolar disk
machine, developed in
the U.K. and intended for application in marine propulsion, drew much attention. A
special feature was the
----------------------- Page 219-----------------------
Fig. 6. Homopolar dc generator with segmented armature disc and

superconducting excitation winding.
segmentation of the armature disk, allowing series connection of conductors to

obtain higher voltage values
with the penalty of additional brushes (Fig. 6). See also references 4,5,6.
The construction can have an iron or a non-iron core. Especially
for the latter a construction with a
superconducting excitation winding offers advantages. In the 1970s a homopolar disk
machine developed in
the U.K. which was intended for application in marine propulsion drew much
attention. A special feature was
the segmentation of the armature disk, allowing series connection of conductors to
obtain higher voltage values
with the penalty of additional brushes (Fig. 6). See also Refs. 4,5,6.
Current Transport by Sliding Contacts. A problem known from
synchronous generators of high
rating is the transport of very large currents by slip rings and brushes. This
topic has also drawn attention
in connection with the development of homopolar dc machines for current of more
than 2 kA where problems
with brush wear are encountered with solid material combination, and, moreover,
where the voltage drop is
unfavorably high compared to the machine armature voltage.
Normal sliding contacts combine for example, bronze rings with carbon
brushes. Selecting the material
combination and design features, like surface finish and specific contact pressure,
requires much experience to
obtain an operating life of more than several thousand hours.
For very large current in connection with large gliding speeds, new concepts,
such as metal matrix contact
surfaces and fluid contacts have been investigated. In the first concept wear-
resistant materials are combined
with conductive binders. Good results have been reported with an aluminum bronze
matrix. Liquid contacts
were first investigated in Austria in the 1960s using mercury together with
different contact arrangements. It
was shown that currents of a few thousand amperes can be transferred with only a
small voltage drop (Fig.
7). Due to the toxic properties of Hg, the environmental impact and safety
requirements make this solution
expensive for practical use. Contacts using NaKa have been considered as an
alternative. See also Refs. 7 and
8.
----------------------- Page 220-----------------------
HOMOPOLAR
AND INDUCTOR MACHINES 7
Fig. 7. Principal arrangement with liquid ring

contact.
Inductor Machines
Designs and Construction. An inductor machine is a synchronous

machine in which one member,
usually stationary, carries armature and excitation windings or permanent magnets
suitably disposed relative
to each other, and in which the other member, usually rotating, is without windings
but carries a number of
regular projections.
In the group of homopolar machines the field winding excites a
unipolar flux which is modulated by
rotor teeth to produce an alternating flux linkage of the ac winding. Variable
reluctance is essential for their
performance.
There are both ac windings and dc excitation windings on the stator. In such
machines the stator flux is
unipolar with an alternating component superimposed on the dc excitation field.
Hence the useful amplitude
of the alternating flux component is limited to less than half the maximum
instantaneous flux.
There are two different embodiments of inductor machines. In the configuration
of Fig. 3 the rotor flux
has a unipolar direction. This design is the homopolar inductor machine,
similar to hybrid step motors in
the fractional horsepower range. In the version of Fig. 8, the rotor
teeth flux alternates, and this type is a
heteropolar inductor machine.
Operation. Inductor machines are of the synchronous type. This applies
to the arrangements outlined
in Figs. 3 and 8. For the latter, the operation is illustrated by Fig. 9. The ratio
of rotor teeth z /stator pole pairs
p s is an uneven number. Here we have zr = 6 andp s = 2. The ac winding normally

comprises m = 3 phases. Then
the coil pitch is one-half of the rotor tooth pitch. The phases U, V, W, and the
field winding F are magnetically
coupled. A damper winding can be introduced if required.
The properties of such a machine can be described analogously to a
synchronous machine using the well-
known equations and circuit representations according to Park’s theory. The
variable reluctance is described
by
and
----------------------- Page 221-----------------------
Fig. 8. Versions of an inductor machine: (a) homopolar;

(b) heteropolar.
where λ0 is the average specific magnetic permeance of the air gap and γ is the
modulation factor.
Because the period of λ is the rotor tooth pitch, it consists only of uneven
components. Note that in Eq. (4)
field harmonics above the fundamental were neglected. Then the machine model is
similar to a conventional
turbo rotor synchronous machine, except that the armature reaction needs special
consideration when describ-
ing the mutual inductance of the field and the ac winding. The direct and quadrature
axis (d,q)-components of
the armature flux are given in terms of the armature current d,q-components and the
field current by
where Lm is the phase inductance corresponding to the average magnetic permeance.

The relevant reactance X m = ωLm consists of two parts. One represents the
magnetizing component X mµ ,
and the rest X ms has the character of a stray field component. This is described
by
----------------------- Page 222-----------------------

HOMOPOLAR AND INDUCTOR MACHINES 9
Fig. 9. Arrangement of poles and winding coils in a

heteropolar inductor machine.
Fig. 10. Circuit diagram of inductor machine

to describe ac operation.
So the ac circuit representation of the machine in two-axis components

without damper winding can be
#
described by Fig. 10, where Up = γX m · i f is the equivalent no-
load voltage, and X σ is the conventional stray
reactance comprising slot-, overhang- and tooth-tip stray field components. Although
γ is not very different
from 1, X ms ≈ 0.5 X m . The short-circuit reactance of the inductor machine
cannot be made significantly lower
than the synchronous reactance. The values reported in Ref. 9 for a 62 kVA, 3000
rpm machine were, in the
# ##
per-unit system, X d = 0.59, X d = 0.42, and X d = 0.39 p.u.
Variable-Speed Inductor Machine Drive
Inverter with dc Side Commutation. An

adjustable speed drive for large power ratings is a load-
commutated inverter (LCI ) synchronous motor drive having a dc link converter.
Because of the load-commutated
operation of the machine-side thyristor inverter, the subtransient
inductance of the machine is part of the
commutation circuit, limiting the firing angle, and hence the maximum speed and
power factor of the machine.
This is why machines with small stray inductance values are preferred.
Some years ago a new type of inverter with backward conducting valves in
the form of thyristors with
parallel diodes in the reverse direction was proposed. This inverter operates with
external commutation on
the dc side, and the machine inductance is no longer a part of the commutation
circuit (Ref. 10). This draws
attention to such types of synchronous motors which, because of their design, are
suited for high speeds but
have large stray inductances. Among them are homopolar and claw-pole type machines.
In the following the
----------------------- Page 223-----------------------
Fig. 11. Six-pulse bridge converter with dc

side commutation.
steady-state operation of an inductor motor supplied by a six-pulse bridge inverter

of the dc side commutated
type (DCI ) is investigated and compared with a motor of standard design in a
conventional LCI drive.
The inverter considered is characterized by external commutation on the dc
side. The circuit of Fig. 11
contains thyristors and parallel diodes in reverse direction in each of
the six branches. These bidirectional
valves determine the properties of the inverter which in this article will be
called DCI. Compared with the
common LCI bridge circuit with unidirectional thyristor switches, we can recognize
dual behavior. The latter
is best understood by assuming an alternating fundamental voltage on the ac side
and constant current on the
dc side, while the new converter reveals its basic performance when there is a
fundamental ac current source
on the ac side and a constant dc voltage impressed on the dc side.
Under these assumptions Fig. 12 shows the waveforms of phase voltage and
current on the ac side when
the control angle α = 150◦ and the energy flows from the intermediate circuit to
the ac load. In the absence of an
overlap, the fundamental power factor cos φ is determined by φ = π − α. We can

conclude that to attain a good
power factor, the control angle α should be as close as possible to 180◦ . Except
for the hold-off interval assigned
to the thyristors, α is limited only by the inductance in the dc commutation

circuit which is very small. Hence
the overlap angle is almost negligible. This is different in the conventional LCI
inverter where considerable
values of the overlap angle appear. Figure 13 allows a comparison.
Performance. In this section the steady-state operation of a
converter drive with an inductor type
machine is investigated. The complete power circuit diagram comprises a controlled
six-pulse rectifier supplied
by the utility system, an intermediate circuit with a capacitor acting as a voltage
source, the dc side commutated
inverter, and the synchronous motor.
As an example a 12-pole motor is considered with a rated speed of 3000 min −
1 assigned to 300 Hz (Ref.
11). It is characterized by a relatively small synchronous reactance and

relatively large subtransient and
# ##
short-circuit reactance values (X d = 0.59, X d = 0.42, X d = 0.39,
and X k = 0.47 p.u.).
The steady state of a converter drive with an inductor motor is simulated.
The motor is operated under
rated load in an overexcited state. This means that the armature current
is ahead of the terminal voltage.
Integration by the Runge–Kutta method was used to obtain high accuracy. In Fig. 13
simulation results are
presented for the motor running at rated speed with rated power supplied by the
inverter with commutation
----------------------- Page 224-----------------------
HOMOPOLAR AND INDUCTOR MACHINES 11
Fig. 12. Idealized voltage and current waveforms of the dc

side commutated inverter (DCI ).
◦
◦
on the dc side. The firing angle was α = 143.4 . The overlap of u = 0.4
was almost negligible. This results in a
margin angle of γ = 37◦ with an equivalent of 342 µs for the thyristors to
recover. Under these circumstances
the fundamental power factor cos φ = 0.8.

For comparison, a standard salient-pole synchronous motor supplied
by a conventional LCI inverter
##
#
was considered. Its subtransient reactance was X d =
0.15, the transient reactance was X d = 0.25, and the
----------------------- Page 225-----------------------
Fig. 13. Comparison of motor voltages and currents (a) for dc side commutated
inverter (DCI ); (b) for load-commutated
inverter (LCI ).
synchronous reactance was X d = 0.9 p.u. The firing angle was adjusted to
yield the same power factor cos
φ = 0.8.
The analysis of the harmonic contents of machine current and
terminal voltages shows remarkable
differences between the LCI and the DCI performance. In Table 1 the
fifth- and seventh-order harmonics
reveal only a remarkably small deviation from a sinusoidal shape with the DCI.
Consequently, the pulsating
torque component generated by a DCI drive is also smaller than with a LCI
configuration.
A deeper insight into the characteristics of the converter fed machine can be
obtained by inspecting the
trajectories of voltages and current in the space-phasor plane. It is well known
that the steady-state, 120◦ block
output waveform of a three-phase, six-step inverter is associated with a regular

hexagon in the phase plane.
Its corners correspond to discrete values of the space phasor during one-sixth of
the period.
In Fig. 14 trajectories are plotted with respect to the stator reference
frame. With the DCI the voltage
trajectory forms a hexagon. The current trace indicates the presence of
some harmonics. In the case of a
purely sinusoidal waveform, there would be a circle. Note that the angular velocity
of the space phasor is not
----------------------- Page 226-----------------------
HOMOPOLAR AND
INDUCTOR MACHINES 13
constant (parameter values of synchronous angle ωt are not indicated here). It is

interesting to consider the
phase diagrams of the same quantities when the reference frame rotates with
synchronous speed, as shown
in the lower part of Fig. 14. Now deviations from a sinusoidal waveform are
indicated by a loop around the
fundamental phasor.
To make a comparison with the performance of the conventional LCI, refer to
Fig. 15. According to the dual
behavior, the hexagon here represents the nearly 120◦ block waveform
current. The motor terminal voltage
resembles a circle except for the notches due to commutation.

The aforementioned converter drive is suitable for high speeds. It
is similar to the load-commutated
inverter (LCI ) drive with a synchronous motor but features a voltage source link
and an inverter with dc side
commutation (DCI ). This inverter matches favorably with machines which, because of
their design, have a large
short-circuit inductance. Among these are machines of the inductor type with their
excitation winding on the
stator. The inverter circuit is simple because normal thyristors for line
commutation are used and it does not
need switch-off semiconductor elements like gate-turn-off thyristors (GTOs). This
simplicity may compensate
for the lower rating of inductor machines as compared with standard design
synchronous machines.
Hybrid Stepping Motors
Designs and Construction. Stepping motors are known in different

configurations:
Permanent magnet motors with heteropolar excitation, especially claw-pole motors
Variable reluctance motors, with double-salient structure

Hybrid motors with one permanent magnet combined with double-salient structure
Of these types the hybrid motor allows for small step angles and displays
large holding torques. Hence
it is preferred for industrial applications. Figure 16a shows the principle of
construction. Note that the cogged
rotor disks embracing the axially magnetized permanent magnet are displaced by one-
half of the tooth pitch.
The step angle is given by
◦
The relatively simple motor depicted in Fig. 16(b) has m = 2 phases and zr = 9
rotor teeth. Hence α = 10 . Phase
numbers up to m = 5 are in use.
Because a hybrid motor is also a synchronous machine, the relevant analysis
can be adapted. Assuming
a two-phase symmetrical system at constant synchronous speed with
sinusoidal waveforms of voltages and
----------------------- Page 227-----------------------
Fig. 14. Space phasor trajectories for steady-state operation of a DCI: (a,b) in
stator frame; (c,d) in synchronous rotating
frame.
currents, the voltage equations in d,q-components can be written as
where
τ = L/R is the winding time constant with L = (Ld +Lq )/2
q = (Ld −Lq )/(Ld +Lq ) is the reluctance factor
ω = zr · dα/dt is the motor speed in electrical rad/s
----------------------- Page 228-----------------------
HOMOPOLAR
Fig. 15. Space phasor trajectories for steady-state operation of

a LCI: (a,b) in stator frame.
ϑ = is the load angle

#0 = is the flux linkage due to permanent magnet excitation
ˆ
Us = is the amplitude of a sinusoidal terminal voltage of frequency ω
In Fig. 17 the circuit representation in d,q-quantities is depicted. The

torque is made up of two components,
one dependent on the magnet flux and the other on the reluctance modulation factor.
This is illustrated by the
torque equation
This model does not take into account iron losses due to alternating magnetization.
It needs the addition of
damper windings to describe the motor behavior in constant current operation.
Stepping motors are electromagnetic energy converters that render a
certain angle increment on the
application of one impulse. The number of steps which the rotor proceeds
in 1 s corresponds to the step
frequency f z . Two-phase motors can be supplied in full-step or half-step
operation (Fig. 18).
In practice the maximum torque decreases with increasing step
frequency. It is usual to specify the
performance of a motor by three limiting torque curves which imply synchronous
operation on a sinusoidal
supply (see Fig. 19):
(1) maximum torque for continuous operation assuming constant speed

(2) maximum load torque for starting (no additional inertia)
(3) maximum load torque TL for starting with a load inertia JL
The transient behavior can be investigated by determining the trajectory of

speed versus position in the
phase plane. Successful operation following a set-point sequence is indicated by a
stable end position, after
decay of oscillations, with the proper number of steps accomplished (Fig. 20).
----------------------- Page 229-----------------------
Fig. 16. Principle construction of two-phase hybrid motors: (a) section;

(b) example with 10◦ step angle.
Stepping Motor Drives. The transistor switches energizing the

windings can be arranged in bipolar
or unipolar circuits. The latter require splitting up the phase windings into two
coils each, but need only half
the number of switches compared with the bipolar circuit. Hence the unipolar
version prevails.
Manufacturers offer a variety of supply equipment, frequency generators, and
program units. With suit-
able motors, subdivision of the step number per turn can be made in
the so-called microstep scheme. For
----------------------- Page 230-----------------------
HOMOPOLAR
Fig. 17. Circuit representation of a hybrid motor in

d,q-components.
Fig. 18. Step rate and excitation pattern for two-phase, full-
step and half-step operation.
◦
instance a 200 step (1.8 ) motor can be run with five intermediate positions per
full step by controlling the
phase currents with a supplementary unit.
The standard operation of stepping motors is in open-loop control.
When loaded above the maximum
available torque, the motor loses steps. In recent times strategies for a closed-
loop operation have been intro-
duced by which the performance is made similar to an electronically commutated
(EC ) motor and the main
disadvantage of stepping motors is overcome. This is especially true when available
methods for sensorless
angle determination are applied.
----------------------- Page 231-----------------------
Fig. 19. Torque limits in dependence on step

rate.
Fig. 20. Phase-plane representation (angular speed over angle) of a sequence of

steps: (a) two steps; (b) four steps with
step loss.
Appendix
Definitions from the International Electrotechnical Vocabulary
Homopolar Machine A machine in which the magnetic flux passes in the same
direction from one member to
the other over the whole of a single air-gap area (411-31-02)
----------------------- Page 232-----------------------
HOMOPOLAR AND
INDUCTOR MACHINES 19
Acyclic Machine A direct current homopolar machine (411-31-03)
Heteropolar Machine A machine having successive physical or effective poles
of opposite polarity (411-31-04)
Inductor Machine A synchronous machine in which one member, usually
stationary, carries armature and
excitation windings or permanent magnets effectively disposed relative to each
other, and in which the other
member, usually rotating, is without windings but carries a number of regular
projections (411-31-11)
BIBLIOGRAPHY
1. G. Wenqi et al., Development and test of a 300 kW superconducting homopolar

generator, IEEE Trans. Magn. , 32 (4):
part 1, 2280–2283, 1996.
2. G. Wenqi, T. Shaodong, Superconducting homopolar machine, 13th Int.
Cryogenic Eng. Conf., 1990, Beijing, China,
Cryogenics, 30: p. 743.
3. R. C. Zowarka, J. P. Kajs, 25 GW homopolar generator experiment, 8th IEEE Int.
Pulsed Power Conf., San Diego, 1991,
pp. 263–267.
4. C. J. Mole et al., Superconducting electrical machinery, Proc. IEEE , 61: 95–
105, 1973.
5. M. Hippner et al., High speed homopolar synchronous generator no load
calculations with the help of a finite element
package, Int. Conf. SM’100 , ETH Zurich, 1991, pp. 812–816.
6. A. J. Micham et al., The self-excited homopolar generator, part 1: Theory and
electrical design, IEEE Trans. Magn. , 25:
362–368, 1989.
7. F. Caricchi, New mercury-gap ring-collector system for double-armature
generator, IEEE Trans. Ind. Appl. , 29: 1129–
1134, 1993.
8. D. Maribo, N. A. Sondergaard, Further studies of a low melting point alloy
used in a liquid metal current collector,
IEEE Trans. Compon. Hybrids Manuf. Technol. , CHMT-10: 452–455, 1987.
9. W. Bedorf, Ermittlung der feldverteilung und berechnungsmethoden fuer
induktormaschinen im hinblick auf die an-
wendung als stromrichtermotor, Ph.D. Dissertation, RWTH, Aachen, 1977.
10. H. P. Beck, M. Michel, Die sechspulsbrueckenschaltung mit
gleichspannungsseitiger kommutierung, Arch. Elektrot. ,
66: 49–56, 1983.
11. M. Stiebler, G. Kezhong, Homopolar synchronous machines for converter
fed drives, Proc. Int. Conf. Evolution and
Modern Aspects of Synchronous Machines , part 3, Zuerich, 1991, pp. 848–851.
12. H. Sequenz, Elektrische Maschinen , 8th ed., Wien: Springer-Verlag, 1971.
13. A. D. Appleton, Performance of a large superconducting motor, Proc. 3rd Int.
Cryogenic Eng. Conf. , Berlin, 1970, pp.
443–446.
14. J. He, F. Lin, A high frequency high power IGBT inverter drive for 45
HP/16,000 rpm brushless homopolar inductor
motor, IEEE Ind. Appl. Conf. , Orlando, 1995, pp. 9–15.
15. T. Kenyo, Stepping Motors and Their Microprocessor Controls, 2nd ed., Oxford:
Clarendon Press, 1994.
16. M. Jufer, G. Heine, Hybrid stepping motors-25 years of development,
Proc. 25th Annu. Symp. Incremental Motion
Control Syst. Soc., San Jose, 1996, pp. 307–316.
17. H. Weh, Transverse flux machines in drive and generator applications, Stockholm
Power Tech, Stockholm, June 1995.
18. IEC 60050-411:1996 International Electrotechnical Vocabulary, Chapter 411
Rotating Machines.
MANFRED H. STIEBLER
Technical University of Berlin
KEZHONG GUO
Shanghai Jiaotong University
----------------------- Page 233-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
HVDC Power Converters

Standard Article
Dennis A. Woodford1
1Manitoba HVDC Research Centre, Winnipeg, Manitoba,
Canada
Search
reserved.
DOI: 10.1002/047134608X.W3022
❍ Acronym Finder
Abstract
Why Use DC Transmission?
Configurations
Applications of HVDC Converter Bridges
DC Converter Operation
Control and Protection
Areas for Development in HVDC Transmission

%20Conversion/W3022.htm17.06.2008 16:40:44
----------------------- Page 234-----------------------

c
HVDC POWER CONVERTERS
Electric power transmission was originally developed with direct current.

The availability of transformers
and the development and improvement of induction motors at the beginning of the
twentieth century led to
greater appeal and use of ac transmission. Through research and development in
Sweden at Allmana Svenska
Electriska Aktiebolaget (ASEA ), an improved multielectrode grid-controlled mercury
arc valve for high power
and voltages was developed in 1929. Experimental plants were set up in the 1930s in
Sweden and the USA to
investigate the use of mercury arc valves in conversion processes for transmission
and frequency changing.
Then dc transmission became practical when long distances were to
be covered or where cables were
required. The increase in the need for electricity after the Second World War
stimulated research, particularly
in Sweden and in Russia. In 1950, a 116 km experimental transmission line at 200 kV
was commissioned from
Moscow to Kasira. The first commercial HVDC line built in 1954 was a 98 km
submarine cable with earth
return between the island of Gotland and the Swedish mainland.
Thyristors were applied to dc transmission in the late 1960s, and solid-state
valves became a reality. In
1969, a contract for the Eel River dc link in Canada with awarded as the first
application of solid-state valves
for HVDC transmission. Today, the highest functional dc voltage for dc transmission
is ±600 kV for the 785 km
transmission line of the Itaipu scheme in Brazil. Dc transmission is now an
integral part of electricity delivery
in many countries throughout the world.
Why Use DC Transmission?
The question is often asked, “Why use dc transmission?” One response is that losses
are lower, but this is not
strictly correct. The level of losses is designed into a transmission system and is
regulated by the size of the
conductor selected. Dc and ac conductors, as overhead transmission lines or
submarine cables, can have lower
losses but at higher expense because the larger cross-sectional area generally
results in lower losses but costs
more.
When converters are used for dc transmission in preference to ac
transmission, it is generally by economic
choice driven by one of the following reasons:
(1) An overhead dc transmission line with its towers can be designed to cost less
per unit of length than an
equivalent ac line designed to transmit the same level of electric power.
However the dc converter stations
at each end are more costly than the terminating stations of an ac
line, and so there is a breakeven
distance above which the total cost of dc transmission is less than
its ac transmission alternative. The
dc transmission line can have a lower visual profile than an
equivalent ac line and so contributes to a
lower environmental impact. Other environmental advantages to a dc transmission
line are dc instead of
ac electric and magnetic fields.
(2) If transmission is by submarine or underground cable, the breakeven distance
is much less than overhead
transmission. It is not practical to consider ac cable systems exceeding 50 km,
but dc cable transmission
----------------------- Page 235-----------------------
2 HVDC POWER CONVERTERS

Fig. 1. Standard graphical symbols for valves and
bridges.
systems in service are as long as hundreds of kilometers. Distances

of 600 km or greater have been
considered feasible.
(3) Some ac electric power systems are not synchronized with neighboring networks
even though the physical
distances between them are quite small. This occurs in Japan where
part of the country has a 60 Hz
network and the other part has a 50 Hz system. It is physically impossible to
connect the two together by
direct ac methods to exchange electric power between them. However, if a dc
converter station is located in
each system with an interconnecting dc link, it is possible to transfer the
required power even though the
ac systems so connected remain asynchronous.
Configurations
The integral part of an HVDC power converter is the valve or valve arm. It is
noncontrollable if constructed
from one or more power diodes in series and controllable if constructed from one or
more thyristors in series.
Figure 1 depicts the International Electrotechnical Commission (IEC) graphical
symbols for valves and bridges
(1). The standard bridge or converter connection is defined as a two-way connection
comprising six valves or
valve arms which are connected as illustrated in Fig. 2. Electric power flowing
between the HVDC valve group
and the ac system is three-phase. When electric power flows into dc valve group from
the ac system, then it is
considered a rectifier. If power flows from the dc valve group into the ac system, it
is an inverter. Each valve
consists of many series-connected thyristors in thyristor modules. Figure 2 depicts
the electric circuit network
for the six-pulse valve group. The six-pulse valve group was usual when valves were
mercury arc.
Twelve-Pulse Valve Group. Nearly all HVDC power converters with
thyristor valves are assembled
in a converter bridge of twelve-pulse configuration. Figure 3 illustrates the use of
two three-phase converter
transformers with one dc side winding, one as an ungrounded star connection and the
other a delta configura-
tion. Consequently the ac voltages applied to each six-pulse valve group of the
twelve-pulse valve group have
a phase difference of 30◦ which is utilized to cancel the ac side fifth and seventh
harmonic currents and, the
dc side sixth harmonic voltage, thus resulting in a significant saving in harmonic

filters. Figure 3 also shows
----------------------- Page 236-----------------------
HVDC POWER CONVERTERS 3
Fig. 2. Electric circuit configuration of the basic six-pulse valve group with its
converter transformer in star–star connec-
tion.
the outline around each of the three groups of four valves in a single vertical
stack. Known as “quadrivalves,”
they are assembled as one valve structure by stacking four valves in
series. Because the voltage rating of
thyristors is several kilovolts, a 500 kV quadrivalve may have hundreds of
individual thyristors connected in
series groups of valve or thyristor modules. A quadrivalve for a high-voltage
converter is mechanically quite
tall and may be suspended from the ceiling of the valve hall, especially in
locations susceptible to earthquakes.
Thyristor Module. A thyristor or valve module is part of a valve in

a mechanical assembly of series-
connected thyristors and their immediate auxiliaries, including heat sinks
cooled by air, water or glycol,
damping circuits, and valve firing electronics. A thyristor module is usually
interchangeable for maintenance
and consists of electric components, as shown in Fig. 4.
Substation Configuration. The central equipment of a dc
substation (2) consists of the thyristor
converters which are usually housed inside a valve hall. Outdoor valves have been
used as in the Cahora Bassa
dc transmission line between Mozambique and South Africa. Figure 5 is a schematic
of the electrical equipment
required for a dc substation. In this example, two poles are represented, which is
usual and it is known as the
“bipole” configuration. Some dc cable systems have only one pole or “monopole”
configuration and may either
use the ground as a return path, when permitted, or use an additional cable to
avoid Earth currents.
From Fig. 5, the essential equipment in a dc substation, in
addition to the valve groups, includes the
converter transformers. They transform the ac system voltage to which the dc system
is connected so that the
correct dc voltage is derived by the converter bridges. For higher rated dc
substations, converter transformers
for 12-pulse operation are usually comprised of single-phase units which is cost
effective in providing spare
units for increased reliability.
The secondary or dc side windings of the converter transformers are connected
to the converter bridges.
The converter transformer is located in the switchyard. If the converter bridges
are located in the valve hall, the
connection has to be made through its wall in either of two ways, first, with phase-
isolated busbars whose bus
conductors are housed within insulated bus ducts containing oil or SF6 as the
insulating medium or, secondly,
----------------------- Page 237-----------------------
Fig. 3. The twelve-pulse valve group configuration with two converter transformers,
one in star–star connection and the
other in star–delta connection.
with wall bushings. When applied at dc voltages of 400 kV or greater, wall bushings
require considerable design
and care to avoid external or internal insulation breakdown.
Harmonic filters are required on the ac side and usually on the dc side. The
characteristic ac side current
harmonics generated by six-pulse converters are 6n ± 1 and 12n ± 1 for 12-pulse
converters where n equals
all positive integers. Ac filters are typically tuned to the 11th, 13th, 23rd, and
25th harmonics for 12-pulse
converters. Tuning to the fifth and seventh harmonics is required if six-pulse
converters are permitted. Ac side
harmonic filters may be switched with circuit breakers or circuit
switchers to accommodate reactive power
requirement strategies because these filters generate reactive power at
fundamental frequency. A parallel
resonance is naturally created between the capacitance of the ac filters and the
inductive impedance of the
ac system. For the special case where such a resonance is lightly damped and tuned
to a frequency between
----------------------- Page 238-----------------------
Fig. 4. Components of the thyristor modules which make up a

valve or quadrivalve.
the second and fourth harmonics, then a low-order harmonic filter at the
second or third harmonic may be
required, even for 12-pulse converter operation.
Characteristic dc side voltage harmonics generated by a six-pulse converter
are of the order of 6n and
when generated by a 12-pulse converter, are of the order of 12n. Dc side filters
reduce harmonic current flow
on dc transmission lines to minimize coupling and interference with adjacent voice
frequency communication
circuits. Where there is no dc line, such as in the back-to-back configuration, dc
side filters may not be required.
Dc reactors are usually included in each pole of a converter station. They
assist the dc filters in filtering
harmonic currents and smooth the dc side current so that a discontinuous
current mode is not reached at
low load current operation. Because the rate of change of dc side
current is limited by the dc reactor, the
commutative process of the dc converter is more robust.
Surge arresters across each valve in the converter bridge, across each
converter bridge, and in the dc and
ac switchyards are coordinated to protect the equipment from all overvoltages
regardless of their source. They
may be used in nonstandard applications, such as filter protection. Modern HVDC
substations use metal oxide
arresters whose rating and selection are made with careful insulation coordination
design.
Applications of HVDC Converter Bridges
The first application for HVDC converters was to provide point-to-point

electrical power interconnections
between asynchronous ac power networks. The other applications which can
be met by HVDC converter
transmission include the following:
(1) Interconnections between asynchronous systems. Some continental electric power

systems consist of asyn-
chronous networks, such as the East, West, Texas, and Quebec networks
in North America, and island
loads, such as the Island of Gotland in the Baltic Sea, make good use of HVDC
interconnections.
(2) Delivering energy from remote energy sources. Where generation has
been developed at remote sites of
available energy, HVDC transmission has been economical for bringing the
electricity to load centers. Gas-
----------------------- Page 239-----------------------
Fig. 5. Example of an HVDC substation.
fired thermal generation can be located close to load centers and may delay
development of isolated energy
sources in the near term.
(3) Importing electric energy into congested load areas. In areas where new
generation is impossible to put
in service to meet load growth or replace inefficient or
decommissioned plants, underground dc cable
transmission is a viable means of importing electricity.
(4) Increasing the capacity of existing ac transmission by conversion to dc
transmission. New transmission
rights-of-way may be impossible to obtain. Existing overhead ac transmission
lines upgraded to or overbuilt
with dc transmission can substantially increase the power transfer capability
on the existing right-of-way.
(5) Power flow control. AC networks do not easily accommodate desired power-flow
control. Power marketers
and system operators may require the power-flow control capability provided by
HVDC transmission.
(6) Stabilization of electric power networks. Some widespread ac power system
networks operate at stability
limits well below the thermal capacity of their transmission conductors. HVDC
transmission is an option
to increase utilization of network conductors along with the various electronic
power controllers which can
be applied to ac transmission.
----------------------- Page 240-----------------------
HVDC
POWER CONVERTERS 7
Fig. 6. Monopolar and bipolar connection of HVDC converter

bridges.
HVDC Converter Arrangements. HVDC converter bridges and lines or

cables can be arranged in a
number of effective configurations. Converter bridges may be arranged either as
monopolar or bipolar, as shown
in the 12-pulse arrangement in Fig. 6. Various ways of using HVDC transmission are
shown in simplified form
in Fig. 7 and include the following:
(1) Back-to-back. There are some applications where the two ac systems to be
interconnected are in the same
physical location or substation. No transmission line or cable is required
between the converter bridges
in this case and the connection may be monopolar or bipolar. Back-to-back dc
links are used in Japan for
interconnections between power system networks of different frequencies (50 and
60 Hz). They are also
used as interconnections between adjacent asynchronous networks.
(2) Transmission between two substations. When it is economical to transfer
electric power through dc trans-
mission or cables from one geographical location to another, two-terminal or
point-to-point HVDC trans-
mission is used. In other words, dc power from a dc rectifier terminal is
dedicated to one other terminal
operating as an inverter. This is typical of most HVDC transmission systems
----------------------- Page 241-----------------------
(3) Multiterminal HVDC transmission system. When three or more HVDC

substations are geographically
separated and have interconnecting transmission lines or cables, the HVDC
transmission system is multi-
terminal. If all substations are connected to the same voltage, then the system
is parallel multiterminal dc.
If one or more converter bridges are added in series in one or both poles, then
the system is series multi-
terminal dc. Parallel multiterminal dc transmission has been applied when the
capacity of any substation
exceeds 10% of the total rectifier substation capacity. It is expected that a
series multiterminal substation
would be applied when its capacity is small (less than 10%) compared to the
total rectifier substation capac-
ity. A combination of parallel and series connections of converter bridges is a
hybrid multiterminal system.
Multiterminal dc systems are more difficult to justify economically
because of the cost of the additional
substations.
(4) Unit connection. When dc transmission is applied right at the point of
generation, it is possible to connect
the converter transformer of the rectifier directly to the generator terminals
so that the generated power
feeds into the dc transmission lines. This might be applied with hydro and wind
turbine generators so that
the turbine’s maximum efficiency is achieved with speed control. Regardless of
the turbine speed, the power
is delivered through the inverter terminal to the ac receiving system at its
fundamental frequency of 50 or
60 Hz.
(5) Diode rectifier. It has been proposed that in some applications where dc power
transmission is only in one
direction, the valves in the rectifier converter bridges can be constructed from
diodes instead of thyristors.
Power-flow control would be achieved at the inverter, and where the unit
connection is used, ac voltage
control by the generator field exciter could be applied to regulate dc power.
This connection may require
high-speed ac circuit breakers between the generator and the rectifier
converter bridges to protect the
diodes from overcurrents resulting from a sustained dc transmission line short
circuit.
Economic Considerations. A study for Oak Ridge National

Laboratory (3) surveyed three suppliers
of HVDC equipment for quotations of turnkey costs to supply two bipolar substations
for four representative
systems. Each substation requires one dc electrode and interfaces to an ac system
with a short circuit capacity
four times the rating of the HVDC system. The four representative
systems are summarized in Table 1.
Table 2 provides a major component breakdown based on avarage values derived from
the responses of the
suppliers. The turnkey costs are in 1995/96 US dollars and are only for
one terminal, assuming that both
terminals are provided by the same supplier. The back-to-back dc link cost is for
the complete installation.
Transmission line costs cannot be so readily defined. Variations depend on the
cost of land use, the width
of the right-of-way required, labor rates for construction, and the
difficulty of the terrain to be crossed. A
simple guideline is that the cost of dc transmission line is 80% to 100% of the
cost of an ac line whose rated line
voltage is the same as the rated pole-to-ground voltage of the dc line. The cost
advantage of dc transmission
for traversing long distances is that it may be rated at twice the power-flow
capacity of an ac line of the same
voltage.
----------------------- Page 242-----------------------
HVDC
POWER CONVERTERS 9
Fig. 7. HVDC converter bridge arrangements.
When electricity must be transmitted by underground or undersea cables, ac

cables become impractical
because of their capacitive charging current if they are longer than a critical
length which is less than 50 km
for undersea applications. For distances longer than this critical length, dc
cables are required. The choice is
system-specific, and economic considerations prevail.
The electrical environmental effects from HVDC transmission lines can be

characterized by field, ion, and
corona effects (4,5). The electric field arises from both the electrical charge on
the conductors and from charges
on air ions and aerosols surrounding the conductor in a HVDC overhead transmission
line. These give rise to
dc electric fields due to the ion current density flowing through the air from or to
the conductors and due to the
----------------------- Page 243-----------------------
ion density in the air. A static magnetic field is produced by dc current flowing
through the conductors. Air ions
produced by HVDC lines form clouds which drift away from the line when blown by the
wind and may come in
contact with humans, animals, and plants outside the transmission line’s right-of-
way or corridor. The corona
effects may produce low levels of radio interference, audible noise, and ozone.
Field and Corona Effects. The field and corona effects of
transmission lines largely favor dc over ac
transmission. The significant considerations are as follows:
(1) For a given power transfer requiring extra high voltage transmission, the dc
transmission line has a tower
profile smaller than the equivalent ac tower carrying the same level
of power. This can also lead to a
narrower right-of-way for dc transmission.
(2) The steady and direct magnetic field of a dc transmission line near or at the
edge of the transmission right-
of-way has about the magnitude as the earth’s naturally occurring magnetic
field. For this reason alone, it
seems unlikely that this small contribution by HVDC transmission lines to the
background geomagnetic
field is a basis for concern.
(3) The static and steady electric field from dc transmission at the
levels experienced beneath lines or at
the edge of the right-of-way have no known adverse biological effects.
There is no theory or mechanism
to explain how a static electric field at the levels produced by dc
transmission lines could affect human
health. The electric field level beneath a HVDC transmission line has a
magnitude similar to the naturally
occurring static field which exists beneath thunder clouds. Electric fields from
ac transmission lines have
undergone more intense scrutiny than fields generated by dc transmission lines.
(4) The ion and corona effects of dc transmission lines contribute a
low level of ozone production to higher
naturally occurring background concentrations. Exacting long-term measurements
are required to detect
such concentrations. The measurements taken at cross-sections across the Nelson
River dc lines in Canada
failed to distinguish background from downwind levels (4). Although solar
radiation influences the produc-
tion of ozone even in a rural environment, thereby maintaining its level, any
incremental contribution from
a dc line is subject to breakdown, leading to a resumption of background levels
downwind from the line.
Investigations of ozone in indoor conditions indicate that in well-mixed air,
the half-life of ozone is 1.5 min
to 7.9 min. Increases in temperature and humidity increase the rate of decay
(4).
----------------------- Page 244-----------------------

(5) If ground return is used with monopolar operation, the resulting dc

magnetic field can cause error in
magnetic compass readings in the vicinity of the dc line or cable. This impact
is minimized by providing a
conductor or cable return path (known as a metallic return) close to the main
conductor or cable to cancel
the magnetic field. Another concern with continuous ground current is that some
of the return current may
flow in metallic structures, such as pipelines, and intensify corrosion if
cathodic protection is not provided.
When pipelines or other continuous metallic grounded structures are in the
vicinity of a dc link, a metallic
return may be necessary.
DC Converter Operation
The six-pulse converter bridge of Fig. 2, the basic converter unit of HVDC
transmission, is used equally well
for rectification when electric power flows from the ac to the dc side and for
inversion when the power flow is
from the dc to the ac side. Thyristor valves operate as switches which turn on and
conduct current, when fired
on receiving a gate pulse, and are forward-biased. A thyristor valve conducts
current in one direction and, once
it conducts, turns off only when it is reverse-biased and the current falls to
zero. This process is known as line
commutation and is discussed in more detail below.
An important property of the thyristor valve is that once its conducting
current falls to zero, when it is
reverse-biased and the gate pulse is removed, too rapid an increase in the
magnitude of the forward-biased
voltage causes the thyristor to turn on inadvertently and conduct. The
design of the thyristor valve and
converter bridge must ensure that such a condition is avoided in useful inverter
operation.
Commutation. Rectification or inversion for HVDC converters
is accomplished through a process
known as line or natural commutation. The valves act as switches so
that the ac voltage is sequentially
switched to provide always a dc voltage. With line commutation the ac voltage at
both the rectifier and inverter
must be provided by the ac networks at each end and should be three-phase and
relatively free of harmonics, as
depicted in Fig. 8. As each valve switches on, it begins to conduct current while
the current begins to fall to zero
in the next valve to turn off. Commutation is the process of transferring current
between any two converter
valves, and both valves carry current simultaneously during the process.
Consider the rectification process. Each valve switches on when it receives a
firing pulse to its gate and
its forward-bias voltage becomes more positive than the forward-bias
voltage of the conducting valve. The
current flow through a conducting valve does not change instantaneously as it
commutates to another valve
because the transfer is through transformer windings. The leakage reactance of the
transformer windings is
also the commutative reactance so long as the ac filters are located on the primary
or ac side of the converter
transformer. The commutative reactance at the rectifier and inverter is shown as an
equivalent reactance X C
in Fig. 8. The sum of all of the valve currents transferred to the dc side and
through the dc reactor is the direct
current, and it is relatively flat because of the inductance of the dc reactor and
converter transformer.
At the inverter, the three-phase ac voltage supplied by the ac system
provides the forward- and reverse-
bias conditions of each valve in the converter bridge to allow commutation of
current between valves, as in
the rectifier. The inverter valve can only turn on and conduct when the positive
direct voltage from the dc line
is greater than the back negative voltage derived from the ac
commutative voltage of the ac system at the
inverter.
Because of the line-commutative, valve-switching process, a nonsinusoidal
current is taken from the ac
system at the rectifier (Ivr in Fig. 8) and is delivered to the ac system at the
inverter (Ivi in Fig. 8). Both Ivr and
Ivi lag the alternating voltage. This nonsinusoidal current waveform consists
of the fundamental frequency
ac component plus higher harmonics taken from and injected into each ac
system. The ac filters divert the
harmonics from entering the ac system by offering a low impedance bypass path that
allows the commutative
voltage to be relatively harmonic-free ( ULr and ULi in Fig. 8).
----------------------- Page 245-----------------------
Fig. 8. Voltage and current waveshapes associated with dc

converter bridges.
Reversal of power flow in a line-commutated dc link is not possible by

reversing the direction of the direct
current. The valves allow conduction only in one direction. Power flow can be
reversed only in line-commutated
dc converter bridges by changing the polarity of the direct voltage. The dual
operation of the converter bridge
as either a rectifier or inverter is achieved through firing control of the grid
pulses.
Converter Bridge Angles. Figure 8 shows the various electrical
angles which define the operation
of converter bridges. These angles are measured on the three-phase,
valve-side voltages and are based on
steady-state conditions with a harmonic-free and idealized three-phase commutative
voltage. They apply to
both inverters and rectifiers.
Delay Angle α. The time expressed in electrical angular measure from
the zero crossing of the idealized
sinusoidal commutating voltage to the starting instant of forward-current
conduction. This angle is controlled
◦
◦
by the gate firing pulse and if less than 90 , the converter bridge is a rectifier,
and if greater than 90 , it is an
inverter. This angle is often referred to as the firing angle.
Advance Angle β. The time expressed in electrical angular measure
from the starting instant of forward-
current conduction to the next zero crossing of the idealized sinusoidal
commutating voltage. The angle of
advance β is related in degrees to the angle of delay α by
Overlap Angle µ. The duration of commutation between two converter

valve arms expressed in electrical
angular measure.
Extinction Angle γ. The time expressed in electrical angular measure
from the end of current conduction
to the next zero crossing of the idealized sinusoidal commutating voltage. γ
depends on the angle of advance β
and the angle of overlap µ and is determined by the relationship
----------------------- Page 246-----------------------
Steady-State dc Converter Bridge Equations. It is useful

to express the commutative reactance of
a six-pulse converter bridge per unit of the converter transformer rating SN
as follows:
where I dN is the rated direct current and UVN is the rated phase-to-phase
voltage on the valve or secondary side
of the converter transformer. Usually the dc converter bridge power rating is known
from its rated dc current
I dN and rated dc voltage UdN . The valve and converter bridge design are very
dependent on the commutative
reactance X C , and so consequently its value is established and known. In modern
HVDC converter bridges it
2
is usually in the range 0.1 <X C < 0.15 per unit where 1.0 per unit is (UVN ) /SN
#.
A reasonably good approximation for the power factor of a converter bridge at
the ac commutating bus is
given by the following expression for a rectifier. Note that the delay angle α is
usually known or determined.
◦ ◦
For example, the normal steady-state range of the delay angle for a
rectifier may be 10 < α < 18 , and the
◦
lowest normal operating power factor occurs when α = 18 :
and for an inverter,
where I d is the dc load current, I dN is the rated dc current, and θ is the

power factor angle. For the inverter,
the normal rated extinction angle is established in the converter bridge design,
usually at γ = 18◦ . Ignoring the
losses in the converter bridge, the power flowing through the bridge Pd is given by
where I d is the operating direct current through the converter bridge and Ud
is the operating direct voltage
across the converter bridge. Having calculated the power factor angle θ from Eq.
(4) or (5) and the throughput
power of the converter bridge from Eq. (6), the reactive power QL
demanded by the converter bridge at the ac
commutating voltage busbar at either the rectifier or inverter is given by
It may be that the rated phase-to-phase voltage UVN on the valve or secondary side
of the converter transformer
is not known. It is possible to compute what it should be if the power factor cos θ
from Eq. (4) or (5) is known
at the converter bridge rating. Then a good estimate of UVN is given
by
Once UVN is known, it is possible to find the converter transformer rating from
Eq. (3).
It may be necessary to determine the overlap angle µ. At the rectifier, the
following approximate expression
can be applied when the delay angle α per unit of commutating reactance X C
and the dc load current I d are
known:
----------------------- Page 247-----------------------
Similarly at the inverter, the extinction angle γ is usually known for steady-state
operation so that
The delay angle α at the inverter may not be inherently known, but once the
extinction angle γ and the overlap
angle µ have been determined, then
It is also possible to determine the nominal turns ratio of the

converter transformer once the rated
secondary (dc valve side) voltage UVN is known and if the primary side
rated phase-to-phase ac bus voltage
ULN is also known. Based on phase-to-phase voltages, the nominal turns ratio of
the converter transformer
TRN is given by
During the operation of a converter bridge, the converter transformer on-line

tap changer adjusts to keep
the delay angle α at a rectifier in its desired normal operating range. Similarly at
the inverter, the on-line tap
changer adjusts to maintain the inverter operation at its desired level of dc
voltage Ud or extinction angle γ.
Knowing the desired levels of dc voltage (Ud ), the dc current I d , the nominal
turns ratio TRN of the converter
transformer, the operating level of the primary side ac voltage UL , and the
extinction angle γ (if an inverter)
or delay angle α (if a rectifier), the per unit turns ratio TR of the
converter transformer is found from the
expression
where X C is the commutating reactance for the converter bridge per unit, φ = α
for a rectifier, and φ = γ if an
inverter. I dN is the rated dc current for the converter bridge and UdN is its
rated dc voltage.
Equations 1 to 13 are the steady-state and reasonably accurate expressions
that define the state of a six-
pulse converter bridge under ideal conditions. More exacting expressions can be
found in Refs. (6,7,8,9,10,11)
and can be used if the network data are known with sufficient accuracy to justify
precise mathematical formu-
lation. Defining the performance and operation of a converter bridge under dynamic
or transient conditions
requires a suitable electromagnetic transients simulation program with
capability of modeling the valves,
converter transformer, the control system which produces the firing pulses to the
valves, and the associated ac
and dc networks.
Short Circuit Ratio. The strength of an ac network at the bus of a
HVDC substation can be expressed
by the short circuit ratio (SCR), defined as the relationship between the short
circuit level in MVA at the HVDC
substation bus at 1.0 per unit ac voltage and dc power in MW.
The capacitors and ac filters connected to the ac bus reduce the short circuit
level. The effective short
circuit ratio (ESCR ) is the ratio between the short circuit level
reduced by the reactive power of the shunt
capacitor banks and ac filters connected to the ac bus at 1.0 per unit voltage and
the rated dc power.
A lower ESCR or SCR means more pronounced interaction between the
HVDC substation and the ac
network (9,10). Ac networks can be classified in the following categories according
to strength:
----------------------- Page 248-----------------------
strong systems with a high ESCR: ESCR > 3.0

systems with a low ESCR: 3.0 > ESCR > 2.0
weak systems with a very low ESCR: ESCR < 2.0
In high ESCR systems, changes in the active/reactive power from the HVDC
substation lead to small or
moderate ac voltage changes. Therefore the additional transient voltage control at
the busbar is not normally
required. The reactive power balance between the ac network and the HVDC substation
is achieved by switched
reactive-power elements.
In low and very low ESCR systems, the changes in the ac network or in the
HVDC transmission power
could lead to voltage oscillations and a need for special control strategies.
Dynamic reactive-power control at
the ac bus at or near the HVDC substation by some form of electronic reactive-power
controller, such as a static
var, compensator (SVC) or static synchronous compensator (STATCOM), may be
necessary (12). In earlier
times, dynamic reactive-power control was achieved with synchronous compensators.
Commutation Failure. When a converter bridge operates as an
inverter, as represented at the receiving
end of the dc link in Fig. 8, a valve turns off when its forward current commutates
to zero and the voltage
across the valve remains negative. The period for which the valve stays
negatively biased is the extinction
angle γ, the duration beyond which the valve then becomes forward-biased. Without a
firing pulse, the valve
ideally stays nonconductive or blocked, even though it experiences a forward bias.
All dc valves require removing the internally stored charges
produced during the forward-conducting
period (defined by period α + µ at the inverter in Fig. 8) before the
valve can successfully block a forward
bias. Therefore the dc inverter requires a minimum period of negative
bias or minimum extinction angle γ
for successful forward blocking. If forward blocking fails and conduction
is initiated without a firing pulse,
commutation failure occurs. This also results in an immediate failure to maintain
current in the succeeding
converter arm as the dc line current returns to the valve which was previously
conducting and has failed to
sustain forward blocking (13).
Commutation failure at a converter bridge operating as an inverter
is caused by any of the following
conditions:
(1) When the dc current entering the inverter experiences an increase in magnitude
which causes the overlap
angle µ to increase, the extinction angle γ is reduced and may
reach the point where the valve cannot
maintain forward blocking. Increasing the inductance of the dc current path
through the converter by the
dc smoothing reactor and commutating reactance reduces the rate of change of dc
current. This has the
greatest effect on commutation failure onset.
(2) When the magnitude of the ac side voltage in one or more phases
reduces or is distorted causing an
inadequate extinction angle as commutation is attempted.
(3) A phase angle shift in the ac commutating voltage can cause commutation
failure. However, the magnitude
of ac voltage reduction and not the corresponding phase shift is the most
dominant factor determining the
onset of commutation failures for single-phase faults.
(4) The value of the predisturbance steady-state extinction angle γ also affects
the sensitivity of the inverter
◦
◦ ◦
to commutation failure. A value of γ = 18 is usual for most
inverters. Increasing γ to values of 25 , 30 ,
or higher reduces the possibility of commutation failure (at the expense of
increasing the reactive-power
demand of the inverter).
(5) The value of valve current before commutation failure also affects the
conditions at which a commutation
failure may occur. A commutation failure may more readily happen if the
predisturbance current is at full
load compared to light load current operation.
----------------------- Page 249-----------------------
Fig. 9. Configurations for applying series capacitors at

HVDC substations.
In general, the more rigid the ac voltage into which the inverter feeds and
with an absence of ac system
disturbances, the less the likelihood of commutation failures.
Series Capacitors with dc Converter Substations. HVDC
transmission systems with long dc cables
are prone to commutation failure when there is a drop in dc voltage Ud at the
inverter. The dc cable has a very
large capacitance which discharges current toward the voltage drop at the inverter.
The discharge current is
limited by the dc voltage derived from the ac voltage of the commutating bus, from
the dc smoothing reactor,
and from the commutating reactance. If the discharge current of the cable increases
too quickly, commutation
failure occurs causing a complete discharge of the cable. Recharging the cable to
its normal operating voltage
delays recovery.
The converter bridge firing controls can be designed to increase the delay
angle α when an increase in
dc current is detected. This may be effective until the limit of the minimum
allowable extinction angle γ is
reached.
Another way to limit the cable discharge current is to operate the inverter
bridge with a three-phase series
capacitor in the ac system on either side of the converter transformer. Any
discharge current from the dc cable
passes into the ac system through the normally functioning converter bridge and in
doing so, passes through
the series capacitor and adds charge to it. As a consequence, the voltage of the
series capacitor increases to
oppose the cable discharge and is reflected through the converter bridge as an
increase in dc voltage Ud . This
acts as a back emf and limits the discharge current of the cable, thereby avoiding
commutation failure.
The proposed locations of the series capacitor are shown in Fig. 9
in single-line diagram form (14,15).
When the capacitor is between the converter transformer and the valve
group, it is known as a capacitor
commutated converter (CCC). When the capacitor is on the ac side of the converter
transformer, it is known as
----------------------- Page 250-----------------------
Fig. 10. Steady-state U –I characteristics for a two-

terminal HVDC system.
d d
a controlled series capacitor converter (CSCC). Each configuration improves the

commutative performance of
the inverter, but the CSCC requires design features to eliminate ferroresonance
between the series capacitor
and the converter transformer if it occurs.
Control and Protection
HVDC transmission systems must transport very large amounts of electric power which
is accomplished only
under tightly controlled conditions. Dc current and voltage are precisely
controlled to effect the desired power
transfer. Therefore it is necessary to measure system quantities continuously and
precisely. These quantities
at each converter bridge include the dc current, its dc side voltage, the delay
angle α, and for an inverter, its
extinction angle γ.
Two-terminal dc transmission systems are more usual, and they have in common
a preferred mode of
control during normal operation. Under steady-state conditions, the inverter is
assigned the task of controlling
the dc voltage. It may do this by maintaining a constant extinction angle γ which
causes the dc voltage Ud
to droop with increasing dc current I d , as shown in the minimum constant
extinction angle γ characteristic
A-B-C-D in Fig. 10. The weaker the ac system at the inverter, the steeper the
droop.
Alternatively, the inverter may normally operate in a dc voltage-controlling
mode which is the constant
Ud characteristic B-H-E in Fig. 10. This means that the extinction angle γ must
increase beyond its minimum
◦
setting, depicted in Fig. 10 as 18 .
If the inverter operates at a minimum constant γ or constant Ud
characteristic, then the rectifier must
◦
control the dc current I d . It can do this so long as the delay angle α is not at
its minimum limit (usually 5 ). The
steady-state constant current characteristic of the rectifier is shown in Fig. 10 as
the vertical section Q-C-H-R.
The operating point of the HDVC system is where the rectifier and inverter
characteristic intersect, either at
points C or H.
The operating point is reached by the action of the on-line tap changers of
the converter transformers. The
inverter must establish the dc voltage Ud by adjusting its on-line tap changer to
achieve the desired operating
level if it is in constant minimum γ control. If in constant Ud
control, the on-line tap changer must adjust its
----------------------- Page 251-----------------------
tap to achieve the controlled level of Ud with an extinction angle equal to or

slightly larger than its minimum
setting of 18◦ in this case.
The on-line tap changers on the converter transformers of the rectifier are
controlled to adjust their tap
settings so that the delay angle α has a working range between approximately 10◦
and 15◦ to maintain the
constant current setting I order (see Fig. 10). If the inverter is operating in
constant dc voltage control at the
operating point H and if the dc current order I order is increased so that
the operating point H moves toward
and beyond point B, the inverter mode of control reverts to constant extinction
angle γ control and operates on
characteristic A-B. Dc voltage Ud is less than the desired value, and so the
converter transformer on-line tap
changer at the inverter boosts its dc side voltage until dc voltage control is
resumed.
Not all HVDC transmission controls have a constant dc voltage control, such
as depicted by the horizontal
characteristic B-H-E in Fig. 10. Instead, the constant extinction angle γ control
of characteristic A-B-C-D and
the tap changer provide the dc voltage control.
Current Margin. The dc current order I order is sent to both
the rectifier and inverter. It is usual to
subtract a small value of current order from the I order sent to the inverter.
This is known as the current margin
Imargin and is depicted in Fig. 10. The inverter also has a current controller, and
it attempts to control the dc
current I d to the value I order −Imargin , but the current controller at the
rectifier normally overrides it to maintain
the dc current at I order . This discrepancy is resolved at the inverter
in normal steady-state operation as its
current controller cannot keep the dc current at the desired value of I order
−Imargin and is forced out of action.
The current control at the inverter becomes active only when the current control at
the rectifier ceases because
its delay angle α is pegged against its minimum delay angle limit. This is readily
observed in the operating
characteristics of Fig. 10 where the minimum delay angle limit at the rectifier is
characteristic P-Q. If for some
reason or other, such as a low ac commutating voltage at the rectifier end, the P-Q
characteristic falls below
points D or E, the operating point shifts from point H to somewhere on the vertical
characteristic D-E-F where it
is intersected by the lowered P-Q characteristic. The inverter reverts to current
control, controls the dc current
I d at the value I order −Imargin , and the rectifier effectively controls dc
voltage as long as it is operating at its
minimum delay angle characteristic P-Q. The controls can be designed so that the
transition from controlling
the current by the rectifier to controlling the current by the inverter is automatic
and smooth.
Voltage-Dependent Current-Order Limit (VDCOL). During
disturbances when the ac voltage at
the rectifier or inverter is depressed, it is not helpful to a weak ac system if the
HVDC transmission system
attempts to maintain full-load current. A sag in ac voltage at either end also
results in a reduced dc voltage.
The dc control characteristics shown in Fig. 10 indicate that the dc current order
is reduced if the dc voltage is
lowered. This can be observed in the rectifier characteristic R-S-T and in the
inverter characteristic F-G in Fig.
10. The controller which reduces the maximum current order is known as a voltage-
dependent current order
limit or VDCOL (sometimes called a VDCL). If invoked by an ac system disturbance,
the VDCOL control keeps
the dc current I d to the lowered limit during recovery which aids the
corresponding recovery of the dc system.
Only when the dc voltage Ud has recovered sufficiently does the dc current return
to its original I order level.
Figure 11 is a schematic diagram showing how dc transmission system controls
are usually implemented.
Ac Voltage Control. It is desirable to maintain the ac system and

commutating bus voltage rigidly at
a constant value for best operation of an HVDC transmission system. This is more
easily achieved when the
short circuit ratio is high. With low or very low short circuit ratio systems,
difficulties may arise following load
changes. With fast load variation, there can be an excess or deficiency of reactive
power at the ac commutating
bus which results in over- and undervoltages, respectively. When the ac system is
weak, changes in converter
ac bus voltage following a disturbance may be beyond permissible limits. In such
cases, an ac voltage controller
is required for the following reasons:
(1) To limit dynamic and transient overvoltage to within permissible limits defined
by substation equipment
specifications and standards.
----------------------- Page 252-----------------------
Fig. 11. HVDC control system layout.
(2) To prevent ac voltage flicker and commutation failure due to ac voltage

fluctuations when load and filter
switching occurs.
(3) To enhance HVDC transmission system recovery following severe ac system
disturbances.
(4) To avoid control system instability, particularly when operating in the
extinction angle control mode at the
inverter.
The synchronous compensator has been the preferred means of ac voltage

control because it increases
the short circuit ratio and is a variable reactive-power source. Its
disadvantages include high losses and
maintenance which add to its overall cost. Other ac voltage controllers available
include the following:
(1) Static compensators which utilize thyristors to control current through

inductors and switch various levels
of capacitors in or out. By this means, fast control of reactive power is
possible to maintain ac voltage within
desired limits. The main disadvantage is that it does not add to the short
circuit ratio.
(2) Converter control through delay angle control is possible to regulate the
reactive-power demand of converter
bridges. This requires that the measured ac voltage be used as a feedback
signal in the dc controls and
delay angle α is transiently modulated to regulate the ac commutating bus
voltage. This form of control is
limited in its effectiveness, particularly when there is little or no dc
current in the converter when voltage
control is required.
(3) Specially cooled metal oxide varistors together with fast mechanical switching
of shunt reactors, capacitors,
and filters. The metal oxide varistors protect the HVDC substation equipment
against transient overvolt-
ages, and switching of reactive-power components achieves the reactive-power
balance. The disadvantage
of this system is that voltage control is not continuous, reactive-power
control is delayed by slow mechanical
switching, and the short circuit ratio is not increased.
(4) Saturated reactors to limit overvoltages and achieve reactive-power balance.
Shunt capacitors and filters
are required to maintain the reactors in saturation. Ac voltage control is
achieved without controls on a
droop characteristic. Short circuit ratio is not increased.
(5) Series capacitors in the form of CCC or CSCC to increase the short circuit
ratio and improve the regulation
of ac commutating bus voltage.
----------------------- Page 253-----------------------
(6) The static synchronous compensator or STATCOM to use gate turn-off thyristors
in the configuration of the
voltage source converter bridge. This is the fastest responding voltage
controller available and may offer
limited capability for an increased short circuit ratio.
Because each ac system with its HVDC application is unique, the voltage
control method applied is subject
to study and design.
Special Purpose Controls. There are a number of special purpose
controllers which can be added to
HVDC controls to take advantage of the fast response of a dc link and help the
performance of the ac system.
Ac System Damping Controls. An ac system is subject to power
swings due to electromechanical
oscillations. A controller can be added to modulate the dc power order or
dc current order to add damping.
The frequency or voltage phase angle of the ac system is measured at one or both
ends of the dc link, and the
controller is designed to adjust the power of the dc link accordingly.
Ac System Frequency Control. A slow responding controller can also
adjust the power of the dc link to
help regulate power system frequency. If the rectifier and inverter are in
asynchronous power systems, the dc
controller can draw power from one system to the other to assist in frequency
stabilization of each.
Step Change Power Adjustment. A noncontinuous power adjustment
can be implemented to take
advantage of the ability of a HVDC transmission system to reduce or
increase power rapidly. If ac system
protection determines that a generator or ac transmission line is to be tripped, a
signal can be sent to the dc
controls to change its power or current order by an amount which compensates for
the loss. This feature is
useful in helping maintain ac system stability and easing the shock of a
disturbance over a wider area.
Ac Undervoltage Compensation. Some portions of an electric
power system are prone to ac voltage
collapse. If a HVDC transmission system is in such an area, a control can be
implemented which on detecting
the ac voltage drop and the rate at which it is dropping, can effect a fast power
or current order reduction of
the dc link. The reduction in power and reactive power can remove the undervoltage
stress on the ac system
and restore its voltage to normal.
Subsynchronous Oscillation Damping. A steam turbine and
electric generator can have mechanical
subsynchronous oscillation modes between the various turbine stages and the
generator. If such a generator
feeds into the rectifier of a dc link, supplementary control may be required on the
dc link to ensure that the
subsynchronous oscillation modes of concern are positively damped to limit
torsional stresses on the turbine
shaft.
Areas for Development in HVDC Transmission
The thyristor as the key component of a converter bridge continues to be

developed so that its voltage and
current rating are increasing. Gate-turn-off thyristors (GTOs) and insulated gate
bipole transistors (IGBTs)
are required for the voltage source converter (VSC) converter bridge configuration.
The VSC converter bridge
is being applied in new developments (12). Its special properties include the
ability to control real and reactive
power independently at the connection bus to the ac system. Reactive power can be
either capacitive or inductive
and can be controlled to change quickly from one to the other.
A voltage source converter as an inverter does not require an
active ac voltage source into which to
commutate as does the conventional line-commutated converter. The VSC inverter can
generate an ac three-
phase voltage and supply electricity to a load as the only source of
power. It requires harmonic filtering,
harmonic cancellation, or pulse-width modulation to provide an acceptable ac
voltage waveshape.
Two applications are now available for the voltage source converter. The first
is for low-voltage dc con-
verters used in dc distribution systems. The first application of a dc distribution
system in 1997 was developed
in Sweden (16) and is known as “HVDC Light.” Other applications for a dc
distribution system may be (1) in a
dc feeder to remote or isolated loads, particularly if underwater or underground
cable is necessary and (2) for
----------------------- Page 254-----------------------
a collector system of a wind farm where cable delivery and optimum and individual
speed control of the wind
turbines is desired for peak turbine efficiency.
The second immediate application for VSC converter bridges is in a back-to-
back configuration. The back-
to-back VSC link is the ultimate transmission and power flow controller. It can
control and reverse power flow
easily and control reactive power independently on each side. With a suitable
control system, it can control
power to enhance and preserve ac system synchronism and act as a rapid phase-angle
power-flow regulator
with a 360◦ range of control.
There is considerable flexibility in the configuration of VSC converter

bridges. Many two-level converter
bridges can be assembled with appropriate harmonic canceling properties to generate
acceptable ac system
voltage waveshapes. Another option is to use multilevel converter bridges to
provide harmonic cancellation.
Additionally, both two-level and multilevel converter bridges can utilize pulse-
width modulation to eliminate
low-order harmonics. With pulse-width modulation, high-pass filters may still be
required because PWM adds
to the higher order harmonics.
As VSC converter bridge technology develops for higher dc voltage
applications, it will be possible to
eliminate converter transformers. This is possible with the low-voltage
applications in use today. It is expected
that the exciting developments in power electronics will continue to provide
exciting new configurations and
applications for HVDC converters.
BIBLIOGRAPHY
1. IEC Committee SC22F, Terminology for high-voltage direct current

transmission, IEC reference number 22F/37/CDV.
2. Physical Layout of Recent HVDC Transmission Projects in North America , IEEE
Special Publication 87TH0177-6-PWR,
September 1986.
3. R. L. Hauth et al., HVDC Power Transmission Technology Assessment Report
ORNL/Sub/95-SR893/1, Oak Ridge, TN:
Oak Ridge National Laboratory, April 1997.
4. O. C. Norris-Elye, N. Tarko, C. V. Thio, Nelson River HVDC Transmission Lines
External Electrical Effects , Final Report
No. E4.55.4, Manitoba, Canada: Manitoba HVDC Research Centre, June 1995.
5. W. H. Bailey, D. E. Weil, J. R. Stewart, HVDC Power Transmission Environmental
Issues Review , Report ORNL/Sub/95-
SR893/2, Oak Ridge, TN: Oak Ridge National Laboratory, April 1997.
6. E. W. Kimbark, Direct Current Transmission , New York: Wiley, 1971, vol. 1.
7. E. Uhlman, Power Transmission by Direct Current , New York: Springer-Verlag,
1975.
8. J. Arrillaga, High Voltage Direct Current Transmission , London: Peregrinus,
1983.
9. K. R. Padiyar, HVDC Transmission-Technology and System Interactions , New
York: Wiley, 1990.
10. Guide for Planning DC Links Terminating at AC Locations Having Low
Short Circuit Capacities, Part 1: AC/DC
Interaction Phenomena , CIGRE Tech. Brochure No. 68, 1992.
11. High-Voltage Direct Current Handbook , 1st ed., Palo Alto: Electric Power Res.
Inst., 1994.
12. FACTS Overview , IEEE and CIGRE joint publication 95 TP 108, April 1995.
13. CIGRE Working Group 14-05, Commutation failures-causes and consequences,
ELECTRA , 165: 1996.
14. J. Reeve, J. A. Baron, G. A. Hanley, A technical assessment of artificial
commutation of HVDC converters, IEEE Trans.
Power Appar. Syst. PAS-87: 1830–1840, 1968.
15. D. A. Woodford, Solving the ferroresonance problem when compensating a dc
converter station with a series capacitor,
IEEE Trans. Power Syst. , 3: 1325–1331, 1996.
16. G. Asplund, K. Eriksson, K. Svensson, Dc transmission based on voltage source
converter, CIGRE SC14 Colloquium,
South Africa, 1997.
DENNIS A. WOODFORD
Manitoba HVDC Research Centre
----------------------- Page 255-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Hydraulic Turbines
Standard Article
Dan Levy1
1University of Limerick, Limerick, Ireland

reserved.
Search
DOI: 10.1002/047134608X.W3023
❍ Acronym Finder
Abstract
Hydro-Plants
Turbine and Generator Coupling
Electrical System
Speed Against Size
Stand-Alone Generators ()
Protection
Cost
Conclusion
List of Symbols

%20Conversion/W3023.htm17.06.2008 16:41:02
----------------------- Page 256-----------------------

c
HYDRAULIC TURBINES
Hydro-Plants
Large hydroelectric installations run at almost constant hydraulic head and grid
synchronous speed. Opti-
mizing the power output under these conditions is relatively easy. Small
installations, such as “run of river
systems,” on the other hand are subject to much more variable hydraulic conditions,
and optimum power (or
efficiency) from such an installation is much more difficult to achieve.
In most large installations, flow of water through the turbine is varied using
a governor system to maintain
constant generator speed. This is achieved by (1) varying the guide vane angle in
Francis and other reaction
turbines, (2) adjustment of blade angle and possibly entry vane angle to Kaplan and
propeller turbines, and (3)
adjustment of capacity in the cross-flow turbine. In impulse turbines such as the
Pelton wheel a variable inlet
nozzle is used. With this control method at constant head, the delivered torque to
the generator is proportional
to the flow, and the turbine speed is that required for synchronous generation at
the particular grid frequency.
This type of installation is a constant speed, constant frequency (CSCF) system;
and ideally, optimum power
(measured in real time) corresponds to constant speed operation.
In smaller installations in which the hydraulic head is relatively low and
variable over a wide range,
normally it is not possible to obtain optimum power at constant speed; and, for
coupling to a grid, the variable
speed but constant generated frequency (VSCF) requirement has, in the past, made
such small sites nonviable
for mains electricity generation. Power produced in these sites has been suitable
only for local heating which
has a much lower commercial value than grid frequency electricity.
Today, however, small hydroelectric systems will become more financially
attractive by virtue of modern
developments in low-cost power converters (from 100 W upwards), special VSCF
generators, and cheap com-
puting units for on-line power measurement and optimizing control. All these will
lead to different approaches
to system design, and this article discusses some of the relevant techniques.
Turbine and Generator Coupling
Turbine Characteristics (23,24). As an example, power–speed,

efficiency–speed, and torque–speed
characteristics of a Francis turbine with head as a parameter are shown in Fig. 1,
Fig. 2, and Fig. 3, respectively.
It can clearly be seen that optimal power is available from the turbine at
different speeds corresponding to
different heads.
If optimal efficiency is required rather than optimal power, then Fig. 2 shows
the turbine should run again
at variable speed depending on the head. These speeds are slightly higher than in
an optimal power operation
and correspond to smooth flow to the inlet of the turbine (1,2,17,18). In both
cases, the optimal load line for
the operation is the locus of the peak power or peak efficiencies shown in Fig. 1
and Fig. 2. The torque–speed
----------------------- Page 257-----------------------
2 HYDRAULIC TURBINES
Fig. 1. Power–speed theoretical characteristic of a Francis turbine. Head as a

parameter. This figure is useful for deter-
mining the optimal speed for maximum power production from the turbine.
characteristic, Fig. 3, which is derived from Fig. 1, is important for stability

analysis of the system, discussed
in the following section.
System Equations. The torque produced by a turbine, T , depends
on hydraulic head, H , flow rate, Q,
t
blade setting, θ, and speed, w, while the generator torque, T , is a function of

voltage, E , current, I , and speed.
g
The generator torque–speed characteristics are discussed in detail in the section

entitled “Electrical System,”
but at this point the system equations can be written as
where J is the moment of inertia the rotating assembly and K is a frictional

torque coefficient.
The current supplied by the generator depends on the load impendance, Z ,
the supply frequency, f , and a
control variable, ue , as well as on voltage and speed:
while the turbine flow depends on load, speed, vane setting, and a control variable,
um (which has in the past
been supplied by a mechanical governor):
Assuming no control action, the solution of Eq. (1) may or may not be stable
depending on the relative slopes
of the turbine and generator torque characteristics at the operating condition, and
this can be shown by either
of the following:
----------------------- Page 258-----------------------
HYDRAULIC TURBINES 3
Fig. 2. Efficiency–speed theoretical characteristic of Francis turbine. Head

as a parameter. This figure is useful for
determining the optimal speed for maximum efficiency of the turbine.
(1) Linearizing Eq. (1) for small disturbances, δw, around the mean operating
speed assuming all other pa-
rameters are constant, that is (20),
Ignoring friction, δw is stable if the term on the right-hand side is negative.
(2) Graphical static stability analysis—that is, superposition of the steady-state

turbine and generator char-
acteristics. The operating speed will be stable if the generator slope is
greater than the turbine slope at the
intersection (operating point).
Notice that in each case the mean or steady-state operating speed, w0 , is

given by
which corresponds to the point of intersection of the curves Tt and Tg (for the
case K = 0).
Under closed-loop control conditions where control actions um
and or ue are applied, system stability is
much more difficult to determine, particularly when several generating systems are
supplying a common grid
since Eq. (1) becomes a multivariable equation.
Optimal Operation. The optimal operating speed, wopt , for maximum
power (or maximum efficiency if
so desired) will vary with hydraulic conditions (e.g., head), and the control
system on the turbine or generator
(or both) should be designed to maintain this optimal operation. Such a
closed-loop control system must
----------------------- Page 259-----------------------
Fig. 3. Torque–speed theoretical characteristic of a Francis turbine. Head as a

parameter. This figure is important for
determining the stability of turbine-generator coupling.
continuously monitor the generator output power (or system efficiency) and use the
control variables ue and/or
um to adjust the speed so that maximum power (or efficiency) under the prevailing
conditions is obtained.
In older installations a mechanical governor was used to provide control
action, um , so that constant
(synchronous) speed was maintained. This governor could be replaced by a variable-
speed (or range-speed)
governor so that the speed set point corresponded to the optimal speed
as determined by the optimizing
computer. However, it is more satisfactory in modern installations to minimize the
mechanical adjustments
and to use the computer to provide control action ue on the generator or
electrical side, although care should
be taken to ensure that overspeed protection is still provided on the mechanical
side.
Electrical System
Synchronous Alternator and Induction Generator. Hydraulic

power is not constant but varies
seasonally or, for a “run of river” system, almost continually, depending on
rainfall, but when connected to a
power grid of capacity much larger than that of the individual hydro-station,
fluctuations in supply can be
easily accommodated.
In conventional systems, the generator is usually a dc excited alternator in
which the generated frequency
is proportional to the speed of rotation. As such, this may not match grid
frequency (if the turbine tends to
run at nonsynchronous speed), but fortunately the electrical constraints provided
by the grid, such as voltage,
frequency, impedance, and so on, force the turbine to run at a constant speed ws
(called synchronous speed),
producing therefore the same grid frequency. The torque–speed characteristic of
this synchronous alternator
is shown in Fig. 4, its vertical line indicating that the torque can vary up to the
maximum value, Tmax , at
synchronous speed. The unit may also run as a synchronous motor. If the turbine
torque increases above Tmax ,
----------------------- Page 260-----------------------
Fig. 4. Torque–speed characteristic of synchronous machine. This figure is

important for determining the operating point
of the system and stability (turbine-gear box-generator coupling).
Fig. 5. Torque–speed characteristic of asynchronous

(induction) machine.
then the alternator will not be able to absorb all the turbine power, speed runaway
will occur, and the system
will be said to be out of synchronism.
Induction machines with small slips can also be considered as constant-speed
systems. (Note: Percent slip
is the difference between synchronous and actual speed. For a four-pole 50 Hz
induction machine, synchronous
speed is 1500 rev/min. Used as a motor at 1485 rev/min, the “positive slip” = (1500
− 1485)/1500 × 100 = 1%).
Used as a generator and run at 1515 rev/min, the “negative slip” is 1%). An
induction generator can operate
on an infinite bus at slip speeds 1% to 5% above synchronous speed, a small
variation compared with typical
shifts in turbine speed. Hence, induction systems, which maintain constant speed
due to electrical constraints
imposed by a grid irrespective of turbine speed, are sometimes also classified as
CSCF systems even though
they are in fact VSCF in a narrow range above synchronous speed. Slip
higher than 5% is not practical,
since the electrical losses increase substantially and the efficiency deteriorates.
In particular, this is true for
large generators. The torque–speed characteristic of an induction generator
is shown in Fig. 5. Similar to
the synchronous machine, if the turbine torque increases above the peak, Eq. (3)
will no longer be satisfied
and the system becomes unstable, speed runaway occurs, and the machine generates no
power. In induction
machines, this maximum torque is also called “runaway” or “pull-out” torque. The
induction generator can also
be classified as a variable-torque, constant-frequency machine since at synchronous
speed its torque is equal
to zero and increases with speed.
Since the characteristic in Fig. 5 is constant and cannot be varied for a
squirrel cage generator, the control
variable ue is constant. An important characteristic of an induction generator is
that it uses reactive power
usually drawn from the grid in order to enable it to generate real power. This
means that grid failure would
----------------------- Page 261-----------------------
prevent generation unless the reactive power was supplied by a capacitor station
system. Since capacitors
always exist on the machine (from grid lines, for example), self-excitation is
still possible and there is a danger
of power being supplied from the induction machine in small hydro-stations.
Fatalities have occurred among
power-line workers due to lines being supplied from unknown sources, and protection
at the generator must
be provided to prevent this. Therefore, even in the induction generator case, there
is no automatic protection,
contrary to what many people may think. In the case of power loss, overspeed
protection of the turbine must
be provided.
If used on a wind station rather than a hydro-station, runaway would be more
acute and although a
speed controller is not necessary, some mechanism to dump excess wind power would
be needed. However,
in the wind station, a useful characteristic of the induction generator is that the
same machine runs as a
motor, drawing power from the grid, when below synchronous speed. This may be used
to advantage on wind
turbines which may not be self-starting and need to be motored up to self-
sustaining speeds. If the rotor of the
induction generator is a wire-wound type, then its torque–speed characteristic can
be controlled with variable
resistances in series with the rotor windings. The result will be a variable-
torque, variable-speed (narrow-
range) characteristic similar to the isosynchronous generator (see the
section entitled “The Double-Stator
(Cascade or Isosynchronous) Induction Machine”).
Although the technical merits of synchronous and induction generators have
been extensively argued in
the literature, they deserve at least a brief summary here. A synchronous generator
runs at constant speed
and hence requires costly speed controls if not run in conjunction with a grid.
However, it can supply negative
reactive power to the grid system, and, if necessary, it operates as a
synchronous condenser, whereas an
induction generator lacks this capability and also needs additional power factor
improvement equipment. On
the other hand, stability problems can arise where the synchronous generator of the
hydro-system is “thinly”
connected to other synchronous machines because of its location in the power
network (a large amount of
reactance exists between the main fossil-fuel generator and the hydro-power
generator). Any additional series
electrical impedance with the output of the machine will alter Te and reduce the
stability range of the operating
point. Induction machines have a distinct advantage in this matter.
It is worth mentioning that the instability may not be in speed (or
frequency) but in power supplied to
the grid, if long power lines with large series impedance are used.
Variable-Speed, Constant-Frequency (VSCF) Systems. In most
cases, variable-frequency power
must be converted to constant-frequency power. The conversion can be made with the
aid of rotary electrome-
chanical machinery or with solid-state converters using transistors and thyristors.
In some cases, nonconven-
tional generators produce constant frequency from a variable speed source. As a
result, generation schemes
involving variable-speed rotors are normally more complicated than constant-speed
systems.
Variable-speed, constant-frequency systems can be summarized as follows:
Ac Generator (Alternator). As shown in Fig. 6, the alternator is a
variable-output-frequency generator
when coupled to a varying-speed prime mover. Its field is derived from a permanent
magnet or dc excited coil.
The output of this machine is a variable-frequency, variable-voltage waveform. If
this voltage is not used for
heating or lighting, it may be processed in oder to be converted to a constant
frequency output. Therefore, the
generator output of varying frequency is rectified to give direct current, which is
converted to ac by an inverter.
The ac–dc–ac conversion is an established technology in high-voltage dc
transmission, and the operation of
inverter technology on the ac power grid is well understood within certain
parameters, including system short-
circuited and voltage stability at the point of interconnection. Application of
this scheme to hydropower is
limited only by the cost of power-processing equipment.
For this system, the control variable ue can be taken as equal to the
field current and may be used to
reduce the field losses at low output power. In small permanent magnet machines, the
field is constant and
therefore ue does not exist.
Field Modulation Techniques, Frequency Down Conversion. The
machine shown in Fig. 7 is basically
a three-phase wire-wound rotor slip rings machine excited by a single-phase grid
frequency (10). The output
----------------------- Page 262-----------------------
Fig. 6. Alternator and frequency conversion system for producing constant output
frequency from a variable speed prime
mover.
Fig. 7. Field modulation system for producing constant output frequency from
a variable speed prime mover.
from the slip rings is at grid frequency, amplitude-modulated by the speed

frequency of the machine which
consists of double-sideband frequency components without carrier. The output is
then demodulated in a bridge
rectifier, inverted by a SCR inverter and filtered to get grid frequency output.
The unique feature of this scheme is that it uses three separate phase
windings for the alternator to
produce single-phase output. The output phase voltages are rectified and
added together. This causes the
ripple component to be dramatically reduced, an effect similar to increasing the
pulse number (sampling rate)
in converter circuits. To produce a three-phase output, three sets of such a
scheme have to be used while
maintaining the proper phase difference between corresponding single-phase outputs.
To keep dc and lower
sidebands content down, the rotational frequency of the machine must be much higher
than its synchronous
speed (or at least 10 times the grid frequency divided by the number of pole
pairs).
The control variable ue in this system is the single-phase field.
This can be adjusted by a variable
transformer, for example.
Ac Commutator Generator (ACCG). Using an ACCG for hydro-energy is
perhaps the simplest means
of deriving constant-frequency output from a variable-speed machine (see Fig. 8),
and the suggestion was first
made by M. P. Kostenko in 1984 (21). Employed in a hydro-generating station
interconnected with a power
grid, ACCG field excitation can be derived from the grid (11,12). The machine output
frequency is equal to the
excitation frequency and can be connected back directly to the grid. The ACCG
provides line frequency output
----------------------- Page 263-----------------------
Fig. 8. Ac commutator generator (ACCG) system for producing constant output

frequency from a variable speed prime
mover.
due to its commutator and rotational voltage only, at any speed (without
transformer voltage), since feedback
compensation for its armature voltage is used. Therefore, the characteristics of
this machine (mechanical and
electrical) are linear and very similar to dc machines. The output current of the
machine is not affected by the
rotor inductance due to compensation which is practically equal to the difference
between grid voltage and the
generated voltage divided by the total resistance in the circuit. In ACCG there is
another parameter which can
modify the characteristic of the machine other than the excitation voltage, and
this is the phase between the
stator and grid voltages. If this phase is adjusted (for example, by adjusting the
position of the commutator
relative to the stator), then the torque magnitude and direction can be controlled.
The machine can work both
as a generator and a motor.
For this machine, the control variable ue can be regarded as a vector—that is,
field voltage and field phase.
The disadvantage is of course the commutator. The problem of reliable
commutation in ac machines is
serious. Several proven techniques are available to improve commutation, including
high-resistance brushes or
----------------------- Page 264-----------------------
Fig. 9. Squirrel cage induction machine with a four-quadrant power converter used
for producing constant output fre-
quency from a variable-speed prime mover.
Fig. 10. Double-output machine with slip power recovery with the aid of
a frequency converter. The figure shows a
two-quadrant converter used for negative slip (supersynchronous) power generation.
ue = # = constant is the triggering
angle.
commutators with many segments. The latter techniques have already been used to
build large ac commutation
motors successfully. These additional features make the machine costly and
according to the best of this author’s
knowledge, no commutator generators are available for sale today on the market.
Squirrel Cage Induction Machine with a Converter. A VSCF
system can be designed from a squirrel
cage induction machine and a converter acting as a frequency changer. In Fig. 9
here the induction generator
works as a CSCF generator in small slip; however, its self-synchronous speed is
dictated and can be adjusted
by the converter.
Although a reliable squirrel cage machine is used, the disadvantage of the
system is that the converter
is a four-quadrant type (i.e., able to drive the machine as a motor as well as a
generator—in other words, able
to absorb and feed power in both directions). Also, the converter should accept
variable frequency at its input
and produce constant frequency (grid frequency) at its output. All solid-state
converters produce undesirable
frequencies, which may be unacceptable if they cannot be filtered properly before
the output is fed to the grid
system (16).
The reactive power in this case is dictated by the converter characteristic
but not by the machine itself. The
control variables in this case are the converter output voltage and frequency.
Therefore ue can be regarded as
a two-variable vector which modifies the characteristic in Fig. 5. This type of
converter is relatively expensive
today and is produced by few companies. An alternative solution is the double-
output induction generator
system.
The Double-Output Induction Generator (DOIG) System (4,5,6,7,8,9).
Figure 10 illustrates the ma-
chine used as a generator is a wire-wound-rotor slip-rings induction machine (or a
cascade induction machine
if a contactless machine is desired without slip rings).
----------------------- Page 265-----------------------
Fig. 11. Double-stator (cascade, isosynchronous) induction machine. The torque–

speed characteristic can be varied by
adjustment of the electrical angle u ◦
e = a(0 ≤ a ≤ 180 ) between the two stators.
As a generator, the output is connected to two different frequency supplies.

One of them is the grid system,
where power is fed from the machine, and the other is supplying slip power to the
machine. Therefore, one port
of the machine is used to deliver power, while the other is used to absorb power at
positive slip (subsynchronous
speed). The vector sum is obviously equal to the mechanical power supplied by the
prime mover plus the losses.
At negative slip (supersychronous speed operation), both supplies absorb
power from the machine (Fig.
10). If the machine is used as VSCF generator at subsynchronous and
supersynchronous speeds, then a four-
quadrant power converter should be used in order to supply and absorb power in both
directions. Since two
different frequencies are involved, the machine is operating in the synchronous
mode at speeds which consist
of the sum or difference of these input frequencies. Therefore, similar to any
synchronous machine, a governor
(mechanical or electrical) is needed to control the turbine speed in order to be
connected to the grid. It is worth
mentioning that the connection to the grid is more complex in this case, since the
converter frequency, voltage,
and phase should be measured and adapted (equalized) to the rotor frequency,
voltage, and phase for minimum
current at the appropriate synchronized speed before connection. The advantage of
this installation is that the
converter will be responsible for the slip power (around 50%) only, in contrast to
the squirrel cage induction
machine with converter (Fig. 9) where the converter was responsible for all of the
power. The disadvantage
is stability problems (since two frequencies are involved). The reactive power is
not better than an ordinary
induction machine.
In this case ue = θ, which is the triggering angle of the inverter. Figure 10
shows a family of curves for
different values of θ.
In asynchronous mode, variable resistors are connected to the rotor port in
order to control its variable-
torque, variable-frequency characteristic in narrow range similar to the
isosynchronous machine.
The Double-Stator (Cascade or Isosynchronous) Induction Machine.
The turbine mechanical power
is equal to the torque multiplied by speed (see Fig. 11). Variable-speed, constant-
frequency generators are
therefore matching the turbine power to the load power on the generator. In the
same way, matching turbine
power to the load can be carried out with variable-torque, constant-speed
generators. The cascade induction
machine characteristic is similar to the wire-wound-rotor induction machine
characteristic which is controlled
by variable resistors in its rotor’s ports.
Therefore, this machine is also an induction type with controllable
characteristics in motoring and gener-
ating region (13,14). It can be used as a variable-speed motor as well as a
variable-speed generator in narrow
range at negative slip. Therefore, it can be also classified in the variable-torque,
constant-frequency family. The
machine consists of two polyphase (or single-phase) wire-wound rotors machines with
their rotors windings
----------------------- Page 266-----------------------
connected in cascade. One stator winding is fixed and the other can be manually or
automatically adjusted to
match the mechanical power to the electrical power when used as a generator or as a
motor. The machine itself
is self-protecting, and no current limiting device is needed. It can be coupled
directly to the power grid and to
the prime mover without an interface network or a governor.
In application as a generator, the machine is connected directly to the grid
and to the prime mover through
an appropriate gear ratio, which is designed according to the optimal output of
electrical power to the grid,
which corresponds to an almost constant speed (variable speed in narrow range).
This speed is equivalent to
the ordinary squirrel cage machine optimal speed in the generator mode (for
negative slip) and is independent
of the angles between the two stators. However, the torque is a function of the
angle between the stators, and
it is zero for zero and maximum for an electrical angle of 180 degrees. Therefore,
in this case ue = α, the angle
between the two stators.
This machine is equivalent to an ordinary induction machine with an ideal
variable transformer turn
ratio (Variac) introduced between the machine and grid (ue = turns ratio).
Simulation of the machine can be
made easily in the laboratory using a Variac and an ordinary induction machine. In
practice, this arrangement
is not practical. As stated before, it is also equivalent to a wire-wound-rotor
induction machine controlled by
variable resistors at its rotor ports.
As the turbine power changes due to changes in the flow rate, the generator
angle needs to be varied in
such a way that the power fed to the grid is maximum. This will correspond to
optimal power output from
the turbine. The adjustable stator is normally geared to a single-phase ac
servomotor which permits manual
adjustment as well as automatic closed-loop adjustment with the aid of a computer.
The two machine systems
can also be designed as a single-unit machine with a special-shaped squirrel cage
rotor, but with two, or a
single winding of two, different poles pairs in order to avoid the coupling between
the two ports (4). In an
asynchronous mode, one stator is connected to the grid while the other is connected
to variable resistances
which control its characteristic, as in the case with a wire-wound-rotor induction
machine.
An example is the 15 kVA hydro-plant which incorporates a
double-stator induction generator
(Fig. 12). This plant was commissioned in 1985 by the author in an abandoned water
mill site (Liffey Mills),
Nenagh, County Tipperary, Ireland, as a demonstration small-scale hydro-plant.
Found on the site was an
old nonadjustable gate, “horizontal Francis turbine” in relatively good condition,
and therefore it was used for
this project without any civil work investment. All the required safety devices
were installed and approved
by the power authority. The feature of the installation is as follows: As the
turbine power changes due to
changes in the flow rate, the generator angle (angle between the two stators—a
control variable) is varied by
a microprocessor in such a way that the power fed to the grid is maximum. This will
correspond to optimal
output power operation of the Francis turbine used to drive the generator. The
adjustable stator is geared to a
single-phase ac servomotor which permits manual adjustment as well as automatic
adjustment with the aid of
a microprocessor control circuit. The microprocessor measures the output power from
time to time and adjusts
the servomotor for maximum output power. For the rest of the time, the
microprocessor is used for monitoring
purposes such as in the case of electrical or mechanical power failure, and it will
adjust the angle between the
stators to zero for protection. This cannot be done immediately due to power
failure, but it is done after the
power is again restored and before the machines are connected back to the grid. In
this manner, minimum grid
current is assured which is equal to the magnetization current of the iron only
before the rotation is again
restored. When this happens, the microprocessor adjusts smoothly the angle between
the stators and increases
its speed as it works as a motor to drive the turbine up to the synchronous speed
where the latter in turn takes
over for the negative slip. The torque–speed characteristic of the generator is
shown in Fig. 13, the mechanical
power is shown in Fig. 14, and the electrical power is fed to the grid in Fig. 15.
The effect of adjustment of the angles between the two stators is clearly
indicated in Fig. 15. This effect
is similar to the adjustment of variable resistances in the rotor ports of a wire-
wound-rotor induction machine.
It is worth mentioning that this site cannot justify hydro-installations
other than the present one, due to
uncontrollable turbine, water level fluctuation, and relatively low output power.
----------------------- Page 267-----------------------
Fig. 12. A 15 kVA double-stator induction machine in Liffey Mills,

Nenagh, County Tipperary, Ireland. The system
provides optimal output turbine power at constant frequency from a variable torque.
The Programmed Pole Machine (Roesel Generator). The machine

in Fig. 16 is basically a synchronous
machine with a continuous variation of the number of poles to match the rotor
speed, and therefore it is a
“synchronous machine” at wide range of speeds. This machine is very
flexible and has the advantages of
the standard synchronous machine as well as the induction machine. The salient
advantage of the machine
becomes obvious when it is used as a self-excited alternator or as a VSCF generator
to feed power to the grid
from any prime mover such as in hydro-, wind, or sea wave turbines without any
interface, synchronization
network, or governor (15,19). The machine can be constructed as a single-phase or
polyphase unit. The rotor
is made from a layer of hard magnetic material with a rectangular hysteretic loop
(typically barium ferrite or
ceramic) with no windings and therefore has no slip rings or commutators.
The basic operation of the machine can be summarized as follows:
The output frequency of any syn-
chronous machine is given by the angular speed divided by the number of pole pairs.
(All conventional syn-
chronous electrical machines have an even number of poles determined by physical
windings.) This means
that the output frequency varies with rotational speed. The Roesel generator is
different in that the number of
poles can be varied continuously; and when varied in inverse proportion to the
speed, the output frequency is
maintained at a constant value.
The variation of the number of poles is achieved with the aid of an
excitation coil wrapped around an
exciter head inside the stator, in addition to the usual output windings.
Therefore, the shape of the stator
is different from that of an ordinary machine, and it is noncylindrical. The mutual
inductance between the
----------------------- Page 268-----------------------
Fig. 13. Torque–speed characteristic of the cascade machine. The figure is

important for determining the operating point
and stability of the system.
excitation coil and the output winding is practically equal to zero in order to
prevent direct power feeding from
the exciter to the output windings. In this way, the two coils are practically
decoupled from each other.
A reference frequency (or line frequency) is applied to the excitation coil
and magnetizes a pole on the
rotor as it turns. This is called “writing” a pole. The pole then induces a voltage
in the stator windings at the
same frequency. The output frequency then has the same value as the input
frequency, independent of shaft
speed at steady state (zero acceleration). As rotor speed decreases, the length of
the poles shorten, so more of
them are written around the circumference of the rotor. As rotor speed increases,
the circumferential length of
the poles increases and fewer of them are written around the circumference of the
rotor. On one extreme, there
will be an even-pole synchronous speed where an even number of equal-length poles
are spaced uniformly
around the rotor. At the other extreme, there will be an odd-pole synchronous
speed, where an odd number of
equal-length poles are equally spaced around the rotor. Between these extremes
there will be fractional poles in
the vicinity of the exciter head as poles are being partially rewritten. At the
even-pole synchronous speed, the
poles remain in the same position from one revolution to the next, so no rewriting
of poles actually takes place.
There will be no rotor hysteresis loss in this case, since the rotor iron
magnetization does not change with time.
At the odd-pole synchronous speed, however, every positive pole is being exactly
replaced with a negative pole
during each revolution, so rotor hysteresis losses will be maximum at this speed.
This loss can be minimized
with the proper exciter shape design. For stand-alone power generation, the output
voltage can be controlled by
regulating the exciter head current. For this case uε = exciter current. Figure 17
shows the open circuit output
voltage as a function of speed at constant exciter current. The stator in this case
was compensated at lower
----------------------- Page 269-----------------------
Fig. 14. Mechanical power–speed characteristic of the cascade machine. The figure
is useful for determining the mechan-
ical output power from the turbine.
speed. For application of parallel to the grid power production, an additional

control facility of the machine
characteristic is available. By controlling the electrical phase angle between the
exciter head current and the
output winding voltage, the machine characteristic can be controlled over a wide
range.
The study of the machine characteristics for application in wind energy power
generation was sponsored
by the European Union “Joule” Programme during 1989–1992. A prototype of the
machine was developed by
the author for that purpose.
Speed Against Size
For any generator, irrespective of whether it is CSCF or VSCF, the output power per
unit volume of the machine
is proportional to the square of the flux density and the machine speed. This flux
density is constant for a given
lamination material. Therefore, in order to reduce the machine size, the operating
speed should be as high as
possible or even above synchronous speed. For that reason, a gear ratio is always
needed for low-speed turbines
even if in principle a machine can produce power at low speed. The nominal
operating speed of the generator
should therefore be a compromise between the turbine speed and generator size. A
gear ratio of 1:7 is normally
acceptable in small installations and can be used as a guideline. For example, when
coupling a 150 rev/min
Francis turbine to a three-pole-pair induction generator at 5% slip, the generator
will run at 1050 rev/min.
----------------------- Page 270-----------------------
Fig. 15. Electrical power–speed characteristic of the cascade machine. The figure
is useful for determining the optimal
electrical power output from the generator.
In very large hydro-power stations such as on the Niagara River in

Ontario and James Bay in Que-
bec, a 1:1 gear ratio is used in order to reduce maintenance and increase
system reliability on account of
size. A typical 76,475 kW, 13,800 V, 95% power factor synchronous generator on the
Niagara River with a
head of 292 ft, running at 150 rev/min, is designed with 24 pole pairs
in order to produce 60 Hz output
frequency.
The size of this generator would be much smaller with a better power factor
if it was designed with many
fewer pole pairs.
It is worth mentioning that the power of a VSCF generator can never be
constant but proportional to its
speed. For the CSCF generator, its power also can never be constant but
proportional to its torque or square
of the flux density. Therefore for both cases, the optimal generator size is
dictated by its maximum operating
power.
----------------------- Page 271-----------------------
Fig. 16. Programmed pole (Roesel) machine. The torque–speed characteristic can be
varied by adjustment of the electrical
angle ue = µ0 between the writing head (exciter) current and the grid (output)
voltage in parallel grid connection. At stand-
alone operation, the machine produces constant output frequency from a variable
speed prime mover.
Stand-Alone Generators (25)
Ideal stand-alone generators are almost-constant power machines with a decreasing

torque–speed character-
istic for a given output voltage and load. This is true for all types of generators
(dc or ac), including induction
generators. Prime movers which are not regulated are generally unstable with stand-
alone operation. However,
any practical generator can offer an increased torque–speed characteristic due to
its losses. This characteristic
is stable for electric power generation from most prime movers. A stand-alone
system using a squirrel cage
induction generator was developed by the University of Toronto (22) where excess
power generated due to load
or speed variation was dampened using variable resistance heating elements. In this
way, constant output
voltage at constant frequency was achieved by variable-speed operation.
Protection
A hydro-plant, independent of size, must have minimal protection devices by law.

This is of particular im-
portance if the plant is operating parallel with the power authority system. Before
a license is granted, the
following protection devices should be installed and approved by the power
authority:
(1) Over and under voltage relay

(2) Over and under frequency relay
(3) Over current relay
(4) Single (or more)-phase failure and ground fault relay
----------------------- Page 272-----------------------

Fig. 17. Open circuit output voltage of the programmed pole generator compensated
at lower speed. The frequency of this
voltage is constant and equal to the exciter head frequency. The voltage shown is
for constant exciter current, which can
be regulated by variation of the exciter head current.
(5) Over and under speed protection (trip for faults on hydro-feeder)
(6) Fuses
(7) Reverse power relay
Extra optional protection devices are
(1) Phase balance relay

(2) Phase sequence relay
All installations should prevent islanding. For synchronous

generators, additional safety devices are
required as follows:
(1) Devices which provide a dedicated synchronizing and fault interrupting circuit
breaker for each unit
(2) A synchro-check relay or autosynchronizing on each synchronous unit
A power factor correction, normally below 90% to 95%, is required. However,

the connection of capacitors
in parallel with the generator need special approval from the power authority, and
in any case it should be
switched off together with the generator when the latter is disconnected for any
reason.
Voltage flicker or other voltage problems on the grid should be avoided. This
can happen in synchronous
generators, and therefore it should include special control of the excitation
system and loading/unloading ramps
rate.
For connection to the grid system at voltages above 50 kV, special isolating
arrangements should be made
with the power authority. If a frequency converter is used, then a facility should
be designed to prevent objec-
tional harmonics or voltage distortion on the grid system or on other customers’
electrical or communication
equipment.
----------------------- Page 273-----------------------
For induction machines operating in parallel with the grid, voltage and
frequency protection are still
necessary since in case of grid failure the induction generator may resonate
(become self excited) with the grid
impedance and produce different voltages at different frequencies (possibly also
harmonics).
Cost
It is evident that the cost of computers and electronic systems is

progressively declining in terms of real
and constant dollars. However, the cost of mechanical systems is rising. Therefore,
it is clear that the most
economical system will be the less mechanically complicated one, disregarding its
electronic complexity, such
as the cost of closed-loop real-time optimization. The cost of power
electronics processing systems such as
solid-state converters and inverters is also declining, which leaves the market
highly competitive. Therefore,
for the present and near future, the induction generator driven by a four-quadrant
converter as a VSCF optimal
system looks to be the most reliable and economically sound relative to other
options.
In the intermediate future, it is possible that the Roesel generator will
also be competitive in the market
when more research and development has taken place.
The synchronous machine will continue to dominate the large power stations.
However, many of these
machines will be replaced with induction machines.
Conclusion
Many run-of-river sites which were considered nonrelevant in the past can be
considered relevant today due
to availability of new types of generators and solid-state power processing
systems. These allow the plant to
operate from a low-variable head at variable speed in order to produce constant
frequency output at optimal
power efficiency. Constant speed and constant frequency optimization are used by
large hydro-plants which
are able to maintain a constant head by switching turbines or by adjusting of the
flow entrance to the turbine.
In the latter, the load on the turbine is varied according to the flow in order to
keep the speed constant.
In plants using closed-loop installation, a sensor may be used to measure
the output power and adjust an
independent control variable in the electrical system in such a way that this power
is always maximum.
There are minimum safety procedures required by law on the operation of a
hydro-plant—in particular,
when it is connected to the grid.
List of Symbols
E Voltage (V)
f Grid frequency (Hz)
H Head (m)
I , I (E , Z , f , w, ue ) Current (A)
J Inertia of the rotating parts (kg ·
m2 )
K Friction constant (Nm · s)

Q Flow rate (m3/s)
T , T (E , I , w) Electromechanical torque of the

generator (N · m)
g g
T , T (Q, H , w, #) Torque of a turbine (N · m)

t t
----------------------- Page 274-----------------------
ue An independent electrical control variable. A vector

in general.
um An independent mechanical control variable. A vector
in general.
Z Grid and machine impedance (#)
α Electrical angle between two stators (rad)
δ Small variation
µ0 Electrical angle (rad)
# Blade sitting angle (rad)
θ Electrical angle of triggering (rad)
w Speed (rad/s)
ws Synchronous speed (rad/s)
w0 Equilibrium speed (rad/s)
BIBLIOGRAPHY
1. C. C. Warnick, Hydro Power Engineering , Englewood Cliffs, NJ: Prentice-Hall,

1984.
2. Vennard, Street, Elementary Fluid Mechanics , 5th ed., New York: Wiley, 1976.
3. J. D. Russell, Jr., The induction generator in today’s industry, Westinghouse
Electric Corp., East Pittsburgh, PA 15112,
presented at IEEE Petroleum and Chemical Conference, Milwaukee, WI, September
1975.
4. A. R. W. Broadway, L. Burbridge, Self cascade machine: A low speed motor or
high frequency brushless alternator,
Proc. IEE 117, (7): 1277–1290, 1970.
5. A. Kusko, C. Samuah, Speed control of single frame cascade induction motor
with slip-power pump back, IEEE Trans.
Ind. Appl. , IA-14 (2): 97–105, 1978.
6. W. Shepherd, J. Stanway, Slip recovery in an induction motor by the use of a
thyristor inverter, IEEE Trans. Ind. Gen.
Appl. IGA-5 (1): 74–82, 1969.
7. J. Noda, Y. Hiro, T. Hori, Brushless Scherbius control of induction motors, in
IEEE Conf. Ind. Appl. Soc. 9th Annu. , Oct.
1974, Part 1, pp. 111–118.
8. K. Oguchi, H. Suzuki, Speed control of a brushless static Kramer system, IEEE
Trans. Ind. Appl. , 1A-17 (1): 22–27,
1981.
9. B. H. Smith, Sychronous behaviour of doubly fed twin stator induction machine,
IEEE Trans. Power Appar. Syst. ,
PAS-86 (10): 1227–1236, 1967.
10. T. S. Devaiah, R. S. Smith, Generation schemes for wind power plants, IEEE
Trans. Aerosp. Electron. Syst. , AES-11
(4): 543–550, 1975.
11. B. Adkins, W. J. Gibbs, Polyphase Commutator Machines , Cambridge, UK:
Cambridge University Press, 1951.
12. R. Smith, Analysis of polyphase commutator generators for wind power
applications, IEEE Trans. Aerosp. Electron.
Syst., AES-12 : 39–41, 1976.
13. D. Levy, The isosynchronous machine, characteristics and applications, in
Proc. IEEE Montech Conf. AC Power Syst. ,
`
Palais des Congres, Montreal, Canada, October 1–3, 1986, pp. 1–6.
14. D. Levy, Analysis of double stator induction machine used for VSCF small scale
hydro/wind power generator, Electric
Power Syst. Res. (USA) , 11 (3): 205–223, 1986.
15. D. Levy, Characteristics of the Roesel machine used for small scale hydro/wind
or sea wave electric power generation
and storage, in Proc. ASCE Waterpower ’87 Conf. , Portland, OR, August 1987,
pp. 1645–1654.
16. D. Levy, E. McQuade, Analysis and synthesis of static power converters, IEEE
Proc: Part G on Electric Circuits and
Systems, 133 (1): 39–57, 1986.
17. R. Daugherty, J. B. Franzini, Fluid Mechanics with Engineering Applications ,
18. E. H. Lewitt, Hydraulics and Fluid Mechanics , London: Pitman, 1966.
19. G. L. Johnson, Wind Energy Systems, Englewood Cliffs, NJ: Prentice-Hall, 1985,
p. 169.
20. N. Minorsky, Theory of Nonlinear Control Systems, New York: McGraw-Hill, 1969,
p. 80.
21. T. S. Jayadev, Windmills stage a comeback, IEEE Spectrum , November 45–49,
1976.
22. R. Bonert, G. Hoops, Stand alone induction generator with terminal impedance
controller and no turbine controls,
IEEE Trans. Energy Convers. , 3 (1): 28–31, 1990.
----------------------- Page 275-----------------------
23. D. Levy, Derivation of the torque-speed characteristic of a reaction turbine,

Proc. Inst. Mech. Eng. , 207, No. A3, Part A,
J. Power Energy, 1993, pp. 165–172.
24. Discussion on reference 23, Proc. Inst. Mech. Eng. , 208, No. A2, Part A, J.
Power Energy, 1994, pp. 151–153.
25. D. Levy, Stand alone induction generators, Electric Power Syst. Res. , 41:
191–201, 1977.
DAN LEVY
University of Limerick
----------------------- Page 276-----------------------
} { { } }
HOME
●
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Hydroelectric Generators
Standard Article

J. G. Campos-Barros1 and Lindsay Ingram2
1Independent Consulting Engineer, Rio de Janeiro, Brazil
2Manitoba HVDC Research Centre, Winnipeg, Manitoba,

Canada
Search
reserved.
❍ Acronym Finder
DOI: 10.1002/047134608X.W3005
Abstract
HVDC Unit-Connection Schemes
Hydraulic Turbines
Other Considerations
Acknowledgments
%20Conversion/W3005.htm17.06.2008 16:41:23
----------------------- Page 277-----------------------

c
HYDROELECTRIC GENERATORS
Hydraulic generating units for electric power generation normally run at a fixed
speed to provide an electrical
frequency of 50 or 60 Hz for conventional system needs. The turbines operate most
efficiently at only one oper-
ating point, which is the best combination of head, speed, and discharge. Any
deviation from these conditions
can cause a marked reduction in hydraulic efficiency because of flow separation,
secondary flows, turbulence,
and friction, all of which in turn, can cause pressure and shaft torque fluctuations
and cavitation.
High voltage direct current (HVDC) transmission of power for electric
utilities was introduced some 40
years ago. Electrical system planners, however, over the last few years
have reexamined its application to
remote hydro sites to reduce costs associated with the expensive sending-end
conversion equipment. Thus an
electrical arrangement known as the “unit connection” has been developed. This
innovation reduces capital
and operating costs considerably and also allows conventional generating
units to operate at an adjustable
speed/frequency, generally in the range of ±10%, with no increase in turbine cost
and only a modest increase
in generator and converter transformer costs.
When there is no necessity for fixed frequency at the generating station,
hydraulic turbines can operate
without the previous restriction of constant speed, thereby improving
hydraulic characteristics throughout
the normal range of operating conditions. Depending on river hydraulic regimes in
relation to plant size, this
opportunity could represent a large revenue improvement. Then adjustable
speed, in this context, depends
wholly on the conditions that the transmission system is HVDC and that
there is no requirement for fixed
frequency, such as an interconnection with another ac system.
When high voltage ac and dc systems are compared at the planning stage,
recognition should be given
to the additional economic benefit that could result from adjustable speed
operation. Utilities in Canada and
Brazil are exploring the possibility of profiting from the benefits of the HVDC unit
connection with adjustable
speed operation for future hydroelectric developments.
This article describes the basic concept of HVDC unit-connection
schemes leading to the efficiency of
adjustable speed turbine operation (sometimes called “variable speed” in referenced
literature). Then it moves
into a detailed analysis of the effects of HVDC and adjustable speed/frequency on
the hydroelectric generator
that is the basis of the article. Finally, the article continues with a brief
discussion of other related considerations
involved with this method of hydroelectric generation and transmission.
HVDC Unit-Connection Schemes
Conventional HVDC transmission schemes in remote hydroelectric sites are

similar to Fig. 1, which shows
the Limestone generating station on the Nelson River in Canada. The
basic feature is that the generating
units feed a common, high-voltage ac bus on the secondary side of the generator
transformers. The presence
of tuned ac filters on this bus imposes steady-state operation at a fixed electrical
frequency. As a result, the
loss of filter banks by protection action during faults or power swings becomes a
highly probable contingency
that subjects the station to harmonic distortion usually leading to large
overvoltages which must be recognized
----------------------- Page 278-----------------------
2 HYDROELECTRIC GENERATORS
Fig. 1. Example of a conventional HVDC sending-end converter station

arrangement, as used on the Nelson River,
Canada.
in the overall station design. Furthermore, the large filter capacitances

can cause generator self-excitation
due to load rejection, resulting in substantial overvoltages. For the station to
cope with these conditions, large
overrating factors (1.4 and above) have to be used when coordinating
converter valves, transformers, and
generator stator insulation ratings (see Ref. 1, Chap. 5).
----------------------- Page 279-----------------------
HYDROELECTRIC GENERATORS 3
Fig. 2. Example of a unit-connected HVDC sending-end arrangement with generators

connected in groups. Copyright
International Water Power and Dam Construction. Used with permission.
There are several possible arrangements for unit connection. Figure

2 shows one example. A group
of generators is paralleled directly on a generator bus and connected to
a 12-pulse converter transformer
arrangement, thereby eliminating one level of transformation and the ac switchyard
and ac filters, all of which
represents large capital cost reductions. Generator breakers, in this example, are
necessary but they must be
carefully selected because they must operate successfully throughout a given range
of variable frequencies.
Low frequency increases the duration of overcurrent in the wave, and high frequency
introduces the possibility
of a voltage restrike. In some cases, the opportunity to reduce costs may be
available by combining the converter
valves and transformers within the powerhouse (3).
In addition to the major equipment cost reductions mentioned, there are
significant electrical equipment
energy-loss savings, staff and maintenance cost reductions, and reliability
improvement because of the overall
simplicity of the scheme.
On the downside, the main disadvantage is that the generating station cannot
easily be interconnected
with an adjacent ac system. It may, however, be able to supply a small local load
at some additional cost.
Capital savings in adopting a unit-connected design are site-dependent. They
can be determined only by
examining all of the alternatives for a specific project. Useful guidance of
general validity may be obtained
from the cost breakdown of a conventional HVDC converter sending-end station, as
shown Refs. in 1 (Chap. 2)
and 3.
----------------------- Page 280-----------------------
Hydraulic Turbines
Utility hydraulic turbines can be
• single-regulated (needle stroke on impulse turbines, wicket gates, or runner

blades on reaction turbines)
or
• double-regulated (wicket gates and runner blades on reaction turbines).
The ability to adjust turbine speed introduces an additional regulating

variable with possible advantages
to both single- and double-regulated turbines.
Unlike steam turbines where rated steam conditions are constantly maintained
by the operating staff,
hydraulic turbine water head and flow conditions, subject to the vagaries
of nature, are rarely constant.
Therefore adjustable speed is an intrinsic feature of hydraulic machines because
speed is directly related to
water head. In some existing installations, hydraulic turbines, such as pump-
turbine machines, are designed
to operate at two speeds, each applicable to a different mode of operation (i.e.,
pumping or generating), thereby
providing a sufficiently high hydraulic machine efficiency in each mode.
Present hydraulic machines are designed as compromises between efficiency and
operational flexibility.
Once adjustable speed can fully compensate for head variations, the
hydraulic design may be directed to
obtaining load flexibility.
HVDC unit connection can simplify the operation of the generating station and
also make it possible to
operate at adjustable speed because the generating frequency, decoupled by the dc
link, is free of the obligation
of synchronous operation.
If the turbine output limits are already established by instabilities and the
avoidance of cavitation (i.e.,
the erosion that takes place on the surface of runner blades from the implosion of
air and vapor bubbles at high
pressure), the only remaining concern with head variation is cavitation at the
inlet edge of the runner blade.
This type of cavitation is very strong and comes from incorrect flow angles caused
by extremely high and low
heads. Because speed is closely related to head by flow similarity laws, it can be
shown that speed must vary
with the inverse of the square root of the head variation to maintain flow angles
(Ref. 1, Chap. 3). Consequently,
if speed can be varied freely there is no minimum head for the turbine, and the
maximum limit is imposed only
by mechanical considerations. Although cavitation at partial load is not of much
concern for hydro machines
with Kaplan blades, rough operation can be experienced at extremely low heads,
making adjustable speed also
a possibility for these units.
Speed variation does not affect the mechanical design of the turbine or the
generator because both must
be designed for the maximum theoretical runaway speed (when the generator has lost
its load and the governor
has failed to close the gates in the maximum gate opening position at maximum
head). However, the dynamic
properties must be reviewed to avoid any critical frequency resonances. The first
lateral critical frequencies
of the shaft are normally well above the maximum runaway speed, so this is not a
concern, but others may
exist and should be checked. Furthermore, the typical arrangement for turbine
generator units is such that
most of the inertia is located in the generator rotor. Hence, the air gap torque
harmonics are spent mostly in
accelerating the generator rotor inertia itself, and only a minor fraction of the
energy associated with them is
transmitted through the shaft.
Hydraulic passage excitations are also a consideration. When turbine
low-frequency pulsations due to
vortices in the draft tube resonate with the power plant structure,
severe operational limitations may be
imposed leading to energy losses and occasionally to structural damage.
Such excitations represent actual
constraints in fixed-speed operation when some operating points have to be avoided.
Adjustable speed has the
potential to eliminate the constraints and most of the resulting energy losses by a
relatively small change of
operational speed.
----------------------- Page 281-----------------------
Fig. 3. Normalized hill curves. AA# defines the fixed-speed operating range
for a given head as normalized flow (and
power) change. Copyright International Water Power and Dam Construction. Used with
permission.
Increased energy generation is derived directly from the hill chart curves in
Fig. 3. The vertical line AA#
defines a certain fixed-speed operation range, as flow and output power vary, and Pf
an operating point for that
speed. Adjusting the operating speed to move from point Pf to
operating point Pa increases the power output
in this case by 1% at no cost. The increased efficiency coming from a less turbulent
water flow also results
in inherently quieter operation and makes the unit less prone to vibrations and/or
cavitation. If operation is
constantly optimized for efficiency in this way, net energy gains are obtained (see
Ref. 1, Chap. 3).
In addition to increased efficiency, environmental benefits, such as improved
survival rates for fish passing
through the turbines, are also achieved. Furthermore, a new approach to
reservoir planning and design is
possible if energy gains are traded for flooded areas (see Ref. 1, Chap. 3).
Figure 4 shows normalized turbine performance curves corresponding to an
actual Francis runner design
in the 700 MW range at a net head of 80 m. For adjustable speed operation at rated
head, the turbine efficiency
at part load increases substantially over fixed-speed operation. The peak efficiency
changes only slightly, and
the efficiency for rated conditions is identical. Of course, with a large number of
units it is possible to operate at
near peak efficiency over a large power range by simply scheduling the number of
units in operation. Increased
efficiency with adjustable speed at rated head and part load, in this example, would
be most beneficial for a
station with a small number of large units.
From Fig. 4 it can also be seen that adjustable speed operation at low head
allows the turbine efficiency
at part load and the available peak power to increase quite dramatically compared
to fixed-speed operation.
In addition, the turbine output and efficiency near full load are
virtually identical for both maximum and
minimum head. This efficiency increase is obtained only with adjustable speed. For
fixed-speed operation at
minimum head and part load, draft tube pressure oscillations and
resulting shaft torque fluctuations are
possible. This zone is avoided by adjustable speed operation.
The Limestone generating station on the Nelson River in Canada has 10
propeller turbines rated at 125.4
MW at a rated head of 27.6 m and a speed of 90 rpm. The turbines operate at
constant speed and supply power
to a conventional HVDC system. In view of future developments on the same river,
adjustable speed operation
has been studied in a hypothetical unit-connected system. Figure 5 shows the
normalized performance curves
----------------------- Page 282-----------------------
Fig. 4. Fixed-speed/adjustable speed comparison. Normalized performance curves for

a 700 MW Francis turbine. Copy-
right International Water Power and Dam Construction. Used with permission.
for this plant with maximum speed variations of 5%, 10%, and 15%. The performance
curve of a Kaplan turbine
is superimposed for reference. The efficiency range for the Kaplan turbine is larger
than the adjustable speed
propeller turbine with a 15% speed range, but its capital and operating costs can
be on the order of 30% greater.
The potential advantages of operating over a range of speeds are significantly

different for the various
standard turbine types as follows:
• Very low specific speed (high-head Pelton wheels). Practically no advantage to

be expected from this type
of development.
• Low specific speed (high-head Francis). Head variations in percent are small.
Some efficiency gains can be
obtained. Pump-turbine units can benefit significantly for each mode of
operation.
• High specific speed (low-head Francis and propeller). Head variations in percent
are very significant. Speed
adjustment can compensate for efficiency losses due to head variation and can be
used to avoid or reduce
cavitation.
• Very high specific speed (low-head propeller, Kaplan, and bulb). Head variations
in percent are very signif-
icant, particularly for tidal schemes. Speed adjustment can compensate for
efficiency losses and avoid or
reduce cavitation. May allow fixed blade propeller design to be used instead of
a more expensive (approx.
30% more) Kaplan design.
For more detailed information on hydraulic turbines, the reader is directed to

Ref. 1, Chap. 3.
----------------------- Page 283-----------------------
Fig. 5. Normalized performance curves for a Nelson River turbine at

rated head compared to speed operation with
speed variations of 0%, 5%, 10%, and 15%. The performance curve of a Kaplan turbine
is shown for reference. Copyright
International Water Power and Dam Construction. Used with permission.
Conventional generators can be built for adjustable speed operation with

only minor design modifications
and modest cost increases. This can be done in unit-connected schemes for HVDC
transmission systems, as
previously described in this article, or with a combination of rectifier-
and load-commutated inverter (LCI)
equipment (also known as “back-to-back” HVDC connection). The latter method is used
when the transmission
system is ac and is generally close to the generating station. But
because the LCI has the same rating as
the synchronous machine, the cost is effectively increased by 75% to
110% due to the LCI equipment. In
developments where conventional HVDC transmission is economically competitive or
needed for other reasons
(frequency conversion, for instance), however, an adjustable speed HVDC unit-
connection solution imposes no
overprice. Rather, in addition to all other inherent advantages of the adjustable
speed mode of operation, it
substantially reduces capital and operating costs.
Hydroelectric generators using adjustable speed is not a new
subject. There are several examples of
pumped-storage installations, particularly in Asia, with ac machines specially
designed for adjustable speed
pump-turbine operation (2). Some of these machines are “doubly fed,” that
is, instead of a conventional dc
excited rotor, they have a three-phase ac wound rotor which is fed by a
“cycloconverter.” The purpose of this
device is to provide an adjustable frequency supply to the rotor winding which then
allows the rotor speed to
vary while the rotating field in the stator remains at a standard ac system
frequency. The cost of this type of
machine is considerably higher than a conventional hydro machine (perhaps 50% to
90% more) because of the
high cost of the rotor and excitation system.
This article discusses adjustable speed operation of conventional
hydroelectric generators in HVDC unit-
connection systems.
----------------------- Page 284-----------------------
General Assessment.
Harmonic Electric Loading. Conventional generators feeding ac systems
are subjected to negative se-
quence current loading. The negative sequence current creates low-order harmonics
as a result of rotor saliency.
Hence, generators are designed to have additional thermal capacity to sustain 8% to
12% of negative sequence
current loading. Generators feeding only HVDC converters in the absence of dc line
current second harmonics
(the normal steady-state operating condition) see virtually no negative sequence
loading. Thus the additional
rating can be used to compensate for the 12-pulse harmonic electrical loading (6).
Torque Harmonics. The presence of pulse-number-related frequencies in
the rotor (i.e., multiples of 600
Hz for a 50 Hz fundamental frequency) will cause a small electromechanical torque
of a few percent. Although
high in the context of large machines, such frequencies might excite natural modes
of resonance either in small
components or, because of nonlinearities, might induce subharmonic vibrations in
larger components.
Adjustable Speed Operation of Hydro Generators. At a higher
speed n (rpm), machine operation will
take place at a correspondingly higher frequency f (Hz). Stray losses
due to windage, mechanical friction,
and harmonics increase. However, there is only a minor influence on the overall
electrical efficiency for the
small increases in speed required to optimize energy conversion in hydro turbines.
The case of downward speed
changes requires more attention as it may impose some additional capacity on
excitation circuitry if full voltage
is required at minimum speed.
Basic Design Frequency and Speed. The present day switching speed
of high power electronic devices,
especially the parameters associated with the current extinction process, allows
HVDC converters to operate
at up to about 120 Hz.
Vibration and Noise. The full pole-pitch, air-gap flux components
caused by current harmonics can
interact with each other and with other components of air-gap flux in a
rather complicated way, causing
vibration forces and contributing to increased noise level.
Generator Power Factor. The absence of ac filters leads to
unit-connected generator operation with
very high fundamental frequency power factor values, as seen from its terminals,
typically 0.9 to 0.94. In the
limit, for a hypothetical transformerless unit connection, the generator terminal
fundamental frequency power
factor is reduced, as a first approximation, to the cosine of the firing angle “α.”
This is so, because the ideal
commutation voltage moves inside the machine. Given the comparatively high
frequency of commutations
(12 per cycle), the voltage behind subtransient reactance, with good approximation,
is the new commutation
voltage (see Ref. 1, Chap. 4). In practice, internalizing the commutation
voltage relieves the machine from
about half the total converter reactive power requirement that it would otherwise
have to provide.
Fundamental Frequency-Generator/Converter Interaction.
The commutation process is a short
circuit between two phases of the generator across the transformer
leakage reactance xt that lasts for only
about 20 electrical degrees. The commutation reactance xc for a unit-connection
arrangement is given by Eq. (1).
x## is the actual operating point subtransient reactance of the generator, xt the
transformer leakage reactance,
and xc the commutation reactance in ohms at the transformer base:
The commutation has two main effects on the operation of a unit connection. One is
detrimental in that it
creates an ohmic drop of the dc voltage in which the commutation reactance plays
the part of a nondissipative
resistance. The greater xc , the greater the loss of voltage:
----------------------- Page 285-----------------------
where
#Ud = dc voltage drop in volts
Ud = dc voltage in volts
IG = generator current in amps
UG = generator voltage in volts
The other main effect of commutation is beneficial in that

operation with somewhat longer commuta-
tion periods reduces harmonic distortion and the fast transients that are
detrimental to the stator winding
insulation.
The addition of the subtransient reactances to the commutation circuit
increases the commutation reac-
tances compared with conventional arrangements.
Reactance Variation with Speed.
Operation with Constant Flux. When the generator operates
with constant flux ( U/f = constant), the
voltage and the frequency vary proportionally with the speed, and the generator
current remains constant. In
this case, all of the generator reactances and the transformer leakage reactance
vary proportionally with the
frequency. Therefore, so does the commutation reactance given by
For a given firing angle, the angle of commutation and the voltage drop of the dc
voltage are constant:
Operation with Adjusted Flux (U G = Constant). In this

case, generator power is proportional to speed
#
implying that I has to vary with frequency (I = K f ) and to
the commutation reactance x . The inductive
G G
c
2
voltage drop becomes proportional to f :
The per unit voltage drop increase with the square of the frequency is of little
concern at higher speeds because
generated voltage can easily be increased through modest flux adjustments in such a
situation. At low speeds,
the per unit voltage drop reduction at a quadratic rate is beneficial in that less
overflux is required to keep the
generated voltage at rated levels.
Typical Values of Commutation Reactances for 12-Pulse Unit Connection.
The subtransient
reactance seen by the commutation circuit during operation of a unit connection is
a nonlinear function of direct
and quadrature axis reactances, dependent on the actual air-gap flux magnitude
and spatial orientation at
the specific operating conditions. For planning and station design, the subtransient
reactance x## of hydro units
can be evaluated from the negative sequence reactance x2 . For a generator equipped
with a continuous damper
## ##
winding, the values of the reactances x d and x q are very close (within 10%),
and the following approximation
is suggested:
In practice, the value of the transformer reactance based on the transformer rating
is about half that of the
subtransient reactance, and so with interpolar damper connections the commutation
reactance approximates
----------------------- Page 286-----------------------
the subtransient reactance:

For 12-pulse thyristor converters at rated frequency, the figures shown in
Table 1 apply. For firing angles
◦ ◦
◦
(α) between 10 and 20 , the commutation angle µ can be kept under 30
for standard values of transformer
leakage and generator subtransient reactances. The 30◦
limit must be obeyed in principle in steady state to
avoid the reduction of dc voltage brought about by simultaneous commutations. As

10% leakage is economically
achievable in practice, the commutation reactance can be set around 20%
with ease and the commutation
angle kept well below 30◦ for thyristor bridge unit connections.
When diodes are used, the nominal firing
angles become practically zero, and the commutation angles increase substantially.
In this case, either lower
subtransient and transformer reactances are selected, or operation with forced
retard (i.e., with reduced dc
voltage and a communication angle equal to 30% at full load) is accepted.
Damper Winding. Interpolar damper connections are sometimes
avoided to improve the mechanical
##
reliability of salient pole rotors. However, for a machine equipped with a
noncontinuous damper, x q increases
##
and reaches a value of about 1.5x d . This causes an increase in x2
of about 25%. Therefore, for a machine
constructed with noncontinuous damper connections,
Suppression of the interpolar damper links increases the value of the commutation
reactance xc by about 12.5%.
This also increases subtransient saliency and may allow a higher proportion of
stator harmonics to penetrate
the rotor circuits, resulting in increased losses, vibration, and noise. Therefore
full cage damper circuits are
highly recommended for HVDC unit-connected systems.
Reactance Variation Due To Machine Load.
All reactances in a generator vary with the machine
state of saturation. The corresponding effect on xc , however, is about 1%, which
is negligible.
The variation of frequency does not affect the reluctance of the
magnetic circuit significantly over the
considered range. Therefore, xc is practically constant in the
working range for constant U/f . However, if full
voltage is required at minimum speed, that is, U/f is
increased because of overexcitation, saturation may
increase, and the subtransient and commutation reactances are reduced.
Influence of Damper Resistance. In a generator
connected to a converter, the effectiveness of the
dampers is more important against the stator current harmonics. The skin effect
increases the resistance of
the dampers by about 30% within the range of 45 Hz to 75 Hz. However, their
resistance is less than 1% of the
##
##
value of the reactance x d , and so the increase is equivalent to about 0.3% of x
d . Thus damper resistance can
be disregarded in the context of commutation analysis.
----------------------- Page 287-----------------------
Fig. 6. Bode diagram of Ld for a 60 Hz, 194.4 MVA

hydro generator.
Fig. 7. Bode diagram of Lq for a 60 Hz, 194.4 MVA

hydro generator.
Generator Frequency Response. Figures 6 and 7 are Bode diagrams

of direct [L (s)] and quadrature
d
[Lq (s)] axis operational inductances plotted from the actual direct and quadrature
axis parameters, obtained
by standard testing of a salient-pole, 194.4 MVA, 13.8 kV hydro generator. The
horizontal axis refers to slip
frequency seen by the rotor (see Ref. 1, Chap. 4).
The diagrams can be derived from a larger number of time
constants, if the relevant test information
is available. This corresponds to a larger number of rotor circuits when deriving
the equivalent circuit and
to a correspondingly larger number of operational inductance poles and zeros. The
characteristic asymptotic
behavior consisting of ever lower “plateaus” remains, nevertheless, unaltered in
shape.
It is very relevant to note from this example that stator harmonics coupled
with rotor frequencies of only
about 50 Hz and up already see operational inductances that remain constant over
the speed ranges considered
for HVDC unit-connected stations. The lowest stator winding frequencies associated
with converter operation
induce slip frequencies of at least fundamental frequency. Therefore, it can indeed
be safely considered that all
machine reactances associated with the commutation process vary linearly with
speed.
Generator Excitation. The most commonly used and simplest system for
generator excitation is the
shunt arrangement. This system is well suited to operation under a constant U/f
(voltage/frequency). A field
flashing system from a battery can be used to provide power to start up
the excitation system. When the
----------------------- Page 288-----------------------
generator is required to operate with constant (rated) voltage over the entire
range of speed/frequency, it may
be necessary to oversize the excitation system to allow the generator to operate at
low frequencies.
Auxiliary Three-Phase Generator. Oversizing may be avoided by using
an auxiliary three-phase syn-
chronous generator coupled to the main unit shaft. This auxiliary generator,
provided with a shunt excitation
system and with a field flashing system, supplies the main generator excitation.
Because it operates with a
constant U/f , the excitation transformer does not need to be oversized.
Fully Independent Auxiliary Generators. In large, multiunit hydro
stations, it may be more convenient
to provide smaller independent hydro units to feed the excitation systems, station
auxiliaries, and local loads.
In this way all effects on auxiliaries caused by harmonics in the station supply
and by the frequency variations
are eliminated.
Cycling and Fatigue. Excessive cycling of machines over significantly
varied operating conditions may
lead to premature aging and even to failures. Designers usually consider the
influence of voltage, frequency,
and load variation. The last two have a direct bearing on operating temperature.
The amplitude of the cyclic
variations and its periodicity are the key variables for each parameter considered.
Terminal voltage suffers some comparatively high-frequency distortion

originated by the commutations
over the operating life of the machine. In principle this could lead to premature
aging of insulation. It is of
little concern, however, in filterless, natural commutation, 12-pulse operation.
Speed/frequency cycling could be seen as the major cause of concern due to
adjustable speed HVDC unit-
connection operation. However, it is characterized by periods of at least hours in
a certain load and speed regime
followed by a slow, controlled transition to the new operating point. The duration
of a full operating cycle varies
from a few hours in pump storage stations to several days in most generating
stations, corresponding to about a
thousand cycles or less per year. The amplitude of the speed adjustments are site-
dependent, but simulations of
practical cases indicate that they are likely to remain within ±10% for the
majority. From present experience,
this additional duty does not require major design modifications or change of rotor
materials.
Stator Winding Insulation. When repetitive impulse voltages reach a
winding, the voltage distribution
along the winding is initially nonlinear and overstresses turn and ground
insulation of the first coil. If the stress
is close to the partial-discharge (PD) inception voltage, electrical aging of the
insulation may occur. In the past
the evaluation of insulation aging by impulse voltages has been studied to
determine its endurance against
lightning surges or the fast switching surges originated by restriking breakers,
especially vacuum types. The
number of pulses in this concern is limited, however, by comparison to the number
of impulses inherent in
converter operation. Frequent repetition of relatively low-peak voltages is
the special feature of converter
operation. Therefore it is essential to carefully study and evaluate the transients
and waveforms expected in
the actual station arrangement during the design stage in close cooperation with
equipment manufacturers
for both cold and hot machines, given that dielectric characteristics vary with
temperature (5).
The insulation system of large machines is made of form-wound windings
consisting of turn insulation
and micaceous ground insulation. In conventional ac operation, the voltage
applied to the turn insulation
is far lower than the PD inception voltage because the voltage
distributes itself uniformly along the entire
winding. In high switching frequency converters, the insulation of the
first turn can be highly stressed by
impulses. Fortunately, however, for generator sizes above 30 MVA, single turn coils
are normal, and so turn
insulation problems are of concern only in smaller machines. In addition, large
machines are always operated
with converter transformers. In schemes such as these, the fast component is
blocked, and only a distorted ac
voltage, comprising mainly slow repetitive components, is applied to the windings.
Poor circuit configuration
and faulty station design, however, can allow fast impulses to be
greatly amplified, especially between the
generator and the converter transformer, by successive reflections due to abrupt
changes in surge impedance.
Special design care in this area is necessary.
----------------------- Page 289-----------------------
Generator Losses.
Windage Losses. If the generator speed range is not too large, the unit
can be self-ventilated. Then
windage losses are proportional to the third power of speed. When the generator
cannot be self-ventilated at
low speeds, however, a forced cooling system may be required to support
ventilation.
Iron Losses. This is a common designation for a number of phenomena of
different natures in the iron
core. They can be split up roughly as “eddy current losses” and “hysteresis
losses.” For 0.5 mm laminations
in delivery condition and at 50 Hz, for instance, the eddy current losses measured
according to the Epstein
method are about 33% of the total losses. Consequently, the hysteresis losses
correspond, in this case, to the
remaining 67%.
Eddy Current Losses. This is a collective name for losses caused
by eddy currents within the lami-
nations themselves and by eddy currents circulating between laminations and some
other solid or structural
components of the core. Eddy current losses vary roughly in proportion to the
square of both induction and
frequency (≈B2 , ≈f 2 ).
Hysteresis Losses. These relate to the area within the B/H

characteristic of the core material and
vary roughly in proportion to frequency (≈f ). The relationship between induction
and hysteresis losses is very
complicated as it depends on both space and time variations of the induction. For
strictly pulsating induction,
the hysteresis losses vary roughly with the square of the induction (≈B2 ). For
strictly rotating induction, losses
have a maximum for some critical value of B after which they decrease with
increasing B .
For design and construction, different manufacturers use somewhat
different practices, but now the
actual design of very large machines is supported by sophisticated CAD packages
that usually rely on finite-
element techniques for accurately considering flux patterns and current eddies.
Small machines may still be
adapted from previous, well-known designs, to reduce escalating engineering costs.
Most machine phenomena,
however, can be simplified when doing preliminary estimates and so can iron losses.
The following relationship
is suggested:
where
PFe = the total iron losses
PFe0 = the total iron losses at rated frequency and nominal induction
f = the actual frequency under consideration
f 0 = the rated frequency
B = the actual induction under consideration
B0 = the nominal induction
kh = the hysteresis share of PFe0
This expression is a fair approximation for 0 <B <B0 when f varies between 25
Hz and 65 Hz and f 0 varies
between 40 Hz and 60 Hz. The numerical value for kh depends, among other
things, on machine design and
lamination treatment during the manufacturing process. The value is usually 0.5 for
0.5 mm steel lamination,
because it is typical that the eddy current loss share increases from the
previously mentioned value of 0.33 for
steel laminations before machining processes to about 0.5 after the machine is
completed. The iron losses due
to voltage harmonics are low and are neglected in the present context.
Stray Losses in Stator Copper. In the windings of ac machines, the
transverse flux in the slots creates
supplementary losses by skin effect in the copper strands in addition to joule
losses. When the generator is
connected to a converter station, the phase currents contain many harmonics that
cause supplementary losses.
These can be computed similarly to those corresponding to the fundamental frequency
and have significant
values up to the 25th. These losses are proportional to the square of the frequency
(f 2 ).
----------------------- Page 290-----------------------
Losses in Pole Shoes and Damper Windings. Stator current

harmonics create variation of the in-
duction in the air gap and cause losses in the pole shoes. These losses can be made
negligible in the iron by
using laminated steel sheets to manufacture the rotor poles. The losses created in
the damper windings by the
harmonics are computed as for an asynchronous machine (up to the 25th harmonic).
These losses may be high
enough in certain operating conditions to cause overheating of the damper
windings. Short-circuiting rings
must be designed to allow free damping-bar thermal expansion.
Generator Rating. The generator terminal power factor in unit
connection is typically higher than
that for a conventional HVDC arrangement of similar characteristics over the entire
range of operation. In
addition, as shown in 1, Chap. 4, a generator rated to provide maximum power at
base speed is overrated for
adjustable speed operation duty because maximum power is provided only at maximum
speed.
Generator Rating at Fixed Speed. The generator MVA rating for 12-
pulse operation is given by
where kH is a factor that allows for harmonic VA demand due to the rectifier load.
It varies with operating
parameters (firing and commutation angles, full or partial load) chosen to specify
the rating point but is only
a fraction of a percent for large 12-pulse unit connections. Reference 6 recommends
values of 1.01 for twelve-
pulse and 1.05 for six-pulse operation. Reference 1, Chap. 4, considers these
numbers typical for laminated
pole construction. Pd is the dc power of the rectifier load connected to the
generator, and φt1 is the fundamental
frequency power factor angle seen from the generator terminals. The converter
fundamental frequency power
factor angle should not be used here because now the commutation voltage is
internal to the generator. The
values to be used are approximated within a few percent by
which corresponds to extracting the generator subtransient reactance from the

converter commutation reac-
tance. This approximation is usually acceptable for planning and specifications (see
Ref. 1, Chap. 4 for a more
detailed discussion).
Generator Rating at Adjustable Speed. The usual definition of the
nominal MVA of the generator
is given by
where the subscript N indicates nominal (rated) values, UGN = rated generator
voltage and IGN = rated gener-
ator current. It has to be adapted for operation at adjustable speed by
where nN = generator rated speed. No consideration is given here to the fact that
iron and copper losses caused
by the skin effect are frequency-dependent, with the consequence that at n < nN a
higher induction, UG >n/nN
· UGN , is acceptable in a first approximation.
The generator rating for adjustable speed operation must be derived from the
most demanding conditions
over the range of operating points which are defined by the actual requirements for
the specific development.
The ideal requirements are reduced voltage at the lowest speed and maximum power
only at highest head and
----------------------- Page 291-----------------------
Fig. 8. Runaway and overspeed curves for hydro
generator sets.
speed. This may not be always the case, however, and so the actual operating
conditions must be investigated
to identify all limiting requirements that do not correspond necessarily to just
one operating point. The need
to absorb 12-pulse harmonic loading must be combined with the actual negative
sequence loading expected
from the small imperfections of the actual equipment, and the need for increased
excitation must be checked
against reduced losses at lower speeds.
Generator Voltage. It is generally required in ac and in dc systems
that generators maintain an almost
constant voltage over the full operating range. This requirement cannot be fulfilled
by a generator whose rated
power is required at maximum speed and is operated with constant flux. However, as
mentioned above, higher
induction can be accepted at lower speeds, and if the generator current is also
reduced in proportion to the
voltage increase, a design compromise becomes easier at the lower end of the speed
range, with no or only a
marginal overrating at base speed.
2
Generator Inertia Parameters (GD ). Generator inertia parameters are
important because, together
with the hydraulic circuit and turbine design, they define the overspeed transient
peak. Figure 8 is a diagram
showing the wicket gates closing after a load rejection in conditions of runaway
speed and overspeed. Runaway
occurs when the turbine is left with no load at maximum head and fully open wicket
gates. It is a steady-state
condition with zero efficiency, and speed is directly proportional to the square
root of the head. A total loss
of governor control leads to runaway. The time to achieve runaway speed increases
with the inertia. Design
standards require that hydro generating sets be built to withstand
runaway conditions without structural
damage.
Overspeed, on the other hand, is the peak velocity of a controlled machine
during a load rejection transient.
The slower the wicket gate closing action, the longer the machine accelerates
toward runaway, and the greater
the overspeed. The larger the generating set inertia, the smaller the acceleration,
and the smaller the overspeed.
It is reasonable to question the limits of overspeed that can be specified for
unit-connected generators. The
so-called “natural” inertia of the generator, which results from electrical and
mechanical design considerations
and is adequate for stability of speed regulation, should be checked first. After
that, load rejection simulations
and structural analysis of the generator and exciter mechanical components at the
corresponding overspeed
and runaway conditions must be performed to see if any limit is violated.
Eventually some small increase in
inertia may be required, but it is likely that a lighter and less costly, nearly
“natural inertia” design would be
acceptable.
----------------------- Page 292-----------------------
New Zealand Field Tests. The New Zealand HVDC system was made
available for harmonic tests
during two periods of converter transformer maintenance in 1993 and 1995.
Under these conditions, the
sending-end HVDC poles and ac filters were disconnected from the ac
system. Operation was uneventful
as a “group connected” unit connection (4). Waveform distortion
measurements confirmed the theoretical
predictions, and the generator operating parameters, such as temperature,
vibration, and noise, remained
within normal limits.
Other Considerations
Converter Transformers. The HVDC unit-connected converter

transformer performs the functions of
the generator and HVDC converter transformer in a conventional scheme. In the
absence of an ac system at
the generating station, the unit connection does not have to maintain a constant ac
voltage or frequency at
the generator terminals. Therefore, control of the rectifier ac voltage, with the
function of keeping the firing
angle within the specified range, is achieved directly by generator excitation
control without the need for a
transformer on-load tap changer (OLTC) required in a conventional system.
Eliminating the OLTC reduces
the cost of the transformer by 20% to 30%, improves reliability, and
reduces maintenance. For a frequency
variation of ±10%, the transformer cost increases by about 6.5%, giving a net cost
reduction of approximately
14% to 24% (see Ref. 1, Appendix 8).
When a unit-connection scheme is operated with adjustable speed, the design
of the transformer takes into
account the resulting frequency range. In most cases, the maximum frequency
corresponds to the maximum
power, and hence the maximum MVA rating is determined. At lower frequencies core
saturation might require
some core oversizing. This is beneficially offset in most cases, however, by a
reduced turbine load caused by
reduced head conditions.
Speed Regulation. A HVDC rectifier load is unique in that it lacks
inertia, is frequency insensitive,
and in many cases is also insensitive to the rectifier voltage. In turn, when an
HVDC scheme is operated at a
fixed power order, isolated machines feeding the rectifier are subject to a torque
load that varies inversely with
frequency. To control the speed of the units in a unit-connected system, there must
be a clear understanding of
the intended system operation. The speed control should order the units to operate
at maximum efficiency for
the given machine loading, dictated by the dc power order and the head.
Several requirements have been identified in analyzing speed regulation for a
unit-connected system, and
the following conclusions have been drawn (see Ref. 1, Appendix 5):
• System damping requirements are very critical. Therefore, proper governor
settings are extremely impor-
tant.
• An auxiliary sending-end dc damping control is not required for sending-end
stability.
• An auxiliary frequency control acting through the governor load setter
is required for plant frequency
control and for operation at the highest achievable frequency for the system
load and head.
• Higher overspeed for loss of dc control should be considered.
• A dc capability control would prevent any excessive underfrequency conditions.
Overall, there are no technical difficulties that cannot be overcome for speed
regulation in a unit-connected
HVDC scheme operating within a speed range of approximately ±10%.
Modeling. When investigating new concepts, such as adjustable speed HVDC
unit-connection schemes,
it is essential to model the complete system to evaluate the consequences of system
effects on the generator
and the effect of the generator on the system. This involves developing
modeling techniques to simulate
the generator, the hydraulic system, and the electrical system to which
the generator is connected. Thus,
consideration must be given to the level of modeling required if
meaningful results are to be obtained. For
----------------------- Page 293-----------------------
instance, if the effect of the converter harmonics on generator heating

is required, then a more complex
generator model has to be used than when control strategies or system disturbances
are to be investigated.
The main difference between conventional and unit-connected HVDC systems is that
the generators are much
closer to the converters in the latter.
In nearly all cases simulation tools are available and, in most cases,
provide results that are accurate
within the bounds of accepted practice. For example, if information on localized
internal effects of the HVDC
system on the generator is required, a complex model based on finite elements can be
used.
Overall, it is essential to realize that the major electrical components of a
HVDC unit-connected scheme
form a closely knit system where mutual interactions are very relevant. Hence,
integrated studies choosing
adequate simulation tools are recommended (see Ref. 1, Chap. 8).
Station Auxiliaries. Station auxiliaries may be supplied directly from a
separate power source, such
as small hydraulic generators, in which case power quality is assured.
Alternatively, station auxiliaries may be
supplied from the main generators but will be subject to some voltage distortion
and frequency variation. The
selection of all powerhouse auxiliary equipment should be properly considered in
designing for this option.
Overall Performance. In principle, the stability and operational
reliability of unit-connected HVDC
systems using thyristor valves is inherently better than conventional HVDC systems.
The following are the
main issues (see Ref. 1, Chap. 7):
• The absence of ac filters drastically reduces the amount of energy stored in

electromagnetic fields. This
energy would be liable to oscillate within the station circuitry in the event
of control actions, perturbations,
and faults. Many station oscillation modes at harmonic and
subsynchronous frequencies are, therefore,
eliminated, together with the risk of generator self-excitation.
• The elimination of on-load tap changers. When used in conventional systems,
they induce perturbations
and overvoltages. They also have high maintenance requirements so that their
removal improves reliability
and reduces costs considerably.
• The wide separation in the magnitude of dominating natural frequencies of
turbine, generator excitation,
and converter valve firing controls makes it comparatively easy to reach optimum
adjustment.
• There is a remarkable increase in sending-end controllability compared with
conventional HVDC stations.
This is important when modulating power to improve receiving-end
system damping and/or frequency
control.
• Eliminating ac filters greatly improves the station dynamic behavior and
virtually eliminates overvoltage
stresses due to harmonic distortion. Generator speed/frequency swings
of large magnitudes do not lead
to disturbances of the firing sequence or to voltage distortion. There are no
filters to be lost by protection
action, and consequential faults cannot develop.
• Fast accurate control of power flow in steady state and during power swings is
one of the most important
characteristics of HVDC systems in large interconnected power system operation.
Thus, should this mod-
ulation duty be required as a major characteristic, adequate margins must be
considered and allowed for
at the design stages.
Diode-Equipped Rectifiers. In the foregoing discussion, it has been

assumed that the rectifiers and
inverters of the HVDC unit connected system are equipped with thyristors and that
the sending-end rectifier
controls the direct current in the transmission line. There is, however, no doubt
that direct current control can
also be performed by an inverter. In this case, the rectifier can be equipped with
diodes, which promises further
major cost reductions in equipment and losses and increased reliability and reduced
maintenance. The need
for rectifier controls and associated communications equipment is also drastically
reduced. On the other hand,
an important protection function of the rectifier current control is lost, as well
as the overcurrent limitation
and fast cleaning of line faults.
----------------------- Page 294-----------------------
The application of diode-equipped rectifiers requires further assessing the
need for dc circuit breakers in
the transmission line (see Ref. 1, Chap. 6).
Acknowledgments
In response to the need for an evaluation by generator specialists of the subject

of unit connection and adjustable
´
speed concepts, Conference Internationale Des Grand Reseaux Electriques A Haute
Tension (CIGRE) Study
Committees 11 (Rotating Machines) and 14 (HVDC links and ac power electronic
equipment) created Joint
Working Group (JWG) 11/14.09 in 1989. The following are the JWG’s main objectives:
• Identify the main factors to be considered for specification and design of

unit-connected generators.
• Propose parameters and performance indexes for comparison with conventional
arrangements.
• Review modeling possibilities for the study and design of systems containing
unit-connected stations.
In 1993, the additional objective of defining the characteristics and

advantages of adjustable speed opera-
tion was added. The final CIGRE report was published in August, 1997, entitled Guide
for Preliminary Design
and Specification of Hydro Stations With HVDC Unit Connected Generators, Technical
Brochure No. 116, item
(1) in the bibliography.
The technical brochure recognizes all individuals and corporations, too
numerous to list here, who con-
tributed to the study’s success. Interested readers are encouraged to review this
report for further information.
BIBLIOGRAPHY
1. CIGRE JWG 11/14.09, Guide for Preliminary Design and Specification of Hydro
Stations with HVDC Unit Connected
Generators, Technical Brochure, No. 116, 1997.
2. D. A. Green, Generating more efficiently with adjustable speed machines, Hydro
Rev. , 13: 18–27, April 1994.
3. L. Ingram, A practical design for an integrated HVDC unit-connected
hydroelectric generating station, IEEE Trans.
Power Deliver. , 3: 1615–1621, 1988.
4. S. J. Macdonald et al., Test results from Benmore HVDC station with the
generators operating in group connected mode,
IEEE Winter Power Meet. , 95 WM, New York: SPWRD, 1995, pp. 242–248.
5. C. Mubemba, K. J. Cornick, The analysis of fast-fronted switching surges in
machine windings-effect of dielectric losses,
Trans. South Afr. EE, 86 (2): 57–64, 1995.
6. P. C. S. Krishnayya, Block and double block connections for HVDC power station
infeed, IEEE PES Summer Power Meet.
and EHV/UHV Conf., Vancouver, Canada, 1973.
READING LIST
J. Arrillaga, J. G. Campos-Barros, H. J. Al-Khashali, Dynamic modelling of single

generator connected to HVDC converters,
IEEE/PAS , 97 (4): 1978.
J. Arrillaga et al., Characteristics of unit connected HVDC generator converters
operating at variable speeds, IEE Proc.-C ,
139 (3): 295–299, 1992.
J. Arrillaga et al., Harmonic elimination by dc ripple reinjection in generator-
converter units operating at variable speeds,
IEE Proc.-C , 140 (1): 57–64, 1993.
D. S. Bajwa, R. M. Mathur, Re-rating of synchronous generators supplying HVDC
converters with special reference to unit
connections, IEEE Canadian Conf. Commun. Power , Montreal: 1976.
J. P. Bowles, HVDC Systems development and concepts-the diode rectifier,
CIGRE SC 14 Meet. Colloquium, Winnipeg,
Canada, 1977.
----------------------- Page 295-----------------------
CIGRE JWG 11/14.09, Study of the application of HVDC variable speed unit connected
generators to hydro developments
in the Brazilian Amazon region, CIGRE Int. Colloquium, HVDC and FACTS,
Wellington, NZ, 1993.
CIGRE JWG 11/14.09, Hydro generators reactances and power factor in HVDC unit
connection, Int. Colloquium HVDC
FACTS , Montreal, Quebec, Canada, 1995.
CIGRE JWG 11/14.09, Adjustable speed operation of hydro electric turbine
generators, Electra , 167: August, 1996.
EPRI Report RP3577-01, Application of adjustable speed machines in conventional and
pumped-storage hydro projects,
1995.
K. W. Kanngiesser, Unit connection of generator and converter to be integrated in
HVDC or HVAC energy transmission,
Int. Symp. HVDC Technol. , Rio de Janeiro, Brazil, 1983.
E. Kita et al., A 400 MW adjustable speed pumped storage system, Water Power and
Dam Construction, 43: 37–39, Nov.
1991.
L. Ingram et al., Adjustable speed hydroelectric turbine/generators, Water Power
and Dam Construction, 47: 40–42, July
1995.
G. Merouge, M. Denoncourt, Variable speed hydraulic generators, CIGRE Int.
Colloquium on HVDC Power Transmission,
Recife, Brazil, 1989.
M. Naidu, R. M. Mathur, Evaluation of unit connected variable speed hydropower
station for HVDC power transmission,
IEEE Trans. Power. Syst. , 4: 668–676, 1989.
R. D. Rangel, A. S. Pedroso, J. G. Campos-Barros, Dynamic performance of
salient pole unit-connected generators, Int.
Colloquium on HVDC Power Transmission, CIGRE, Recife, Brazil, 1989.
L. B. Reis, M. A. Saidel, Operation at adjustable speed: Prospectives of higher
energy efficiency and better environmental
integration for hydroelectric plants, Int. Symp. Electr. Power Eng. ,
Stockholm, Sweden, June, 1995.
H. Tanaka, An 82 MW Variable speed pumped-storage system, Water Power and Dam
Construction, 43: 25–26, Nov. 1991.
L. Terens, Schafer, Variable speed in hydro power generation utilizing static
frequency converters, WATERPOWER-An Int.
Conf. Hydropower, Nashville, TN, 1993.
J. G. CAMPOS-BARROS
Independent Consulting Engineer
LINDSAY INGRAM
Manitoba HVDC Research Centre
----------------------- Page 296-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Hydrogen Energy Systems
Standard Article

S. A. Sherif1 2 3
, F. Barbir , T. N. Veziroglu
1University of Florida, Gainesville, FL
2Energy Partners, Inc., West Palm Beach, FL

3University of Miami, Coral Gables, FL
Search
reserved.
❍ Acronym Finder
DOI: 10.1002/047134608X.W3025
Abstract
Fossil Fuels
New Energy System
Comparison of Energy Systems
Gaseous Hydrogen
Liquid Hydrogen
Slush Hydrogen
Conclusions

%20Conversion/W3025.htm17.06.2008 16:41:52
----------------------- Page 297-----------------------

c
HYDROGEN ENERGY SYSTEMS
Utilization of fossil fuels appears to be causing global problems such

as the greenhouse effect, ozone layer
depletion, and acid rain and pollution, which are posing great danger to our
environment and eventually to
life on our planet. Many engineers and scientists agree that a solution to all
these global problems would be
to replace the existing fossil fuel system with the hydrogen energy system.
Hydrogen is a very efficient and
clean fuel. Its combustion produces no greenhouse gases, no ozone layer-depleting
chemicals, little or no acid
rain ingredients, and no pollution. Hydrogen, produced from renewable (solar)
energy sources, would provide
a permanent energy system, which we may never have to change.
However, other energy systems have been proposed for the postpetroleum era
such as the synthetic fossil
fuel system. In this system, synthetic gasoline and synthetic natural gas are to be
produced using the abundant
deposits of coal. In a way, this could ensure the continuation of the present
fossil fuel system.
The two possible energy systems for the postfossil fuel era (solar hydrogen
energy system and synthetic
fossil fuel system) are compared with the present fossil fuel system by taking into
consideration production
costs, environmental damage, and utilization efficiencies. Results indicate
that the solar hydrogen energy
system is the best energy system to guarantee a sustainable future.
Under normal conditions, hydrogen is a gas, and for many applications it will
be used in gaseous form, just
like natural gas. Wherever space considerations are important, such as in aerospace
applications, hydrogen is
used in liquid form. If space saving becomes the number one priority, as in the
case with the earth-to-stationery
orbit single-stage transport, then slush hydrogen could be the best answer.
Fossil Fuels
After the invention of the steam engine in the 1860s, when advances brought about
by the Industrial Revolution
started to replace humans’ and beasts’ toil with nature’s energy sources, a bright
future seemed to be certain
for humankind. More and more of nature’s energy, initially in the form of wood and
coal and later as oil and
natural gas, was being harnessed for the benefit of humans. This resulted in mass
production of goods with a
corresponding reduction in prices, thus raising living standards. Communities asked
for factories, railroads,
highways, seaports, and airports. This meant more jobs, more income, more
goods, and more services. The
world’s standard of living was rising. When the Industrial Revolution started, the
annual gross world product
per capita was in some tens of dollars; today, it is about $6600 and rising
exponentially.
Fossil fuels, which fed this amazing economic growth, were the medicine to
cure deprivation, but it was
an untested medicine. As humans consumed more and more fossil fuels, two important
predicaments started
to emerge: (1) fossil fuels would be depleted in a foreseeable future, and (2)
fossil fuels and their combustion
products were causing global environmental problems.
Depletion. The demand for energy continues to rise for two main reasons:
(a) the continuing increase
in world population and (b) the growing demand by the developing countries in order
to improve their living
standards. At the present time, a large portion (about 70%) of the world energy
demand is met by fluid fossil
----------------------- Page 298-----------------------

2 HYDROGEN ENERGY SYSTEMS
Fig. 1. Estimates of world fossil fuel

production.
fuels (petroleum and natural gas) because of the availability of these resources
and their convenience of use.
However, it is expected that the world fluid fossil fuel production will soon peak
and thereafter begin to decrease
(1,2,3,4). Figure 1 shows estimates of the production rates of fossil fuels and the
world demand. It can be seen
that the fluid fossil fuel production worldwide will continue to rise for the next
15 years and then will start
to decrease. The coal production—because of obvious environmental reasons—
is expected to remain nearly
constant for the next decade and then start to decrease.
Meanwhile, as a result of the growing world population and the
desire of people to better their living
standards, the world demand for fluid fuels is rising (Fig. 1). It is expected that
the world population (which is
5.88 billion as of 1998 and rising at 1.5% per year) growth will slow down and
reach about 10 to 12 billion by the
end of the next century (5). Consequently, the world demand for fluid fuels will
slow down and reach around 1.6
× 1012 GJ per year (1 GJ of energy is approximately equal to the energy contained
in 3.4 gallons of petroleum).
There will be a growing gap, starting within the next 10 years, between the demand
and production of fluid
fuels.
Environmental Damage. The second predicament involving fossil
fuels is the environmental damage
being caused by their combustion products. Technologies for fossil fuel extraction,
transportation, processing,
and particularly their end use (combustion) have harmful effects on the
environment, which cause direct and
indirect negative effects on the economy. Excavation of coal devastates the land,
which has to be reclaimed and is
out of use for several years. During the extraction, transportation, and storage of
oil and gas, spills and leakages,
which cause water and air pollution, occur. Refining processes also have an
environmental impact. However,
most of the fossil fuel environmental impact occurs during end use. End use of all
fossil fuels is combustion,
irrespective of the final purpose (heating, electricity production, or motive power
for transportation). The main
constituents of fossil fuels are carbon and hydrogen in addition to some other
ingredients, which are originally
in the fuel (sulfur) or are added during refining (lead, alcohols). Combustion of
fossil fuels produces various
gases (CO , SO , NO , CH), soot and ash, droplets of tar, and other organic
compounds, which are all released
x x x
into the atmosphere and cause air pollution. Air pollution may be defined as the
presence of some gases and
particulates that are not a natural constituent of the atmosphere, or the presence
of natural constituents in
----------------------- Page 299-----------------------
HYDROGEN ENERGY SYSTEMS 3
an abnormal concentration. Air pollution causes damage to human health, animals,

crops, and structures, in
addition to reducing visibility, among other problems.
Once in the atmosphere, triggered by sunlight or by mixing with water and
other atmospheric compounds,
these primary pollutants may undergo chemical reactions, change their form, and
become secondary pollutants,
like ozone, aerosols, peroxyacyl nitrates, and various acids. Precipitation of
sulfur and nitrogen oxides, which
have dissolved in clouds and in rain droplets to form sulfuric and
nitric acids is called acid rain, but also
acid dew, acid fog, and acid snow have been recorded. Carbon dioxide in
equilibrium with water produces
weak carbonic acid. Acid deposition (wet or dry) causes soil and water
acidification, resulting in damage to the
aquatic and terrestrial ecosystems, affecting humans, animals, vegetation, and
structures.
The remaining products of combustion in the atmosphere, mainly
carbon dioxide, together with other
so-called greenhouse gases (methane, nitrogen oxides, and chlorofluorocarbons),
result in thermal changes by
absorbing the infrared energy the Earth radiates into the atmosphere and by
reradiating some back to Earth,
causing global temperatures to increase. The effects of temperature increase are
melting of the ice caps, sea
level rise, and climate change, which includes heat waves, droughts, floods,
stronger storms, and more wildfires.
Using the studies by scores of environmental scientists, these stated damages
have been calculated for the
different types of fossil fuels listed (6). Table 1 presents the results for each
type of damage (in 1998 US dollars).
It can be seen that the environmental damage for coal is $14.51/GJ of coal
consumed, for petroleum is $12.52/GJ
of petroleum consumed, for natural gas $8.26/GJ of natural gas consumed, and the
weighted mean damage in
the world is $12.05/GJ of fossil fuel consumption. These damage costs are not
included in the prices of fossil
fuels, but they are paid for by the people directly or indirectly through taxes,
health expenditures, insurance
premiums, and a reduced quality of living. In other words, today’s fossil fuels are
heavily subsidized. If the
respective environmental damages were included in the fossil fuel prices, it would
force earlier introduction of
cleaner fuels, such as hydrogen, with many benefits to the economy and the
environment.
In order to see the worldwide dimensions of the fossil fuel
environmental damage, Table 2 has been
prepared. Note that 37% of the total damage is caused by coal, whereas the coal
consumption is 31% of the
total fossil fuel consumption. On the other hand, only 20% of the damage is caused
by natural gas, which has
a market share of 29%. It is clear that increasing the natural gas
consumption at the expense of coal and
petroleum will be environmentally beneficial. This would also pave the way for
greater public acceptance of
gaseous fuels, which may result in a smoother change to hydrogen, also a gaseous
fuel.
Note also from Table 2 that the annual worldwide environmental damage caused
by fossil fuels is $4345
billion, or 11% of the gross world product. This is a very large figure. Conversion
to a cleaner fuel, such as
hydrogen, would enable the world to save this enormous sum and perhaps use it to
improve the quality of life
worldwide.
New Energy System
Because of the environmental problems fossil fuels are causing, energy researchers
are looking at all possible
sources of energy to replace these fuels. There are quite a number of primary
energy sources, such as ther-
monuclear energy, nuclear breeders, solar energy, wind energy, hydropower,
geothermal energy, ocean currents,
tides, and waves, that are available. At the consumer end, about one-quarter of the
primary energy is used as
electricity and three-quarters is used as fuel. The primary energy sources already
mentioned must therefore
be converted to these energy carriers needed by the consumer. In contrast with
fossil fuels, none of the new
primary energy sources can directly be used as a fuel (e.g., for air and land
transportation). Consequently, they
must be used to manufacture a fuel or fuels as well as to generate electricity.
Comparison of Fuels. Because we need to manufacture a fuel for the
post fossil fuel era, we are in a
position to select the best possible fuel. There are many candidates, such as
synthetic gasoline, synthetic natural
gas (methane), methanol, ethanol, and hydrogen. The fuel of choice must satisfy the
following conditions (7):
----------------------- Page 300-----------------------
• It must be convenient for transportation;

• It must be versatile or convert with ease to other energy forms at the user
end;
• It must have high utilization efficiency; and
• It must be safe to use.
In addition, the resulting energy system must be environmentally compatible

and economical.
----------------------- Page 301-----------------------
Transportation Fuel. Surface vehicles and airplanes must carry their

fuel for a certain distance before
replenishing their fuel supply. In the case of space transportation, space vehicles
must carry the fuel, as well as
the oxidant, necessary for their scheduled range. Therefore, it is important that
the transportation fuel be as
light as possible and also take as little space as possible. We can combine these
requirements in a dimensionless
number, termed the motivity factor (8):
where E is the energy generated by the fuel, M is the mass of the fuel, V is the
volume of the fuel, and the
subscript h refers to hydrogen. The higher the motivity factor, the better the fuel
for transportation. Table 3
lists the pertinent properties of some fuels, as well as the motivity factors
calculated using Eq. (1). It can be
seen that among the liquid fuels LH2 has the best motivity factor, whereas
methanol has the lowest motivity
factor. Among the gaseous fuels, GH2 has the best motivity factor.
Consideration of the utilization efficiency advantage of hydrogen further
improves hydrogen’s standing
as the best transportation fuel. Of course, this is one of the reasons why hydrogen
is the fuel of choice for the
space programs around the world, even though presently it is more expensive than
fossil fuels.
Versatility. At the user’s end, all fuels must be converted through a
process (such as combustion) to other
forms of energy (thermal, mechanical, or electrical energies). If a fuel
can be converted through more than
one process to various forms of energy at the user’s end, it becomes more
versatile and more convenient to
use. Table 4 lists various fuels and processes by which they can be converted to
other forms of energy at the
user’s end. It can be seen that most fossil fuels can be converted through one
process only, that of combustion.
In certain cases, natural gas could be used in catalytic combustion and
in fuel cells, in addition to flame
combustion. Hydrogen, however, can be converted to other forms of energy in five
different ways; in addition
to flame combustion, it can be converted directly to steam, converted to heat
through catalytic combustion,
----------------------- Page 302-----------------------
be used as a heat source and/or heat sink through chemical reactions, and be
converted directly to electricity
through electrochemical processes (9). In other words, hydrogen is the most
versatile fuel.
Utilization Efficiency. In comparing the fuels, it is important to
take into account the utilization effi-
ciencies at the user’s end. For utilization by the user, fuels are
converted to various energy forms, such as
mechanical, electrical and thermal. Studies show that in almost every instance of
utilization, hydrogen can be
converted to the desired energy form more efficiently than other fuels. Table 5
presents the utilization efficiency
factors, defined as the fossil fuel utilization efficiency divided by the hydrogen
utilization efficiency for various
applications. It can be seen that hydrogen is the most efficient fuel. This results
in conservation of resources,
in addition to conserving energy.
Safety. The safety aspects of fuels involve their toxicity on the one hand
and the fire hazard properties
on the other. In addition to the toxicity of their combustion products, the fuels
themselves can be toxic.
Toxicity increases as the carbon-to-hydrogen ratio increases. Hydrogen and
its main combustion product,
water or water vapor, are not toxic. However, NOx , which can be produced through
the flame combustion of
hydrogen (as well as through the combustion of fossil fuels) displays toxic
effects.
----------------------- Page 303-----------------------
Table 6 lists the characteristics of fuels related to fire hazards. Lower

density makes a fuel safer because
it increases the buoyancy force for speedy dispersal of the fuel in case of a leak.
For the same reason, higher
diffusion coefficients are helpful. Higher specific heat causes a fuel to
be safer because it slows down the
temperature increases for a given heat input. Wider ignition limits, lower ignition
energies, and lower ignition
temperatures make the fuels less safe, as they increase the limits within which a
fire could commence. Higher
flame temperatures, explosion energy, and/or flame emissivity make a fuel less safe
as well because its fire
would be more damaging.
Table 7 compares the safety of fuels. For each of the toxic element and fire
hazard characteristics, it ranks
the fuels from 1 to 3, 1 being the safest and 3 the least safe. These rankings have
been summed up for each
fuel in order to arrive at an overall ranking. The total rankings have been
prorated to obtain the safety factors,
defined as the ratio of the total ranking for hydrogen to that of a
given fuel. It can be seen that hydrogen
becomes the safest fuel, whereas gasoline is the least safe, and methane is in
between the two.
----------------------- Page 304-----------------------
The Best Fuel. When we look at the fuel options critically under the
criteria given earlier, it becomes
clear that hydrogen is the best transportation fuel, the most versatile
fuel, the most efficient fuel, and the
safest fuel. In summary, hydrogen is the best fuel.
Hydrogen Energy System. As a result of the aforementioned
observations, it becomes clear that it
would be expedient to manufacture hydrogen using any and all primary energy
sources, in order to make up
for their shortcomings. Such an energy system is then called the hydrogen energy
system.
Figure 2 presents a schematic diagram of the proposed hydrogen energy system.
In this system, hydrogen
(and oxygen) is produced in large industrial plants where the primary
energy source (solar, nuclear, and
even fossil) and water (H2 O), the raw material, are available. For large-scale
storage, hydrogen can be stored
underground in ex-mines, caverns, and/or aquifers. Hydrogen is then
transported, by means of pipelines or
super tankers, to energy consumption centers. It is then used in
electricity, transportation, and industrial,
residential, and commercial sectors as a fuel and/or an energy carrier. The by-
product is water or water vapor.
If flame combustion of hydrogen is used, then some NOx is also produced.
Water and water vapor are recycled
back, through rain, rivers, lakes and oceans, to make up for the water used in the
first place to manufacture
hydrogen.
The oxygen produced in the industrial plant making hydrogen could either be
released into the atmosphere
or be shipped or piped to industrial and city centers for use in fuel cells
(instead of air) for electricity generation.
This would have the advantage of increasing the utilization efficiency. The oxygen
could be used by industry
for nonenergy applications and also for rejuvenating the polluted rivers
and lakes, or speeding up sewage
treatment.
It should be noted that in the hydrogen energy system, hydrogen is not a
primary source of energy. It is
an intermediary or secondary form of energy or an energy carrier. Hydrogen
complements the primary energy
sources and presents them to the consumer in a convenient form at the desired
location and time.
Details of the hydrogen energy system, including production, storage,
transportation, distribution, uti-
lization, environmental impact, and economics, can be found in the proceedings of
the THEME and the World
Hydrogen Energy Conferences (10,11,12,13,14,15,16,17,18,19,20,21,22,23).
----------------------- Page 305-----------------------
Fig. 2. A schematic diagram of the hydrogen energy

system.
Fig. 3. Fossil fuel system.
Fossil Fuel System. A simplified version of today’s energy

system is shown in Fig. 3. Fossil fuels
are used for transportation (mostly petroleum products); for heat generation in
residential, commercial and
industrial sectors; and for electric power generation. For transportation, mostly
petroleum products are used
(gasoline, diesel fuel, jet fuel, etc.). Heat generation includes space and
domestic water heating, cooking, steam
generation and direct heating, and/or drying in various industrial processes. All
three forms of fossil fuels are
used for these purposes. In electric power generation, coal is used mainly for the
base load generation, and
natural gas and heating oil are used for peak load. Part of the electric power is
produced by hydro and nuclear
powers.
In comparing with other energy systems, it can be assumed that 40% of the
primary energy (in fossil fuel
equivalent units) will be used for thermal energy generation, 30% for electric
power generation, and 30% for
transportation (2/3 for surface transportation and 1/3 for air transportation)
(24). Energy supplied by hydro
and nuclear power plants (mostly in the form of electric power) and by other
nonfossil fuel sources do not need
to be taken into account because it is assumed that it will be the same for all
three considered systems. Actually,
it is reasonable to expect that in the future even more electrical energy will be
supplied by these sources.
----------------------- Page 306-----------------------
Fig. 4. Coal/synthetic fossil fuel

system.
When one considers early 2000s, it can be expected that about half of the
thermal energy will be supplied
by natural gas, and the rest will be supplied by petroleum fuels (fuel
oil and residual oil) and coal. Coal is
assumed to be the main energy source for electricity generation, gasoline for
surface transportation, and jet
fuel for air transportation. This is of course a simplified version of the fossil
fuel energy system, but it is close
enough to the present patterns of energy consumption and can be used as a basis for
comparisons.
Coal/Synthetic Fossil Fuel System. Reserves of fossil fuels
are finite, particularly those of oil and
natural gas. Known reserves of oil and natural gas are about 8000 EJ (1 EJ = 1018
J), which would be enough for
the next 40 years at the current consumption rate (25). If the exponential
population growth and the demand
for growth are taken into account, they would only last about 25 years.
Even if the estimated additional
undiscovered resources were added, that would satisfy energy needs for fluid fuels
an additional 30 years or
so. Coal reserves are much larger, known reserves are about 20,000 EJ, but
estimated ultimately recoverable
resources add up to 150,000 EJ. These large amounts of coal could eventually be
used to produce synthetic
liquid fuels, allowing society to continue employing the present energy
system. Such a system is called the
coal/synthetic fossil fuel system because coal is to be used to manufacture
synthetic fossil fuels, as well as to
be directly used for electricity generation.
In this case, it can be assumed that the present fossil fuel system will be
continued by the substitution
with synthetic fuels derived from coal wherever convenient and/or necessary.
Patterns of energy consumption
are also assumed to be unchanged (see Fig. 4). Coal will be used extensively for
thermal as well as electric
power generation because it is much cheaper than synthetic fuels.
However, some end uses require fluid
fuels. Therefore, it has been assumed that synthetic natural gas (SNG) will be used
for some thermal energy
generation (primarily in the residential sector) and also as fuel for surface
transportation, where it will share
the market with synthetic gasoline. Synthetic jet fuel will be used in air
transportation.
Solar Hydrogen Energy System. If solar energy, in its direct
and/or indirect forms (e.g., hydro, wind,
etc.), is used to manufacture hydrogen, then the resulting system is called the
solar hydrogen energy system.
In this system, both the primary and secondary energy sources are renewable and
environmentally compatible,
resulting in a clean and permanent energy system. Figure 5 presents a schematic of
the solar hydrogen energy
system.
In this case, it is assumed that the conversion to hydrogen energy
will take place, and one-third of
the hydrogen needed will be produced from hydropower (and/or wind power)
and two-thirds by direct and
indirect (other than hydropower) from solar energy forms. The same percentage of
energy demands sectors
----------------------- Page 307-----------------------
Fig. 5. Solar-hydrogen energy system.
as the above systems will be assumed. It will further be assumed that one-half of
the thermal energy will be
achieved by flame combustion, one-quarter by steam generation with hydrogen/oxygen
steam generators, and
the last quarter by catalytic combustion; electric power will be generated by fuel
cells; one-half of the surface
transportation will use gaseous hydrogen-burning internal combustion engines, and
the other half will use fuel
cells. In air transportation, both subsonic and supersonic, liquid hydrogen will be
used.
Comparison of Energy Systems
These three energy systems will now be compared environmentally and economically.
Pollution. Table 8 lists the pollutants for the three energy systems
described earlier. It can be seen that
the coal/synthetic fossil system is the worst from the environmental point of view,
whereas the solar hydrogen
energy system is the best. The solar hydrogen system will not produce any CO , CO,
SO , hydrocarbons, or
2 x
particulates, except some NO . However, the solar-hydrogen-produced NO
is much less than that produced
x
x
by the other energy systems. This is due to the fact that in the
solar hydrogen energy system only the
flame combustion of hydrogen in air will generate NOx . The other utilization
processes (such as direct steam
generation, use of hydrogen in fuel cells, hydriding processes, etc.) will not
produce any NOx .
Vapor Generation. There is a notion that the hydrogen energy
system would produce more water
vapor than the other energy systems because the fuel is pure hydrogen. When one
considers the problem in
detail, this is not so.
Only the flame combustion of fuels in air or in oxygen will produce water
vapor. In the case of hydrogen,
those other processes mentioned earlier will not produce any water vapor.
Consequently, contrary to popular
belief, the solar hydrogen energy system will produce less water vapor than the
other systems.
Global warming, which is caused by the utilization of fossil fuels,
also causes an increase in water
vapor generation. Assuming that Earth’s mean temperature has increased by 0.5◦C
since the beginning of the
Industrial Revolution (26), this additional water vapor generation and that
produced by the combustion of fuels
have been calculated. The results are presented in Table 9. It can be seen that (1)
the two fossil fuel systems
----------------------- Page 308-----------------------
generate much more additional (above natural) water vapor than the solar hydrogen
energy system; (2) the
additional water vapor generated by global warming is much greater than that
produced by the combustion
of fuels; (3) the amount of water vapor generated by fuels is minimal compared to
that generated naturally;
and (4) the solar hydrogen energy system causes the smallest increase in vapor
generation. Again, when the
additional vapor generation is considered, the solar hydrogen energy
system becomes environmentally the
most compatible system.
Environmental Damage. Table 10 presents the environmental damage
per gigaJoule of the energy
consumed for each of the three energy systems considered and also for
their fuel components in 1998 US
dollars, as well as the environmental compatibility factors, defined as the ratio of
the environmental damage
caused by the hydrogen energy system to that caused by a given energy system. The
environmental damage
for the solar hydrogen energy system is a result of the NOx produced. It can
be seen that the solar hydrogen
energy system is environmentally the most compatible system.
It should be mentioned that hydrogen also has the answer to the
depletion of the ozone layer, which
is mainly caused by chlorofluorocarbons. Refrigeration and air-conditioning systems
based on the hydriding
property of hydrogen do not need chlorofluorocarbons but need hydrogen, and any
hydrogen leak would not
cause ozone layer depletion. Such refrigeration systems are also very
quiet because they do not have any
moving machinery.
Economics. An economical comparison between competing energy
systems should be based on the
effective costs of the services these fuels provide. The effective costs include
the utilization efficiency, the cost
----------------------- Page 309-----------------------
of the fuel, and the costs associated with fuel consumption but that
are not included in its price (so-called
external costs). External costs include the costs of the physical damage done to
humans, fauna, flora, and the
environment as a result of harmful emissions, oil spills and leaks, and coal strip
mining, as well as governmental
expenditures for pollution abatement and expenditures for military protection of
oil supplies.
In economic considerations, it is also important to compare the future costs
of hydrogen (which will be
considerably lower than they are today because of the assumed market and technology
development) with the
future costs, both internal and external, of fossil fuels (which will unavoidably
be higher than today’s prices as
a result of depletion, international conflicts, and environmental impact).
The effective cost of a fuel can be calculated using the following
relationship:
where C is the internal cost or the conventional cost of the fuel, C

the external cost including the environmental
i e
damage caused by the fuel, η the fossil fuel utilization efficiency for
the application k, and η the synthetic
fk
sk
fuel (including hydrogen) utilization efficiency for the same application or end
use.
In order to evaluate the overall cost to society Co , the three scenarios
considered earlier will be used. This
cost can be calculated from the relationship
where αn is the fraction of energy used by the energy sector n, such as

electricity generating, heat producing,
surface transportation, and subsonic and supersonic air transportation. Because αn
is a fraction, their sum is
----------------------- Page 310-----------------------
Substituting Eq. (2) into Eq. (3), one obtains
Using Eqs. (2)–(5), Tables 11, 12, 13 have been prepared for the three energy
scenarios (i.e., the fossil
fuel system, the coal/synthetic fossil fuel system, and the solar hydrogen energy
system) in 1998 US dollars.
Comparing the results, it becomes clear that the solar hydrogen energy system is
the most cost-effective energy
system and results in the lowest overall effective cost to society.
Gaseous Hydrogen
Hydrogen exists in three isotopes: protium, deuterium, and tritium. A standard

hydrogen atom (protium) is
the simplest of all the elements and consists of one proton and one electron.
Molecular hydrogen (H2 ) exists
in two forms: ortho- and para-hydrogen. Both forms have identical
chemical properties, but because of a
different spin orientation they have somewhat different physical properties. At
room temperature hydrogen
consists of approximately 75% ortho- and 25% para-hydrogen. Because para-hydrogen
is more stable at lower
temperatures, its concentration increases at lower temperatures, reaching virtually
100% at liquid hydrogen
temperatures.
Hydrogen is an odorless, colorless gas. With a molecular weight of 2.016,
hydrogen is the lightest element.
Its density is about 14 times less than air (0.08376 kg/m3 at standard temperature
and pressure). Hydrogen is
liquid at temperatures below 20.3 K (at atmospheric pressure). Hydrogen has the
highest energy content per
----------------------- Page 311-----------------------
unit mass of all fuels—higher heating value is 141.9 MJ/kg, almost three times
higher than gasoline. Some
important properties of hydrogen are compiled in Table 14.
Hydrogen Production. Hydrogen is the most plentiful element
in the universe, making up about
three-quarters of all matter. All the stars and many of the planets essentially
consist of hydrogen. However, on
Earth free hydrogen is scarce. The atmosphere contains trace amounts of it (0.07%),
and it is usually found in
small amounts mixed with natural gas in crustal reservoirs. A few wells, however,
have been found to contain
large amounts of hydrogen, such as some wells in Kansas that contain 40% hydrogen,
60% nitrogen and trace
amounts of hydrocarbons (27). The Earth’s surface contains about 0.14% hydrogen
(the tenth most abundant
element), most of which resides in a chemical combination with oxygen as water.
Hydrogen, therefore, must be produced. Logical sources of hydrogen are
hydrocarbon (fossil) fuels (C H )
x y
and water (H O). Presently, hydrogen is mostly being produced from fossil fuels
(natural gas, oil, and coal).
2
Except for the space program, hydrogen is not being used directly as a fuel
or energy carrier. It is being
used in refineries to upgrade crude oil (hydrotreating and hydrocracking), in the
chemical industry to synthesize
various chemical compounds (such as ammonia and methanol), and in metallurgical
processes (as a reduction
or protection gas). The total annual hydrogen production worldwide in 1996 was
about 40 million tons (5.6 EJ)
(28). Less than 10% of this was supplied by industrial gas companies; the rest is
produced at consumer-owned
and -operated plants (so-called captive production), such as refineries, and ammonia
and methanol producers.
Production of hydrogen as an energy carrier would require an increase in production
rates by several orders of
magnitude.
Hydrogen Production from Fossil Fuels.
Steam Reforming of Natural Gas. Steam reforming of hydrocarbons
(mainly natural gas) has been the
most efficient, economical, and widely used process for hydrogen production. A
simplified basic flow diagram
of a conventional steam reforming process is shown in Fig. 6 (29). The process
basically consists of three main
steps: (1) synthesis gas generation, (2) water–gas shift, and (3) gas purification.
The feedstock (natural gas) is
mixed with process steam and reacted over a nickel-based catalyst contained inside
a system of alloyed steel
tubes. To protect the catalyst, natural gas must be desulfurized before being fed
to the reformer. The following
----------------------- Page 312-----------------------
reactions take place in the reformer:
The reforming reaction is strongly endothermic, and energy is supplied by

combustion of natural gas or
◦ ◦
fuel oil. The metallurgy of the tubes usually limits the reaction temperature to
700 to 925 C. The synthesis
gas leaving a catalytic reformer is typically a mixture of H , CO, CO , and CH .
After the reformer, the gas
2
2 4
mixture passes through a heat recovery step cooling it down to about 350◦C and
feeding it into a water–gas
----------------------- Page 313-----------------------
Fig. 6. Block diagram of hydrogen production by steam

reforming of natural gas.
shift reactor to produce additional hydrogen. Next, the cold gas passes through gas
purification units to remove
the CO , the remaining CO, and other impurities in order to deliver purified
hydrogen. Several commercial
2
processes can be used for removing CO2 (and CO), such as wet
scrubbing, pressure swing adsorption, and,
recently, membrane processes.
The efficiency of the steam reforming process, expressed as the
ratio of the heating value of produced
hydrogen and energy input (feedstock, fuel and small amounts of electricity) is
about 65% to 75%. The cost of
produced hydrogen is about $6/GJ but is strongly dependent upon the cost and
availability of natural gas or
other fuel feedstock.
Partial Oxidation. Partial oxidation is used for converting hydrocarbons
heavier than naphtha (for which
steam reforming is not applicable). Most recently, partial oxidation reformers are
being developed for on-board
reforming of natural gas, ethanol, and even gasoline, to be coupled with fuel cells
for automobile propulsion. A
simplified basic flow diagram of an industrial partial oxidation process is shown in
Fig. 7. There are three main
steps: (1) synthesis gas generation, (2) water–gas shift reaction, and (3) gas
purification. The partial oxidation
reactions are typically as follows:
----------------------- Page 314-----------------------
Fig. 7. Block diagram of hydrogen production by

partial oxidation.
where n = 1 and m = 1.3 for residual fuel oils.

In the synthesis gas generation step, the hydrocarbon feedstock is partially
oxidized with oxygen, and
carbon monoxide is shifted with steam to produce CO and H . Because
of the difficulties of separating nitrogen
2 2
to produce pure hydrogen, pure oxygen is typically used in this process. If
hydrogen is to be used in fuel cells, air
is an acceptable oxidant for partial oxidation because a relatively high nitrogen
content does not significantly
affect the fuel cell performance.
While the hydrocarbon–oxygen reaction is exothermic, additional energy
required for the endothermic
hydrocarbon–steam reaction, which is similar to steam reforming, is supplied by
burning the additional fuel.
◦ ◦
Operating temperatures are rather high, 1150 to 1315 C, but with
novel catalysts it is possible to accomplish
the partial oxidation process at lower temperatures. The product steam (a very hot
mixture of hydrogen, oxygen,
CO, CO , steam, and a small amount of CH ) is typically quenched with water as it
exits the reactor to solidify
2 4
the slag for downstream removal with cyclones or gas filtration or in a
slurry. The sulfur contained in the
feedstock is converted mainly to H2 S and only a small portion to COS. After sulfur
removal, the gas is treated
through the shift reaction and gas purification in the same way as the product gas
from the steam-reforming
process.
The efficiency of partial oxidation is generally lower than for steam
reforming, usually about 50%. Hydro-
gen can be produced for approximately $10/GJ if inexpensive oil (less than $4/GJ)
is available.
Recently, partial oxidation combined with steam reforming has started to be
used for reforming lighter
hydrocarbons and alcohols, such as gasoline, natural gas, propane, methanol, and
ethanol. These reformers
are being developed for small-scale reforming (50 kW and less) for both
transportation and stationary power
applications (30). They typically use air instead of pure oxygen, which results in
a lower hydrogen content in
the reformatted gas. The CO2 removal process is not necessarily applied.
With proper catalyst selection and
thermal integration, those reformers can reach up to 80% efficiency.
Thermal Cracking of Natural Gas. The thermal decomposition of
natural gas has been practiced for
many years for the production of carbon black for rubber tire vulcanization, for
pigment, and for the printing
----------------------- Page 315-----------------------
Fig. 8. Block diagram of hydrogen production by coal

gasification process.
◦
industry. A methane–air flame is used to heat up firebrick to temperatures in the
order of 1400 C. The air is
then turned off, and the methane alone decomposes on the hot firebrick until the
temperature drops to about
800◦C. The micron-size carbon particles are collected in the effluent gas
stream in bag filters. The process
is typically practiced batchwise in tandem furnaces: while one furnace is producing

carbon black, the other
is being heated up. Attempts have also been made to thermally crack natural gas in
a continuous fixed bed
reactor (31). With appropriate catalysts, it is possible to increase the rate of
decomposition and accomplish an
efficient continuous process (32). Another approach is to use high-temperature
plasma for thermal cracking of
natural gas or even heavier hydrocarbons (33).
Because of a valuable by-product (carbon black), the economics of this
process for hydrogen production
look favorable. It should be noted that this is the only process of hydrogen
production from hydrocarbons that
does not produce CO2 (unless carbon black is later used as a fuel).
Coal Gasification. In the coal gasification process (so called Koppers–
Totzek process), the pulverized coal
is rapidly partially oxidized with oxygen and steam at essentially atmospheric
pressure. The raw gas is then
cooled to recover waste heat, followed by quenching with water to remove entrained
ash particles before going
through the steps of compression, shift conversion, and gas purification (again by
conventional wet scrubbing
or pressure swing adsorption processes). The product is hydrogen at about 2.8 MPa
with purity higher than
97.5%. The simplified block diagram is shown in Fig. 8.
In order to avoid the expensive and highly energy-consuming hydrogen
compression step, Texaco has
developed a coal gasification process at elevated pressures, around 5.5 MPa. By
operating in a direct quench
mode at that pressure, a high steam content in the synthesis gas is desirable to
use in the shift reaction for
additional hydrogen production. The raw gas is then desulfurized, shifted, and
purified. The product hydrogen
is at about 4 MPa with a purity higher than 97%.
The coal gasification process is complicated because of the necessity to
handle solid fuel and to remove
large amounts of ash. The solids-handling problem has a significant impact on costs.
Generally, the lower cost of
coal does not compensate for the higher capital cost of coal gasification systems
(compared to steam reforming),
and the cost of the produced hydrogen is about $12 to $14/GJ (29).
Hydrogen Production from Biomass. Hydrogen can be obtained from
biomass by a pyrolysis/gasification
process (34). The biomass preparation step involves heating of the biomass/water
slurry to a high temperature
under pressure in a reactor. This process decomposes and partially
oxidizes the biomass, producing a gas
product consisting of hydrogen, methane, CO2 , CO, and nitrogen. Mineral matter is
removed from the bottom
of the reactor. The gas stream goes to a high-temperature shift reactor where the
hydrogen content is increased.
A relatively high purity hydrogen is produced in the subsequent pressure swing
adsorption unit. The whole
system is very similar to a coal gasification plant, with the exception of the unit
for pretreatment of the biomass
----------------------- Page 316-----------------------
and the design of the reactor. Because of the lower calorific value per unit mass of
biomass as compared to coal,
the processing facility is larger than that of comparably sized coal gasification
plants.
Hydrogen Production from Water. The most logical source for
large-scale hydrogen production is water,
which is abundant on Earth. Different methods of hydrogen production from
water have been or are being
developed. They include
• electrolysis,
• direct thermal decomposition or thermolysis,
• thermochemical processes, and
• photolysis.
Electrolysis. Electrolysis appears to be the only method developed

to date that can be used for large-
scale hydrogen production in a post-fossil fuel era. Production of hydrogen by
water electrolysis is a 50-year old
technology, as based on a fundamentally simple process, is very efficient, and does
not involve moving parts.
The following reactions take place at the electrodes of an electrolysis cell
filled with a suitable electrolyte
(aqueous solution of KOH or NaOH or NaCl) upon the application of a potential:
The reversible decomposition potential (Erev = DG/nF) of this reaction is
1.229 V at standard conditions.
The total theoretical water decomposition potential is 1.480 V corresponding to
hydrogen’s enthalpy (because
DH = DG + TDS). Because of irreversible processes occurring at the anode and
cathode, including the electrical
resistance of the cell, the actual potentials are always higher,
typically between 1.75 V and 2.05 V. This
corresponds to efficiencies of 72% to 82%, respectively.
Several advanced electrolyzer technologies are being developed. List includes
the following:
• Advanced alkaline electrolysis, which employs new materials for

membranes and electrodes that allow
further improvement in efficiency—up to 90% (35,36).
• Solid polymer electrolytic (SPE) process, which employs a proton-conducting
ion exchange membrane as
electrolyte and as a membrane that separates the electrolysis cell. This type
of electrolyzers can operate
at very high current densities (up to 2 A/cm2 , which is about one order of
magnitude higher than stan-
dard electrolyzers with alkaline liquid electrolyte). The water to be

dissociated does not require dissolved
electrolytes to increase its conductivity and is added solely to the anode
side (36,37).
◦ ◦
• High temperature steam electrolysis, which operates between 700
and 1000 C and which employs oxygen
ion-conducting ceramics as electrolyte. Electrical energy consumption is
reduced because part of the energy
required for water dissociation is supplied in the form of heat. The water to
be dissociated is entered on the
cathode side as steam which forms a steam–hydrogen mixture during electrolytic
dissociation. The O2 -ions
are transported through the ceramic material to the anode, where they are
discharged as oxygen (38).
----------------------- Page 317-----------------------
An electrolysis plant can operate over a wide range of capacity factors and
is convenient for a wide range
of operating capacities, which makes this process interesting for coupling
with renewable energy sources,
particularly with photovoltaics. Photovoltaics generate low voltage–direct
current, which is exactly what is
required for the electrolysis process.
Performance of photovoltaic–electrolyzer systems has been studied
extensively both in theory and in
practice (39,40,41,42). Several experimental PV-electrolysis plants are currently
operating all over the world,
such as
• Solar-Wasserstoff-Bayern pilot plant in Neunburg vorm Wald in Germany (43),

• HYSOLAR project in Saudi Arabia (44),
• Schatz Energy Center, Humboldt State University, Arcata, California (45),
• Helsinki University of Technology, Helsinki, Finland (46),
• INTA Energy Laboratory, Huelva, Spain (47).
Direct Thermal Decomposition of Water (Thermolysis). Water can be

split thermally at temperatures
above 2000 K. The overall thermal dissociation of water can be shown as (48):
The degree of dissociation is a function of temperature: only 1% at 2000 K, 8.5% at

2500 K, and 34% at 3000
K. The product is a mixture of gases at extremely high temperatures. The main
problems in connection with
this method are related to materials required for extremely high temperatures,
recombination of the reaction
products at high temperatures, and separation of hydrogen from the mixture.
Thermochemical Cycles. Thermochemical production of hydrogen involves
the chemical splitting of water
at temperatures lower than those needed for thermolysis, through a series of
cyclical chemical reactions which
ultimately release hydrogen. Since the mid-1960s, research has been performed to
investigate a number of po-
tential thermochemical cycles for the production of hydrogen, and some 2000 to 3000
cycles have been invented.
After examining their practicability in terms of reaction and process technology,
only 20 to 30 remained appli-
cable for large-scale hydrogen production. Some of the more thoroughly investigated
thermochemical process
cycles follow (37,49,50):
sulfuric acid–iodine cycle,

hybrid sulfuric acid cycle,
hybrid sulfuric acid–hydrogen bromide cycle,
calcium bromide–iron oxide cycle (UT-3), and
iron chlorine cycle (Mark 9).
Depending on the temperatures at which these processes are occurring,

relatively high efficiencies are
achievable (40% to 50%). However, the problems related to movement of a large mass
of materials in chemical
reactions, toxicity of some of the chemicals involved, and corrosion at high
temperatures remain to be solved
in order for these methods to become practical.
Photolysis. Photolysis, or direct extraction of hydrogen from water
using only sunlight as an energy
source, can be accomplished by photobiological systems, photochemical
assemblies, or photoelectrochemical
cells (51,52). Intensive research activities are opening new perspectives for
photoconversion, where new redox
catalysts, colloidal semiconductors, immobilized enzymes, and selected
microorganisms could provide means
of large-scale solar energy harvesting and conversion into hydrogen.
----------------------- Page 318-----------------------
Gaseous Hydrogen Storage and Distribution. Hydrogen as

an energy carrier must be stored to
overcome daily and seasonal discrepancies between energy source
availability and demand. Also, to over-
come geographical discrepancies between hydrogen production sites and hydrogen
users, hydrogen has to be
transported.
Gaseous Hydrogen Storage. Hydrogen can be stored either as a
pressurized gas or as a liquid. It also
can be stored in chemical or physical combinations with other materials,
such as metal hydrides, chemical
hydrides, glass microspheres, and cryo-absorbers. Depending on storage size and
application, several types of
hydrogen storage systems may be differentiated:
(1) Stationary Large Storage Systems These are typically storage devices at the
production site or at the start
or end of pipelines and other transportation pathways.
(2) Stationary Small Storage Systems These can be found at the distribution or
final user level, for example, a
storage system to meet the demand of an industrial plant.
(3) Mobile Storage Systems for Transport and Distribution These include
both large-capacity devices, such
as a liquid hydrogen tanker–bulk carrier, and small systems, such as a gaseous
or liquid hydrogen truck
trailer.
(4) Vehicle Tanks These store hydrogen used as fuel for road vehicles.
Because of hydrogen’s low density, its storage always requires relatively

large volumes and is associated
with either high pressures (thus requiring heavy vessels) or extremely low
temperatures and/or a combination
with other materials (much heavier than hydrogen itself). Table 15 shows achievable
storage densities with
different types of hydrogen storage. Some novel hydrogen storage method may achieve
even higher storage
densities, but it has yet to be proven in terms of practicality, cost, and safety.
Large Underground Hydrogen Storage. Future hydrogen supply systems
will have a structure similar to
today’s natural gas supply systems. Underground storage of hydrogen in caverns,
aquifers, depleted petroleum
and natural gas fields, and human-made caverns resulting from mining and
other activities is likely to be
technologically and economically feasible (53). Hydrogen storage systems of the
same type and energy content
will be more expensive by approximately a factor of 3 than natural gas storage
systems because of hydrogen’s
lower volumetric heating value. Technical problems, specifically for
underground storage of hydrogen other
----------------------- Page 319-----------------------
than expected losses of the working gas in the amount of 1% to 3% per year are not
anticipated. The city of
Kiel’s public utility has been storing town gas with a hydrogen content of 60% to
65% in a gas cavern with a
geometric volume of about 32,000 m3 and a pressure of 80 to 160 bar at a
depth of 1330 m since 1971 (54).
Gaz de France, the French National Gas Company, has stored hydrogen-rich refinery
by-product gases in an
aquifer structure near Beynes, France. Imperial Chemical Industries of Great
Britain stores its hydrogen in
the salt mine caverns near Teeside in the United Kingdom (55).
Above-Ground Pressurized Gas Storage Systems. Pressurized gas
storage systems are used today in
natural gas businesses in various sizes and pressure ranges from standard pressure
cylinders (0.05 m3 , 200
3
bar) to stationary high-pressure containers (over 200 bar) or low-pressure
spherical containers ( >30,000 m ,
12 bar to 16 bar). This application range will be similar for hydrogen storage.
Vehicular Pressurized Hydrogen Tanks. Development of ultralight
but strong new composite materials
has enabled storage of hydrogen in automobiles. Pressure vessels that allow
hydrogen storage at pressures
>200 bar have been developed and used in automobiles (such as Daimler-Benz NECAR
II). Storage density of
higher than 0.05 kg H2/kg of total weight is easily achievable (56).
Metal Hydride Storage. Hydrogen can form metal hydrides with
some metals and alloys. During the
formation of the metal hydride, hydrogen molecules are split, and hydrogen atoms
are inserted in spaces inside
the lattice of suitable metals and/or alloys. In such a way, an effective storage
is created comparable to the
density of liquid hydrogen. However, when the mass of the metal or alloy is taken
into account, then the metal
hydride gravimetric storage density is comparable to storage of
pressurized hydrogen. The best achievable
gravimetric storage density is about 0.07 kg H2/kg of metal, for a high
temperature hydride such as MgH2
as shown in Table 16, which gives a comparison of some hydriding substances with
liquid hydrogen, gaseous
hydrogen, and gasoline (57).
----------------------- Page 320-----------------------
During the storage process (charging or absorption), heat is released. This

heat must be removed in order
to achieve the continuity of the reaction. During the hydrogen release process
(discharging or desorption), heat
must be supplied to the storage tank.
An advantage of storing hydrogen in hydriding substances is the safety
aspect. A serious damage to a
hydride tank (such as the one that could be caused by a collision) would not pose a
fire hazard because hydrogen
would remain in the metal structure.
Novel Hydrogen Storage Methods. Hydrogen can be physically
adsorbed on activated carbon and be
“packed” on the surface and inside the carbon structure more densely
than if it has been just compressed.
Amounts of up to 48 g H2/kg of carbon have been reported at 6.0 MPa and 87 K (58).
The adsorption capacity is
a function of pressure and temperature; therefore, at higher pressures and/or lower
temperatures even larger
amounts of hydrogen can be adsorbed. For any practical use, relatively low
temperatures are needed (<100
K). Because adsorption is a surface process, the adsorption capacity of hydrogen on
activated carbon is largely
the result of the high surface area of the activated carbon, although some other
carbon properties affect the
capability of activated carbon to adsorb hydrogen.
Researchers from Northeastern University in Boston, MA, have recently
announced that they have de-
veloped a carbon storage material that can store as high as 75% of hydrogen by
weight (59). This material,
apparently some kind of carbon nanotubes or carbon whiskers, is currently being
researched in several labora-
tories. The best results achieved with carbon nanotubes to date confirmed by the
National Renewable Energy
Laboratory is hydrogen storage density corresponding to about 10% of the nanotube
weight (60).
Hydrogen can be stored in glass microspheres of approximately 50 mm diameter.
The microspheres can
be filled with hydrogen by heating them to increase the glass permeability to
hydrogen. At room temperature,
a pressure of approximately 25 MPa is achieved, resulting in a storage density of
14% mass fraction and 10 kg
H /m3 (61). At 62 MPa, a bed of glass microspheres can store 20 kg H /m3 . The
release of hydrogen occurs by
2
2
reheating the spheres to again increase the permeability.
Researchers at the University of Hawaii are investigating hydrogen storage
via polyhydride complexes.
Complexes that catalyze the reversible hydrogenation of unsaturated
hydrocarbons have been found. This
catalytic reaction could be the basis for a low-temperature hydrogen storage system
with an available hydrogen
density greater than 7% (62).
Hydrogen Transport and Distribution.
Interregional Hydrogen Transport. In the hydrogen energy
system, it is envisaged that, from the
production plants and/or storage, hydrogen will be transmitted to
consumers by means of underground
pipelines (gaseous hydrogen) and/or supertankers (liquid hydrogen).
Presently, hydrogen transportation
through pipelines is used either in links between nearby production and
utilization sites (up to 10 km) or
in more extensive networks (roughly 200 km). Table 17 lists the principal
existing hydrogen pipelines (63).
Future developments will certainly entail greater flow rates and distances.
It would be possible to use the
existing natural gas pipelines with some modifications. For hydrogen pipelines, it
is necessary to use steels
less prone to embrittlement by hydrogen under pressure [particularly for very pure
hydrogen (>99.5% purity)].
Reciprocating compressors used for natural gas can be used for hydrogen without
major design modifications.
However, special attention must be given to sealing (to avoid hydrogen leaks) and
to materials selection for the
parts subject to fatigue stress. Use of centrifugal compressors for hydrogen
creates more problems because of
hydrogen’s exceptional lightness.
As a rule, hydrogen transmission through pipelines requires larger diameter
piping and more compression
power than natural gas for the same energy throughput. However, because of lower
pressure losses in the case
of hydrogen, the recompression stations would need to be spaced twice as far apart.
In economic terms, most
of the studies found that the cost of large-scale transmission of
hydrogen is about 1.5 to 1.8 times that of
natural gas transmission. However, transportation of hydrogen over distances
greater than 1000 km is more
economical than transmission of electricity (64).
----------------------- Page 321-----------------------
Regional Hydrogen Transport and Distribution. To match the

consumption demand, hydrogen can be
regionally transported and distributed, both as a gas and as a liquid, by pipelines
or in special cases in containers
by road and rail transportation. Gaseous (and liquid) hydrogen carriage is subject
to strict regulations ensuring
public safety, which in some countries is very constraining. The transportation of
hydrogen in a discontinuous
mode, whether in gaseous or liquid state, is currently used by
occasional or low-volume users. The cost of
discontinuous transport is very high (it can be as high as two to five times the
production cost). In the future
energy system, discontinuous transportation of hydrogen would see little use,
except for special users (mainly
non-energy-related).
Hydrogen in the gas phase is generally transported in pressurized cylindrical
vessels (typically at 200
bar) arranged in frames adapted to road transport. The unit capacity of these
frames or skids can be as great as
3000 m3 . Hydrogen gas distribution companies also install such frames at the
user’s site to serve as stationary
storage.
Hydrogen Conversion Technologies. Hydrogen as an energy
carrier can be converted in useful
forms of energy in several ways, namely:
• combustion in internal combustion and jet and rocket engines,

• combustion with pure oxygen to generate steam,
• catalytic combustion to generate heat,
• electrochemical conversion to electricity, and
• metal hydrides conversions.
Hydrogen Combustion in Internal Combustion and Jet and Rocket Engines.

Hydrogen is a very good
fuel for internal combustion engines. Hydrogen-powered internal combustion
engines are on average about
20% more efficient than comparable gasoline engines. The ideal thermal efficiency of
an internal combustion
----------------------- Page 322-----------------------

engine is
where
r = compression ratio and
k = ratio of specific heats (C /C ).
p v
Equation (16) shows that the thermal efficiency can be improved by increasing
either the compression
ratio or the specific heat ratio. In hydrogen engines, both ratios are higher than
in a comparable gasoline engine
because of hydrogen’s lower self-ignition temperature and ability to burn in lean
mixtures. However, the use
of hydrogen in internal combustion engines results in the loss of power because of
the lower energy content
in a stoichiometric mixture in the engine’s cylinder. A stoichiometric mixture of
gasoline and air and gaseous
hydrogen and air premixed externally occupy ∼2% and 30% of the cylinder volume,
respectively. Under these
conditions, the energy of the hydrogen mixture is only 85% of the gasoline mixture,
thus resulting in about 15%
reduction in power. Therefore, the same engine running on hydrogen will have ∼15%
less power than when
operated with gasoline. The power output of a hydrogen engine can be improved by
using more advanced fuel
injection techniques or liquid hydrogen. For example, if liquid hydrogen is
premixed with air, the amount of
hydrogen that can be introduced in the combustion cylinder can be increased by
approximately one-third (65).
One of the most important advantages of hydrogen as a fuel for
internal combustion engines is that
hydrogen engines emit by far fewer pollutants than comparable gasoline engines.
Basically, the only products
of hydrogen combustion in air are water vapor and small amounts of nitrogen oxides.
Hydrogen has a wide
flammability range in air (5% to 75% vol.), therefore, high excess air
can be utilized more effectively. The
formation of nitrogen oxides in hydrogen/air combustion can be minimized
with excess air. NOx emissions
can also be lowered by cooling the combustion environment using techniques such as
water injection, exhaust
gas recirculation, or using liquid hydrogen. The emissions of NOx
in hydrogen engines are typically one
order of magnitude smaller than emissions from comparable gasoline
engines. Small amounts of unburned
hydrocarbons, CO2 , and CO have been detected in hydrogen engines due to
lubrication oil (65).
The low-ignition energy and fast-flame propagation of hydrogen have led to
problems of preignition and
backfire. These problems have been overcome by adding hydrogen to the air
mixture at the point where
and when the conditions for preignition are less likely, such as delivering the
fuel and air separately to the
combustion chamber and/or injecting hydrogen under pressure into the combustion
chamber before the piston
is at the top dead center and after the intake air valve has been
closed. Water injection and exhaust gas
recirculation techniques are also used in hydrogen engines to help control
premature ignition. Note that most
of the research on hydrogen combustion in internal combustion engines has been
conducted with modifications
of existing engines designed to burn gasoline. Redesign of the combustion
chamber and coolant systems to
accommodate hydrogen’s unique combustion properties could be the most
effective method of solving the
problems of preignition and knocking (65).
Hydrogen use in turbines and jet engines is similar to the use of
conventional jet fuel. The use of hydrogen
avoids the problems of sediments and corrosion on turbine blades which prolongs
life and reduces maintenance.
Gas inlet temperatures can be pushed beyond normal gas turbine temperatures of
800◦C, thus increasing the
overall efficiency. The only pollutants from the use of hydrogen in turbines and jet
engines are nitrogen oxides.
Steam Generation by Hydrogen/Oxygen Combustion. Hydrogen
combusted with pure oxygen results
in pure steam; that is,
----------------------- Page 323-----------------------
Fig. 9. Schematic representation of catalytic

burner.
◦
This reaction would develop temperatures in the flame zone above 3000 C; therefore,
additional water has to
be injected so that the steam temperature can be regulated at a desired level. Both
saturated and superheated
vapor can be produced.
The German Aerospace Research Establishment (DLR ) has developed a compact
hydrogen/oxygen steam
generator (66). The steam generator consists of the ignition, combustion, and
evaporation chambers. In the
ignition chamber, a combustible mixture of hydrogen and oxygen at a low
oxidant/fuel ratio is ignited by means
of a spark plug. The rest of the oxygen is added in the combustion chamber to
adjust the oxidant/fuel ratio
exactly to the stoichiometric one. Water is also injected in the combustion chamber
after it has passed through
the double walls of the combustion chamber. The evaporation chamber serves to
homogenize the steam. The
steam temperature is monitored and controlled. Such a device is close to 100%
efficient because there are no
emissions other than steam and little or no thermal losses.
The hydrogen steam generator can be used to generate steam for spinning
reserve in power plants, for
peak load electricity generation, in industrial steam supply networks, and
as a micro steam generator in
medical technology and biotechnology applications (66).
Catalytic Combustion of Hydrogen. Hydrogen and oxygen in the
presence of a suitable catalyst may be
combined at temperatures significantly lower than flame combustion (from ambient to
500◦C). This principle
can be used to design catalytic burners and heaters. Catalytic burners

require considerably more surface
area than conventional flame burners. Therefore, the catalyst is typically
dispersed in a porous structure.
The reaction rate and resulting temperature are easily controlled by controlling
the hydrogen flow rate. The
reaction takes place in a reaction zone of the porous catalytic sintered
metal cylinders or plates in which
hydrogen and oxygen are mixed by diffusion from opposite sides. A combustible
mixture is formed only in the
reaction zone and assisted with a (platinum) catalyst to burn at low temperatures
(Fig. 9). The only product of
catalytic combustion of hydrogen is water vapor. Because of low temperatures, no
nitrogen oxides are formed.
The reaction cannot migrate into the hydrogen supply because no flame and hydrogen
concentration is above
the higher flammable limit (75%).
Possible applications of catalytic burners are in household appliances such
as cooking ranges and space
heaters. The same principle is also used in hydrogen sensors.
Electrochemical Conversion (Fuel Cells). Hydrogen can be
combined with oxygen without combustion
in an electrochemical reaction (reverse of electrolysis) and produce electricity
(dc). The device where such a
reaction takes place is called the electrochemical fuel cell or just fuel cell.
Depending on the type of the electrolyte used, there are several types of
fuel cells:
• Alkaline fuel cells (AFC ) use concentrated (85 wt %) KOH as the electrolyte
for high-temperature opera-
◦
◦
tion (250 C) and less concentrated (35 wt % to 50 wt %) for lower-temperature
operation (<120 C). The
----------------------- Page 324-----------------------
electrolyte is retained in a matrix (usually asbestos), and a wide range of

electrocatalysts can be used (such
as Ni, Ag, metal oxides, and noble metals). This fuel cell is intolerant to
CO2 present in either the fuel or
oxidant (67).
• Polymer electrolyte membrane or proton exchange membrane fuel cells
(PEMFC ) use a thin polymer
membrane (such as perfluorosulfonated acid polymer) as the electrolyte.
Membranes as thin as 12 µm to
20 µm have been developed; they are excellent proton conductors. The catalyst
is typically platinum with
loadings about 0.3 mg/cm2 , or, if the hydrogen feed contains minute amounts
of CO, Pt–Ru alloys are used.
◦
◦ ◦
Operating temperatures are usually below 100 C, more typically between 60
and 80 C.
• Phosphoric acid fuel cells (PAFC ), use concentrated phosphoric acid (∼100%)
as the electrolyte. The matrix
used to retain the acid is usually SiC, and the electrocatalyst in both the
anode and cathode is Platinum
◦
◦
black. Operating temperatures are typically between 150 C and 220 C (67,68).
• Molten carbonate fuel cells (MCFC ) have the electrolyte composed of a

combination of alkali (Li, Na, K)
◦
carbonates, which are retained in a ceramic matrix of LiAlO2 . Operating
temperatures are between 600 C
and 700◦C where the carbonates form a highly conductive molten salt, with
carbonate ions providing ionic
conduction. At such high operating temperatures, noble metal catalysts are

typically not required (67,68).
• Solid oxide fuel cells (SOFC) use a solid, nonporous metal oxide,
usually Y O -stabilized ZrO as the
2 3 2
◦ ◦
electrolyte. The cell operates at 900 to 1000 C where ionic
conduction by oxygen ions takes place (67,68).
A typical fuel cell consists of the electrolyte in contact with porous

electrodes on both sides. A schematic
representation of a fuel cell with reactant and product gases, along
with ions flow directions for the major
types of fuel cells are shown in Fig. 10. The electrochemical reactions
occur at the three-phase interface—
porous electrode/electrolyte/reactants. The actual electrochemical reactions that
occur in these types of fuel
cells are different (as shown in Table 18), although the overall reaction is the
same (i.e., H + 1/2O → H O).
2 2 2
Low-temperature fuel cells (AFC, PEMFC, PAFC) require noble electrocatalysts to
achieve practical reaction
rates at the anode and cathode. High-temperature fuel cells (MCFC and SOFC) can
also utilize CO and CH4 as
fuels. The operating temperature is high enough so that CO and CH4 can be
converted into hydrogen through
the water–gas shift and steam reforming reactions, respectively.
The electrolyte not only transports dissolved reactants to the electrode,
but it also conducts ionic charge
between the electrodes and thereby completes the cell electric circuit, as shown in
Fig. 10.
----------------------- Page 325-----------------------
Fig. 10. Operating principle of various types

of fuel cells.
◦
The reversible potential of the above electrochemical reactions is 1.229 V
(at standard conditions, i.e., 25 C
and atmospheric pressure), and it corresponds to the Gibbs free energy according to
the following equation:
where
0 ◦
DG = Gibbs free energy at 25 C and atmospheric pressure
n = number of electrons involved in the reaction
F = Faraday’s constant
0 ◦
E = reversible potential at 25 C and atmospheric pressure (V)
The reversible potential changes with temperature and pressure; in general,

it is lower at higher temper-
atures (reaching ∼1.0 V at 1000 K), and it is higher at higher pressures or higher
concentrations of reactants.
The actual voltage of an operational fuel cell is always lower than the
reversible potential resulting from
various irreversible losses, such as activation polarization, concentration
polarization, and ohmic resistance.
Even though ohmic resistance is directly proportional to the current, activation
polarization is a logarithmic
function of current, and thus more pronounced at very low current densities, and
concentration polarization
is an exponential function of current and thus becomes a limiting factor at high
current densities. Figure 11
----------------------- Page 326-----------------------
Fig. 11. Typical fuel cell

polarization curve.
Fig. 12. Polarization curves of some representative fuel cells. AFC—

United Technologies’ fuel cell operating in Space
◦ ◦
Shuttle, with H2 and O2 at 80 to 90 C, 410 kPa. PEMFC—Ballard’s 25 kW fuel
cell operating in Daimler-Benz vehicle
(NECAR II) with H2 and air (1996). PAFC—IFC’s fuel cell operating in ONSI PC25
200 kW power plant, operating on re-
formed natural gas and air at atmospheric pressure (1996) (dashed line in IFC’s
PC23 operating at 820 kPa). MCFC—ERC’s
atmospheric pressure fuel cell operating with natural gas (1994). SOFC—Westinghouse
fuel cell operating at atmospheric
pressure using natural gas (1991).
shows a typical fuel cell polarization curve with pronounced regions of predominant
irreversible losses. Figure
12 shows actual polarization curves of some representative fuel cells. The fuel
cells are typically operated in a
range between 0.6 V and 0.8 V. The Space Shuttle fuel cell (alkaline) is designed
to operate at 0.86 V and 410
mA/cm2 (67). PEM fuel cells have the highest achievable current densities,
between 1 and 2 mA/cm2 at 0.6 V
with pressurized hydrogen and air.

The fuel cell efficiency is a function of cell voltage. The theoretical fuel
cell efficiency is
----------------------- Page 327-----------------------
where DH is hydrogen’s enthalpy or heating value (higher or lower). The theoretical

fuel cell efficiency, defined
as a ratio between produced electricity and the higher heating value of hydrogen
consumed, is therefore 83%.
The lower heating value of hydrogen results in an efficiency of 98%. Because the
actual voltage of an operational
fuel cell is lower than the reversible potential, the fuel cell efficiency is always
lower than the theoretical one.
Generally, the fuel cell efficiency is a product of several efficiencies:
where
hTh = thermal efficiency, which is the ratio between Gibbs free energy of
the reaction and the heating
value of the fuel, DGr/DHfuel (similar to internal combustion engines, the
fuel cell efficiency is often
expressed in terms of the lower heating value);
hv = voltage efficiency, defined as a ratio between the actual voltage (V )
and thermodynamic voltage (E )
(i.e., V/E );
hF = Faradaic efficiency, or the ratio between the actual current and the
current corresponding to the
rate at which the reactant species are consumed [I/nFm, where m is the rate (in
moles/s) at which the
reactants are consumed];
hU = fuel utilization, or the ratio between the amount of fuel actually
consumed in the electrochemical
reaction and fuel supplied to the fuel cell.
For a hydrogen/oxygen or hydrogen/air fuel cell operating with 100% fuel

utilization, the efficiency is a
function of cell voltage only. For such a fuel cell, the efficiency in an operating
range between 0.6 V and 0.8 V
is between 0.48 and 0.64.
In order to get usable voltages (i.e., tens or hundreds of volts),
the cells are combined in a stack. The
cells are physically separated from each other and electrically connected in series
by a bipolar separator plate.
Figure 13 shows a schematic representation of a typical fuel cell stack.
Alkaline fuel cells have been used in the space program (Apollo
and Space Shuttle) since the 1960s.
Phosphoric acid fuel cells are already commercially available in container packages
for stationary electricity
generation. PEM fuel cells are a serious candidate for automotive applications, but
also for small-scale dis-
tributed stationary power generation. High-temperature fuel cells, such as molten
carbonate and solid oxide
fuel cells, have been developed to a precommercial/demonstration stage for
stationary power generation.
Energy Conversions Involving Metal Hydrides. Hydrogen’s
property to form metal hydrides may be
used not only for hydrogen storage (as described in the section entitled
“Metal Hydride Storage”) but also
for various energy conversions. When a hydride is formed by the chemical
combination of hydrogen with a
metal, an element, or an alloy, heat is generated (i.e., the process is
exothermic). Conversely, in order to release
hydrogen from a metal hydride, heat must be supplied. These processes can be
represented by the following
chemical reactions:
where M represents the hydriding substance, a metal, an element, or an

alloy. The rate of these reactions
increases with an increase in the surface area. Therefore, in general,
the hydriding substances are used in
powdered form to speed up the reactions.
----------------------- Page 328-----------------------
Fig. 13. Schematic representation of a typical fuel

cell stack.
Elements or metals with unfilled shells or subshells are suitable hydriding

substances. Metal and hydro-
gen atoms form chemical compounds by sharing their electrons in the unfilled
subshells of the metal atom and
the K shells of the hydrogen atoms.
Ideally, for a given temperature, the charging or absorption process and the
discharging or desorption
process take place at the same constant pressure. However, there is a hysteresis
effect and the pressure is not
absolutely constant; for a given temperature, charging pressures are higher than
the discharging pressures.
The heat generated during the charging process and the heat needed for
discharging are functions of the
hydriding substance, the hydrogen pressure, and temperature at which the
heat is supplied or extracted.
Using different metals and by forming different alloys, different hydriding
characteristics can be obtained. In
other words, it is possible to make or find hydriding substances that are more
suitable for a given application,
such as waste heat storage, electricity generation, pumping, hydrogen purification,
and isotope separation.
Heat Storage. Any heat, such as waste heat, solar heat, and geothermal
heat, can be stored as potential
hydrogen hydriding energy by supplying the heat to a hydride as shown in the
following relationship:
Whenever heat is needed, hydrogen released can be supplied back to the

hydride, releasing the hydriding
reaction heat; namely,
If hydrogen is supplied to the hydriding substance at the same pressure

as it was released, then the heat
released will be at the same temperature (or slightly lower as a result of the
hysteresis effect) as that of the
----------------------- Page 329-----------------------
heat supplied. However, by increasing the pressure of the hydrogen

supplied, the temperature of the heat
released can be increased; and conversely, by reducing the hydrogen
pressure, the temperature of the heat
released can be reduced. This means that metal hydrides can be used as heat pumps.
Electricity Storage. Hydriding substances can be used for electricity
storage in two ways. In one of the
methods, electricity (direct current) is used to electrolyze the water, and the
hydrogen produced is stored in a
hydriding substance. When electricity is needed, hydrogen is released from the
hydriding substance by adding
heat and using it in a fuel cell to produce direct current electricity. Heat from
the fuel cell can be used to release
hydrogen from the metal hydride. In the second method, one electrode is covered
with a hydriding substance
(e.g., titanium nickel alloy). During the electrolysis of water, hydrogen
produced on the electrode surface is
immediately absorbed by the hydriding substance covering the electrode. Then, when
electricity is needed, the
electrolyzer operates in a reverse mode as a fuel cell producing electricity using
the hydrogen released from
the metal hydride.
Heating/Cooling. Hydrogen together with hydriding
substances can be used for heating or air-
conditioning the buildings. Figure 14 shows how one of the proposed
systems works. The system consists
of four hydride tanks, a solar collector (or a heat source), and a number of heat
exchangers. Hydride tank 1 is
connected to hydride tank 3 with a hydrogen pipe in order to allow the movement of
hydrogen from one tank to
the other. Similarly, hydride tanks 2 and 4 are connected in the same fashion.
Tanks 1 and 2 contain the same
hydriding substance (CaNi ), and tanks 3 and 4 contain another hydriding material
(LaNi ). Heat exchangers
5
5
and the hydride tanks are connected by water-carrying pipe circuits or loops,
equipped with a set of switches
and valves, so that a hydride tank in a given water loop can be replaced by another
hydride tank.
When the system works as a heater, the heat from a heat source (e.g., solar
collector) is carried to tank
1 at about 100◦C. The heat drives the hydrogen from tank 1 to tank 3, where
hydrogen is absorbed forming
a hydride, and heat is released at 40◦C. The water loop carries this heat to the
building heat exchangers and
heats the air in the building. At the same time, water in the other loop absorbs
heat from the ambient and
carries it to tank 4. This heat drives off the hydrogen from tank 4 to tank 2 where
hydride is formed and heat
is generated at 40◦C. The whole operation of driving hydrogen from tanks 1 and 4 to
tanks 3 and 2 takes about
2 minutes. At the end of this cycle, the hydride tanks are switched from one loop
to the other in cycle II (as
shown in Fig. 14). Now, the solar and ambient heats are used to drive off the
hydrogen in tanks 2 and 3 to tanks
4 and 1, respectively. The heat produced during the absorption processes in tanks 1
and 4 is used for heating
the building. After this, the cycles are repeated.
When the system works as an air conditioner, the building heat exchangers are
placed in the 8◦C water
loop, while the outside heat exchangers are placed in the 40◦C water loops, and the
operation proceeds in two
cycles as described earlier.

Electricity Generation. If passed through a turbine or expansion
engine, hydrogen moving from one
hydride tank to another could produce mechanical and electrical energy,
as shown in Fig. 15. The system
is somewhat similar to the one proposed for heating and cooling.
However, it consists of only three tanks
containing the same kind of hydriding substance (in this case LaNi5
alloy). During the first cycle, hydrogen
driven off from the desorption tank (tank 1) by means of solar heat
(or heat from any other source) passes
through the expansion turbine producing mechanical energy and electricity and then
at a lower pressure is
absorbed by the hydriding substance in tank 2 producing heat at 40◦C. In this case,
the heat is produced at a
lower temperature than the temperature of desorption because hydrogen is at a lower

pressure after passing
through the turbine. The heat produced in the absorption tank (tank 2) is rejected
to the environment through
the water cooling system. The same water cooling system is also used to
cool down the cooling tank (tank
◦ ◦
3) from 100 to 40 C because it has served as the desorption tank in the
previous cycle. In the second cycle
through a system of switches and valves, the tanks are displaced one step to the
right in the diagram (i.e., the
cooling tank becomes the absorption tank, the absorption tank becomes the
desorption tank, and the desorption
tank becomes the cooling tank). Then, the cycles are repeated. Using this method,
low-quality heat could be
converted to electricity.
----------------------- Page 330-----------------------
Fig. 14. Hydrogen/hydride heating–cooling system.
Pumping or Pressurizing. The relationship between the hydrogen pressure (p

) and the hydride temper-
ature (T) in the absorption or desorption processes is given by the following
equation:
----------------------- Page 331-----------------------

Fig. 15. Electricity generation via hydrogen and

hydrides.
where the numerator in the exponent term is the hydriding reaction enthalpy, R is
the gas constant, and S is
a constant depending on the hydriding substance. If hysteresis is neglected, Eq.
(25) would hold for both the
absorption and desorption processes. By using heat, it would be possible to
increase hydrogen pressure by a
hundred or more times. The pressure ratio would be
where PHi is the hydrogen pressure corresponding to the hydride temperature

THi , and PLo is the hydrogen
pressure corresponding to the hydride temperature TLo .
Hydrogen Purification. In many applications, such as the
hydrogenation of vegetable oils and some
laboratory experiments, high-purity hydrogen is needed. One property of
hydriding substances is that they
absorb hydrogen but not the other elements or molecules which may be
present as impurities. Using this
property of the hydriding substances, it is possible to purify hydrogen
simply and inexpensively. A hydride
tank is first charged with impure hydrogen while heat is being removed. Only
hydrogen forms hydrides, which
allows impurities to be flushed from the hydride tank. Once the impurities are
flushed from the tank, heat
may be applied, and pure hydrogen is released from the hydride. For continuous
operation, the system may be
equipped with three tanks (absorption, flushing, and desorption), and their roles
can be switched in cycles.
Deuterium Separation. The property of hydrides used for hydrogen
purification can also be used for
deuterium separation. The system is similar to the hydrogen purification system.
Deuterium is flushed from
the hydriding tank, while hydrogen remains bonded in the metal hydride. Using
titanium nickel alloy as the
----------------------- Page 332-----------------------
hydriding substance, it is possible to obtain 85% pure deuterium after

only two steps. With three and four
steps, it is possible to obtain 99.9% and 99.999% pure deuterium, respectively.
Safety Aspects of Hydrogen Use. Like any other fuel or energy
carrier, hydrogen poses risks if not
properly handled or controlled. The risk of hydrogen, therefore, must be considered
relative to the common fuels
such as gasoline, propane, or natural gas. The specific physical characteristics of
hydrogen are quite different
from those common fuels. Some of those properties make hydrogen potentially less
hazardous, whereas other
hydrogen characteristics could theoretically make it more dangerous in certain
situations.
Because hydrogen has the smallest molecule, it has a greater tendency to
escape through small openings
than other liquid or gaseous fuels. Based on properties of hydrogen such as
density, viscosity, and diffusion
coefficient in air, the propensity of hydrogen to leak through holes or
joints of low-pressure fuel lines may
be only 1.26 to 2.8 times faster than a natural gas leak through the same hole (and
not 3.8 times faster as
frequently assumed based solely on diffusion coefficients). Experiments have
indicated that most leaks from
residential natural gas lines are laminar (69). Because natural gas has over three
times the energy density per
unit volume, the natural gas leak would result in more energy release than a
hydrogen leak.
For very large leaks from high-pressure storage tanks, the leak rate is
limited by the sonic speed. Because
of the higher sonic speed in hydrogen (1308 m/s), hydrogen would initially escape
much faster than natural
gas (sonic speed in natural gas is 449 m/s). Again, because natural gas has more
than three times the energy
density than hydrogen, a natural gas leak will always contain more energy.
Some high-strength steels are prone to hydrogen embrittlement. Prolonged
exposure to hydrogen, par-
ticularly at high temperatures and pressures, can cause these steels to lose
strength, eventually leading to
failure. However, most other construction, tank, and pipe materials are not prone
to hydrogen embrittlement.
Therefore, with the proper choice of materials, hydrogen embrittlement
should not contribute to hydrogen
safety risks.
If a leak should occur for whatever reason, hydrogen will disperse much
faster than any other fuel, thus
reducing the hazard levels. Hydrogen is both more buoyant and more diffusive than
either gasoline, propane,
or natural gas. Table 19 compares some properties and leak rates for hydrogen and
natural gas.
Hydrogen/air mixture can burn in relatively wide volume ratios, between 4%
and 75% of hydrogen in air.
The other fuels have much lower flammability ranges (viz., natural gas 5.3% to 15%,
propane 2.1% to 10%,
and gasoline 1% to 7.8%). However, the range has a little practical value. In many
actual leak situations, the
key parameter that determines if a leak would ignite is the lower flammability
limit, and hydrogen’s lower
flammability limit is 4 times higher than that of gasoline, 1.9 times higher than
that of propane, and slightly
lower than that of natural gas.
----------------------- Page 333-----------------------
Hydrogen has a very low ignition energy (0.02 mJ), about one order of
magnitude lower than other fuels.
The ignition energy is a function of fuel/air ratio, and for hydrogen it reaches a
minimum at about 25% to 30%.
At the lower flammability limit, hydrogen ignition energy is comparable with that of
natural gas (70).
Hydrogen has a flame velocity seven times faster than that of natural gas or
gasoline. A hydrogen flame
would therefore be more likely to progress to deflagration or even
detonation than other fuels. However,
the likelihood of a detonation depends in a complex manner on the exact fuel/air
ratio, the temperature, and
particularly the geometry of the confined space. Hydrogen detonation in the open
atmosphere is highly unlikely.
The lower detonability fuel/air ratio for hydrogen is 13% to 18%, which is
two times higher than that of
natural gas and 12 times higher than that of gasoline. Because the lower
flammability limit is 4%, an explosion
is possible only under the most unusual scenarios (e.g., hydrogen would first have
to accumulate and reach 13%
concentration in a closed space without ignition, and only then an ignition source
would have to be triggered).
Should an explosion occur, hydrogen has the lowest explosive energy per unit
stored energy in the fuel,
and a given volume of hydrogen would have 22 times less explosive energy than the
same volume filled with
gasoline vapor.
Hydrogen flame is nearly invisible, which may be dangerous because people in
the vicinity of a hydrogen
flame may not even know there is a fire. This may be remedied by adding some
chemicals that will provide
the necessary luminosity. The low emissivity of hydrogen flames means that near-by
materials and people will
be much less likely to ignite and/or be hurt by radiant heat transfer. The fumes
and soot from a gasoline fire
pose a risk to anyone inhaling the smoke, whereas hydrogen fires produce only water
vapor (unless secondary
materials begin to burn).
Liquid hydrogen presents another set of safety issues, such as the risk of
cold burns and the increased
duration of leaked cryogenic fuel. A large spill of liquid hydrogen has some
characteristics of a gasoline spill;
however, it will dissipate much faster. Another potential danger is a
violent explosion of a boiling liquid
expanding vapor in case of a pressure relief valve failure.
In conclusion, hydrogen appears to pose risks of the same order of magnitude
as other fuels. In spite of
public perception, in many aspects hydrogen is actually a safer fuel than gasoline
and natural gas. As a matter
of fact, hydrogen has a very good safety record, as a constituent of the “town gas”
widely used in Europe and
the United States in the 19th and early 20th century, as a commercially used
industrial gas, and as a fuel in
space programs. There have been accidents, but nothing that would characterize
hydrogen as more dangerous
than other fuels has occurred.
One of the most remembered accidents involving hydrogen is the Hindenburg
dirigible disaster in 1937.
However, hydrogen did not cause that accident, and hydrogen fire did not directly
cause any casualties. The
accident appears to have been caused by a static electricity discharge,
and it was the balloon’s lining that
caught fire first (71). Once hydrogen that the balloon was filled with for buoyancy
(instead of helium as it was
originally designed to be filled with) was ignited, it burned (as any fuel is
supposed to). However, hydrogen
fire went straight up, and it did not radiate heat so the people in the gondola
underneath the balloon were
not burned or suffocated. As a matter of fact, 56 survivors walked out of gondola
when it landed and all the
hydrogen and balloon structure burned out. Therefore, even in a worst case scenario
accident, hydrogen proved
to be a safe fuel.
Liquid Hydrogen
Hydrogen may be stored in gaseous, liquid, slush, and solid forms (metal
hydrides). Every mode of storage
has its advantages and disadvantages in terms of cost, stability,
convenience of usage, and energy density.
Liquid hydrogen has several advantages over other storage modes, especially in
terms of energy density and
convenience of utilization. However, several problems stand in the way of its
large-scale production and long-
term storage. This includes problems associated with energy consumed in
liquefaction as well as boil-off losses
----------------------- Page 334-----------------------
during storage and handling. Boil-off losses occur because of a variety

of mechanisms such as ortho–para
conversion, thermal stratification and self pressurization, heat leaks, sloshing,
and flashing. This section will
review some of these problems and discuss some of the prospects of large-scale
liquid hydrogen production and
utilization.
Introduction. Liquid hydrogen has many different uses, and its usefulness
in a number of areas has
been demonstrated. Presently, liquid hydrogen is primarily used as a rocket fuel
and is predestined for super-
sonic and hypersonic space vehicles primarily because it has the smallest boiling
point density and the highest
specific thrust of any known fuel. Its favorable characteristics include its high
heating value per unit mass, its
wide ignition range in hydrogen/oxygen or air mixtures, as well as its large flame
speed and cooling capacity
due to its high specific heat which permits very effective engine cooling and
cooling the critical parts of the
outer skin (72,73). Liquid hydrogen has some other important uses such
as in high-energy nuclear physics
and bubble chambers. The transport of hydrogen is vastly more economical
when it is in liquid form, even
though cryogenic refrigeration and special dewar vessels are required. Although
liquid hydrogen can provide
many advantages, its uses are restricted in part because liquefying hydrogen by
existing conventional methods
consumes a large amount of energy (around 30% of its heating value). Also, boil-off
losses associated with the
storage, transportation, and handling of liquid hydrogen can consume up to 40% of
its available combustion
energy. It is therefore important to search for ways that can improve the efficiency
of the liquefiers and di-
minish the boil-off losses. Improving liquefier efficiency not only helps to recover
some of the aforementioned
30% losses but can also contribute to reducing the other 40% losses caused by boil-
off. This can be partially
achieved by reliquefaction of the evaporated hydrogen. A significant improvement in
the reliquefier efficiency
is required, however, in order to provide an economical reliquefaction process.
Liquid Hydrogen Production. The production of liquid hydrogen
requires the use of liquefiers that
use different principles of cooling. In general, hydrogen liquefiers may be
classified as conventional, magnetic,
or hybrid. Many types of conventional liquefiers exist such as the Linde–Hampson
liquefiers, the Linde dual-
pressure liquefiers, the Claude liquefiers, the Kapitza liquefiers, the Heylandt
liquefiers, and those liquefiers
using the Collins cycle, just to name a few. Conventional liquefiers
generally are made up of compressors,
expanders, heat exchangers, and Joule–Thomson valves. Magnetic liquefiers, on the
other hand, use the mag-
netocaloric effect. This effect is based on the principle that some magnetic
materials experience a temperature
increase upon the application of a magnetic field and a temperature drop upon
lifting the magnetic field. The
magnetic analog of several conventional liquefiers include the Brayton liquefiers,
the Stirling liquefiers, and the
active magnetic regenerative (AMR ) liquefier. Additional information on liquid
hydrogen production methods
can be found in Sherif et al. (74).
Liquid Hydrogen Storage. As reported earlier, long-term storage of
liquid hydrogen is associated with
several problems primarily related to boil-off losses. Research on the mechanisms
contributing to boil-off losses
in cryogenic hydrogen storage systems exists to varying degrees of
sophistication depending on the nature
of the boil-off mechanism. These mechanisms are (1) ortho–para conversion;
(2) heat leak (shape and size
effect, thermal stratification, thermal overfill, insulation, conduction, radiation,
cooldown), (3) sloshing, and
(4) flashing. These mechanisms will be described in what follows.
Boil-Off Resulting from Ortho–Para Conversion. The hydrogen
molecule is made up of two identical
atoms, each of which is composed of a nucleus of one proton and one neutron and an
electron orbiting around
the nucleus. The rotational band in the vibration–rotation spectrum of hydrogen was
early observed to show
a weak–strong alteration in intensity. This led to the idea that two varieties of
hydrogen exist. This idea was
rationalized by Heisenberg and Hund in terms of the possible combinations of the
nuclear spins in the hydrogen
molecule (75).
At high temperatures, the equilibrium mixture consists of 75% ortho-hydrogen
and 25% para-hydrogen.
Hydrogen of this composition is termed normal and is in equilibrium at room
temperature and above. At 0 K,
on the other hand, all the molecules must be in a rotational ground state at
equilibrium. Therefore, all the
hydrogen molecules at equilibrium should be in the para state at absolute zero.
Pure para-hydrogen may be
----------------------- Page 335-----------------------

HYDROGEN
ENERGY SYSTEMS 39
prepared at equilibrium, but the maximum possible concentration of ortho-hydrogen

is 75% under equilibrium
conditions. Many of the thermophysical properties of normal and para-hydrogen are
significantly different from
each other. These differences have often been taken advantage of in analyses
involving hydrogen. For instance,
at 120 K to 190 K, the thermal conductivity of para-hydrogen is 50% more than that
of ortho-hydrogen. The
thermal conductivity can, therefore, be used to determine the ortho–para ratio of a
hydrogen sample (76).
By cooling hydrogen gas from room temperature to the nbp (normal boiling
point), the ortho-hydrogen
converts to para-hydrogen spontaneously. The decrease of ortho
concentration should be from 75% to 0.2%.
Because the percentage of the ortho–para composition is temperature-dependent, and
because, in the lique-
faction process, the temperature continuously decreases, one of the questions that
should arise pertains to the
rate of conversion. The conversion of a nonequilibrium ortho–para composition to an
equilibrium composition
is a very slow process in the absence of a catalyst. This mode of conversion (i.e.,
in the absence of a catalyst) is
called self-conversion.
The self-conversion rate is also a function of temperature. At the nbp of
air, the half-life of conversion is
greater than one year, whereas at 923 K and a 0.0067 MPa the half-life is on the
order of 10 min. The half-life
period is defined as the amount of time required to convert half of the excess ortho
(or para) composition present
at the starting time (76).
The conversion of ortho-hydrogen to para-hydrogen is an exothermic reaction.
The heat of conversion is
related to the change of momentum of the hydrogen nucleus when the direction of
spin changes. The amount
of heat given off in this conversion process is temperature-dependent. The heat of
conversion is greater than
the latent heat of vaporization of normal and para-hydrogen at the nbp. If the
unconverted normal hydrogen
is placed in a storage vessel, the heat of conversion will be released within the
container, which leads to the
evaporation of the liquid. Because of these peculiarities in the physical
properties of hydrogen, the boil-off of
the stored liquid will be considerably larger than what one would be able to
determine from calculations based
on ordinary heat leak to the storage tank.
In order to minimize the storage boil-off losses, the conversion rate of
ortho-hydrogen to para-hydrogen
should be accelerated with a catalyst that converts the hydrogen during the
liquefaction process (77,78,79,80).
In the presence of a catalyst that is well mixed with hydrogen, the reaction
approaches a first-order one for the
gas phase. If the catalyst is added to the liquid phase, the reaction becomes a
zero-order one (81).
The transformation can be catalyzed by a number of surface active (or
paramagnetic) species. A convenient
way is to absorb normal-hydrogen on active charcoal cooled with liquid
hydrogen and then pump off the
equilibrium mixture. The conversion may take only a few minutes if a highly active
form of charcoal is used
(75), but the conversion effectiveness appears to depend on the presence of an
absorbed layer of oxygen, which,
for thoroughly purified hydrogen, may not be available (77). Other suitable ortho–
para catalysts are metals such
as tungsten or nickel, any paramagnetic oxides like chromium or gadolinium oxides,
and other paramagnetic
substances like NO2 , ferric hydroxide gel, and nickel silicate. These catalysts
are capable of enabling almost
complete conversion to be achieved in a few minutes. In paramagnetic fields during a
homogeneous conversion
process (using paramagnetic gases or by passing hydrogen through solutions
containing paramagnetic ions),
the nuclear spin is reversed without breaking the H H bond. This suggests that
strongly paramagnetic species
such as Gd3+ are highly effective. It has been shown that paramagnetic catalysis
conversion is accelerated by
the action of an externally applied magnetic field. This effect tends to achieve
saturation in a field of about
1000 Oersted (79577 A/m) (75). In a magnetocaloric liquefier, there is a good chance
of taking advantage of
this phenomenon primarily because both the magnetic ions and the external magnetic
field are available, thus
making a good design the only prerequisite for achieving a fast and economic
conversion.
The use of a catalyst usually results in a larger refrigeration load and
consequently in an efficiency penalty
primarily because the heat of conversion must be removed. The time for which
hydrogen is to be stored usually
determines the optimum amount of conversion. For use within a few hours, no
conversion is necessary. For
example, large-scale use of liquid hydrogen as a fuel for jet aircraft is one of
those cases where conversion is not
necessary because utilization of the liquid is almost a continuous process and
long-term storage is therefore not
----------------------- Page 336-----------------------
needed (80). For some other uses, a partial conversion might be required to create
more favorable conditions. It
should be noted that, for every initial ortho concentration, there exists a unique
curve for boil-off of hydrogen
with respect to time.
The heat of conversion at higher temperatures is small (at 300 K, the heat of
conversion is 270 kJ/kg) and
increases as the temperature decreases, where it reaches 519 kJ/kg at 77K, the nbp
for liquid nitrogen. For
temperatures lower than the nbp of nitrogen, the heat of conversion remains
constant (523 kJ/kg). However, the
equilibrium percentage of para-hydrogen at the nbp is 50%. Therefore, in order to
produce equilibrium liquid
hydrogen, 50% of hydrogen must be converted from ortho to para at temperatures
lower than liquid nitrogen
temperatures. Because the heat of conversion is almost constant for temperatures
lower than those of liquid
nitrogen, the refrigeration load required to remove the heat of conversion should
be almost constant at different
temperatures. Ideal refrigeration requires that the energy needed for the same
amount of refrigeration load
be temperature-dependent. This suggests that, in order to operate at an
optimum power, it is necessary to
have continuous conversion, especially for temperatures lower than those
of liquid nitrogen. By continuous
conversion, up to 15% energy saving may be achieved.
Boil-Off Resulting from Heat Leak.
Reservoir Shape and Size Effect. Hydrogen is usually transported in
large quantities by truck tankers
of 30 m3 to 60 m3 capacity, by rail tank cars of 115 m3 capacity, and by
barge containers of 950 m3 capacity
(82). Liquid hydrogen storage vessels are usually available in sizes ranging from
one liter dewar flasks used
in laboratory applications to large tanks of 5000 m3 capacity. The National
Aeronautics and Space Adminis-
tration (NASA ) typically uses large tanks of 3800 m3 capacity (25 m in
diameter) (83). The total boil-off rate
from such dewars is approximately 600,000 LPY (liters per year), which is vented
to a burn pond. The heat
leakage losses are generally proportional to the ratio of surface area to the
volume of the storage vessel (S/V ).
The most favorable shape is therefore spherical because it has the least
surface to volume ratio. Spherical
shape containers have another advantage. They have good mechanical strength because
stresses and strains
are distributed uniformly. In large sizes, spherical containers are expensive
because of their manufacturing
difficulty. They are often used for on-site large volume storage, however.
Storage vessels may also be constructed in other shapes such as cylindrical,
conical, or any combination of
these shapes. Cylindrical vessels are usually required for transportation of liquid
hydrogen by trailers or railway
cars because of limitations imposed on the maximum allowable diameter of the
vessel. For normal highway
transportation, the outside diameter of the vessel cannot exceed 2.44 m.
From an economics standpoint,
cylindrical vessels with either dish, elliptical, or hemispherical heads are very
good, and their S/V ratios are
only about 10% greater than that of the sphere (84).
Because boil-off losses resulting from heat leak are proportional to S/V as
reported earlier, the evapora-
tion rate will diminish drastically as the storage tank size is increased. For
double-walled vacuum-insulated
spherical dewars, boil-off losses are typically 0.3% to 0.5% per day for containers
having a storage volume of
50 m3 , 0.2% for 103 m3 tanks, and about 0.06% for 19,000 m3 tanks (85).
Obviously, the larger the size of the
dewar, the smaller the cost per unit volume of storage. Interestingly enough, the
rate of evaporation does not
substantially decrease with increasing the size of the container for cylindrical
vessels of constant diameter.
Thermal Stratification and Thermal Overfill. Stratification is the
existence of a fluid at different tem-
peratures within one vessel. Heat transfer from the sidewalls and bottom to a
liquid hydrogen storage vessel
will always result in an increase of the sensible enthalpy of the liquid. When the
heat flux is small enough and
the inside wall of the vessel is smooth, boiling is unlikely to occur (77). In this
case, the added energy tends to
get distributed nonuniformly because the imposed heat flux usually produces a free
convective flow. The warm
liquid hydrogen becomes buoyant and rises to the top. With the passage of time, the
lower boundary of the
stratified region moves toward the bottom of the vessel. Because the warm liquid
rises to the surface, it is not
uncommon to have the temperature at the top layers rise much more rapidly than that
of the bulk liquid (84).
As a consequence, the upper layers are more likely to experience a relatively large
vapor pressure increase,
whereas the lower layers experience little or no vapor pressure change. This
thermal stratification creates a
----------------------- Page 337-----------------------
vessel pressure corresponding to a temperature greater than the average liquid

temperature because the tank
pressure is mostly determined by the vapor pressure of the free surface. This in
turn greatly reduces the length
of liquid storage time without having to vent the vapor. Naturally, this leads to a
decrease in the lock-up time.
The stratification in temperature is stable because the warmer liquid has the lower
density, in addition to the
fact that liquid hydrogen is a poor thermal conductor (84).
Thermal overfill is a phenomenon that occurs when one of the following
conditions exists.
(1) The average specific enthalpy of the liquid is greater than that of the
saturation temperature, in which
case the liquid boil-off from the vented dewars is found to be nonuniform. This
can be attributed to the
unstable stratification which occurs in vented dewars as a result of
a sudden lowering of atmospheric
pressure (generally, the liquid becomes superheated in the lower
region, while the surface temperature
remains saturated, which leads to generating instant vapor of large
quantities).
(2) The operating pressure of the tank is less than that corresponding to the
saturation temperature of the
feed liquid, in which case only the surface layer is able to respond to the new
operating pressure. If the
surface layer is distributed by some means, for instance by top-loading more
liquid, the underlying liquid
may be brought into rapid equilibrium, and that, in turn, would cause a rapid
boil-off (77,86).
One way to decrease boil-off losses resulting from stratification and

thermal overfill is by employing
high-conductivity plates (conductors) installed vertically in the vessel. The
plates produce heat paths of low
resistance between the bottom and top of the vessel and can operate most
satisfactorily in eliminating tem-
perature gradients and excessive pressures. Another way is to pump the
heat out and maintain the liquid
at subcooled or saturated conditions. An ideal refrigeration system to
perform this task can be an efficient
magnetic refrigerator. The magnetic refrigerator is very suitable for this job
because of its relatively higher
efficiency, compactness, lower price, and reliability (87,88,89,90,91,92,93,94,95).
Barclay (88) proposed a magnetic refrigerator that can be used for
the reliquefaction of the hydrogen
vapor. This design can be modified and used for the preceding cases. In
order to perform the refrigeration
effectively, a thermal valve is needed to permit heat flow when there is superheat
and to stop heat flow when
there is no superheat (86).
Insulation and Conduction Through Supports. Liquid hydrogen
containers are usually of three types:
double-jacketed vessels with liquid nitrogen in the outer jacket, superinsulated
vessels with either a reflecting
powder or multilayer insulation, and containers with vapor-cooled shields
employing super insulation. Al-
though multilayer insulation (MLI ) provides for a low boil-off rate, the addition
of a vapor-cooled shield (VCS)
will lower the boil-off losses even further. A VCS is a type of insulation that
takes the vapor boil-off and passes
it past the tank before being reliquified or vented. Published data indicate that a
reduction of more than 50%
in boil-off may be achieved for a 45,045 kg liquid hydrogen cryogenic facility with
a VCS than without one.
Brown (96) showed that locating the VCS at half the distance from the tank to the
outer surface of a 10.16 cm
MLI of a 45,045 kg facility would reduce the boil-off by 10%. A dual VCS system on
the tank would improve
the performance by 40% over a single VCS. Brown (96) also showed that the preferred
locations for the inner
and outer shields in a dual VCS system are 30% and 66% of the distance from the
tank to the outer surface of
the MLI.
Mechanical supports that connect the inner and outer vessel of
double-walled vacuum tanks are an
integral part of the insulation problem, primarily because of the heat influx that
can be conducted to the inner
vessel. One way to diminish these boil-off losses is to cool the support employing
the produced hydrogen vapor
continuously in a counterflow fashion.
In practice, the thermal conductivity of the support varies with temperature,
and there exists difficulty
in manufacturing a continuous heat exchange surface in the temperature
range of interest. In this case, a
succession of discrete cooling stations provides for a better design. Bejan
(97,98,99) introduced an optimization
----------------------- Page 338-----------------------

method for finding the location and the temperature of the cooling
stations, and he used that procedure in
order to optimize helium boil-off losses.
Another optimization of boil-off losses can be considered by changing the
value of the heat sink temper-
ature from 20.4 K to 33 K (critical temperature) and examining the
temperature at which the boil-off rate
is minimum. This is primarily because, as the storage temperature increases, the
amount of heat leak and,
consequently, the boil-off losses decrease. On the other hand, the latent heat of
vaporization decreases with
decreasing storage temperature which contributes to increasing the boil-
off. Thus, there should be an opti-
mum storage temperature or pressure that balances the aforementioned effects.
Employing this procedure, the
optimum temperature and pressure of the hydrogen vessel may be determined.
Boil-Off Resulting from Sloshing. Another process that leads to boil-
off during liquid hydrogen trans-
portation by tankers is sloshing. Sloshing is the motion of liquid in a
vessel caused by acceleration or de-
celeration. Because of different types of acceleration and deceleration, there
exist different types of sloshing.
Acceleration causes the liquid to move to one end and then reflect from that end,
thus producing a hydraulic
jump. The latter then travels to the other end, thus transforming some of its
impact energy to thermal energy.
The thermal energy dissipated eventually leads to an increase in the evaporation
rate of the liquid (77,84). The
insertion of traverse or antislosh baffles not only restrains the motion of the
liquid, thus reducing the impact
forces, but also increases the frequency above the natural frequency of the tanker
(77,81).
Boil-Off Resulting from Flashing. Another source of boil-off is
flashing. This problem occurs when liquid
hydrogen, at a high pressure (2.4 to 2.7 atm), is transferred from trucks and rail
cars to a low-pressure dewar
(1.17 atm). This problem can be reduced if transportation of liquid
hydrogen is carried out at atmospheric
pressures. Furthermore, some of the low-pressure hydrogen can be captured and
reliquefied. One of the best
systems that can be selected for reliquefaction is one that employs a magnetic
refrigerator as an external cycle.
Another suitable cycle can be the Claude cycle with a saturated-vapor compressor
and a wet expander. The
system operates best if the sink is a liquid nitrogen bath.
One other method, which is proposed by NASA, is to capture some of this low-
pressure hydrogen by metal
hydrides and then reliquefy the hydrogen vapor. According to this scheme, a
liquefier with a capacity of 40 tons
per day (TPD) would be necessary to make up for the losses. This idea was
considered expensive and was later
rejected.
Concluding Remarks on Liquid Hydrogen. This section
addressed the production, storage, and
utilization of liquid hydrogen. More research is probably needed to address some of
the fundamental issues
that are relevant to the efficient utilization and storage of liquid hydrogen.
Additional information on liquid
hydrogen storage can be found in Gursu et al. (100,101).
Slush Hydrogen
Slush hydrogen is a mixture of liquid and frozen hydrogen in equilibrium

with the gas at the triple point.
Slush hydrogen is expected to be the primary propellant for the National Aero-Space
Plane (NASP ) because
of its potential in reducing its physical size and significantly cutting the
projected gross liftoff weight. This
section provides an overview of slush hydrogen production methods and reviews the
current state of knowledge
regarding slush hydrogen utilization.
Introduction. Hydrogen has become an important rocket fuel despite the
problems associated with its
use. Some of these problems include its low density, temperature stratification,
short holding time caused by
its low latent heat, hazards associated with high vent rates, and unstable flight
conditions caused by sloshing
of the liquid in the fuel tank. Rapial and Daney (102) report that two techniques
can completely eliminate some
of these undesirable features and partially eliminate the others. One is the
production of liquid–solid hydrogen
mixtures “slush hydrogen” by further refrigeration of the liquid. The other is the
gelation of liquid hydrogen
by the addition of a fine particle gelling substance.
----------------------- Page 339-----------------------
Slush hydrogen is a mixture of liquid and frozen hydrogen in equilibrium with

the gas at the triple point,
13.8 K. The density of the icelike form is about 20% higher than that of the
boiling liquid. To obtain the icelike
form, one must remove the heat content of the liquid at 20.3 K until the triple
point is reached and then remove
the latent heat of fusion. The “cold content” of the icelike form of hydrogen is
some 25% higher than that of
the saturated vapor at 20.3 K. Justi (103) reports that in producing the slush
hydrogen, in comparison with
the icelike form, one loses part of the increase in density and cold storage, but
in mixing the icelike form with
liquid hydrogen, one obtains a medium that is transportable in pipelines. Slush
hydrogen can be produced from
liquid hydrogen with the help of a vacuum pump and expansion through a jet (103).
Baker and Matsch (104) gave a tabulated comparison among the properties of
liquid and slush hydrogen.
According to them, the smaller enthalpy of slush hydrogen reduces the evaporation
losses during storage and
transport and as a consequence permits longer storage without venting. The
increased density permits ground
transport equipment to carry 15% greater loads; for space vehicles, a given fuel
load can be carried in tankage
that is 13% smaller in volume.
Kandebo (105) reported that slush hydrogen is expected to be the primary
propellant for the NASP. He
mentioned that using slush instead of liquid hydrogen reduces the
physical size of the NASP and cuts the
projected gross liftoff weight by up to 30%. The freeze–thaw process, which is the
most used slush production
process, and the auger process, which is relatively new, were discussed and
compared. One expensive alternative
method to pressurize tanks was also mentioned. This alternative is to pressurize
tanks with a combination of
helium and hydrogen because helium does not condense at slush hydrogen storage
temperatures and is not
very dense.
Justi (103) reported that the slushlike form of hydrogen has the advantage of
a greater “cold content”
compared to liquid hydrogen. The disadvantage of slush hydrogen lies in the present
state of the technology
for its production, which is not yet well advanced, so that it is impractical to
calculate a realistic price. Justi
argues, however, that the production cost of slush hydrogen now and in the near
future is greater than the
liquefaction costs of hydrogen because of the larger energy use involved.
Sindt (106) studied the characteristics of slush hydrogen preparation,
storage, transfer and instrumenta-
tion. In that study, slush preparation by intermittent vacuum pumping was
discussed. Slush was aged 100 h,
during which time the solid particle size and structure were observed. The solid
particle structure was found
to change dramatically during aging, even though the changes in the particle size
were insignificant. Sindt
also mentioned that slush with over 50% solid content could be transferred and
pumped with losses similar to
those in triple-point liquid hydrogen if the Reynolds numbers were high enough.
Voth (107) reported on a slush hydrogen production method with an auger
rotating inside a brass tube
refrigerated with liquid helium. The auger was reported to produce small
particles from the cryogens on a
continuous basis so that the resulting slush mixture could be transferred and
stored. He also reported that the
auger is capable of producing slush at pressures higher than the triple-point
pressure of the cryogen.
Methods for determining the quality of liquid–solid hydrogen mixtures, along
with relevant thermody-
namic analyses were discussed by Daney and Mann (108) and Mann et al. (109).
Properties of liquid and slush
hydrogen and their applications were discussed by Smith (110) and Angus (111). The
flow characteristics of
slush hydrogen with different solid contents were examined by Sindt and Ludtke
(112) and by Sindt et al. (113).
The latter study reported on experimental results in which a second component was
added to liquid hydrogen
to determine the effect produced by the freeze–thaw process on the size of the
solid particles. Perrel and Haase
(114) developed a computer simulation method to describe the self-catalysis of
ortho-hydrogen molecules. The
simulations were performed on a small computer using Monte Carlo techniques. Roder
(115) studied the phase
transitions in solid hydrogen and suggested that, except for the lambda transition
that occurs at low tempera-
tures in solid hydrogen, and involves a change from hexagonal close-packed (hcp) to
face-centered cubic (fcc ),
there may be an additional phase transition in solid hydrogen.
Although current interest in liquid–solid mixtures of hydrogen as a potential
rocket propellant has lead to
theoretical and experimental investigations of its characteristics, there is still
a lot of potential for additional
----------------------- Page 340-----------------------
investigations. For example, the problem of transferring heat to slush

hydrogen along with solid density
stratification in large containers needs to be investigated more thoroughly.
Production of Slush Hydrogen. Several successful methods for
producing solids from liquid–solid
mixtures were reported in the literature. However, if the preparation method is to
be usable, it must work in
systems of all sizes and still meet certain basic criteria that create a process
that is thermodynamically efficient
and reproducible. The process should also be able to produce a mixture with enough
fluidity to transfer through
existing systems and with sufficient solid content so that the density and heat
capacity are significantly greater
than those for the normal boiling point liquid.
Freeze-Thaw Method. Sindt (106) summarized the basic principles of
the freeze–thaw method and its
feasibility. Two of the methods reported by Sindt involve the formation of solid
particles by spraying precooled
liquid into an environment well below the triple-point pressure and
inducing solid formation by allowing
precooled helium gas to flow through a triple-point temperature liquid. The spray
technique produced a very
low density solid particle, which had to be partially melted to form a
liquid–solid mixture. The technique
of allowing helium to flow through the liquid-produced solid by reducing
the partial pressure of hydrogen
below its triple-point pressure which resulted in a clear, rigid tube of solid
forming around the helium bubble
train. The solid was found not to be convertible to a mixture without
further treatment such as crushing.
Sindt found that a more desirable preparation method was to evacuate the
ullage over the liquid, which
created a solid layer at the liquid surface. The solid layer produced had a texture
that was very dependent
upon the rate of vacuum pumping. Low pumping rates produced a dense, nearly
transparent layer of solid.
Subsequent breaking of the solid left large rigid pieces that formed inhomogeneous
mixtures even with vigorous
mixing. Higher pumping rates produced a solid layer that was porous and consisted
of agglomerates of loosely
attached fine solid particles. The solid layer was found to be easily broken and
mixed with the liquid to form a
homogeneous liquid–solid mixture that could be defined as slush. Sindt discovered,
however, that solids could
not be continuously generated at the liquid surface without vigorous mixing.
Sindt (106) also reported on a continuous preparation method, which he
developed, capable of forming
solid layers in cycles by periodic vacuum pumping. During the pumping (pressure
reduction) portion of the
cycle, the solid layer formed as previously described. When the pumping was
stopped, the solid layer melted and
settled into the liquid, forming slush. Although mixing was not necessarily
required, the preparation method
was accelerated by some mechanical mixing which helped break the layer of solids.
This technique has been
called the freeze–thaw process.
In the same study, Sindt was able to determine the acceptable limits of
vacuum pumping rates for freeze–
thaw slush production. He found that low pumping rates were unacceptable because of
the texture of the solid
produced. Very high pumping rates were also found to be undesirable
because liquid droplets were carried
into the warm regions of the system by the velocity of the evolving
gas. The evaporation of these droplets
was, therefore, considered a loss to the preparation of solid. Sindt
also determined the rate that produces
solids that are easily transformable into slush by measuring the evolved gas during
many cycles of freeze–
thaw preparation. This rate was determined to be 0.9 m3 s − 1/m2 of
liquid surface area at vacuum pump inlet
conditions of 300 K and 6.9 kNm − 2 . This was also found to be the pumping rate
during the evacuation phase
of the freeze–thaw cycle. He concluded that the range of the vacuum

pumping rate for satisfactory slush
3 − 1 2 3 − 1 2
production was 0.8 m s /m to 1.2 m s /m of liquid surface area.
Pumping rates in that range were found
to result in an efficient exchange of energy between the evaporating and freezing
liquids. The efficiency of this
exchange of energy is best illustrated by the experimentally determined
irreversibility reported by Daney and
Mann (108), which was found to be less than 3.4%. Pumping rates within this range
have been used to produce
slush of solid fractions to 0.45 in vessels 10 cm to 76 cm in diameter.
According to Voth (107), although the freeze–thaw production method proved to
be technically feasible and
fully developed, it was shown to have disadvantages. These disadvantages include
the fact that the freeze–thaw
process is a batch process and that it operates at the triple-point pressure of the
cryogen. For hydrogen, this
pressure was estimated to be 0.07 bars absolute, a pressure capable of drawing air
through inadvertent leaks
----------------------- Page 341-----------------------
in the system. The air could be thought of as presenting a potential safety problem
in a hydrogen system. Voth
also showed that the freeze–thaw process required either costly equipment to
recover the generated triple-point
vapor or the loss of approximately 16% of the normal boiling-point liquid hydrogen
and approximately 24% of
the normal boiling-point liquid oxygen if the vapor was discarded.
Auger Method. Voth (107) reported on a study for producing liquid–solid
mixtures of oxygen or hydrogen
using an auger. He described an auger used to scrape frozen solid from the inside
of a refrigerated brass tube
in order to produce slush hydrogen. He showed that slush hydrogen could be
continuously produced by this
method, and because it could be immersed in liquid, slush was produced at pressures
above the triple-point
pressure. Voth was also able to produce the increased pressure pneumatically or by
generating a temperature
stratification near the surface of the liquid. He observed that the auger system
produced particles in the size
realm of the particle produced by the freeze–thaw method so that, like the freeze–
thaw-produced slush, the
auger-produced slush could be readily transferred and stored.
According to Voth (107), the surface freezing occurring in the freeze–thaw
production process is thermo-
dynamically reversible. In contrast, the freezing process in the auger is
irreversible because a temperature
difference must exist between the refrigerant and the freezing cryogen, and the
energy added to scrape the
solid cryogen out of the brass tube must be removed by the refrigerant.
Voth argued that in spite of these
irreversibilities, the auger system was found to require less energy to produce
slush hydrogen than a particle
freeze–thaw system. He reported that the temperature difference required
refrigeration temperatures below
the triple-point temperature of hydrogen so that a gaseous or liquid helium
refrigerator was required for the
auger.
Although this method was determined to offer the advantage of operating above
atmospheric pressures
(an important consideration in lowering the possibility of oxygen intrusion
into the slush), it was shown to
have the disadvantage of using expensive helium coolant.
Quality Determination. A thermodynamic analysis of the freeze–thaw
production process for forming
mixtures of slush developed by Mann et al. (109) showed that the measurement of two
quantities—(1) the mass
fraction pumped during the production process, and (2) the heat leak per unit mass
into the production dewar—
were sufficient to determine the quality of the resulting mixture. Because
irreversibilities were observed to
exist in the freeze–thaw process, a comparison of the predicted quality with the
qualities determined by the
measurement of thermodynamic properties was found necessary.
According to Daney and Mann (108), in the freeze–thaw method of
forming a triple-point mixture of
hydrogen, the latter was partially evaporated under the reduced pressure obtained
by a vacuum pump, while a
refrigeration effect (approximately equal to the latent heat of vaporization) was
experienced by the remaining
liquid or liquid–solid mixture. They suggested, by specifying the initial
state and the process or path to be
followed, that the quality of the liquid–solid mixture could be predicted as a
function of the mass of the vapor
removed. According to them, therefore, the problem of quality prediction consisted
of two parts. The first was
the determination of when a specific initial state, triple-point liquid was reached,
whereas the second was the
prediction of the quality once this standard initial state was achieved. They also
reported that the arrival at the
triple-point liquid condition may be determined in several ways: (1) vapor pressure
measurement, (2) visual
observation of the formation of a few solid particles, and (3) prediction from
the mass fraction of the vapor
pumped off in the cooling-down process. Daney and Mann argued that the first two
methods provided a more
accurate means of determining the arrival at triple-point liquid conditions, and
hence a more accurate quality
determination could be realized. However, they also argued that it may be of
interest to see what accuracy in
quality determination could be expected by starting the mass accounting with a
normal boiling-point liquid
instead of a triple-point liquid.
Energy Requirements. A study to determine the energy requirements
of the freeze–thaw and auger
methods was reported by Voth (107) who mentioned that producing slush appears to be
a reasonable way of
increasing the density and heat capacity of cryogens. He observed that when large
quantities of slush were
produced, the energy required to produce them became significant. He therefore
concluded that the production
----------------------- Page 342-----------------------
energy depended on both the thermodynamic reversible energy requirements to produce

slush and the energy
required to overcome irreversibilities in a practical system. He
determined the energy required to produce
slush with a 0.5 solid mass fraction for four practical slush-hydrogen-producing
systems. Three of the systems
employed the freeze–thaw production method, whereas the fourth used an auger and a
helium refrigerator to
produce slush from normal boiling-point liquid hydrogen.
According to Voth (107), the thermodynamic reversible energy required to
produce slush hydrogen was
equal to the thermodynamic availability of the slush. Also using normal hydrogen at
a temperature of 300 K
and a pressure of 1.013 bars as the base fluid, the reversible energy required to
produce normal boiling-point
para-hydrogen was 3971.4 W·h/kg, whereas the reversible energy required to produce
slush with a solid mass
fraction of 0.5 was 4372.8 W·h/kg. Voth explained that the energy required by
practical systems was higher
because of component inefficiencies. The calculated energies for the four
cases were based on liquefier and
refrigerator efficiencies of 40% of that of Carnot. He assumed that the vacuum pumps
required by the freeze–
thaw production had an efficiency of 50% of the isothermal efficiency. However, he
did not include the increased
production energy resulting from heat leak into the containers and transfer lines
because it was difficult to
estimate without a firm system definition and because the heat leak would have been
nearly equal for all the
systems studied. Because heat leak was not included, the calculated energies for
the four cases were lower
than those of an actual system, but the results did allow a comparison among the
various slush production
systems.
Voth also found that the auger method introduced additional irreversibilities
because of the rotational
power added to the slush generator and the temperature difference between the
refrigerant and the freezing
hydrogen. He concluded that a value of 1.05 for the ratio of total supplied
refrigeration to the refrigeration
available for the freezing hydrogen in the auger system was usually
needed to determine the refrigeration
required to produce slush with a solid fraction of 0.5. The input
energy to the refrigerator connected to
the auger was observed to depend on the refrigeration temperature; the
lower the temperature, the higher
the input energy. For the auger system, he assumed that the lowest
refrigeration temperature was 10 K.
Because a refrigerator was required to cool the normal boiling-point liquid to a
triple-point liquid, the highest
refrigeration temperature was assumed to be 19.76 K. He noted that these
refrigeration temperatures were
within the capabilities of a closed-cycle helium refrigerator. Voth also
reported that even though the power
required to produce slush hydrogen using the auger was relatively easy to
calculate, the power required to
produce slush hydrogen using the freeze–thaw production method was very system
dependent.
Solid Particle Characteristics. Because applications of slush
Hydrogen always include some storage,
Mann et al. (109) observed the aging effects on solid particles in slush as solids
aged as much as 100 hours.
Particles of slush in freshly prepared solids were observed to be in
the form of agglomerates of very small
subparticles loosely attached. These agglomerates were remnants of the porous layer
that forms during the
pumping phase of preparation. They observed that as slush aged, the
solid particle configuration changed
dramatically and the loosely attached subparticles filled in with solids to form a
more smoothly rounded particle
of higher density. They asserted that the exact mechanism of the change was not
completely understood but
suggested that it involved a mass energy exchange that occurred in such a way as to
assume a more stable
spherical shape with less surface area. According to them, the major portion of the
change occurred within the
first 5 h. They remarked that accompanying the change in particle configuration was
an increase in the density
of slush that had settled to the bottom of the vessel. This tended to increase
density partly from the increase
in the bulk density of each particle and partly from a closer packing of particles
because of their more rounded
shape after aging. They thus concluded that the aging phenomenon of the
particles and the accompanying
increase in settled slush density were apparently a result of heat flow to the
slush. Freshly prepared slush was
concluded to have a maximum settled solid fraction of 0.35 to 0.45.
During the experimental study of Mann et al. (109), the slush particle size
and particle settling velocity
were measured using photo instrumentation. They noted that these
characteristics of particles should be
determined because they generally affect the fluid mechanics of liquid–solid
mixtures. They also reported that
----------------------- Page 343-----------------------
particle size determines the minimum restrictions permissible in systems using the
mixtures. They concluded
that although particle configuration changed with age, particle size did not change
significantly and that the
distribution of particle size could be satisfactorily described by a modified
logarithmic function. They reported
that size ranged from 0.5 mm to 10 mm, with 2 mm being the most frequently
occurring size. They also reported
over 170 settling velocity measurements on both fresh and aged slush particles and
concluded that the settling
velocity of fresh particles was lower than that of aged particles and was extremely
dependent on the shape of
individual particles.
Sindt (106) reported that mixing of slush hydrogen to maintain homogeneity
was required in most of the
potential applications, and if mixers were to impart velocities to the fluid greater
than the settling velocities,
particles would have been carried along and mixing would have resulted. Sindt used
several methods to mix
the slush in a vessel. One method used a simple propeller with blades designed to
impact velocity to the fluid
with lift. Another method used high-velocity streams of fluid expelling from
openings in a duct. According to
him, both methods of mixing had been used to mix slush of a solid fraction as high
as 0.6. Also, the solid content
in slush could be maintained and even increased by transferring slush
into the vessel and removing liquid
through a screen with 0.6 mm openings. Sindt explained that by using this method
the solid fraction could be
increased until the solids protrude above the liquid–gas interface.
Transfer Characteristics. Most of the studies on transfer
characteristics of slush hydrogen have
been carried out at the Cryogenics Division of the National Institute of
Standards and Technology (NIST ).
Experimental results pertaining to slush hydrogen fluidity will be reported in this
section.
Slush Hydrogen Flow in Pipes. Sindt (106) reported that transfer and
pumping characteristics of slush
hydrogen should be similar to those of liquid hydrogen if slush is to be used in
the existing liquid hydrogen
system and if slush is to be handled as a fluid. During the
experimental study of Sindt et al. (113), slush
of solid fractions as high as 0.6 was transferred through vacuum-insulated pumps of
16.6 mm to 25 mm in
diameter. Transfer pressure losses were determined in triple-point liquid and in
slush of solid fractions up to
0.55 in a 16.6 mm diameter pipe. They estimated slush data to have an uncertainty
occurring at the high solid
fractions and the lowest flow rates. According to them, pressure loss data for
flowing slush revealed that, at
low flow rates, pressure losses in flowing slush were double those of triple-point
liquid if solid fractions were
high. The data also revealed that at high flow rates, pressure losses in flowing
slush were less than the losses
in triple-point liquid if the solid fraction was near 0.35. They also carried out a
more complete analysis of the
losses in flowing slush hydrogen by calculating a friction factor and plotting it
versus the Reynolds number. In
the calculations, they used triple-point liquid viscosity because the viscosity of
slush hydrogen was generally
unknown.
Sindt and Ludtke (112) observed that the characteristics of slush
hydrogen flow losses at the higher
Reynolds numbers were more evident when displayed in terms of the
friction factors than if displayed as
pressure losses. They reported that friction losses in flowing slush of solid
fractions to 0.4 were as much as 10%
less than in liquid and that the friction-loss curve for 0.5 solid fraction slush
appeared to be following a slope
such that it would be below triple-point liquid losses at high Reynolds numbers.
They reported these data for
a 16.6 mm diameter pipe and concluded that low losses in slush aged 10 hours were
4% to 10% higher than
losses in 1-hour old slush.
The experiments of Sindt et al. (113) revealed that a gradient in
the solid concentration may develop
when the slush velocity is 0.5 ms − 1, whereas at a velocity of 0.15 ms − 1, the
solids were likely to settle forming
a sliding bed. Considering these observations, it was concluded that the critical
velocity of slush in the pipe
was about 0.5 m/s.
Slush Hydrogen Flow in Restrictions. Sindt et al. (113)
experimentally determined flow losses in re-
strictions such as valves, orifices, and venturies using slush hydrogen of solid
fractions to 0.5. They determined
the losses in a 3/4 inch nominal copper globe valve with the valve fully opened and
in cases when it was partially
opened and expressed the losses as a resistance coefficient independent of the slush
solid fraction. According
to their study, with the valve partially opened, the losses seemed to be less with
increasing the solid fraction,
----------------------- Page 344-----------------------
whereas with the valve fully opened, the trend was reversed. Also, the loss
coefficients compared favorably
with the typical valve coefficient for water.
Sindt and Ludtke (112) remarked that liquid–solid mixtures of hydrogen flow
through orifices of 6.4 mm
and 9.5 mm in diameter with no plugging and with a pressure drop of
the same magnitude as that of the
triple-point liquid. They observed that the slush mass flow rate was dependent on
the pressure differential
across the orifice as would have been expected for a nearly incompressible,
Newtonian fluid and that when
the minimum pressure in the orifice reached the triple-point pressure, cavitation
with accompanying choking
occurred. At that point, the mass flow became dependent on the upstream pressure and
independent of the
downstream pressure. Flow characteristics of slush flowing through a venturi have
also been determined by
Sindt et al. (113). According to them, flow losses through the venturi were not
affected by the solid fraction of
the slush.
Concluding Remarks on Slush Hydrogen. This section
presented an overview of the state-of-the
art of slush hydrogen production and utilization technologies. It is apparent that
the use of slush hydrogen
should be considered only for cases in which higher density and greater solid
content are really needed. This
is mainly because the production costs of slush hydrogen now and in the
near future are greater than the
liquefaction costs of hydrogen resulting from the larger energy use involved.
Conclusions
As a result of the aforementioned facts, it can be seen that hydrogen as a fuel and
the Solar Hydrogen Energy
System have unmatched advantages when compared with fossil fuels and the Fossil
Fuel System, respectively.
`
Hydrogen. The advantages of hydrogen vis-a-vis fossil fuels can be
summed up as follows:
(1) Liquid hydrogen is the best transportation fuel when compared with liquid
fuels such as gasoline, jet fuel,
and alcohols; and gaseous hydrogen is the best gaseous transportation fuel.
(2) Hydrogen can be converted to useful energy forms (thermal, mechanical, and
electrical) at the user end
through five different processes, whereas fossil fuels can be converted through
only one process (i.e., flame
combustion). In other words, hydrogen is the most versatile fuel.
(3) Hydrogen has the highest utilization efficiency when it comes to conversion to
useful energy forms (thermal,
mechanical, and electrical) at the user end. Overall, hydrogen is 39% more
efficient than fossil fuels. In other
words, hydrogen will save primary energy resources. It could also be termed as
the most energy-conserving
fuel.
(4) When fire hazards and toxicity are taken into account, hydrogen becomes the
safest fuel.
`
Solar Hydrogen Energy System. The advantages of the Solar
Hydrogen Energy System vis-a-vis the
present Fossil Fuel System and the Synthetic Fossil Fuel System can be summed up as
follows:
• 5.When the environmental impact is taken into consideration, the Solar Hydrogen
Energy System becomes
the most environmentally compatible energy system. It will not produce
greenhouse gases, ozone-layer-
damaging chemicals, oil spills, and climate change and will produce little or no
acid rain ingredients and
pollution. It will actually reverse the global warming and bring the Earth back
to its normal temperatures
by decreasing the CO2 in the atmosphere to its level before the Industrial
Revolution.
• 6.The Solar Hydrogen Energy System has the lowest effective cost when
environmental damage and the
higher utilization efficiency of hydrogen are taken into account. In other words,
the Solar Hydrogen Energy
System will cost society least when compared with the present Fossil Fuel System
and the Synthetic Fossil
Fuel System.
----------------------- Page 345-----------------------
BIBLIOGRAPHY
1. M. A. Elliot, N. C. Turner, Estimating the future rate of production of the

world’s fossil fuels,Amer. Chem. Soc., 163rd
Natl. Meet., Div. Fuel Chem. Symp. Non-Fossil Chem. Fuels , Boston, 1972.
2. D. Root, E. Attanasi, Amer. Assoc. Pet. Geol. Bull. , 1978.
3. J. D. Parent, A Survey of the United States and Total World Production,
Proved Reserves, and Remaining Recoverable
Resources of Fossil Fuels and Uranium as of December 31, 1977 , Chicago:
Institute of Gas Technology, 1979.
4. J. O. M. Bockris, T. N. Veziroglu, A solar hydrogen economy for U.S.A., Int.
J. Hydrogen Energy , 8: 323, 1983.
5. T. N. Veziroglu, O. Basar, Dynamics of a universal hydrogen fuel
system, Hydrogen Energy, Part B , 1309–1326,
1974.
6. F. Barbir, T. N. Veziroglu, H. J. Plass, Jr., Environmental damage due to
fossil fuels use, Int. J. Hydrogen Energy , 10:
739, 1990.
7. T. N. Veziroglu, F. Barbir, Hydrogen: The wonder fuel, Int. J. Hydrogen
Energy , 17: 391, 1992.
8. T. N. Veziroglu, F. Barbir, Transportation fuel-hydrogen, Energy Technology
and the Environment , New York: Wiley-
Interscience, 1995, vol. 4, pp. 2712–2370.
9. T. N. Veziroglu, Hydrogen technology for energy needs of human settlements,
Int. J. Hydrogen Energy , 12 (2): 99–129,
1987.
10. T. N. Veziroglu (ed.), Hydrogen Energy , parts A and B, New York: Plenum,
1975.
11. T. N. Veziroglu (ed.), Proc. 1st World Hydrogen Energy Conf. , Coral Gables,
FL: Clean Energy Research Institute, Univ.
Miami, 1976.
12. T. N. Veziroglu, W. Seifritz (eds.),Hyrogen Energy Progress, Proc. 2nd World
Hydrogen Energy Conf. , Oxford: Pergamon,
1979.
13. T. N. Veziroglu, K. Fueki, T. Ohta (eds.), Hydrogen Energy Progress, Proc.
3rd World Hydrogen Energy Conf. , Oxford:
Pergamon, 1981.
14. T. N. Veziroglu, W. D. Van Vorst, J. H. Kelley (eds.), Hydrogen
Energy Progress, Proc. 4th World Hydrogen Energy
Conf., Oxford: Pergamon, 1982.
15. T. N. Veziroglu, J. B. Taylor (eds.), Hydrogen Energy Progress,
Proc. 5th World Hydrogen Energy Conf. , Oxford:
Pergamon, 1984.
16. T. N. Veziroglu, N. Getoff, P. Weinzierl (eds.), Hydrogen Energy
Progress, Proc. 6th World Hydrogen Energy Conf. ,
Oxford: Pergamon, 1986.
17. T. N. Veziroglu, A. N. Protsenko (eds.), Hydrogen Energy Progress, Proc. 7th
World Hydrogen Energy Conf. , Oxford:
Pergamon, 1988.
18. T. N. Veziroglu, P. K. Takahashi (eds.), Hydrogen Energy Progress, Proc. 8th
World Hydrogen Energy Conf. , Oxford:
Pergamon, 1990.
19. J. O. M. Bockris, T. N. Veziroglu, with D. Smith, Solar Hydrogen Energy: The
Power to Save the Earth, London: Optima,
1991.
20. T. N. Veziroglu, C. Derive, J. Pottier (eds.), Hydrogen Energy Progress,
Proc. 9th World Hydrogen Energy Conf. , Coral
Gables, FL: International Assoc. Hydrogen Energy, 1992.
21. D. L. Block, T. N. Veziroglu (eds.), Hydrogen Energy Progress, Proc. 10th
World Hydrogen Energy Conf. , Coral Gables,
FL: International Assoc. Hydrogen Energy, 1994.
22. T. N. Veziroglu et al. (eds.), Hydrogen Energy Progress, Proc. 11th World
Hydrogen Energy Conf. , Coral Gables, FL:
International Assoc. Hydrogen Energy, 1996.
23. T. N. Veziroglu, F. Barbir, Hydrogen Energy Technologies , Vienna; UNIDO,
1998.
24. T N. Veziroglu, F. Barbir, Hydrogen energy system: A permanent solution to
the global energy-environmental problems,
35th IUPAC Congr. Proc., Istanbul, Turkey, 1995.
25. W. Fulkerson, R. J. Judkins, M. K. Sanghvi, Energy from fossil fuels, Sci.
Amer., 263 (3): 129–135, 1990.
26. T. N. Veziroglu, I. Gurkan, M. M. Padki, Remediation of greenhouse problem
through replacement of fossil fuels by
hydrogen, Int. J. Hydrogen Energy , 14 (4): 257–266, April, 1989.
27. E. Goebel et al., Geology, composition, isotopes of naturally occurring H2/N2
rich gas from wells near Junction City,
Kansas, Oil Gas J., pp. 217–222, May, 1994.
28. B. Heydorn, SRI Consulting Chemical Economics Handbook, Menlo Park, CA: SRI,
1998.
----------------------- Page 346-----------------------
29. M. Steinberg, H. C. Cheng, Modern and prospective technologies for hydrogen

production from fossil fuels, in T. N.
Veziroglu and A. N. Protensko (eds.), Hydrogen Energy Progress VII , Oxford:
Pergamon, 1988, vol. 2, pp. 699–740.
30. W. L. Mitchell, Development of a partial oxidation reformer for
liquid fuels, Proc. Fuel Cells for Transportation
TOPTEC: Addressing the Fuel Infrastructure Issue, Alexandria, VA, Warrendale,
PA: SAE, 1996.
31. J. G. Pohleny, N. H. Scott, Method for hydrogen production by catalytic
decomposition of a gaseous hydrocarbon atom,
U.S. Patent No. 3,284,161, 1966.
32. N. Muradov, Production of hydrogen by thermocracking of natural gas, Proc.
U.S. DOE Hydrogen Program Review ,
Golden, CO: NREL, 1996, vol. 1, pp. 513–535.
33. S. Lynum, CO2 -free hydrogen from hydrocarbons—The Kvaerner CB&H process,
10th World Hydrogen Energy Conf.,
Cocoa Beach, FL, 1994.
34. National Hydrogen Association, The Hydrogen Technology Assessment, Phase I,
Report for NASA, Washington, DC,
1991.
35. M. Bonner et al., Status of advanced electrolytic hydrogen production in the
United States and abroad, Int. J. Hydrogen
Energy , 9 (4): 269–275, 1984.
36. S. Dutta, Technology assessment of advanced electrolytic hydrogen production,
Int. J. Hydrogen Energy , 15 (6): 379–
386, 1990.
37. H. Wendt, Water splitting methods, in C.-J. Winter and J. Nitsch (eds.),
Hydrogen as an Energy Carrier , New York:
Springer-Verlag, 1988, pp. 166–238.
38. M. A. Liepa, A. Borhan, High-temperature steam electrolysis: Technical and
economic evaluation of alternative process
designs, Int. J. Hydrogen Energy , 11 (7): 435–442, 1986.
39. O. G. Hancock, Jr., A photovoltaic-powered water electrolyzer: Its
performance and economics, in T. N. Veziroglu and
J. B. Taylor (eds.), Hydrogen Energy Progress V , New York: Pergamon, 1984,
pp. 335–344.
40. C. Carpetis, An assessment of electrolytic hydrogen production by means of
photovoltaic energy conversion, Int. J.
Hydrogen Energy , 9 (12): 969–992, 1984.
41. A. Siegel, T. Schott, Optimization of photovoltaic hydrogen
production, Int. J. Hydrogen Energy , 13 (11): 659–678,
1988.
42. H. Steeb et al., Operation experience of a 10 kW PV-electrolysis system in
different power matching modes, in T. N.
Veziroglu and P. K. Takahashi (eds.), Hydrogen Energy Progress VIII , New
York: Pergamon, 1990, vol. 2, pp. 691–700.
43. H. Blank, A. Szyszka, Solar hydrogen demonstration plant in Neunburg vorm
Wald, in T. N. Veziroglu, C. Derive, and
J. Pottier (eds.), Hydrogen Energy Progress IX , vol. 2, Paris: M. C. I.,
1992, pp. 677–686.
44. W. Grasse, F. Oster, H. Aba-Oud, HYSOLAR: The German-Saudi Arabian
program on solar hydrogen-5 years of
experience, Int. J. Hydrogen Energy , 17 (1): 1–8, 1992.
45. P. Lehman, C. E. Chamberlain, Design of a photovoltaic-hydrogen-fuel cell
energy system, Int. J. Hydrogen Energy ,
16 (5): 349–352, 1991.
46. P. D. Lund, Optimization of stand alone photovoltaic system with hydrogen
storage for total energy self-sufficiency,
Int. J. Hydrogen Energy , 16 (11): 735–740, 1991.
47. A. G. Garcia-Conde, F. Rosa, Solar hydrogen production: A Spanish experience,
in T. N. Veziroglu, C. Derive, and J.
Pottier (eds.), Hydrogen Energy Progress IX , vol. 2, Paris: M. C. I., 1992,
pp. 723–732.
48. S. Z. Baykara, E. Bilgen, An overall assessment of hydrogen production by
solar water thermolysis, Int. J. Hydrogen
Energy , 14 (12): 881–889, 1989.
49. H. Engels et al., Thermochemical hydrogen production, Int. J. Hydrogen Energy
, 12 (5): 291–295, 1987.
50. S. Yalcin, A review of nuclear hydrogen production, Int. J. Hydrogen Energy ,
14 (8): 551–561, 1989.
51. S. R. Bull, Hydrogen production by photoprocesses, Proc. Int. Renewable
Energy Conf. , Honolulu, 1988, pp. 413–426.
52. I. Willner, B. Steinberger-Willner, Solar hydrogen production through
photobiological, photochemical and photoelec-
trochemical assemblies, Int. J. Hydrogen Energy , 13 (10): 593–604, 1988.
53. J. B. Taylor et al., Technical and economic assessment of methods for the
storage of large quantities of hydrogen, Int.
J. Hydrogen Energy, 11 (1): 5–22, 1986.
54. C. Carpetis, Storage, transport and distribution of hydrogen, in C.-J. Winter
and J. Nitsch, (eds.), Hydrogen as an
Energy Carrier , New York: Springer-Verlag, 1988, pp. 249–289.
55. J. D. Pottier, E. Blondin, Mass storage of hydrogen, NATO ASI Ser., Ser. E
295: 167–180, 1995.
----------------------- Page 347-----------------------
56. F. Mitlitsky, Development of an advanced, composite, lightweight, high

pressure storage tank for on-board storage
of compressed hydrogen, Proc. Fuel Cells for Transportation TOPTEC:
Addressing the Fuel Infrastructure Issue ,
Alexandria, VA, Warrendale, PA: SAE, 1996.
57. T. N. Veziroglu, Hydrogen technology for energy needs of human settlements,
1987.
58. J. A. Schwartz, K. A. G. Amankwah, Hydrogen storage systems, Geol. Surv. Prof.
Pap. (U.S.), 1570: 725–736, 1993.
59. P. Hoffmann, Boston team claims development of 5,000-mile range onboard
hydrogen storage method, Hydrogen Fuel
Cell Lett., 12 (2): 1, 1997.
60. A. C. Dillon, K. M. Jones, M. J. Heben, Carbon nanotube materials for hydrogen
storage, Proc. U.S. DOE Hydrogen
Program Rev. , Golden, CO: NREL, 1996, vol. 2, pp. 747–763.
61. G. Rambach, C. Hendricks, Hydrogen transport and storage in engineered glass
microspheres, Proc. U.S. DOE Hydro-
gen Program Rev. , Golden, CO: NREL, 1996, vol. 2, pp. 765–772.
62. C. Jensen, Hydrogen storage via polyhydride complexes, Proc. U.S. DOE Hydrogen
Program Rev. , Golden, CO: NREL,
1996, vol. 2, pp. 787–794.
63. J. D. Pottier, Hydrogen transmission for future energy systems, NATO ASI Ser.,
Ser. E 295: 181–194, 1995.
64. F. Oney, The comparison of pipeline transportation of hydrogen and
natural gas, M.S. Thesis, Univ. Miami, Coral
Gables, 1991.
65. J. M. Norbeck et al., Hydrogen Fuel for Surface Transportation , Warrendale,
PA: SAE, 1996.
66. H. J. Sternfeld, P. Heinrich, A demonstration plant for the hydrogen/oxygen
spinning reserve, Int. J. Hydrogen Energy ,
14: 703–716, 1989.
67. K. Kinoshita, F. R. McLarnon, E. J. Cairns, Fuel Cells: A Handbook ,
Morgantown, WV: U.S. Department of Energy,
1988, DOE/METC88/6069.
68. L. J. M. J. Blumen, M. N. Mugerwa (eds.), Fuel Cell Systems , New York:
Plenum, 1993.
69. C. E. Thomas, Preliminary Hydrogen Vehicle Safety Report , prepared
for Ford Motor Company, Contract No. DE-
AC02-94CE50389, Washington, DC: U.S. Dept. of Energy, 1996.
70. M. R. Swain, M. N. Swain, A comparison of H , CH , and C H
fuel leakage in residential settings, Int. J. Hydrogen
2 4 3 8
Energy , 17 (10): 807–815, 1992.

71. M. DiChristina, What really downed the Hindenburg, Pop. Sci. , November, 1997,
pp. 71–76.
72. G. D. Brewer, The prospects for liquid hydrogen fueled aircraft, Int. J.
Hydrogen Energy , 7: 21–41, 1982.
73. C. J. Winter, J. Nitsch, Hydrogen as an Energy Carrier , Berlin: Springer-
Verlag, 1988.
74. S. A. Sherif, M. Lordgooei, M. T. Syed, Hydrogen liquefaction, in T. N.
Veziroglu (ed.), Solar Hydrogen Energy System,
Final Tech. Rep., Coral Gables, FL: Clean Energy Research Institute, Univ.
Miami, 1989, pp. C1–C199.
75. K. M. Mackay, The element hydrogen, ortho- and para-hydrogen, atomic hydrogen,
Comprehensive Inorganic Chem-
istry, New York: Pergamon, 1973, pp. 1–22.
76. R. D. McCarty, J. Hord, H. M. Roder, Selected Properties of Hydrogen
(Engineering Design Data), Washington, DC:
U.S. Dept. Commerce, Natl. Bureau Standards, 1981. Monog. No. 168.
77. B. A. Hands, Cryogenic Engineering, New York: Academic Press, 1986.
78. C. L. Newton, Hydrogen production, liquefaction and use, Cryog. Eng. News,
Part I, No. 8, pp. 50–60, 1967.
79. C. L. Newton, Hydrogen production, liquefaction and use, Cryog. Eng. News,
Part II, No. 9, pp. 24–29, 1967.
80. C. R. Baker, R. L. Shaner, A study of the efficiency of hydrogen liquefaction,
Int. J. Hydrogen Energy , 3: 321–334,
1978.
81. R. F. Barron, Cryogenic Systems, Oxford, UK: Oxford Univ. Press, 1985.
82. E. L. Huston, Liquid and solid storage of hydrogen, Hydrogen Energy Progress V
, pp. 1171–1186, Toronto, Ontario,
Canada: Pergamon, 1984.
83. J. B. Taylor et al., Technical and economical assessment of methods for the
storage of large quantities of hydrogen,
Int. J. Hydrogen Energy , 11 (1): 5–22, 1986.
84. R. B. Scott, Cryogenic Engineering, Princeton, NJ: Van Nostrand, 1962.
85. H. H. Ewe, H. J. Selbach, The storage of hydrogen, in W. E. Justi (ed.), A
Solar Hydrogen Energy System , London:
Plenum, pp. 243–263, 1987.
86. A. Kalanidhi, Boil-off in long term stored liquid hydrogen, Int. J. Hydrogen
Energy , 13 (5): 311–313, 1988.
87. J. A. Barclay, A comparison of the efficiency of gas and magnetic
refrigerators, Proc. 22nd Natl. Heat Transfer Conf. ,
Niagara Falls, NY, 1984.
----------------------- Page 348-----------------------
88. J. A. Barclay, Magnetic refrigeration for low-temperature applications, Proc.

Cryocooler Conf. , Colorado, 1984.
89. J. A. Barclay, W. C. Overton, Jr., W. F. Stewart, Phase I Final Report;
Magnetic Reliquefication of LH2 Storage Tank
Boil-off , Ref. No. PT-SPD/6037/2511 C/000000/04/82, (NASA-Defense
Purchase Req. CC-22163B), NASA-Kennedy
Space Center, 1983.
90. J. A. Barclay, The Theory of an Active Magnetic Regenerative Refrigerator,
NASA-R-2287, Washington, DC: NASA,
1983.
91. J. A. Barclay, W. A. Steyert, Materials for magnetic refrigeration
between 2 K and 20 K, Cryogenics, 22: 73–79,
1982.
92. J. A. Barclay, W. F. Stewart, The effect of parasitic refrigeration on the
efficiency of magnetic liquefiers, Proc. 17th
Intersoc. Energy Convers. Eng. Conf. (IECEC ’82) , Los Angeles, 1982, p.
1166.
93. J. A. Barclay, Use of a ferrofluid as the heat-exhange fluid in a magnetic
refrigerator, J. Appl. Phys., 53 (4): 2887,
1982.
94. J. A. Barclay, S. Sarangi, Selection of regenerator geometry for magnetic
refrigerator applications, Proc. ASME Winter
Annu. Meet. , New Orleans, LA, 1984.
95. J. A. Barclay et al., Experimental results on a low-temperature magnetic
refrigerator, Cryog. Eng. Conf., Boston, 1985.
96. N. S. Brown, Advanced long term cryogenic stoage systems, NASA-
Marshal Space Flight Center, Huntsville, AL,
NASA Conf. Publ. 2465P: 7–16, 1986.
97. A. Bejan, J. L. Smith, Jr., Thermodynamic optimization of mechanical supports
for cryogenic apparatus, Cryogenics,
pp. 158–161, March, 1974.
98. A. Bejan, Discrete cooling of low heat leak supports to 4.2 K, Cryogenics, 15
(1): 290–292, May, 1975.
99. A. Bejan, Entropy Generation Through Heat and Fluid Flow , New York: Wiley,
1982.
100. S. Gursu et al., An optimization study of liquid hydrogen boil-off losses,
1992.
101. S. Gursu et al., Analysis and optimization of thermal stratification and self
pressurization effects in liquid hydrogen
storage systems, Am. Soc. Mech. Eng. [Pap.] , 91-WA-HT-4: 1991.
102. A. S. Rapial, D. E. Daney, Preparation and characterization of slush and
nitrogen gels, NBS Tech. Note (U.S.) , 378:
1969.
103. E. W. Justi, A Solar Hydrogen Energy System , New York: Plenum, 1987.
104. C. R. Baker, L. C. Matsch, Production and distribution of liquid hydrogen,
Adv. Pet. Chem. Refin. , 10: 37–81, 1965.
105. S. W. Kandebo, Researchers explore slush hydrogen as fuel for national aero-
space plane,Aviat. Week Space Technol. ,
pp. 37–38, June 26, 1989.
106. C. F. Sindt, A summary of the characterization study of slush hydrogen,
Cryogenics, 10: 372–380, 1970.
107. R. O. Voth, Producing liquid-solid mixtures (slushes) of oxygen or hydrogen
using an auger, Cryogenics, 25: 511–517,
1985.
108. D. E. Daney, D. B. Mann, Quality determination of liquid-solid hydrogen
mixtures, Cryogenics, 7: 280–285, 1967.
109. D. B. Mann et al., Liquid-solid mixtures of hydrogen near the triple point,
Adv. Cryog. Eng. , 11: 207–217, 1966.
110. E. M. Smith, Slush hydrogen for aerospace applications, Int. J. Hydrogen
Energy , 14 (3): 201–213, 1989.
111. H. C. Angus, Storage, distribution and compression of hydrogen, Chem. Ind.
(London), 16: 68–72, 1984.
112. C. F. Sindt, P. R. Ludtke, Slush hydrogen flow characteristics and
solid fraction upgrading, Adv. Cryog. Eng. , 15:
382–390, 1970.
113. C. F. Sindt, P. R. Ludtke, D. E. Daney, Slush hydrogen fluid characterization
and instrumentation, NBS Tech. Note
(U.S.), 377: 1969.
114. L. R. Perrel, D. G. Haase, Self-catalysis in solid hydrogen, a computer
simulation, Amer. J. Phys. , 52 (9): 831–833,
1984.
115. H. M. Roder, A new phase transition in solid hydrogen, Cryogenics, 13 (7):
439–440, 1973.
S. A. SHERIF
University of Florida
F. BARBIR
Energy Partners, Inc.
T. N. VEZIROGLU
University of Miami
----------------------- Page 349-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Hydrothermal Power Systems
Standard Article

Xiaohong Guan 1 1 2 3
, Peter B. Luh , Ernan Ni , Renhou Li
1Xian Jiaotong University, Xian, Shaanxi, China
2University of Connecticut, Storrs, CT

3Xian Jiaotong University, Xian, Shaanxi, China
Search
reserved.
❍ Acronym Finder
DOI: 10.1002/047134608X.W3026
Abstract
Problem Description
Solution Methodologies
Conclusions

%20Conversion/W3026.htm17.06.2008 16:42:15
----------------------- Page 350-----------------------

c
HYDROTHERMAL POWER SYSTEMS
A modern power system usually contains various types of generating sources

including thermal (nuclear, fossil
steam, gas turbine), hydro (run-of-the-river and regular plants), and pumped-
storage units, as well as other
renewable sources such as wind and solar. The primary goal of
hydrothermal power system operation is to
provide reliable and economical power (1) for electricity users. To realize this
goal, many operational issues
have to be addressed, including those at a snapshot such as automatic generation
control and optimal power
flow (see Power system control) and those over a time horizon. One of
the most important daily tasks is
hydrothermal scheduling or hydrothermal coordination. This determines when and at
what levels units should
be run to meet the system demand and reserve requirements while satisfying
individual unit constraints with
a minimum total generation cost. The economic consequence of operation
scheduling is significant. A small
percentage of reduction in generation costs may result in savings of millions of US
dollars for a large utility
company (2). With deregulation of the electric utility industry a global trend, the
daily scheduling activities
may be changed for the new environment. The task of coordinating generating
resources to minimize costs or
to maximize profits, however, will become even more important.
The operating characteristics and generation costs of generators
differ significantly from one unit to
another. Identifying these characteristics and properly coordinating their
operations can make full use of their
advantages to reduce total generation cost. Among thermal units, the
operating costs of nuclear units are
cheaper than others. These units are thus often used as “base load units” with
generation levels fixed at the
full capacity in view of economic, safety, and environmental reasons. Most large
fossil steam units are also base
load units, whereas others operate in cycles following the system demand with
additional costs and constraints
associated with their startups and shutdowns. Gas turbines can be quickly started
but have high fuel costs and
are often used to satisfy peak demand. Compared to the costs of thermal units,
generation costs of hydro units
are usually insignificant. However, their “fuel,” the natural water inflow, is
limited within a time frame, and
their operations may be coupled because of the hydraulic connection of reservoirs
within a river catchment.
A pumped-storage unit pumps water into its upper reservoir during the
period of low demand, using the
economical energy generated by base load units, and generates at the time of high
demand and high costs. It
can smooth peak load, provide reserve, and play an important role in reducing the
overall generating costs.
Based on the goals of operations and the time horizon under consideration,
hydrothermal scheduling can
be classified into several levels of hierarchy as shown in Fig. 1.
Nuclear units need to be refueled once every few years, and the
plants must be shut down several
weeks during refueling for maintenance. Preventive maintenance is also needed for
other units and is usually
scheduled over a horizon of one to several years. Midterm hydro scheduling or
seasonal hydro planning produces
hydro schedules over a time frame from several months to a year, and determines how
to distribute the total
water resource among the weeks under consideration. In the process, it is necessary
to estimate natural water
inflows, system demand, and unit unavailability due to unscheduled unit
outages based on historical and
forecast data. Maintenance scheduling in conjunction with midterm hydro scheduling
ensures that units and
hydro energy are available at the times of high demand. Taking weekly hydro energy
allocated by the midterm
scheduling as an input, the short-term hydrothermal scheduling generates schedules
for thermal and hydro
----------------------- Page 351-----------------------

2 HYDROTHERMAL POWER SYSTEMS
Fig. 1. Hierarchies of hydrothermal

scheduling.
units on an hourly basis over a horizon from one day to ten days. It encompasses
two interrelated problems:
(1) the simultaneous determination of the “unit commitment” (unit up and
down) status and preliminary
hourly generation levels for all units and (2) the “economic dispatch”
of power (i.e., the determination of
hourly generation levels) given the unit commitment status. The second
problem is a special case of the
first. Although hydrothermal scheduling provides a preliminary power output for each
unit, actual generation
levels are decided by an economic dispatch calculation on an hourly basis using the
latest demand and cost
information. The focus of this article is on short-term hydrothermal scheduling.
The approaches developed for
short-term scheduling, however, can often be used for midterm and longterm
scheduling problems.
Problem Description
Short-term hydrothermal scheduling has to consider many factors including
• unit generation costs

• system-wide constraints
• individual unit operating constraints
Generation Costs. Generation costs typically include fuel (running)

costs, startup costs, and mainte-
nance costs. Generally only fuel and startup costs of thermal units are considered
in the short-term hydrother-
mal scheduling. The fuel cost of a thermal unit depends on its heat-rate
characteristics and can be expressed
as a load-dependent incremental cost plus an overhead called no-load cost existing
all the time as long as the
unit is running—resulting from the heat loss, wear and tear, etc. This no-load cost
is different from capital
charge, which is incurred whether the unit is running or not and is irrelevant to
scheduling. Thus if a unit is
shut down, its generation cost is assumed to be zero. Typically when a unit is
running, its generation cost can
be expressed by a polynomial function:
----------------------- Page 352-----------------------
HYDROTHERMAL POWER SYSTEMS 3
Fig. 2. Piece-wise linear cost function.
where p (t) is unit power output at time t, and C0 the no-load cost. For many
units, C [p (t)] can be well approxi-
mated by a quadratic, piecewise quadratic, or by an n-segment piecewise linear and
convex function:
as shown in Fig. 2, where p˜ is the generation capacity and is the minimum
generation. The cost curve in Eq.
(1) is smooth, and the cost in Eq. (2) is computationally convenient for economic
dispatch. The forms of the cost
functions, however, are not essential for solving hydrothermal scheduling problems.
The startup cost of a thermal unit is made up of the fuel and operating costs
to bring the unit on-line; it
is a time-varying functuion because it depends on the unit status (i.e., the time
since last shutdown). In the
literature, this cost is often assumed to be an exponential function of unit down
time as
where t is the time since last shutdown, Cs a positive constant representing the
cold startup cost, and d is the
boiler cool-down coefficient. For certain units, the startup cost is modeled as a
linear function of unit downtime,
while remaining constant after the cold startup time (3). The startup or
shutdown costs for a hydro or a
pumped-storage unit are usually ignored because they are insignificant as compared
to those of a thermal unit.
Frequent startups and shutdowns, however, may increase maintenance costs (4).
Therefore, penalty terms can
be added to discourage excessive startups and shutdowns of hydro and pumped-storage
units.
System-Wide Constraints. The system demand or load balance
equation is written as
where p ti (t), p hj (t), p pk (t) are generation levels of thermal unit i, hydro
unit j , and pumped-storage unit k at
time t, respectively; Pd (t) the total system demand at time t containing the
aggregated customer demand and
losses (approximately accounted for by solving an optimal power flow problem). The
generation of a pumped-
storage unit may take negative values representing pumping. If transactions among
utilities are considered,
the system demand in Eq. (4) should also add the net power sold and deduce the net
power purchased.
The system reserve requirements include spinning and supplemental reserves
which are the extra power
that the system should be able to provide within specified times guard against
disturbances such as unscheduled
----------------------- Page 353-----------------------
outage of a unit or unforeseen load changes. The spinning reserve is the reserve
that on-line units can provide.
Generally, the spinning reserve contribution of an on-line unit is the amount of
additional power that the unit
can provide within, say, 10 min {i.e., r [p (t)] = min[p˜ −p (t), r˜], where p (t)
is the unit’s power output, p˜ is the
generating capacity, and r˜ is the maximum reserve contribution determined by the
unit’s ramping capability}.
The reserve is zero if the unit is down, except for some specifically designated gas
turbines that can be started
very quickly. The system-wide spinning reserve constraints require that the total
reserve contribution should
be no less than a specified requirement Pr (t) at any time; in other words,
Supplemental reserves such as 10 min nonspinning reserve or 30 min reserve can be

defined in the similar
way.
In short-term scheduling, the capacity of transmission networks is often
assumed to be large enough, and
transmission constraints are generally ignored. With the deregulation of the
electric power industry and open
access of transmission networks, there is an increasing pressure not only to
optimize the use of generating
resources but also to utilize transmission networks optimally (see Electricity
supply industry). Therefore the
capacity of transmission networks, viewed as power flow limits on branches and/or
voltage bounds on buses, is
becoming important in operation scheduling. Integrated resolution of the scheduling
and transmission problem,
however, is complicated and difficult (see Electricity supply industry and
Electricity supply industry) in view
of the sizes of practical transmission networks. Adding a large number of nonlinear
transmission constraints
in hydrothermal scheduling would drastically increase computational requirements.
Therefore, transmission
networks are usually simplified to a dc power flow model (5,6), or
transmission constraints are indirectly
considered by limiting the generations of individual regions (7). Nevertheless,
methods for solving integrated
scheduling and optimal power flow problems have been presented with limited success
(8,9).
Individual Unit Operating Constraints. Individual unit constraints
may be quite different from one
unit to another in view of the differences in unit characteristics and principles
of operation. A nuclear unit
seldom changes its generation level unless absolutely necessary because of economic
and safety considerations.
Run-of-the-river hydro units are not scheduable (i.e., the natural water inflows are
used instantaneously for
generation). Almost all units are subject to minimum generation and capacity
constraints as in Eq. (6)
The minimum up (down) time constraints specify how long a unit must be
kept on (down) before it can be
shut down (started up). They are imposed on large thermal units because of heat
stress on the equipment, and
on hydro units to prevent frequent startups and shutdowns. Ramp rate constraints
specify a unit’s maximum
allowable generation change between two consecutive time instances, as described by
where # is the unit’s ramp rate. The ramp rate constraints couple the
unit’s generation levels over the
scheduling horizon and are difficult to deal with. In addition, some
units can be “must-run,” or “must-not-
run,” or must generate at predetermined levels during specified periods for
scheduled maintenance or testing
purposes.
The power output of a hydro unit is a function of the amount of water
discharge and reservoir head—the
elevation difference between the upstream and downstream water surfaces of a
reservoir (see Hydroelectric
----------------------- Page 354-----------------------
Fig. 3. A typical hydro power output versus discharged

water and reservoir head.
power stations). A typical power output versus discharged water and head is shown
in Fig. 3. The minimum and
maximum water discharge, and thus the minimum and maximum power output, can also be
head-dependent.
Because the reservoir head is a function of the volume of water stored given the
shape of the reservoir bed,
the power output is thus generally modeled as a function of the volume of water
stored and water discharge.
Forbidden operating regions are common for many units (i.e., a unit may
not operate in certain regions or
may even be limited to operate at discrete points for safety or efficiency reasons).
Pumped-storage units have
characteristics similar to those of hydro units (see Hydroelectric power stations)
and can significantly contribute
to system reserve in both generating and pumping modes.
Scheduling hydro units is generally more difficult than scheduling thermal
units because the reservoirs
in a river catchment or water shed are hydraulically coupled. The water discharged
by an upstream unit affects
the levels of the downstream reservoirs. The water flow in a river catchment can be
described by the following
the water balance equation
where v (t), w (t), and s (t) are, respectively, the reservoir content, water
discharge, and spillage of hydro unitj
j j j
at time t, ζ (t) is the sum of natural inflows to reservoirj at t, τ is the time
required for water discharged fromj to
j j
reach its directly downstream reservoir, and U is the set of direct upstream
reservoirs of reservoirj . Therefore
j
the schedules of hydro units in a river catchment are coupled not only
through system-wide demand and
reserve requirements in Eqs. (4) and (5) but also the water balance in Eq. (8). The
hydro scheduling problem
can also be viewed as a resource allocation problem—there is limited
hydro energy in a cascaded reservoir
system, and the problem is to determine how to allocate the energy to maximize the
benefits or savings.
The water volume in a reservoir is also constrained by the reservoir
capacity and its minimum (or dead)
volume as
From this analysis, the short-term hydrothermal scheduling involves discrete

decision variables such as unit
on/off and continuous decision variables such generation levels. It can thus be
formulated as a mixed-integer
----------------------- Page 355-----------------------
programming problem:
where Sti (t), Shj (t), and Spk (t) are, respectively, the startup costs
for thermal, hydro, and pumped-storage
units, subject to system-wide generation, reserve, and possibly transmission
constraints, water flow balance
constraints of each river catchment, and individual unit constraints. This
problem mathematically belongs
to the class of NP (Nonpolynomial) hard combinatorial optimization problems
(i.e., the computational time
required to find an optimal solution escalates exponentially with problem size).
Consistently obtaining optimal
schedules has been proven to be extremely difficult for systems of practical sizes.
Nevertheless, because the
cost to be minimized is the sum of individual unit costs and the system-wide demand
and reserve requirements
are also unit-wise additive, the problem is “separable,” and efficient near-optimal
approaches can be developed
as will be presented later.
For midterm and longterm scheduling problems, since many parameters
are stochastic in nature, un-
certainties may have to be considered. Examples include demand, unit
availability, and natural inflows of
reservoirs. For instance, if a possible flood is not appropriately considered, hydro
energy may be forced to spill
out during the flood. On other hand, if an unusual draught is not properly taken
into account, the hydro energy
may not be used when it is most needed. By dividing a year into several seasons, it
is possible to approximate
the demand and inflows as stationary series using Markov chains, and the problem is
to minimize the expected
value of the cost function in Eq. (10). The resulting solution is a “decision
policy” detailing what to do under
each possible circumstance, as opposed to a sequence of decisions for a
deterministic problem. The decision
policy can, in turn, provide targets or constraints for the short-term scheduling
problem.
Solution Methodologies
The approaches for solving hydrothermal scheduling problem can be classified into
five categories (2):
• Partial enumeration such as branch and bound,

• Dynamic programming,
• Bender’s partitioning,
• Heuristics,
• Near-optimal methods such as Lagrangian relaxation and its extensions.
Branch-and-bound is a tree search method with various branching and bounding

(pruning) techniques
carried on a tree as in Fig. 4.
A parent node within the tree represents a problem simpler than the original
one with some constraints
(typically integrality requirements on some discrete variables) relaxed and some
variables fixed at specified
values. The child nodes are a set of problems with more variables specified. The
cost of the optimal schedule
represented by a parent node is therefore a lower bound on those for
its child nodes. If the cost of any
feasible schedule is lower than the optimal cost of a node, the schedules
corresponding its children should be
discarded. This “bounding process” may significantly reduce the space to be
searched. The three elements of
the branch-and-bound method are thus the representation of a problem in terms of a
search tree, the method
for solving problems associated with each node, and the method for searching
(branching and bounding) the
----------------------- Page 356-----------------------
Fig. 4. Search tree of the branch-and-bound

method.
Fig. 5. The state transition diagram. For clarity, only the transitions from the
state “unit 2 up” are shown, and transitions
from other states are similar to those from state “unit 2 up.”
tree. Many methods can be used to solve problems associated with individual nodes.
One such method is to
allow all 0–1 commitment decision variables to take in real values between 0 and 1
for the problem associated
with the root node. Then, for each child of a node, gradually set
certain commitment variables to either 0
or 1. The optimization problem at each node is thus reduced to a standard linear or
nonlinear programming
problem that can be efficiently solved by using an appropriate continuous
variable optimization algorithm.
Many approximations and modifications have been developed to improve the efficiency
of this method.
Dynamic programming based on backward (or forward) searching of a state space
has been extensively
used in many areas. In its applications to hydrothermal scheduling, stages
correspond to hours, and states are
associated with unit commitment status. Figure 5 illustrates the states of two
consecutive stages for a simple
system with 3 thermal units without considering their minimum up and down times.
Calculating the optimal
fuel cost C [p ti (t)] for a given commitment status requires solving an economic
dispatch problem, typically a
constrained nonlinear programming problem, to produce the generation levels of all
units. The optimal “cost-to-
go” functions are then established for all possible states by working backward (or
forward) along the scheduling
horizon.
The optimal decision at each state is then obtained by traversing from the
initial (or terminal) state. It
can be seen that the number of states at each time unit for this simple
configuration is 2I , a very large number
if number of units I is large. When minimum up and down times come into the
picture, a single up state as
in Fig. 5 cannot fully represent the status of a unit because the unit may not be
able to shut down within the
next few hours. The number of states required to describe the problem will thus be
increased to MI , where
M is the sum of minimum up and down times. Dynamic programming can also be used to
include hydro and
pumped-storage units. Many constraints such as discontinuous operating regions and
even discrete generation
levels can be incorporated. However, because the continuous reservoir levels have
to be discretized, a very large
state space is required.
The idea of Bender’s partitioning or Bender’s decomposition is to
decompose a problem into a “master
problem” involving “complicating” (usually discrete) variables and a “slave
problem” involving other (usu-
ally continuous) variables (2). The slave problem typically corresponds to economic
dispatch with fixed unit
----------------------- Page 357-----------------------
commitment status. The marginal costs obtained from the slave problem are used to
limit the commitment of
the units in the master problem, which can be further decomposed into a set of
subproblems associated with
individual units. The master problem then provides a new set of commitment
decisions. The two problems are
iteratively solved until the solutions converge.
For branch-and-bound, dynamic programming, and Bender’s decomposition, the
“curse of dimensionality”
is apparent when there are many units in the system, and these methods
can hardly be used to schedule
problems of practical sizes. Approximations and heuristics have to be
applied at the sacrifice of optimality.
One typical method to simplify the problem is to decompose it into two subproblems:
hydro subproblem and
thermal. The marginal costs obtained from thermal scheduling are used as a basis to
allocate hydro energy over
the hours. Then in thermal scheduling, hydro contributions are deducted from the
system demand and reserve
requirement. Hydrothermal coordination is realized by iteratively updating the
hydro and thermal solutions.
This iterative coordination scheme is heuristic in nature, thus optimality and
sometimes convergence cannot
be guaranteed.
Heuristic methods are still widely used by many electric utilities because of
their simplicity and compu-
tational efficiency. These methods are often developed based on the
knowledge of experienced operators for
specific systems. One method is the so-called priority list commitment. Thermal
units are first ranked based
on their full-load average costs, and units are committed in the ascending order of
the cost to meet system
demand and reserve requirements. Hydro energy is then allocated to
“shave” the peak of system demand
(having highest incremental costs). Manual adjustment is needed to satisfy
individual unit constraints. The
disadvantages of heuristics are that the methods are ad hoc for specific systems,
manual adjustment can be
very time consuming, and the schedule obtained may be far from optimal.
Lagrangian relaxation and its extensions have consistently been proven to
provide near-optimal solutions
in a computationally efficient manner and can be easily adapted to different problem
configurations (3,5,6,7,8,
10,11,12). They are thus among the most successful and most widely used methods and
deserve more attention
here.
Lagrangian relaxation has been fundamental for deriving necessary and
sufficient conditions for optimal-
ity. Over the decades, it has been the corner stone of resource allocation theory
in economics. In mathematics, it
has also shed much light on the development of efficient algorithms for constrained
optimization. The method is
particularly powerful for combinatorial optimization of “separable problems” where
the objective function and
system-wide coupling constraints are additive in terms of basic decision variables.
For the scheduling problem
under consideration, because the total cost to be minimized is the sum of
individual unit costs, and the demand
and reserve requirements are also unit/transaction-wise additive, the problem is
“separable,” and Lagrangian
relaxation can be effectively applied.
The key idea of the approach, in a nut shell, is decomposition and
coordination. “Hard” system demand
and reserve requirements are first “softened” or “relaxed” by having to pay a
“price” or “penalty” for constraint
violations. The per unit violation penalty at a particular hour is the
hour’s “Lagrange multipliers,” or the
“shadow prices” in the economics literature. Because the original problem is
separable, the “relaxed problem”
can be decomposed into many smaller subproblems, one for each unit or transaction.
Given a set of multipliers,
each unit determines its generation levels across the time horizon to
minimize its cost and subject to its
own constraints, and similarly for transactions. These subproblems are not
NP-hard, and can be efficiently
solved by using “dynamic programming.” The multipliers are then iteratively
adjusted based on the levels of
constraint violation following the market economy concept (i.e., increase the
multipliers for undergenerated
hours and reduce the multipliers for overgenerated hours). At the
termination of such multiplier-updating
iterations, system-wide constraints may still be violated at a few hours.
Simple heuristics are then applied
to adjust subproblem solutions to form a feasible schedule satisfying all
constraints. Because the value of the
dual problem is a lower bound of the optimal cost, the quality of the feasible
schedule can be quantitatively
evaluated.
----------------------- Page 358-----------------------

Fig. 6. Framework of the Lagrangian relaxation
method.
The resolution of the original problem is thus performed through a two-level

approach as shown in Fig. 6,
where the low level consists of solving individual subproblems. Coordination of
subproblem soluotions is done
through the iterative updating of Lagrange multipliers at the high level to ensure
near optimality of the overall
solution. In optimization terminology, the nondifferentiable concave “dual
function is iteratively maximized
during the multiplier updating process.
To be more specific, the “Lagrangian” is first formulated as follows:
where λ(t) and µ(t) are, respectively, the “Lagrange multipliers” associated with
system demand and reserve
requirements at time t. By using the duality theory (15) and the decomposable
structure of Eq. (11), a two-level
maximum–minimum optimization problem can be formed. Given the multipliers, the low
level consists of the
following subproblems:
Thermal subproblems (Pt − i), i = 1, 2, . . ., I : min Lti , with
subject to individual thermal constraints.
----------------------- Page 359-----------------------
Fig. 7. The state transition diagram for a thermal subproblem. The minimum down
time is assumed to be 3 hours, and
the minimum up time, 5 hours.
Hydro subproblems (Ph −j ),j = 1, 2, . . ., J : min Lhj , with
subject to individual hydro constraints.

Pumped-storage subproblems (Pp −k), k = 1, 2, . . ., K : min Lpk , with
subject to individual pumped-storage constraints.
Let L∗ti [λ(t), µ(t)], L∗hj [λ(t), µ(t)], and L∗pk [λ(t), µ(t)] denote
the optimal sub-Lagrangians for (Pt − i),
(Ph −j ), and (Pp −k), respectively. Then the high-level dual problem is max
#[λ(t), µ(t)], with
subject to µ(t) ≥ 0. This is a two-level optimization framework. The

solution process thus consists of (1)
solving individual subproblems, (2) optimizing high-level multipliers, and (3)
obtaining a near-optimal feasible
schedule. The thermal subproblems, one for each unit, are typically
solved by using dynamic programming
(3). Stages correspond to hours within the scheduling horizon, and states
within a stage correspond to the
number of hours that the unit has been up or down as shown in Fig.
7, where the minimum up and down
times are assumed as 5 and 3 hours, respectively. The optimal generation cost at a
particular time instance for
a particular state can be obtained by optimizing a single variable function, with
startup and shutdown costs
modeled as state transition costs. The optimal schedule of the unit for a given set
of multipliers can then be
obtained by using dynamic programming with a few states and well-structured state
transitions. The “curse of
dimensionality” caused by the coupling among units as mentioned earlier does not
exist anymore because the
units have been decoupled by using Lagrange multipliers.
----------------------- Page 360-----------------------
Units with ramp rate constraints are harder to handle because generation
levels of two consecutive hours
are now coupled. Ramp rates can be handled through relaxation by introducing
another set of multipliers to
be updated at the intermediate level.
The hydro and pumped-storage subproblems are usually more difficult to solve
because of the hydraulic
coupling among the units within a river catchment. Network flow is an efficient
algorithm for hydro subprob-
lems. However, it cannot handle discontinuous operating regions (e.g., forbidden
regions) nor discrete operating
states (e.g., no generation). In this case, additional multipliers can be
used to relax hydraulic coupling and
updated at the intermediate level (16). Individual hydro subproblems can
then be solved by dynamic pro-
gramming similar to that for thermal subproblems, and a nonlinear network
flow can be applied when the
commitment status of hydro units is fixed.
The high-level dual problem is to update the multipliers so as to
maximize the dual function in Eq.
(15). Because discrete variables are involved at the low level, the dual function
may not be differentiable at
some points. The subgradient method is commonly used to update λ(t) and µ(t)
because of its simplicity and
performance (9,17,18):
where
is the subgradient of #[λ(t), µ(t)] with respect to λ(t), and
is the subgradient of #[λ(t), µ(t)] with respect to µ(t), l is the high-level

iteration index, and α is the step size
which may be adaptively adjusted to speed up convergence (5,10). Other methods have
recently been developed
to update the multipliers with improved performance, including the reduced
complexity bundle method (RCBM )
(19). According to the duality theory, the multiplier λ(t) is the shadow price or
the marginal generation cost
(i.e., the cost for generating an additional MW of power at time t). This is a by-
product of Lagrangian relaxation
and can be used to perform “what-if” analysis.
Subproblem solutions, when put together, are usually infeasible (i.e., the
once relaxed system-wide con-
straints are generally not satisfied). Heuristic methods are thus used to modify
subproblem solutions to form
near-optimal feasible schedules. The generation levels of hydro, pumped-
storage, and units with energy or
ramp rate constraints are difficult to adjust because their generation levels are
coupled across time through
various unit-wise constraints. The schedules of these units are therefore first
modified to satisfy their individ-
ual constraints if these constraints were violated because of the mid-level
relaxation; then they remain fixed.
The generation levels of the thermal units without these constraints are
then adjusted to meet the system
demand and reserve requirements based on subproblem solutions using heuristics
(3,13,14).
----------------------- Page 361-----------------------
An advantage of Lagrangian relaxation is that the dual cost #(t), µ(t)]

defined in Eq. (11) is a lower bound
to all feasible costs including the optimal one. Therefore the relative duality gap
ε defined as
where C is the total generation cost of a feasible schedule as defined in Eq. (10)
can be used to quantitatively
measure the solution quality. If ε is small, the schedule is near optimal because
the optimal cost C∗ must lie
between any feasible cost and the dual cost; in other words,
Another advantage of Lagrangian relaxation is its computational efficiency.

Because the number of dual
iterations generally does not increase as the number of units grows,
computational requirements increase
almost linearly as the problem size grows. A drawback of this approach is that a
heuristic method is most likely
needed to modify the dual solution to obtain a feasible schedule.
In summary, heuristic approaches such as priority-list have been widely used
in practice because of their
simplicity and insensitivity to modeling accuracy as compared to other approaches.
Their disadvantages are
obvious, and there is much room for improvement. At the early stage in developing
optimization-based algo-
rithms, rigorous approaches including branch-and-bound, dynamic programming, and
Bender’s partitioning
were extensively investigated. These approaches, however, suffer from the “curse of
dimensionality” and can
hardly be put into practical use without approximation and simplification. Near-
optimal methods such as Lan-
grangian relaxation and its extensions have been developed for more than two
decades and have become the
main stream because of their near-optimal solution quality and computational
efficiency. Because heuristics
are involved to obtain feasible schedules, there is still room for
improvement. Recent efforts include using
evoloutionary computing approaches such as genetic algorithms to improve feasible
schedules (20,21).
Solving scheduling problems with uncertainties usually requires stochastic
optimization (17,18,22,23,24,
25). One of the approaches is stochastic dynamic programming. The idea is to extend
dynamic programming to
include states with probabilistic transitions and to use expected costs as the
minimization objective. The direct
consequence is that the state space and/or the number of possible transitions would
increase significantly. For
example, when stochastic dynamic programming is used to solve hydro scheduling
problems with stochastic
inflows, one more dimension will be added to the space to include probable inflows in
addition to reservoir levels.
To reduce computational complexity, reservoirs in a river catchment can
be aggregated. The “aggregation–
decomposition” method is to retain one reservoir and aggregate all the others into
an equivalent one (18). The
head effect on water-power conversion, however, can hardly be considered. Another
method uses successive
approximation in which reservoirs are solved one at a time by using stochastic
dynamic programming while
fixing the schedules of other reservoirs. The procedure is repeated until
convergence is obtained; however,
global optimality cannot be guaranteed. Lagrangian relaxation can also be
applied where individual hydro
subproblems with stochastic inflows are solved by using stochastic dynamic
programming.
Another approach is optimization based on scenario analysis (17,25). The
uncertain demand, unit avail-
ability, water inflows, etc., are modeled by a limited number of “scenarios” as
shown in Fig. 8.
----------------------- Page 362-----------------------
Fig. 8. Scenario tree.
Each scenario is associated with a probability of occurrence γm . The

objective is to minimize the expected
cost over M possible scenarios
Because the number of possible scenarios and consequently the computational

requirements increase drasti-
cally as the number of uncertain factors increases, this approach can treat only a
limited number of scenarios.
Conclusions
Hydrothermal scheduling is faced daily by electric power producers. It is a

difficult mixed-integer programming
problem with significant economic impact. Currently heuristic methods such as
priority-list are still widely used
in practice because of their simplicity. Near-optimal methods—Lagrangian relaxation
and its extensions—seem
to be most common among recent scheduling software packages.
Future challenges include the following two aspects. First, more and more
constraints have to be con-
sidered. As mentioned in the section entitled “Problem Description,”
transmission constraints are being in-
corporated into scheduling (5,6,8). Other types of constraints might
emerge or have emerged, for example,
the environmental constraints limiting the amount of pollution from electricity
generation (12). System pa-
rameters such as fuel prices and system demand are usually considered
deterministic, but some of them may
have to be modeled as stochastic or fuzzy in view of the uncertain nature of
competitive markets. All these
would complicate the scheduling problems. The challenge is to develop appropriate
models and corresponding
solution methodologies that capture the essence of the constraints while
maintaining solution quality and
computational requirements. Second, worldwide deregulation of the electric power
industry would profoundly
change the operations of hydrothermal systems therefore the associated scheduling
problems. As presented
earlier, a utility company is traditionally responsible for generating and
delivering power to industrial, busi-
ness, and residential customers in its service area. Under the new market
structure, the utilities participate
in electric markets. The resource scheduling problem still exists but has
to be transformed and integrated
with the market bidding structure. Market participants, including utilities and
Independent Power Producers
(IPP ), have to decide how to submit bids and schedule generating resources to
maximize their profit subject to
various constraints, with bid selection determined by an Independent System
Operator (ISO ). The scheduling
----------------------- Page 363-----------------------
problem will not disappear, but will be transformed into a more complicated bidding
and resource allocation
problem. The results and insights obtained by researchers and practitioners over
the decades on hydrothermal
scheduling have established a solid foundation for us to tackle the new
issues confronting the deregulated
industry.
BIBLIOGRAPHY
1. A. Wood, B. Wollenberg, Power Generation and Control , Chichester, UK: Wiley,

1984.
2. A. Cohen, V. Sherkat, Optimization-based methods for operations
scheduling, Proc. IEEE , 75: 1574–1592,
1987.
3. X. Guan et al., An optimization-based method for unit commitment, Int. J.
Electr. Power Energy Syst. , 14 (1): 9–17,
1992.
4. O. Nilsson, D. Sjelvgren, Hydro unit start-up and their impact on
the short term scheduling strategies of Swedish
power producers, IEEE Trans. Power Syst. , 12: 38–44, 1997.
5. J. J. Shaw, A Direct Method for Security-Constrained Unit Commitment,
IEEE Trans. Power Syst. , 10: 1329–1342,
1995.
6. S. J. Wang et al., Short-term generation scheduling with transmission
constraints using augmented Lagrangian relax-
ation, IEEE Trans. Power Syst. , 10: 1294–1301, 1995.
7. A. F. M. Ferreira et al., Short-term resource scheduling in multi-area
hydrothermal power systems, Elec. Power Energy
Syst., 11 (3): 200–212, 1989.
8. R. Baldick, Generalized Unit Commitment, IEEE Trans. Power Syst. , 10: 465–
473, 1995.
9. H. P. Wolf, H. P. Crowder, Validation of subgradient optimization, Math.
Programming , 6: 62–88, 1974.
10. J. J. Shaw, D. P. Bertsekas, Optimal scheduling of large hydrothermal power
systems, IEEE Trans. Power Appar. Syst. ,
PAS-104: 286–293, 1985.
11. A. Renaud, Daily generation management at Electricite de France: From
planning towards real time, IEEE Trans.
Autom. Control , 38: 1080–1093, 1993.
12. El-Kaib, H. Ma, J. Hart, Environmetally constrained economic dispatch using
Lagrangian relaxation method, IEEE
Trans. Power Syst., 9: 1723–1729, 1994.
13. X. Guan et al., Optimization-based scheduling of hydrothermal power systems
with pumped-storage units, IEEE Trans.
Power Syst. , 9: 1023–1031, 1994.
14. X. Guan, P. B. Luh, L. Zhang, Nonlinear approximation method in
Lagrangian relaxation-based algorithms for hy-
drothermal scheduling, IEEE Trans. Power Syst. , 10: 772–778, 1995.
15. G. L. Nemhauser, L. A. Wolsey, Integer and Combinatorial Optimization ,
Chichester, UK: Wiley, 1988.
16. X. Guan et al., An optimization-based scheduling algorithm for scheduling
hydrothermal power systems with cascaded
reservoirs and discrete hydro constraints, IEEE Trans. Power Syst. , 12: 1775–
1780, 1997.
17. P. Carpentier et al., Stochastic optimization of unit commitment: A new
decomposition framework, IEEE Trans. Power
Syst., 11: 1067–1073, 1996.
18. A. Turgeon, Optimal operation of multireservoir power systems with stochastic
inflows, Water Resource Res., 16 (2):
274–283, 1980.
19. X. Guan, R. Baldick, E. Liu, Integrating power system scheduling and
optimal power flow, Proc. 1996 Power Syst.
Comput. Conf., Dresden, Germany, 1996, pp. 717–723.
20. T. T. Mansfield, G. B. Sheble, Genetic-based unit commitment, IEEE Trans. Power
Syst. , 11: 1996.
21. X. Guan et al., An Genetic Algorithm for Power System scheduling, Proc. 2nd
Chinese World Cong. Intell. Control Intell.
Autom. ,Xian, China, 1997, pp. 1105–1110.
22. V. Quintana, A. Chikhani, A stochastic model for mid-term operation planning
of hydro-thermal systems with random
reservoir inflows, IEEE Trans. Power Appar. Syst. , PAS-100: 1119–1125, 1985.
23. C. Li, R. Yan, J. Zhou, Stochastic optimization of interconnected
multireservoir power systems, IEEE Trans. Power
Syst., 5: 1487–1494, 1990.
----------------------- Page 364-----------------------
24. G. Contaxis, S. Kavatza, Hydrothermal scheduling of a multireservoir

power system with stochastic inflows, IEEE
Trans. Power Syst., 5: 766–772, 1990.
25. S. Takriti, J. Birge, E. Long, A stochastic model for the unit
commitment problems, IEEE Trans. Power Syst. , 11:
1497–1508, 1996.
XIAOHONG GUAN
PETER B. LUH
Xian Jiaotong University
ERNAN NI
University of Connecticut
RENHOU LI
Xian Jiaotong University
----------------------- Page 365-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Magnetohydrodynamic Power Plants
Standard Article

James N. Chapman1
1Tennessee University Space Institute, Tullahoma, TN

reserved.
Search
DOI: 10.1002/047134608X.W3028
❍ Acronym Finder
Abstract
The Mhd Topping Cycle
The Steam Bottoming Cycle

%20Conversion/W3028.htm17.06.2008 16:42:36
----------------------- Page 366-----------------------
260 MAGNETOHYDRODYNAMIC POWER PLANTS
Gas
flow
Figure
2. Illustration of MHD generator concept in showing electri-
cal
conducting gas moving through magnetic flux lines to generate an
electric field. Note that electrodes must be placed in contact with the
gas to
permit the current to flow through an external circuit.
case,
as illustrated in Fig. 2, the electrical field is generated
in
the conducting gas which is passing through a magnetic
field.
In order to utilize this power, electrodes must be in
physical contact with the gas to permit current flow. The
cathode emits electrons and the anode collects them, the same
convention as in an electron tube.
The conducting gas used in power generation is a partially
ionized gas. In the regime of central power generation, it is
described as slightly ionized, having free electron densities on

the
order of 1016 3
electrons/m . The method of producing the
conducting gas, or plasma, as it is more correctly called, leads

MAGNETOHYDRODYNAMIC POWER PLANTS
to a
classification of type of MHD generators: equilibrium and
nonequilibrium. In the equilibrium case, the gas is heated to

Magnetohydrodynamic (MHD) power generation refers to a a
temperature such that a small fraction (around 1 in 10,000
technique in which a conducting gas is moved through a mag- of the
atoms in the gas) are thermally ionized. In nonequilib-
netic field to produce an electric field, which is utilized to pro- rium
ionization, energy is selectively added to the electrons,
duce power. A conventional generator is illustrated in Fig. 1,
resulting in a fraction of ionization that exceeds that due to
in which an electrical conductor is moved through a magnetic
thermal excitation alone. In the latter case, the physics of the
field and an electric field is generated in the wire. In the MHD gas
(especially collision cross section of electrons with heavier
particles and the mean time between collisions) must be such

that
the ionization rate is greater than the recombination
rate.
The conditions that meet this requirement are very low
pressures for combustion gases or the use of noble gases that

Gas have
low collision cross sections, preferably also at reasonably
flow low
pressures. The low-pressure condition severely limits ap-
plication of combustion gases for nonequilibrium applications.

The
use of a noble gas, such as argon, forces the constraint
that
the gas be reused and, hence, is called closed-cycle MHD.
Gas
By
contrast, the equilibrium ionization technique has been
flow
applied primarily to
combustion products which are utilized
N N once
and then exhausted to the atmosphere. Thus, it is classi-
fied as
open-cycle MHD.
To
produce the plasma for open-cycle MHD power genera-
tion,
a fossil fuel such as coal, char, oil, or gas is burned at a
S S high
enough temperature to produce the required ionization.
This
temperature required for ionization is too high for a
practical power system if just the normal combustion prod-

ucts
are present. The lowest possible temperature is about
4500
K, the temperature at which nitrogen oxide (NO) begins
Piston engine generator Turbo-generator to
ionize. However, an element that has a lower ionization
Figure 1. Illustration of conventional generator concepts showing a

potential can be added to reduce this temperature to more
copper conductor moving through magnetic flux lines to generate an
manageable levels. Potassium, with an ionization potential of
electric field. 4.34
eV, is the most practical choice [cesium (with an ioniza-

----------------------- Page 367-----------------------
MAGNETOHYDRODYNAMIC POWER PLANTS 261
tion potential of 3.89 eV) and rubidium (with an ionization

14
potential of 4.16 eV) are the only elements that are lower,
but both are very expensive]. With 1% potassium in the flow,
) 12 P = 5 atm
efficient MHD power production requires a flame temperature
m P = 7 atm
S
of about 2800 K. To achieve this high a flame temperature
( P = 9 atm
y 10
t
with fossil fuel requires significant preheating of the combus-
i
i
tion air or oxygen enrichment of the air or some combination
t
u
of the two techniques.
d 8
n
Fuels that have been considered for MHD power plants
o
include coal, natural gas, oil, bitumen, char, and refinery

l
c 6
i
coke. Fuels that have a high carbon-to-hydrogen ratio produce
r
c
better electrical conductivity at the same conditions. This is
e
l
because the water that is formed when hydrogen burns disas-
E 4
sociates into monatomic hydrogen and an OH radical. The lat-

ter has an affinity for electrons. When it joins with an elec-
tron to become a negative ion, its mass is so much larger than
1100 1200 1300 1400 1500 1600 1700 1800
an electron that it effectively removes that electron from the
Air preheat temperature (K)
conductivity process. However, fuels that are high in hydro-
(a)
gen, such as natural gas, can be used but require a few tens
of degrees Kelvin higher temperature to achieve the same
2850
electrical conductivity. Calculated electrical conductivity and
flame temperature for coal with 1% potassium added as po-
P = 5 atm
tassium carbonate is shown in Fig. 3(a) and (b), respectively,
2800
P = 7 atm
for a range of air preheat temperatures and pressures.
These )
K
calculations are all for a fuel-rich condition, 85% of theoretical
( P = 9 atm
e 2750
r
oxygen. The reasons for choosing this fuel-rich condition are
u
a
twofold: It optimizes the electrical conductivity, and it is es-
r
p 2700
sential for the control of nitrogen oxide emissions to a very
m
e
low level.
t
e
The actual configuration of the MHD generator, in its sim-
m 2650
a
plest form, is shown in Fig. 4(a). This arrangement is called
l
F
a continuous electrode Faraday generator because the elec-
2600
trodes are continuous along the generator length. This con-
figuration is used in some short-duty time applications, such
as the Russian PAMIR, which generates a pulse of several
2550
1100 1200 1300 1400 1500 1600 1700 1800

megawatts for several seconds. However, it is inefficient be-
Air preheat temperature (K)
cause there is a completed path for circulating axial currents
to flow through the plasma and back along the length of the
(b)
electrode, causing electrical losses. This problem is circum-
Figure
3. (a) Electrical conductivity versus air preheat temperature
vented in the segmented Faraday generator as shown in Fig. for
Illinois #6 coal dried to 2% moisture with 1% potassium added
4(b). In this design, the electrodes are narrow strips with inas
potassium carbonate. (b) Flame temperature versus air preheat
sulators between the electrodes, preventing a return path for
temperature for Illinois #6 coal dried to 2% moisture with 1% potas-
circulating currents through the electrodes in the axial direc- sium.
Note the preheat temperature required to achieve 2800 K
tion. The segmented Faraday generator is the most efficient flame
temperature.
configuration, but it suffers from one disadvantage for appli-
cation to commercial power generation: It requires a separate
loading circuit for each pair of electrodes, as shown in Fig. field and
EH is the Hall electric field. Electrodes at this wall
4(b). angle
lie along equipotential lines at design conditions be-
The diagonally conducting generator, as illustrated in Fig. cause
the electric field vector is perpendicular to the elec-
4(c), avoids this problem and its efficiency approaches that of trode.
This generator can be loaded from one end to the other,
the segmented Faraday generator when properly designed or it
can be loaded with more than one load connection along
and loaded. In an MHD generator, when current flows in the the
length of the generator. For a generator designed for cen-
direction mutually perpendicular to the velocity and the mag- tral
power application, where high efficiency and high relia-
netic field, it is called the Faraday current and is in the y - bility
are desired, it would normally be loaded with multiple
direction on the coordinate system shown in Fig. 4. When cur- loads to
approximate the optimum current. This would result
rent flows in a magnetic field, there is a force on moving in
something like four to six separate load circuits versus sev-
charged particles that tends to separate the electrons from eral
hundred for the segmented Faraday configuration. Inso-
the positive ions. This effect is called the Hall effect. The re- far as
known to the author, the diagonal connection is the
sult is an electric field along the length of the generator, only
configuration considered for commercial power genera-
called the Hall field. The diagonal generator loads the vector tion in
a linear open-cycle system.
sum of the Faraday field and the Hall field, and the wall angle A
principal attraction of the MHD generator for central
is determined as tan#1 E /E , where E is the Faraday electric power
generation is its high conversion efficiency, as mea-
F H F
----------------------- Page 368-----------------------
Anode less
than 100 kPa, the conductivity has decreased to around
1 S/m
and MHD generation is no longer attractive. However,
the
working gas still contains considerable sensible energy.
This leads to the concept of a combined cycle MHD steam

Ey Jy power
plant in which a MHD topping cycle is used on a steam
Rankine bottoming cycle. This concept is shown schematically
y Load
in
Fig. 5. Note that there are numerous opportunities to inte-
x Insulator grate
the topping and bottoming cycles for improved efficiency
J × B u and
resource allocation. The requirement for heating the com-
z I
B
bustion air can be incorporated into the bottoming cycle fur-
Cathode
nace
and result in a large efficiency improvement because the
energy is recycled from the less efficient bottoming cycle to

(a) the
more efficient topping cycle. Also, the mechanical power
to
run the compressors for the topping cycle combustion can
be
met with a steam turbine economically, especially in large
Anode
plants. The cooling water from the topping cycle can be incor-
porated into the boiler feedwater heating design to utilize

that
energy. The potassium used as seed in the topping cycle
needs
to be recovered in the bottoming cycle and processed
Load
Ey Jy for
reuse.
The low levels of environmental intrusion from the MHD

y
steam
combined cycle power plant are one of its attractive
Insulator
x
features. Department of Energy studies (1) have shown that
J × B u such
plants have the lowest levels of any of the possible coal-
z
B Cathode fired
combined cycles. Sulfur dioxide emissions are completely
I
eliminated by combination of the sulfur in the fuel with potas-

sium
seed, which is collected as solid potassium sulfate in the
(b)
particulate control system. Nitrogen oxides are formed in
large
amounts (10 kppm to 12 kppm) in the high-temperature
Load
combustors of MHD plants, even at fuel-rich stoichiometries.
I
However, by cooling the gases in a radiant furnace before

E net
completing the combustion, the nitrogen oxides (actually, al-
E net + EH most
completely nitric oxide, NO, at fuel-rich conditions) re-
lax
toward equilibrium at lower temperatures and at a sec-
EH
ondary combustion temperature of about 1000 K (1350#F),

y very
little NO remains (2). The secondary combustion must
x be
made in a way that avoids much temperature rise to avoid
J × B u Conducting sidewalls lying along the
formation of additional nitrogen oxides.
z
B equipotential planes which are
The MHD plant has a higher particle loading in the ex-
perpendicular to net electric field haust
because of the potassium seed added. Furthermore, the
J size
distribution tends to be smaller due to the higher temper-
(c)
atures having vaporized some of the solids. However, the re-
Figure 4. (a) Continuous electrode Faraday MHD generator. (This is

sistivity of the ash is lower due to the high potassium content,
the simplest MHD generator.) (b) Segmented electrode Faraday MHD and
electrostatic precipitators (ESPs) are very efficient in re-
generator. The electrodes are segmented in the X -direction to avoid a
moving the solids, although more capability is required than
return path for circulating currents. (c) Diagonal conducting wall would
be required for a normal coal plant. In addition to need-
MHD generator. The sidewalls are perpendicular to the net electrical ing
to remove the particulate to meet Clean Air Act stan-
field and lie along equipotential lines.
dards, it needs to be removed in the MHD case in order to
recycle the potassium seed. MHD pilot plants have operated

with
a completely invisible stack plume. Volatile organic com-
sured on the basis of the first law of thermodynamics (energy pound
(VOC) emissions, which are essential components for
conservation). That efficiency is defined as the electrical
formation of fine particulate in the atmosphere, are almost
power output divided by the energy lost by the gas stream
completely eliminated in the MHD plant because combustion
while passing through the generator. Alternatively, it may be
efficient enough to produce electrically conducting plasma has
calculated as the electrical power generated divided by the been
shown to contain no unburned elemental or organic com-
sum of the electrical power generated plus the heat loss to the pound
carbon (3). This is an important advantage because
walls. Efficiency as so defined can typically be around 90%, power
plants seek to control particulates in the atmosphere
depending on the size and pressure ratio (thus length) of the
having mean diameters of 2.5 #m and less because the trou-
generator. The MHD generator is only useful in the tempera-
blesome part of these particulates are formed in the atmo-
ture range in which there is sufficient electrical conductivity.
sphere from unburned VOCs, sulfur oxides, and nitrogen
At about 2200 K and at exhaust pressures typically slightly
oxides.
----------------------- Page 369-----------------------
MHP power
Alternating
current Inverter
current
Direct
Electric power
Generator current
Coal handling
Air heater
Radiant
Regenerated MHD channel
boiler ESP
seed Combustor
Superheater
Magnet
Secondary Particles
Slag
combustor Steam Air
dump
To seed regeneration facility

Disposal
Slag
Air Comp
Steam turbine Generator
Electric
power
Figure 5. Schematic of
MHD/steam combined cycle power plant.
3
THE MHD TOPPING CYCLE
typically over 120 MW/m . The walls of the combustor rapidly
become covered with coal slag, solidified next to the wall and
The MHD topping cycle consists of a means of generating the
molten on the flame side. In spite of insulation effects of this
plasma (generally a combustor), a nozzle to accelerate the
slag layer, heat fluxes to the walls average about 100 W/cm2 .
plasma, an MHD generator with associated control, power

The potassium seed is pulverized with the coal and fed to-
take-off system, inverters to convert the generated direct cur-
gether in this configuration. The solids, consisting of coal ash
rent power to alternating current, a magnet to provide a mag-
and seed, are molten at the exit of the combustor and are
netic field in the generator volume, and a diffuser to reduce
blown over with the gas for separation later in the steam bot-
the plasma velocity and recover static pressure before it en-
toming plant. Scale-up of this combustor concept is feasible
ters the radiant furnace. As previously noted, there are sev-
by using clusters of these burners, firing into a common
eral streams that are integrated with the bottoming cycle in-
chamber.
cluding air preheat and/or oxygen production, recovery and
The advantages of this type of combustor are simplicity,
reprocessing of the potassium seed, and use of boiler feedwa-
low cost, and higher, more uniform electrical conductivity be-
ter for cooling of all topping cycle components.
cause of the time available for the seed to reach chemical and
thermal equilibrium with the gas.

Combustor
The other type of combustor that has been developed is the
two-stage, slag-rejecting combustor (5). The concept for this

The combustor used in a MHD power plant faces some ex-
treme conditions compared with the burners in other fossil
plants. It must generate a plasma at about 2800 K and de-
liver it at pressures of perhaps 0.2 MPa to 1.0 MPa. In order
Cooling water tubes
to reach this temperature, it is necessary to have a small vol-
ume, with good mixing of fuel and oxidant and extremely
rapid heat release. When burning coal, the volume must be a
compromise (optimization) whereby time is allowed for carbon
Coal
burnout while keeping the combustor small enough to limit
injector
tube
the heat losses to the walls. There have been two successful
coal combustors demonstrated that differ substantially
in Oxidizer
design.
jet ports
Oil
One of these is a single-stage, full-slag-carryover combus-
Disperser nozzle
tor modeled somewhat after rocket engine technology (4). A
plate
sketch of this combustor is shown in Fig. 6. In this design,
the pulverized coal is injected in a conical pattern from the
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
###################################################################################
############################yyyyy
Refractory
middle of the injector plate. Air is fed through a number of
small holes in the injector plate where a pressure drop is
Figure 6. Single stage MHD coal combustor concept. This design is
##yy#######y#y#y#y#y#y####y#y#y
taken to accelerate the air. These jets of air mix very rapidly
optimized for rapid turbulent mixing and for minimum heat loss to
with the coal and produce high heat release concentrations,
the walls.
----------------------- Page 370-----------------------
design is to produce mixing in the first stage by injecting the The

Nozzle
air tangentially to produce swirl for mixing. The air preheat
A gas
dynamic nozzle is needed between the primary combus-
temperature and stoichiometry are chosen so that the coal
tor and
the MHD generator to accelerate the flow to the de-
ash is liquid in this stage and it is tapped out at the bottom
sired
entrance condition. The voltage generated in the MHD
of this first stage. The gases are taken off tangentially so as
generator is the product of gas velocity, magnetic field flux
to get a deswirl effect going into the second stage. The seed
density, and the distance over which it acts. From this view-
and additional air are added in the second stage. The objec-
point
alone, the velocity should be high. Other considerations
tive of this design is to remove as much coal slag as possible
such as
convective heat transfer coefficient, size of generator,
before adding the potassium seed to minimize the amount of
and
total amount of power that can be generated from a unit
potassium seed that chemically combines with the slag and
of
fuel, however, place upper limits on the desired entrance
may be more difficult to recover. Slag removal efficiencies of
velocity. In practice, a high subsonic flow (Mach number 0.8

up to 50% were demonstrated experimentally with coal
to
0.9) or low supersonic flow (Mach number 1.1 to 1.3) is
ground to 70% through 200 mesh (6). Additional slag removal
normally chosen for central power generation. It is desirable

can be achieved with larger particles of coal, but the carbon
to
choose the velocity and area lofting to avoid the formation
burnout suffers as the coal particle size increases. Also, this
of a
shock in the generator. Thus, a subsonic generator should
combustor has problems in scale-up because as the size of the
be
lofted to remain subsonic throughout its length, and a su-
combustor increases, slag removal decreases unless the air
personic generator should be designed to remain supersonic

injection velocity is increased accordingly. As noted, the ad-
throughout its length and under all operating conditions. This

vantage of this design is that it removes some of the coal ash
places
some difficult restrictions on turn-down of the genera-
before the potassium seed is injected. The disadvantages are
tor to
operate at part load conditions.
larger size, more heat loss, and possibly lower electrical con-
The
isentropic one-dimensional gas dynamic relations can
ductivity due to insufficient residence time for the potassium
be used
to design the nozzle. The supersonic nozzle cross-sec-
seed to reach thermal and chemical equilibrium.
tional
area converges to a throat, and then it diverges to the
Regardless of which type of combustor is used, the primary
area
required for the desired entrance Mach number. The
stoichiometry should be chosen to meet two objectives: to
subsonic nozzle must be lofted to provide a throat (a short
maximize the electrical conductivity and minimize the nitro-
supersonic region) and then it converges to produce the de-

gen oxides emitted by the plant. A plot of flame temperature
sired
subsonic entrance Mach number. Heat fluxes to the walls
and electrical conductivity versus primary stoichiometry is
are
normally higher in the nozzle throat than in any other part
shown in Fig. 7. It shows that the primary combustor should
of the
plant. Design of the cooling for the nozzle, especially at
be operated under fuel-rich conditions at about 85% to 90%
the
throat, is a challenge. The best success has been demon-
of theoretical oxygen to maximize temperature and electrical
strated
with channeled cooling water passages with high-
conductivity. This happily turns out to be compatible with
velocity water to ensure good convective heat transfer between

very low nitrogen oxide emissions as well.
the
inside of the cooling water passage and the water.
The MHD
Generator
2880 11.6
Analysis and characterization of the MHD generator involves
the
consideration of both Maxwell’s equations for electromag-
netic
fields and the conservation of mass, energy, and momen-
2870 11.4 tum
equations for fluid flow (7). In addition, of course, an
equation of state is required for the fluid. In MHD calcula-

2860 11.2 ) tions,
these sets of equations are coupled. The electrical ef-
m fect, J
B , is a body force which must be considered in the
/
) S
K ( fluid
flow. Similarly, the electric field produced by the motion
(
y
e 2850 11.0 t
r i of the
electrically conducting gas in the magnetic field must
u v
i
t t be
considered in the electromagnetic solution. This is illus-
a c
r u
e d trated
in Fig. 8. There are no cases of practical interest for
p 2840 10.8 n
m o which
the equations can be solved in closed form. Thus, the
e c
T l
equations must be solved numerically for analysis or design
a
c
i
10.6 r of the
MHD generator. In cases where the interaction is low,
2830 t
c
e it may
be a satisfactory approximation to neglect the effect of
l
E
the
body force, J
B , in the flow calculations. In those cases
2820 10.4
where
the body force is neglected in the calculation of flow
quantities, the flow is said to be uncoupled. We consider here

2810 10.2 only
the regime of MHD generation for central power plants.
0.75 0.80 0.85 0.90 0.95 1.00 In
this regime, the J
B body force is the major sink for
oxidant
momentum and cannot be neglected in the design calculations
Primary stoichiometry ( )
theoretical oxidant for the
MHD generator. Full solutions to this set of equations
Figure 7. Flame temperature and electrical conductivity versus pri- are
available. See, for example, Ref. 8, which is a public do-
mary stoichiometry; coal with oxygen-enriched air, 1% potassium, main
code published by Argonne National Laboratory.
pressure # 7 atm. Note that the electrical conductivity peaks at a The
problem can be simplified for preliminary design cal-
stoichiometry of #0.85.
culations by considering the problem as one-dimensional (9).
----------------------- Page 371-----------------------
Conservation of Mass or Continuity Equation:
Body force
ρuA = m˙ (3)
F = J × B where
A # cross-sectional area of duct and
m˙ = mass flow rate through duct
Ohm’s
Law (including only significant terms for this case):
Conservation
Electric Maxwell's of mass, Fluid
#
and momentum and flow
J E u B E J B
magnetic equations equations
JJ = σ (EE + uu ×BB + EE d ) − B (JJ ×BB ) (4)
energy plus
BB
fields
equation of state
where
Ed is an electrical field loss due to plasma to electrode
voltage drop. The electrical field corresponding to a potential

drop,
#, defined as
E V
d d
# =
=
uB uBd
Ohm’s # =
Hall parameter = electron cyclotron frequency times
law
mean time between collision of electron with
heavy particles
J = (u × B × E )
σ
From
Ohm’s law, an equation for J #E can be derived for use
J is current density
in the
energy equation (10)
B is magnetic field flux density
F is force
J 2
u is velocity
J E − J
JJ ·EE = y B (1 − #) (5)
E is electrical field
σ
σ is electrical conductivity
An
equation of state is required. That is, given two thermody-
Figure 8. Illustration of coupling of the gas-dynamic and electro-
namic
variables such as T and P , a relationship is required
magnetic relations.
such
that the other thermodynamic variables can be calcu-
lated.
This is typically a table of values calculated from a
chemical equilibrium code such as the NASA SP-273 Code or

This simplification involves the assumption that the magnetic one of
its variants. An alternative that is now feasible due to
field is aligned with one axis—for example, the z-axis as
the
greatly expanded capability of modern computers is to use
shown in Fig. 4. This implies that induced magnetic fields due
a
chemical equilibrium code to calculate the other variables
to currents in the generator are negligible. This is certainly
at
each point.
the case in the central power generation regime. We also need
The
electrical conductivity, #, must also be available from
to assume that the current density and flow quantities are
the
table or calculation. The method used by the author was
uniform in a plane perpendicular to the direction of flow, the
originally described by Frost (11).
yz -plane, and are represented by the average value in the
Equations (1), (2), (3), and (5) can be simplified to three

plane. With such assumptions, the simplified equations
independent differential equations in four unknowns, P , T,

become:
U, and
A . If one of these variables is fixed or known, a solu-
tion
for the other variables can be computed along the length
Conservation of Momentum Equation: of the
generator. Once the flow variables are calculated, the
electrical variables can be computed via the relations tabu-

u dP
duu Pw
J B lated
in Table 1 of Ref. 10.
ρu + = −JJy BB = 4 (1)
dx dx D
For
a practical design calculation, constant velocity is the
case
of most interest because this is very near to the maxi-
where
mum
power generation point for a given gas. The equations
can
be modified to perform a constant Mach number case
#W # shear stress
which
is useful in designing for the absence of shocks in the
D # hydraulic diameter
generator. Either case then gives all the electrical and ther-
# # density of the plasma
modynamic variables along the generator and the cross-sec-

tional
area at each point along the length. A loading factor,
Conservation of Energy Equation:
which
relates the impedance of the load to that of the genera-
tor,
is also required. This is defined as
u
dh 2 duu qw
u u J E
ρuu + ρuu = JJ ·EE + 4 (2)
dx dx D
R V
K = L or L
where q # heat flux to walls.
RL + Rg VG
w
----------------------- Page 372-----------------------
where a
diverging duct, with (a) the wall angles chosen to be small
enough to avoid flow separation at the walls and (b) the

RL is load resistance
length sufficient to recover as much static pressure as possi-
Rg is internal generator resistance ble
but no longer so as to minimize the heat loss to the walls.
VL is load voltage For
a supersonic generator, a constant area section is placed
VG is generated voltage at
the generator outlet to cause the flow to shock to subsonic,
then
the same type diffuser is used as for the subsonic dif-
There is an elementary theorem in electrical engineering that
fuser. A common measure of performance of the diffuser is
called the pressure recovery coefficient, C , which is the ratio

says the maximum power transfer from a circuit to a load
p
occurs when the load resistance equals the generator internal of
the recovered static pressure to the dynamic pressure avail-
resistance. Thus, one is tempted to set K # 0.5. This is not a
able. It may be expressed as
good choice, however, because it means that half the gener-
P − P
ated power is dissipated by the current flowing through the
Cp = 2 1
1 2
gas. If a higher K is chosen, the generator becomes longer and
2 ρu
more efficient. The correct choice, in general, will depend on
the overall plant economics in terms of levelized cost of elec-
where
tricity being minimized. The MHD generator tends to be rela-
tively cheap, but making the magnet longer adds considerable P2
is static pressure at the outlet
expense. A loading factor, K , of 0.88 to 0.90 is typical. P1
is static presure at the inlet
For a given load factor and gas conditions, the amount of #
is fluid density at the inlet
power that can be generated depends on the pressure ratio u
is fluid velocity at the inlet
across the generator. The exit pressure must be determined
by the conditions needed at the diffuser entrance in order to The
coefficient of performance for practical diffusers in MHD
recover to the pressure needed at the furnace inlet (it may be
topping cycles are expected to be 0.5 to 0.6. Heat loss in the
slightly less than atmospheric for induced draft designs).
diffuser walls is incorporated into the boiler feedwater heat-
Some common efficiency definitions that apply to the MHD ing
chain and, in some aggressive designs, may be used for
generator are given below
boiling surface.
Electrical power output

Enthalpy extraction = THE
STEAM BOTTOMING CYCLE
Sensible enthalpy of gas
= Pout The
steam bottoming cycle takes the hot gases from the MHD
(h0in − href )m˙
diffuser and uses the sensible heat in them to generate elec-
trical power by a Rankine cycle in a manner that is similar to

where a
conventional coal-fired steam power plant. There are some
differences in the MHD bottoming cycle application, and

h0 is total enthalpy at generator entrance
these will be emphasized in this treatment.
in
href is static enthalpy at reference condition of 1 atm, 298 K

(77#F)
Boiler Design
The
first difference from a conventional steam plant is that
Note that since the only energy loss from the gas is electrical the
hot gases enter from a single duct rather than distributed
power output and heat transfer to the walls, we obtain
burners. The gas is at a higher temperature (2000 K to 2200
K,
3140#F to 3500#F) and is reducing. The gas is reducing
Generator efficiency = Electrical power output
because the gases exiting the MHD cycle must be cooled be-
Decrease in enthalpy across generator fore
completing the combustion. At this point they contain
P0
carbon monoxide, hydrogen, and hydrogen sulfide, which are
=
(h0 − h0 )m˙
known to be corrosive to power plant materials. The basic
in out
structure can be a boiler fabricated of membrane boiler tube,

where h0 # total enthalpy at generator exit. Also, since the but
the interior metal surface must be protected from cor-
out
rosion.
only energy loss from the gas is electrical power output and
A
standard technique is to weld studs to the interior sur-
heat transfer to the walls, we obtain
face
of the boiler tube and refractory coat all metal surfaces
P that
will come into contact with the combustion gases (12).
Generator efficiency = out
This technique is well established commercially because it
Pout + Qw
is
used in cyclone-fired boilers, in some gasifier applications,
and
in the black liquor boilers of paper plants. The refractory
when Qw # total heat transferred from gas to walls.
selected must be rated to withstand the expected gas temper-
atures. It can be anticipated that the refractory will burn

The Diffuser
back
to an equilibrium thickness, thinner in the lower, hotter
The function of the diffuser is to slow the gas down and re-
regions and thicker in the upper, lower-temperature regions.
cover (convert) as much of the dynamic pressure as possible The
effect of the total flow entering the boiler at a single point
to static pressure. The diffuser for a subsonic generator is just
primarily affects the desirability of distributing the flow of
----------------------- Page 373-----------------------
hot combustion gases uniformly over the interior surface of

viewpoint, it is desirable to preheat the secondary combustion
the boiler so as to utilize all the boiling surface efficiently. air
to some modest temperature such as 589 K (600#F). This
can
be done as an alternative to economizer surface in the
Air Heating
plant. Another design issue is the overall plant stoichiometry,
or
the amount of excess air added. The impact of increasing
As noted previously, preheating combustion air is one of the
excess air is increased stack losses. Experimental work with

techniques used to attain a flame temperature sufficient to
a
pilot plant showed that 5% to 10% excess air is a reasonable
get adequate equilibrium ionization of the potassium seed.
design point with control by measurement of CO emissions in

This temperature needs to be around 2800 K (4580#F). With
the
stack. Conventional coal plants typically operate at much
standard atmospheric air, a preheat temperature of 1900 K
higher excess air levels. Staged combustion and stack gas

(2960#F) to 2000 K (3140#F) is needed to achieve this flame
monitoring permit operating at this low level of excess air to

temperature. A method of heating air to this temperature eco-
gain
the benefit of increased efficiency.
nomically is one of the biggest remaining development prob-
lems for the MHD steam combined cycle power plant. (It
should also be noted that there are a myriad of other power
Particulate Loading
plant applications of such an air heater if it were developed.)
The
particulate loading in the steam plant portion of the
It has generally been presumed that a metallic tube air
MHD
steam combined cycle power plant is inherently higher
heater is not applicable to heating air to these temperatures,
than
in a conventional coal-fired power plant. This is because
although it may be the most economical solution for heating
of
the potassium seed added to permit equilibrium ionization
the air to an intermediate temperature. The original MHD
at
reasonable temperatures. Potassium is needed to the ex-
plant concept envisioned the use of recuperative heat ex-
tent
of something like 1% of the total flow. If the potassium
changers such as those used to preheat the air for blast fur-
is
added as a compound such as potassium carbonate or po-
naces. Clearly these are technically feasible, but the huge in-
tassium formate, the total weight percentage of compound

vestments and operational problems inherent in these type
added
is 1.77% and 2.15%, respectively, of the total flow. The
heaters for a power plant make them unlikely final choices.
carbonate or formate disassociates in the combustor of course,

More recently, ceramic materials for heat transfer tubes have
but
it re-forms potassium sulfate or potassium carbonate in
been developed. Tests of individual tubes have shown satis-
the
lower-temperature regions of the steam plant. This turns
factory results for this application, but the problems of mani-
out
to result in typically a doubling of solids content in the
folding, treatment of thermal stress, and so on, remain to be
plant
components compared to operation with just the coal
solved. In the absence of an air heater that will preheat air
ash
particulate. In addition, these potassium compound par-
to these temperatures, the design solution for the plant is to
ticulates tend to be much smaller in size than the coal ash
heat the air to as high as feasible with metallic air heaters
particles. It should be noted that potassium sulfate forms

and use some oxygen enrichment to achieve the required
preferentially as long as any sulfur is available. After all the

flame temperatures (13). A typical solution to this problem is
sulfur is used, the remaining potassium forms carbonate. The

to add oxygen so that the fraction by weight is about 36% and
significance of this observation here is that the carbonate has
preheat the resulting mixture to 922 K (1200#F) in a metallic
a
lower melting point (1170 K, 1646#F) than the sulfate (1342
air heater.
K,
1955#F). Any heat transfer surface having a gas side tem-
perature higher than this will encounter liquid deposits that

Secondary Combustion are
harder to remove by soot-blowing. Since the larger parti-
As previously noted, the primary combustor must operate fuel cles

are more likely to be removed in the furnace or elsewhere
rich for optimum conductivity and low NOx emissions. After in
the plant, the fraction of the total particulate that is com-
the combustion products are cooled to the point that nitrogen posed
of small potassium compounds in the flue gas toward
oxide has decomposed to the desired level, the combustion the
lower temperature part of the plant often approaches 80%
must be completed to recover the remaining fuel energy and to
90%. The design implications of this are threefold: (i) The
prevent emission of unburned and corrosive compounds, espe-
tendency to collect on convective heat transfer surfaces and
cially CO, H , and H S. The design goals are to achieve as block
the gas passages is greatly increased, (ii) heat transfer
2 2
complete a combustion as possible and to avoid a temperature

efficiency of convective sections is reduced (higher fouling fac-
increase that will increase the formation of nitrogen oxides.
tors), and (iii) the removal of sufficient particles to meet stack
These goals are contradictory, of course, and a compromise gas
emission requirements may be more difficult. Early ex-
must be chosen that is acceptable in both respects. The com-
periments with potassium seed resulted in convective sections
pleteness of combustion is largely a function of the efficiency that
were completely plugged to gas flow with deposits. The
of mixing the exhaust gases with the air. This can be modeled
boiler must be designed with wide tube spacing, at least as
numerically or by a laboratory simulation. In order to avoid a wide
as those in used for the worst fouling coals and lignites.
significant temperature increase with the accompanying in- If
the plant will ever burn low sulfur coal, the spacings should
crease in the formation of NOx , the air can be introduced over be
further increased to provide a margin of safety from the
a time period during which heat is removed from the gas. An
expected potassium carbonate deposits. In any case, adequate
alternative technique is to use exhaust gas recirculation, soot-
blowing capability must be provided. The amount of soot-
which involves taking exhaust gases from a lower-tempera-
blowing provided must be balanced against the design fouling
ture region in the plant and, by use of a blower or fan, rein-
factor used in calculating the amount of heat transfer surface
jecting them into the secondary combustor region so as to
required, but it should be noted that fouling factors tend to
keep the mixture temperature from rising. From a combus- be
higher in the MHD case due to the potassium compounds
tion kinetics viewpoint and an overall steam plant efficiency
present.
----------------------- Page 374-----------------------
99.9 99.5%
removal efficiency. With such a fine particulate, a ma-
Data jor
problem with greatly improved efficiency is reintrainment
99.5 of small
particles when the plates are rapped to remove the
99 Spline
95
collected particles. An attractive alternative to the dry ESP
90 seems
to be the wet ESP, in which the plates are washed
% 75
continuously with water. A baghouse is also a feasible choice.
e The bag
fabric must be chosen to avoid blinding. A Gore-Tex
v
i
t 50 fabric,
manufactured by W. L. Gore & Co., Knoxville, TN, was
a
l
u used
successfully in the pilot plant.
m
u If a
dry ESP is used, it has been suggested that if should
C
10 0.81 MMD be
designed to operate at a high enough gas temperature so
5 that it
can be placed before the economizer/low temperature
1 air
heater surface. Since the temperature difference between
0.5 flue gas
and feedwater or air is low in this region, a large
0.1 amount
of convective surface is required and substantial sav-
0.1 1 10 50 100 ings
result from having a cleaner gas and lower fouling factor.
Diameter ( m)
µ
Figure 9. Typical particle size distribution in particulate from MHD Seed

Recovery and Regeneration
pilot plant, upstream of the ESP/baghouse. Note the bimodal distri- The
economics of the MHD steam power plant depend on re-
bution with the smaller particles being potassium compounds and the covery
and reuse of the potassium seed. If the potassium is
larger ones coal ash.
recovered as a sulfur or chlorine compound, chemical pro-

cessing
to convert it to a sulfur and chlorine free compound is
required. Potassium is largely recovered from the exhaust by

Particulate Collection
the ESP
or baghouse, but significant quantities may also be
As previously noted, the principal differences between the
recovered from hoppers under heat transfer surfaces.
particulate in the MHD steam plant and that in a conven- If
the plant has a slag rejecting combustor or wet bottom
tional coal-fired steam plant are the higher mass loading, furnace,
it can be anticipated that some potassium will be
smaller particulate, and different chemical composition in present
with the coal slag from these effluent streams. The
that it is high in potassium compounds. chemical
form of potassium in these streams may be as com-
In pilot plant tests designed to simulate a commercial pounds
with the coal slag components, especially silicates. A
plant (14), the particle loading was measured to be 14 g/m3 to method
has been devised to economically recover this potas-
16 g/m3 3 3
(6 grains/ft to 7 grains/ft ). A typical particle size sium by
leaching with calcium hydroxide to recover potassium
distribution is shown in Fig. 9. The mean mass diameter was
hydroxide, which integrates well with the formate seed regen-
about 0.6 #m. The size distribution is bimodal, with the larger eration
process (15,16). A number of possible processes were
particles tending to be coal ash and the smaller ones potas- studied
during the US Department of Energy MHD develop-
sium sulfate. The resistivity of this mixture for two similar ment
program with the conclusion that the formate process,
cases is shown in Fig. 10. The resistivity for temperatures of which
was used in Germany during World War II, is the best
interest are in the range of 109 # # cm to 1010 # # cm. This is a choice
for converting the potassium sulfate recovered to po-
desirable range for good ESP operation and tends to partially tassium
formate for reuse. The recovered potassium carbon-
compensate for the very heavy mass loading and small size. ate can
be separated from the sulfate and chloride by differ-
A four-field dry ESP with a design specific collection area ences in
their solubility in water. Separation of sulfate from
(SCA) of 475 ft2/ 1000 acfm performed in the pilot plant with
carbonate was performed well in the pilot plant program by a
1011
Illinois #6
1010
)
m
c
Ω- 109
(
y
t
i
v
i
t 108
s
i
s
e
R
107
106
Figure 10. Ash resistivity from MHD pi- 2.6 2.4
2.2 2.0 1.8 1.6 1.4 1000/K
lot plant as a function of temperatures. 111 144
181 227 282 352 441 °C
Resistivities from 109 to 1010 # # cm are de- 233 290
359 440 540 665 826 °F
sirable from ESP operation.

Temperature
----------------------- Page 375-----------------------
MAGNETORESISTANCE 269
rotary vacuum filter working on the water from the wet ESP.
17. Technical assessment guide (TAG), EPRI Report No. P-2410-SR,
The system was saturated with carbonate, and the sulfate
1981 ed., Palo Alto, CA: The Electric Power Research Institute,
was removed by the filter.
1982.
18. J. N. Chapman and W. H. Boss, User’s Manual For The UTSI
MHD/Steam Power Plant Systems Code, DOE/ET/10815-98,

Economic Plant Design
University of Tennessee Space Institute, 1985.

In order to adequately optimize the performance of the MHD
19. J. N. Chapman and J. H. Hollis, The UTSI performance–cost
steam combined cycle power plant, a model of the cost and
model of combined cycle MHD steam power plants, Proc. Int. Liai-
performance is needed. The Electric Power Research Institute
son Group Symp. Modelling MHD Power Stations, Eindhoven Uni-
(EPRI) has developed a methodology for calculating the lev-
versity, The Netherlands, 1986.
elized cost of electricity over the planned lifetime of the plant
20. L. E. Van Biber and J. F. Pierce, A conceptual design of the
(17). The complexities and trade-offs in the MHD plant re-
Scholz MHD retrofit plant, Proc. 26th Symp. Eng. Aspects MHD,
quire a model that calculates performance and cost for use in
Nashville, TN, 1988.
the EPRI methodology. Potential designers should develop

such a model or use one of those listed in the references
JAMES N. CHAPMAN
(9,18–20).
Tennessee University Space
Institute
BIBLIOGRAPHY
MAGNETOOPTIC EFFECT. See FARADAY EFFECT .

1. G. Seikel and L. P. Harris, A summary of the ECAS MHD power
plant results, NASA Technical Memorandum X-73491, 1976.
2. L. W. Crawford and R. A. Attig, Impact of process variables on

NO emissions for Illinois #6 and rosebud coal fuels in MHD
x
power generation, Proc. 30th Symp. Eng. Aspects MHD, Baltimore,

MD, 1992, pp. VIII, 1.1–1.7.
3. K. D. Parks, Priority pollutant analysis of MHD-derived combus-

tion products, Proc. 28th Symp. Eng. Aspects MHD, Chicago, IL,
1980.
4. R. C. Attig et al., Design report for the combustor for the low
mass flow coal fired flow facility, Report No. DR-2.1.2-79-02, Tul-
lahoma, TN: The University of Tennessee Space Institute, 1979.
5. TRW Applied Technology Division, 10th Q. Progress Report, Re-

port No. MHD-ITC-90-311, 1990.
6. A. Grove et al., 50-Mwt combustor performance summary, Proc.

28th Symp. Eng. Aspects MHD, Chicago, IL, 1990.
7. R. J. Rosa, Magnetohydrodynamic Energy Conversion, New York:

McGraw-Hill, 1968.
8. J. X. Bouillard et al., User’s manual for MG MHD: A multigrid

three-dimensional computer code for the analysis of MHD genera-
tors and diffusers, ANL/MHD-89/1, Argonne, IL: Argonne Na-
tional Laboratory, 1988.
9. J. N. Chapman, On the economic optimization of the magnetohy-

drodynamic-steam power plant, PhD dissertation, Eng. Sci.
Mechan. Dept. University Tennessee, Knoxville, TN, 1977.
10. Y. C. L. Wu, Performance theory of diagonal conducting wall

MHD generators, AIAA J., 14 (10): 1362–1368, 1976.
11. L. S. Frost, Conductivity of seeded atmospheric pressure plas-

mas, J. Appl. Phys., 32 (10): 1961.
12. Anonymous, Steam: Its Generation and Use, 39th ed., New York:
Babcock & Wilcox, 1978.
13. J. N. Chapman and W. H. Boss, An economic analysis of the opti-

mum stoichiometry for an early commercial MHD-steam plant,
Proc. 27th Symp. Eng. Aspects MHD, Reno, NV, 1989.
14. R. C. Attig, Illinois No. 6 coal 2000 Hour Proof-Of-Concept Report,

University Tennessee Space Ins., Tullahoma, TN, 1994.
15. J. K. Holt et al., Recovery of insoluble potassium by base extrac-

tion from MHD spent seed and slag, 10th Int. Conf. MHD Electr.
Power Generation, Tiruchirapalli, India, 1989.
16. A. C. Sheth et al., (UTSI) and R. L. Solomon et al. (RCC), Techni-

cal and economical consideration of the formate process for MHD
seed recovery and regeneration, 10th Int. Conf. MHD Electr. Power
Generation, Tiruchirapalli, India, 1989.
----------------------- Page 376-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Multirate Filterbanks

Standard Article
Xiang-Gen Xia 1
1University of Delaware, Newark, DE

reserved.
Search
DOI: 10.1002/047134608X.W7212
❍ Acronym Finder
Abstract
Introduction
Notations, Basic Building Blocks, and Their Properties
M-Channel Multirate Filterbanks
Perfect Reconstruction Fir Multirate Filterbank Factorization and

Construction
DFT and Cosine Modulated Filterbanks
Some Additional and Recent Research Topics
Summary

%20Conversion/W7212.htm17.06.2008 16:43:00
----------------------- Page 377-----------------------
MULTIRATE FILTERBANKS 35
x (t) x (n) = x (nT ) x (n) x (Mn)
a a s
(a)
(b)
Figure 2. An analog-to-digital (A/D) signal converter converts an an-

alog signal to a digital sequence and is implemented before digital
signal processing.
tion) by factor M . By doing so, the new rate for x(Mn) is 1/M
times of the original rate of x(n).
The resampled signal x(Mn) can be expanded to signal

ˆ
MULTIRATE FILTERBANKS
x(n) with the data rate the same as x(n) by inserting M
1
zeros between each two samples of x(Mn), which is denoted

As the computation speed in modern chip design increases in
by #M and called upsampling (or expansion) by factor M
an exponential way, our world is turning rapidly from analog
shown in Fig. 3(b). A discrete signal x(n) may be converted to
ˆ (t) as shown in Fig. 3(a), where f is the

to digital mode, such as digital communications, digital
an analog signal xa 0
audios, and digital televisions and videos. This change has
bandwidth of the original analog signal xa(t). This is called a
ˆ
been providing higher quality services than before, making
D/A converter. The analog signal xa(t) can be mathematically
the information highway possible, and revolutionizing our liv-
represented as
ing standards. In this section, we briefly describe concepts of
analog-to-digital (A/D) and digital-to-analog (D/A) converters,
xâ (t) = #x (n) sin π(f st − n) (2)
rate conversion, some applications of multirate filtering, such

π(f t − n)
n
as transmultiplexing and multiresolution image analysis and
coding, and finally the outline of this article.
which is called the Shannon sampling theorem.
A noninteger (fractional) multiple rate change is shown in
ˆ
INTRODUCTION
Fig. 4, where the data rate of x(n) is N/M times of the data
rate of x(n). For more about rate conversion, see Ref. 1.

A/D and D/A Converters and Rate Conversion
There are many applications of multirate filtering. We next
want to briefly discuss two of them: transmultiplexing and
In A/D transitions, analog signals, such as audio signals and
multiresolution image analysis and coding.

video signals, from our physical world are first converted to
digital signals, then these digital signals are processed using
digital signal processing (DSP) techniques, such as filtering,
Transmultiplexing
detection, and compression. Finally the processed digital sig-
Multiuser communications play important roles in current

nals are converted back to physical analog signals. This pro-
communication systems, where multiple users share a com-

cess is illustrated in Fig. 1.
mon channel. Such examples include phone (wireline and

An A/D converter samples an analog signal xa(t) into a dis-
wireless mobile) networks and satellite communications. In

crete sequence x(n) # x (nT ) with a sampling frequency f #
a s s
multiuser communication systems, there are three common
1/Ts , where Ts is the sampling period length. It is shown in
multiplexing methods: (1) time division multiple access

Fig. 2(a). For the sampled signal x(n) to convey all the infor-
(TDMA), where different users use different time slots; (2) fre-
mation that the original analog signal xa(t) has, it is necessary
quency division multiple access (FDMA), where different us-

and sufficient for the sampling frequency (or sampling rate)
ers use different frequency slots; and (3) code division multi-
f s not to be below the Nyquist frequency, i.e., twice of the
bandwidth (Hz) of the analog signal xa(t):
ple access (CDMA), where different users use different codes.
These multiplexing methods are called transmultiplexing. A

f s ≥ 2f 0 (1)
general transmultiplexing block diagram with P users is
where f is the bandwidth (Hz) of x (t). A sampled signal x(n)

0 a
may be resampled with a reduced sampling rate 1/M shown

in Fig. 2(b), where M # is called downsampling (or decima-
^
^ (t) x (Mn) x (n)
x (n) x
fs fs
–
Analog Analog
2 2 (b)
signal Analog to Digital signal Digital to signal
digital (A/D) processing analog (D/A)
(a)
converter (DSP) converter
Figure 3. A digital-to-analog (D/A) signal converter, which converts

Figure 1. A general signal processing diagram. It is used in current a
digital signal to an analog signal and is implemented after digital
application systems.
signal processing.

----------------------- Page 378-----------------------
36 MULTIRATE FILTERBANKS
^
and reconstruction are built upon tree-structured multirate
x (n) x (n)
N M
filterbanks as illustrated in Fig. 7(a), which are used along
the x-axis and the y -axis separately in two-dimensional image
processing. As an example, an original image is shown in Fig.

Figure 4. Rate conversion changes the data rates of a signal.
8 and its three-level multiresolution decomposition is shown
in Fig. 9. In multiresolution image compression, such as em-
bedded zero-tree wavelet (EZW) coding (4), the correlations

shown in Fig. 5. Each transmitter uses a linear time-invari-
between the images at different resolutions are fully taken
ant (LTI) finite impulse response (FIR) filter Fp (z) and each
into account. It has been shown that this compression method
receiver uses another LTI FIR filter Hp (z).
has better performance than the conventional discrete cosine
The FIR filters in Fig. 5 for TDMA, FDMA, and CDMA
transform (DCT) compression method. For more details about
systems have the following features shown in Fig. 6.
EZW and the comparisons between EZW and DCT ap-
proaches, one may consult Refs. 4 and 5. By using the EZW

1. In TDMA transmultiplexing, each user occupies one
image coding, the original image in Fig. 8 is compressed with
fixed time slot in total M time slots, i.e., Fp (z) # z
p and compression ratio 64, and its decompressed image is
shown
Hp (z) # zp for 0 # p # P
1 [see Fig. 6(a)]. in Fig. 10.
2. In FDMA transmultiplexing, each user occupies one

In manymultiresolution image analyses, such as image/
fixed frequency slot in total M frequency slots, i.e.,
video compression, the perfect reconstruction of the multirate
F (z) and H (z) are both supported mainly in a fre-
filterbank shown in Fig. 7(a) is needed, which is equivalent to
p p
quency band with 1/M of the total bandwidth [see Fig.

the perfect reconstruction of the 2-channel multirate fil-
6(b)].
terbank shown in Fig. 7(b). Certain tree-structured multirate
3. In CDMA transmultiplexing, each user uses a pseudo-
filterbanks also lead to wavelets (6–13).
random sequence to spread the information sequence,
M
1
i.e., F (z) # #n #0 f (n)z
n with a pseudo-random binary Outline
p p
sequence f p (n) for each p , 0 # p # P

1 [see Fig. 6(c)].
From the two multirate filtering applications, one can see the
importance of the perfect reconstruction property of a

Since synchronization for FDMA systems is not as impor-
multirate filterbank. The rest of this article is devoted to de-

tant as the one for TDMA and CDMA systems, in some appli-
veloping a systematic theory of building perfect reconstruc-

cations FDMA is preferred and mixed with other systems,
tion (PR) multirate filterbanks. It should be pointed out that

such as in satellite communications. Because FIR filters can-
the PR property may not be necessary for good performance

not be ideal bandpass filters, it is not obvious that the re-
in some applications. This article is organized as follows. In

ceivers in the FDMA system in Fig. 5 are able to recover the
ˆ
the second section, we study some necessary properties on
original transmitted information sequences, that is, xp (n) #
some basic building blocks for multirate filterbanks. In the

c x (n
n ) for integers n and nonzero constants c . We will
p p p p p
third section, we study perfect reconstruction multirate fil-
see later that this perfect recovery problem is equivalent to
terbanks. In the fourth section, we develop factorization and

the perfect reconstruction of a multirate filterbank. For more
construction for multirate filterbanks with perfect reconstruc-

about multirate filterbanks and transmultiplexers, see, for ex-
tion. In the fifth section, we study two special kinds of
ample, Refs. 2 and 3. More applications include ADSL, DMT,
multirate filterbanks: discrete Fourier transform (DFT) fil-

etc.
terbanks and cosine modulated multirate filterbanks, which
are related to short-time Fourier transforms and Gabor trans-

Multiresolution Image Analysis and Coding
forms. In the sixth section, we describe some recent develop-

Multiresolution image analysis is an important technique in
ments on multirate filterbanks. In the last section, we sum-
digital image processing and has applications in target detec-
marize this article. There have been extensive studies on
tion, image browsing, image/video compression, etc. The basic
multirate filterbanks in the past two decades. Due to the
idea is to decompose an image into different resolutions and
length limitation, this article is by no means a complete sur-
then process these images with different resolutions ac-
vey on this subject but provides some basics on multirate fil-
cording to different needs. The multiresolution decomposition
terbanks. For more comprehensive descriptions and historical
Channel
^
x0 (n)
x0 (n)
F (z) Ho (z) M
M
o
^
x 1 (n)
x 1 (n)
H (Z)
F (z) H (z)
1 M
M
1 1
⋅
⋅ ⋅ ⋅ ⋅ ⋅
⋅
⋅ ⋅ ⋅ ⋅ ⋅
⋅
⋅ ⋅ ⋅ ⋅ ⋅
^
xp –1 (n)
xp –1 (n)
F + H (z)
(z) Fp –1 (Z)
Figure 5. Transmultiplexer which is used in many M
p –1 p –1 M
communication systems including multiuser systems.
----------------------- Page 379-----------------------
2π /M
⋅ ⋅ ⋅ ⋅ ⋅ ⋅
⋅ ⋅ ⋅
1 2 M t 1 2 M ω
2π ω Figure 6. Multi-access characteristics having
three common multi-access methods: (a) TDMA; (b)

(a) (b)
(c) FDMA; and (c) CDMA.
events on multirate filterbanks, we refer the reader to Refs. 1

ters, X (z) and Y(z), denote the z-transforms of scalar-valued
and 8–13.
sequences; bold-faced lowercase letters, x(n) and y(n), denote
vector sequences; and bold-faced capital letters, H(z) and
F(z), denote matrices and matrix polynomials. For a se-

NOTATIONS, BASIC BUILDING
quence x(n), its z-transform is defined by

BLOCKS, AND THEIR PROPERTIES
Notations
X (z ) = #x (n)z −n
Throughout this article, all lowercase letters, x(n) and y (n),
j #
denote scalar values and scalar-valued sequences; capital let-
and X (e ) is its discrete time Fourier transform (DTFT).
x (n)
H0 (z) 2 H0 (z) 2 ⋅ ⋅ ⋅
H0(z) 2
H 1(z) 2
H 1(z) 2
H 1(z) 2
^
x (n)
2 F0 (z) + ⋅ ⋅ ⋅ 2 F0 (z) +
2 F0 (z) +
2 F 1(z) 2 F 1(z)
2 F 1(z)
(a)
^
x (n)
x (n)
H0 (z) 2 2 F0 (z) +
Figure 7. Multiresolution decomposi-

H 1(z) 2 2 F 1(z)
tion and reconstruction, where a signal
is decomposed into different resolutions
and then can be reconstructed from the

(b)
various resolution decompositions.
----------------------- Page 380-----------------------
Figure 8. Original image, which is a test image from compression-

Figure 10. Decompressed image with compression ratio 64, which is
using wavelets.
obtained by using the wavelet-based compression algorithms devel-
oped by Said and Pearlman (5).
A matrix polynomial H(z) is

Decimator and Expander
An
M-fold decimator (or downsampling) and L-fold expander
H (z ) = # H z −n
(or upsampling) are depicted in Fig. 11(a) with an example
n
n in
Fig. 11(b), and Fig. 11(c) with an example in Fig. 11(d),
respectively, where
where Hn are constant matrices with same size. For a square
matrix polynomial H(z), det(H(z)) denotes its determinant.
yD (n) = x (Mn )
T †
For a matrix A, matrices A , A , and A* denote the transpose,
transpose conjugate, and conjugate, respectively. For a matrix
†
˜
polynomial H(z), its tilde operation H(z) denotes H (1/z*), that
is,
0
˜ # † n # −n
H (z ) = H z , if H (z ) = H z (3)
n n
n n
x (n) yD (n)
M
A matrix polynomial H(z) has FIR inverse if and only if there
M = 2
exists a matrix polynomial F(z) such that F(z)H(z) # I, where
I is the identity matrix.
(a)
The matrix WM denotes the M-point DFT matrix, that is,
0
WM = (Wmn )0≤m ,n ≤M −1
M
where Wm # e
j 2#/M .
(b)
x (n) yE (n)
(c)
L = 2
(d)
Figure 11. Decimator and expander. In the decimator, a data rate

Figure 9. Multiresolution decomposition of the original image, which is
reduced M times. In the expander, a data rate is increased L times.
is an example of a multiresolution decomposition of the test image in
These two operations play important roles in multirate signal pro-
Fig. 8, by using two-channel filterbanks (or wavelets).
cessing.
----------------------- Page 381-----------------------
j ω j
ω
Y (e ) Y (e
)
E D
j ω M = 2
X (e ) L = 2
1
1 1
2
–π 0 π ω –π 0 π ω –π 0
π ω
Figure 12.
The frequency domain representation ex-
(a) (b) (c)
amples of the decimator and expander.
and
Polyphase Representations
#
The polyphase representation was first invented by Bellanger
x[n/L], if n is a multiple of L ,
yE (n) =
et al. (14) and Vary (15) and first recognized by Vaidyanathan
0, otherwise.
and Vetterli in the simplifications of multirate filterbank the-
ory studies. It can be briefly described as follows. For any

In the frequency and z-transform domains (1,9),
given integer N , any filter H(z) can be decomposed into
M −1
# #
j ω −j ωn 1 j (ω−2π k )/M
N −1
Y (e ) = y (n)e = X (e ) (4)
# −l N
D D M
H (z ) = z E (z ) (5)
l
n k =0
l =0
and
where
# −n L j ω j ωL
Y (z ) = y (n)z = X (z ), and Y (e ) = X (e )
#
E E E
n
E (z ) = h[Nn + l]z −n
The graphical meaning for the expander is that the DTFT of

the expanded yE (n) is an L-fold compressed version of the un-
and h[n] is the impulse response of H(z). The decomposition,
j #
Eq. (5), is called the Type 1 polyphase representation of H(z).
compressed X (e ) shown in Fig. 12(a,b). The graphical mean-
ing for the decimator is the following [shown in Fig. 12(a,c)]:
Meanwhile, H(z) can be decomposed into
j # j #/M
N −1
#
1. Stretch X (e ) by a factor M to obtain X (e );
−N + 1+l N
H (z ) = z R (z ) (6)
2. Obtain M
1 copies of this stretched version by shifting
l
l =0
it uniformly in successive amounts of 2#;
3. Add all these shifted and stretched versions to the origi-
where Rl(z) # EN
1
l(z), which is called the Type 2 polyphase
j #/M ), and divide
nal unshifted and stretched version X (e
representation of H(z). For l # 0, 1, . . ., N
1, El(z) and
by M .
Rl(z) are called the lth Type 1 and Type 2 polyphase compo-
nents of H(z), respectively. We will see later that the Type 1

j #
The M
1 shifted and stretched versions of X (e ) in Eq. (4)
polyphase representation is for the analysis bank and the
are the aliasing created by the down-sampling.
Type 2 polyphase representation is for the synthesis bank in
a multirate filterbank.
Noble Identities
The main purpose for introducing these polyphase repre-

The following two Noble identities play important roles in the
sentations is to move the decimator from the right side of an
multirate filterbank theory. They tell us when the orders of
LTI filter to the left side (expander from the left side of an
the decimator/expander and an LTI system can be switched.
LTI filter to the right side) by using the Noble identities in
The two Noble identities are shown in Fig. 13, where y 1(n) #
Fig. 13. In the Noble identities, the power of the variable z in
y (n) and y (n) # y (n).
an LTI filter needs to rise, which usually does not hold for an
2 3 4
x (n) x (n)
y 2 (n)
K H (z) H (zk) K
(a)
x (n) y 3 (n) x (n)

y 4 (n)
H (z) N N H (zN
)
Figure 13. The Noble identities, the rules of switch-

(b)
ing the order of LTI system and decimator/expander.
----------------------- Page 382-----------------------
x (n)
x (n)
H (z) M N F (z)
H (z) M M F (z)
0 0 0 0
0 0
H (z) M 1 N F (z)
H (z) M F (z)
1 1 1
1 M 1
⋅ ⋅ ⋅ ⋅ ⋅
⋅ ⋅ ⋅ ⋅ ⋅
⋅ ⋅ ⋅ ⋅ ⋅
⋅ ⋅ ⋅ ⋅ ⋅
⋅ ⋅ ⋅ ⋅ ⋅
⋅ ⋅ ⋅ ⋅ ⋅
^
^
x
(n)
x (n)
HM–1 (z) MM–1 NM–1 FM–1 (z) +
HM–1 (z) M M FM–1 (z) +
Analysis bank Synthesis bank

Figure 14. A general M-channel multirate filterbank.

Figure 16. M-channel maximally decimated multirate filterbank.
Maximally Decimated Multirate Filterbanks: Perfect

LTI filter but does hold for the polyphase representations of
Reconstruction and Aliasing Component Matrix
an LTI filter as shown in Eqs. (5) and (6).
An M-channel maximally decimated multirate filterbank is
shown in Fig. 16. By comparing it with the transmultiplexer
in Fig. 5, one can see that, when P # M, the transmitter side

M-CHANNEL MULTIRATE FILTERBANKS
in Fig. 5 is the same as the synthesis bank in Fig. 16, and the
receiver side in Fig. 5 is the same as the analysis bank in Fig.

A general M-channel multirate filterbank is depicted in Fig.
16. The multiresolution decomposition filterbank in Fig. 7(b)
14, where the left side is an analysis bank and the right side
is a 2-channel maximally decimated filterbank. In both appli-
is a synthesis bank, each of which has M LTI filters. In many
ˆ
cations, it is desired that the output signal x(n) is equal to the

applications, such as FDMA, these M LTI filters occupy M
input x(n) in some sense of the multirate filterbank in Fig. 16,
different frequency bands as shown in Fig. 15.
which is the perfect reconstruction property as follows.
There are several cases for an M-channel multirate fil-
A multirate filterbank in Fig. 16 is called perfect recon-
terbank in Fig. 14:
ˆ
struction (PR) if and only if x(n) # cx(n

n0) for a nonzero
constant c and an integer n0.

1. when M0 # M1 # # # # # MM
1 # N0 # N 1 # # # # # The question now becomes how to
construct a PR multirate
filterbank, in other words, what conditions on Hm(z)

and
NM
1 # M, the filterbank is called maximally decimated;
Fm(z) are for the PR. Since in many applications, such as

2. when M0 # M1 # # # # # MM
1 # N0 # N 1 # # # # # transmultiplexing and image analysis and
coding, FIR filters
NM
1 # M, the filterbank is called nonmaximally deci-
are preferred, in what follows we are only interested in FIR
mated;
filters Hm(z) and Fm(z) in Fig. 16 In this case, the multirate
3. when M0 # M1 # # # # # MM
1 # N0 # N 1 # # # # # filterbank is called FIR. Examples of 2-
channel PR filterbanks
NM
1 M, the filterbank is called over decimated; were
first obtained by Smith and Barnwell (28) and Mintzer
4. when Mk and Nl are not all equal, the filterbank is
(29) independently. Some early studies on multirate filter-
called nonuniformally decimated.
banks can also be found in Refs. 29a–c.
In the z-transform domain, the PR property becomes
Although there are increasing discussions on the cases 2–4

ˆ −n0
X (z ) = cz X (z ) (7)
lately, such as applications of nonmaximally decimated
multirate filterbanks in intersymbol interference (ISI) cancel-

In terms of an input signal X (z), by using Eq. (4) the output
lation (16–21), and studies of nonuniformally decimated
ˆ
X (z) in Fig. 16 can be formulated as follows:

multirate filterbanks (22–27), the first case is the most well
studied and the most important case. In this section, we focus
M −1
ˆ # l
on the first case, that is, maximally decimated multirate fil-
X (z ) = A 0 (z )X (z ) + A (z )X (zWM ) (8)
l
terbanks.
l = 1
H0 H 1 H2 HM – 1
⋅ ⋅ ⋅
Figure 15. M-channel analysis filter frequency
response example.
0 2π
ω
----------------------- Page 383-----------------------

where
ter Fm(z) in Fig. 16, let Rl,m(z) be its lth Type 2 polyphase
component, for 0 # m, k, l # M
1. Let
M −1
#
1
A (z ) = H (zW l )F (z ), 0 ≤ l ≤ M − 1 (9)
l M k M k
E (z ) = (Em ,k (z ))0≤m ,k ≤M −1, and R (z ) = (Rl ,m (z ))0≤l ,m ≤M −1
k =0
Clearly, the second term in the right hand side of Eq. (8) is
which are called the polyphase matrices of the analysis bank
the aliasing term. For the PR property Eq. (7), we need
and the synthesis bank in Fig. 16, respectively. Then, it is not
hard to see that

−n
A 0 (z ) = cz 0 and A (z ) = 0 for 1 ≤ l ≤ M − 1 (10)
l
M ˜ M
h (z ) = E (z )e(z ) and f (z ) = e(z )R (z ), (13)
We now want to simplify the PR condition Eq. (10). To do so,
where h(z) # (H T
1
let
0(z), . . ., HM
1(z)) , e(z) # (1, z , . . .,
M #1 T T
z ) , f (z) # (F (z), . . ., F (z)) , and e(z) is the tilde opera-
0 M
1
t (z ) = [cz−n0 , 0, . . ., 0]T
tion of e(z). Thus, by using the Noble identities, the multirate
T
filterbank in Fig. 16 is the same as the one shown in Fig.
f (z ) = [F (z ), F (z ), . . .,F (z )]
0 1 M −1
17(a), which is called the polyphase representation of the
multirate filterbank in Fig. 16.

and
Let P(z) # R(z)E(z), then the multirate filterbank in Fig.
 
16 is equivalent to the MIMO system in Fig. 17(b) with sys-

H0 (z ) . . . HM −1 (z )
tem transform matrix P(z). Clearly, the PR property is equiv-
 
 
. . .
alent to the invertibility of the polyphase matrix E(z).
H (z ) = . . . (11)
 . . . 
 M −1 M −1 
H0 (zWM ) . . . HM −1 (zWM )
Theorem 2. An FIR multirate filterbank in Fig. 16 is perfect
reconstruction if and only if the polyphase matrix E(z) has an

Then
FIR inverse.
H (z )f (z ) = Mt (z ) (12)
Theorem 1 deals with the AC matrix H(z) while Theorem
2 deals with the polyphase matrix E(z). From Eq. (13), it is

Thus, given an analysis bank, H0(z), . . ., HM
1(z), if the ma-
not hard to see the following relationships between these two

trix polynomial H(z) has an FIR inverse, then the synthesis
matrices:
bank f (z) can be solved from the Eq. (12). In other words, the
multirate filterbank in Fig. 16 with this synthesis bank is PR.
H (z ) = W† D (z )ET (zM ) and E (zM ) = HT (z )WM D (z −1) (14)
The matrix polynomial H(z) in Eq. (11) is called the aliasing
M
component (AC) matrix. In conclusion, we have the following

theorem.
x (n)
Theorem 1. An FIR multirate filterbank in Fig. 16 is perfect
M M
reconstruction if and only if its AC matrix H(z) has FIR in-
z–1 z–1
verse.
–1 M M –1
z z
E (z) R (z)
⋅ ⋅ ⋅
One can see that the AC matrix H(z) is a structured ma-
⋅ ⋅ ⋅
trix, where its components are not free but related. This lim-
⋅ ⋅ ⋅
–1
–1 z
its the study and construction of PR multirate filterbanks. We
z ^
x (n)
next want to use the polyphase representations and Noble
M M +
identities introduced in the second section and convert the
AC matrix to the polyphase matrix in which all components
are free.
(a)
Maximally Decimated Multirate Filterbanks:

Perfect Reconstruction and Polyphase Matrix
x (n)
M M
The analysis in the previous subsection is a direct analysis of
z–1 z–1
the relationship between the input and the output in Fig. 16.
We next want to first simplify the block diagram in Fig. 16 by
z–1 M M z–1
R (z) E (z)
using some properties of building blocks studied in the second
⋅ ⋅ ⋅ ⋅
⋅ ⋅ ⋅ ⋅
section, such as Noble identities and polyphase representa-

⋅ ⋅ ⋅ ⋅
tions, and then study the PR property for the simplified sys-
–1 z–1
tem. The main idea for the simplification is to switch the or-
z x (n)
M M +
ders of decimator/expander and FIR filters, and then convert
the multirate filterbank into a multi-input and multi-output
(b)
(MIMO) system.
For each analysis filter Hm(z) in Fig. 16, let Em,k(z) be its
Figure 17. Polyphase representation of M-channel maximally deci-
kth Type 1 polyphase component, and for each synthesis fil-
mated multirate filterbank.
----------------------- Page 384-----------------------
where WN is the DFT matrix and D(z) is the diagonal matrix

Factorization of FIR Polyphase Matrices with FIR Inverses
polynomial:
The goal of this subsection is to

characterize all FIR causal
−1 −M + 1
M # M matrix polynomials with FIR inverses. Since the de-
D (z ) = diag(1,z , . . .,z )
terminant of the FIR inverse of an M #M matrix polynomial
is the inverse of its determinant, we have the following

From Eq. (14), it is clear that the FIR invertibilities of the AC
lemma.
matrix H(z) and the polyphase matrix E(z) are equivalent.
Unlike matrix H(z), matrix E(z) does not have any relation-
ship between its components, which leads to the systematic
Lemma 1. An FIR matrix polynomial has FIR inverse if and
n
construction and factorization discussed later.
only if its determinant is cz n for a nonzero constant c and an
In some applications, such as the cross-talk cancellation in
integer n0.
transmultiplexers in Fig. 5, PR may not be necessary as long
as the aliasing (cross-talk) is cancelled in a multirate fil-
terbank, i.e., the second term at the right hand side of Eq. (8)
Let H(z) be an FIR causal matrix polynomial with an FIR
is zero. A necessary and sufficient condition on such fil-
inverse. If det(H(z)) # cz
# , then # is called its McMillan de-
terbanks was obtained by Vaidyanathan and Mitra (30),

gree , the minimal number of delay elements to implement the
which is stated as follows. A square matrix polynomial P(z)
MIMO system (9). A matrix polynomial H(z) is called unimod-
is called pseudo-circulant if and only if it has the following
ular if and only if its McMillan degree is 0, that is, its deter-
form:
minant is a nonzero constant. To introduce the complete fac-
torization of
FIR matrix polynomials with FIR inverses, let
 P (z ) P (z ) · · · P (z ) 
us first introduce three types of elementary row (column) op-
0 M −1 1
erations:
 
z −1P (z ) P (z ) · · · z −1P (z )
 
1 0 2
 
 . . . . 
P (z ) = . . . .
 . . . . 
 
Type 1. Interchange two rows (or columns).
 
−1 −1
z P (z ) z P (z ) · · · P (z )
 −1 M −2 −1 M −3 M −1 
Type 2. Multiply a row (or column) with a nonzero con-
z P (z ) z P (z ) · · · P (z )
M −1 M −2 0
stant c.
Type 3. Add a polynomial multiple of a row (or column) to

Notice that when no z
1 appeared on the lower triangular another row (or
column).
components in P(z), it would be circulant.
Theorem 3. An M-channel multirate filterbank in Fig. 16 is

The corresponding matrices of these elementary operations
aliasing free if and only if the polyphase matrix P(z) #
are called elementary matrices, which have the following
R(z)E(z) is pseudo-circulant. Under this condition, the fil-
forms.
ˆ
Let ei be the M dimensional vector with its ith entry 1 and
terbank output and the input are related by X (z) # A0(z)X(z)
as in Eq. (8), where
other entries 0 for i # 1, 2, . . ., M, i.e.,
A (z ) = z −M + 1(P (zM ) + z −1P (zM )

e = (0 · · · 0 1 0 · · · 0)T
0 0 1
i
i
+ · · · + z −M + 1PM −1 (zM ))
A Type 1 elementary matrix A can be written as
PERFECT RECONSTRUCTION FIR MULTIRATE FILTERBANK
†
FACTORIZATION AND CONSTRUCTION
A = I + (e − e )(e − e ) ,
i j j i
From the studies in the previous section, PR FIR M-channel
for certain i # j and 1 # i, j # M .

multirate filterbanks are converted to M #M matrix polyno-
A Type 2 elementary matrix A can be written as
mials. In this section, we focus on M #M matrix polynomials
E(z) # (Ek,l(z))0#k,l#M
1. Without loss of generality, in what fol-
lows we only consider FIR and causal matrix polynomials, i.e.,
A = I + ce e†
i i
L
E (z ) = #E z −n ,L is a nonnegative integer
for certain c #
1 and a certain i, 1 # i # M .
n
n=0
A Type 3 elementary matrix U(z) can be written as
As studied in the third section, when the polyphase matrices

U (z ) = I + α(z )e e†
i j
of analysis banks have FIR inverses, the corresponding syn-
thesis banks can be obtained by using the inverses for the PR
multirate filterbanks, i.e., PR FIR multirate filterbanks are
where #(z) is a polynomial of z
1 and i # j with 1 # i, j # M .
constructed. In this section, we first study general E(z) with

With these three elementary operations/matrices, any
FIR inverses and then study paraunitary matrix polynomials
M # N matrix polynomial can be diagonalized and the re-
E(z) that are corresponding to paraunitary multirate fil-
sulted decomposition is called the Smith–McMillan decompo-
terbanks.
sition, which is stated as follows. An M # N matrix polyno-
----------------------- Page 385-----------------------
mial has the following Smith–McMillan decomposition; see,

Theorem 4. A causal FIR M #M matrix polynomial H(z) has
for example, Ref. (9):
an FIR inverse if and only if H(z) has the following form
 
H (z ) = Vρ (z ) · · ·V (z )Aσ Uσ (z ) · · ·A U (z ) (16)
γ0 (z ) 0 · · · 0 0 · · · 0
1 1 1
 0 γ (z ) · · · 0 0 · · · 0 
 
1
where # is the McMillan degree of H(z),
is a certain nonneg-
 
 
. . . . . . .
. . . . . . .
ative integer, V (z) # O for i # 1, 2, . . ., #, A # A
and
 
i i
. . . . . . .
 
U (z) # U for i # 1, 2, . . .,
.
 
H (z ) = W (z ) 0 0 · · · γ (z ) 0 · · · 0 U
(z ) i
 p 
 
 
0 0 · · · 0 0 · · · 0
 
Factorization of Paraunitary FIR Matrix Polynomials

 . . . . . . . 
 . . . . . . . 
. . . . . . .
 
In this subsection, we
introduce paraunitary matrix polyno-
0 0 · · · 0 0 · · · 0
mials and corresponding multirate filterbanks, which are spe-
(15) cial FIR multirate filterbanks with FIR inverses.
An M # N matrix polynomial H(z) is called paraunitary if
and only if
where W(z) and U(z) are products of some elementary matrix
polynomials with sizes M #M and N # N , respectively, #i(z)
˜
are polynomials of z
1, # (z) divides # (z), for i # 0, 1, . . .,
H (z )H (z ) = dIN , for all complex values z
i i #1
p
1, i.e.,
where d is a positive constant and IN is the N # N identity
matrix. When we restrict the complex value z on the unit cir-

γ (z ) |γ (z ), i = 0, 1, . . .,p − 1
i i+ 1
cle, i.e., in the Fourier transform domain, it becomes the con-
cept of lossless matrices. An M # N matrix polynomial H(z)

and
is called lossless if and only if
#i+ 1 (z )
H† (ej ω )H (ej ω ) = dIN , for all real values ω
γ (z ) =
i # (z )
i
where d is a positive constant. When H(z) is FIR, lossless is
j #
equivalent to paraunitary. When H(z) exists for all z # e
but
where (z) # 1, (z) for i 0 is the greatest common divisor
0 i
not all complex values z , lossless is not equivalent to parauni-
of all the i # i minors of H(z).
tary, while paraunitary always implies lossless. Since we are

When H(z) is a square causal matrix polynomial with an
interested in FIR H(z), we only consider paraunitary matrix

FIR inverse, the diagonal matrix in the Smith–McMillan de-
n
n polynomials in this subsection. An example of paraunitary
composition has the form of diag(c z 1, . . ., c z M )
with
1 M
matrix polynomial is:
nm # 0 and cm # 0 for 1 # m # M, and this is a complete
characterization of all square causal matrix polynomials with
# −1 + 1 z −1 − 1#
z
FIR inverses. The factorization in Eq. (15) is, however, not
H (z ) = −1 −1
convenient to be incorporated in the optimal design studied
z − 1 z + 1
later. We next want to introduce another factorization. For

more details, see Refs. 31 and 32.
In this case
We define three kinds of basic matrices.
# #
˜ z + 1 z − 1
H (z ) =
Class I.
z − 1 z + 1
O # V(z): V(z) # I
vv† # z
1vv† where v is an M # 1
constant vector with unit norm

. ˜
. A paraunitary multirate filterbank is
and H(z)H(z) # 4I2

Let V(z) # I
vv† # z
1vv† # O . Then, its inverse shown in Fig. 18.
V
1 † †
(z) # I
vv # zvv .
Class II.
U # U(z): U(z) # I # #z
me e† where # is a constant, m is
i j
a nonnegative integer, and i # j with 1 # i, j # M
. x (n) *
*
H0 (z) M M H0(1/z )
Let U(z) # I # #z
me e† # U . Then, its inverse U
1(z) #
i j
I
#z
m †
e e .
i j
M M H * *
Class III.
H 1 (z) 1(1/z )
A # A : A # I # (e
e )(e
e )† for certain i # j and 1
i j j i
†
# i, j # M or A # I # ce e for certain c #
1 and a ⋅ ⋅ ⋅
⋅ ⋅
i i
⋅ ⋅ ⋅ ⋅ ⋅
certain i, 1 # i # M
. ⋅ ⋅
⋅ ⋅ ⋅
Let A # I # (e
e )(e
e )† # A . Then, its inverse
^
i j j i
x (n)
1 †
1 H (z) M M H * *
+
A # A . Let A # I # ce e # A . Then, its inverse A
# M–1 (1/z )
i i
M–1
I
c/(c #1)e †
e .
i i
With these three cases of matrices, we have the following

Figure 18. Paraunitary M-channel maximally decimated multirate
complete factorization.
filterbank.
----------------------- Page 386-----------------------
Similar to orthogonal transformations (matrices), the ad-

Table 1. Optimized Analysis Filter Impulse Responses of
vantages of paraunitary multirate filterbanks include that
3-Channel Paraunitary Filterbank
they preserve signal energies in the decompositions (or trans-
n h (n) h (n) h (n)

formations) and the synthesis banks (or inverse transforma-
0 1 2
tions) are simply the tilde operations of the analysis banks.

0
0.0429753
0.0927704 0.0429888
In this sense, paraunitary multirate filterbanks are general-
1 0.0000139 0.0000008
0.0000139
izations of orthogonal transformations, such as DFT, by add-
2 0.1489104 0.0087654
0.1489217
ing delay variables (or memory) into the transformations. In
3 0.2971954 0.0000226 0.2972354
4 0.3537539 0.1864025
0.3537496
the following, we want to present a complete characterization
5 0.2672266
0.0000020 0.2672007
of all paraunitary matrix polynomials obtained by Vaidyana-
6 0.0870758
0.3543303
0.0870508
than (9,33–35).
7
0.0521155
0.0000363
0.0520909
8
0.0875973 0.3564594 0.0875756
Theorem 5. An M # M causal FIR matrix polynomial H(z)
9
0.0427096
0.0000049
0.0427067
is paraunitary if and only if it can be factorized as
10 0.0474530
0.1931082
0.0474452
11 0.0429618 0.0000230 0.0429677

H (z ) = dVρ (z ) · · ·V 1 (z )H0 (17)
12 0.0 0.0 0.0
13
0.0232765
0.0000026
0.0232749
where d is a positive constant, # is the McMillan degree of
14 0.0000022 0.0 0.0000022
H(z), H is an M # M unitary constant matrix, V (z) # O for
0 i
i # 1, 2, . . ., #, and if # # 0 then H(z) # dH .

0
For a nonsquare paraunitary matrix polynomial, the fol-

Perfect Reconstruction Multirate Filterbank Design
lowing similar factorization holds (9,36–38).

After the complete characterizations of PR multirate fil-
terbanks, the next important issue is the design of a desired

Theorem 6. An M # N causal FIR matrix polynomial H(z)
PR multirate filterbank. In this subsection, we want to briefly
is paraunitary if and only if it can be factorized as
describe a method for the design. The goal here is to design
an M-channel PR multirate filterbank such that all analysis

H (z ) = dVρ (z ) · · ·V 1 (z )H0
filters have good filter properties, i.e., with good passband and
stopband attenuation properties. What one can do is to use

where d is a positive constant, H0 is an M # N unitary con-
the factorizations Eqs. (16) and (17) to parameterize these

stant matrix, Vi(z) # O for i # 1, 2, . . ., #, and if # # 0 then
H(z) # dH0.
analysis filters and then formulate the minimization problem
for the parameters:
For 2-channel paraunitary matrix polynomials, the above
M −1
factorization is simplified as the following lattice representa-
# j ω 2
min |Hm (e ) | dω (18)
tion (9).
m=0 mth stopband
Corollary 1. A 2 # 2 causal FIR matrix polynomial H(z) is

The following is a design example obtained from Ref. 9.
paraunitary if and only if it can be factorized as
Consider a 3-channel causal FIR paraunitary filterbank, that
is, M # 3 in Fig. 17 and Eq. (17). The maximal length of all

H (z ) = dRρ (z ) · · · R (z )R #1 0 #
analysis filters Hm(z) equals 15. By implementing this minimi-
1 0 0 ±1
zation, the optimized 3 analysis filter impulse responses are
listed in Table 1 and their frequency responses are shown in

where d is a positive constant, # is the McMillan degree of
Fig. 20, where the coefficients in Table 1 are from Ref. 9.
H(z), and
Since all filter coefficients are real, their Fourier spectra are
symmetric about the origin. For more details on the design

Ri = # cos θi sin θi # , and (z ) = #1 0 #
issues, see Refs. 9, 12, and 12a.
− sin θ cos θ 0 z −1
i i
DFT AND COSINE MODULATED FILTERBANKS

and # is an angle for i # 0, 1, 2, . . ., #.
i
In the previous section, we have studied general M-channel

A lattice realization of a 2-channel paraunitary analysis
maximally decimated multirate filterbanks, where the Fou-
bank is shown in Fig. 19, where # # #d, ci # cos #i and si #
rier spectra of M analysis filters may not necessarily have the
sin # for i # 0, 1, . . ., #, where # delays are needed.
same shape. In many applications, such as FDMA communi-
i
α –c –c –c
x (n)
0 1 ρ
–s –s –s
z–1
0 1 ρ
s0 s 1 s
ρ
Figure 19. Lattice realization of 2-channel paraunitary

2 z–1 z–1
+ c0 c 1 c
analysis bank.
–α ρ
----------------------- Page 387-----------------------
which is shown in Fig. 21 with P(z) # diag(P (z), P (z), . . .,
0 1
0
PM
1(z)).
jω jω jω
H (e ) H (e ) H (e )
0 1 2
By the study in the third section, it is clear that the DFT
–10
filterbanks are PR if and only if
–20
Pm (z ) = cmz −nm , cm =# 0, 0 ≤ m ≤ M − 1,
–30
for some integer nm
–40
or
–50
M −1
–60
P (z ) = # cmz −Mnm −l , cm =# 0 (20)
0.0 0.1 0.2 0.3 0.4 0.5

m=0
Normalized frequency ( /2 )
ω π
The paraunitariness of this DFT filterbank forces #cm # # c # 0

Figure 20. Frequency responses of 3 analysis filters. (Illustration
for all 0 # m # M
1, which is basically equivalent to the
courtesy of Prentice-Hall, Englewood Cliffs, NJ  1993 by Prentice-
DFT. When Pm(z) # 1/#M for 0 # m # M
1, the DFT
Hall.)
filterbank is precisely reduced to the DFT as shown in Fig.
22, which is the reason for the name of the DFT filterbanks.
One can see that the condition on the prototype filter P(z)
cation systems, it is however quite often that all M analysis in
Eq. (20) for the PR property is very restrictive and usually
filters are derived from a single prototype filter and therefore
limits their applications. There are three ways to get around
have the same shape of their Fourier spectra. The advantage
this condition. The first one, which is also the most intuitive
of such systems is the implementation simplicity. In this sec-
one, is to design P(z) with excellent lowpass property. Then,
tion, we introduce two kinds of such filterbanks. One is the
the DFT filterbank is almost PR because the whole frequency
discrete Fourier transform (DFT) filterbank, where analysis
band is almost divided with a wall-cut manner by M analysis
filters are single-sided shifts of a prototype filter in the fre-
filters. The second way is to use nonmaximally decimated
quency domain (or exponential modulation). The other is the
DFT filterbanks, that is, the decimation factor is less than the
cosine modulated filterbank, where analysis filters are dou-
number of channels (or users), which corresponds to
ble-sided shifts of a prototype filter.
oversampled short-time Fourier transforms or discrete Gabor
transforms. For more details, see, for example, Refs. 1, 39–42,
42a. The third way is to use double-sided shifts instead of

DFT Filterbanks
the single-sided shifts as in Eq. (20), which leads to cosine

DFT filterbanks form a class of the simplest multirate fil-
modulated filterbanks as we shall see in the next subsection.
terbanks, where all analysis filters are shifted from a single
prototype filter in the frequency domain. The question then
Cosine Modulated Filterbanks
becomes when DFT filterbanks are PR. To study this ques-
tion, let us formulate the analysis filters. Let
The DFT filterbanks in the previous subsection 5.1 have two
disadvantages. One is that analysis filter coefficients are com-

L −1
plex-valued and the other is that the PR condition is too re-
P (z ) = # p (n)z −n
strictive. We now want to use double-sided shifts or cosine
n=0
modulations to construct M analysis filters with real coeffi-
cients and better filter properties in PR multirate banks.

be an FIR filter with length L , which is usually a good low-
Let P(z) be a prototype filter with length L as before. Let
pass filter. The M analysis filters are
U (z ) = P (zW m+0.5) and V (z ) = P (zW −(m+0.5)),

m 2M m 2M
L −1
Hm (z ) = P (zW m ) = # p (n)W −mnz −n , 0 ≤ m ≤ M − 1 (19)
0 ≤ m ≤ M − 1
M M
n=0
and for 0 # m # M
1,
where their Fourier spectra are illustrated in Fig. 15. Notice
mn H (z ) = a U (z ) + a V (z )
that M analysis filter coefficients hm(n) # WM p (n) are no
m m m m m
longer real even when the prototype filter coefficients p (n)
L −1
are real.
= # 2 real(amW −(m+0.5)n )p (n)z −n
2M
Let P (z), 0 # l # M
1, be the Type 1 polyphase compo- n=0
l
nents of the prototype filter P(z) with total M components.

Then it is not hard to see that the polyphase matrix of the
This tells us that the analysis filter coefficients are all real.
analysis bank Hm(z), 0 # m # M
1, in Eq. (19) is Furthermore, let
E (z ) = W† diag(P (z ), P (z ), . . .,P (z ))
am = W (m+0.5)(L −1)/2+ (−1)mπ /4
M 0 1 M −1
2M
----------------------- Page 388-----------------------
x
(n) x (n)
M P0(z) M
–1
–1
z
z
–1
M P 1(z) z–1 M
z
WM P (z) WM
⋅
⋅ ⋅ ⋅
⋅
⋅ ⋅ ⋅
⋅
⋅ ⋅ ⋅
–1
–1
z
M PM–1 (z) z M
Figure 21. DFT analysis bank, a special multirate fil-

terbank.
Analysis bank Analysis bank
Then, the analysis filter coefficients are

then condition Eq. (23) is equivalent to
# π # L − 1# π #
˜ (z )Q (z ) = α, m = 0, 1, 2, . . .,M − 1
l
Qm m
hm (n) = 2 cos (m + 0.5) n − + (−1) p (n)
M 2 4
(21) i.e., all Q (z) are 2 # 1 paraunitary matrix polynomials. They
have been completely characterized in the previous section

The corresponding synthesis filter coefficients are
and the factorization in Theorem 6 can be used to construct
# # # #
π L − 1 π
optimal cosine modulated filterbanks similar to those studied
f m (n) = 2 cos (m + 0.5) n − − (−1)l p (n)
(22) in the same section. The following example is from Ref. 12.
M 2 4
Figure 23(a) shows the frequency response of an optimized
prototype filter of length 129, and Figure 23(b) shows 8 fre-

From these filters, one can see why they are called cosine
quency responses for the 8-channel PR cosine modulated fil-

modulated filterbanks. We now present a necessary and suf-
terbank with the prototype filter in Fig. 23(a). One can see
ficient condition for the PR property (9,10,12,43).
that their stopband attenuations are about 80dB.

Theorem 7. Let Pl(z), 0 # l # 2M
1, be the Type 1 poly- Several updated research results
on cosine modulated fil-
terbanks are:
phase components of a prototype filter P(z) of length L with
2M total components. If L # 2KM # 1 for some positive inte-
1. Notice that the previous paraunitary cosine modulated

ger K and M analysis filters and synthesis filters are defined
filterbanks require the prototype filters of length 2KM .

in Eqs. (21) and (22), respectively, then the cosine modulated
This requirement is relaxed in Ref. 44.

filterbank is paraunitary if and only if
2. Notice that in the cosine modulated filterbanks, the

˜ ˜
analysis filters may not be linear phase although the
Pm (z )Pm (z ) + Pm+M (z )Pm+M (z ) = α, 0 ≤ m ≤ M − 1
(23)
prototype filter is. All linear phase analysis and synthe-
sis filters in cosine modulated filterbanks are obtained

for a constant # 0.
in Ref. 45 by using two prototype filters simultaneously.

This theorem suggests the following method to construct
3. Two dimensional cosine modulated filterbanks are stud-
paraunitary cosine modulated filterbanks. Define
ied and constructed in Refs. 46–48.
Qm (z ) = # Pm (z ) # , m = 0, 1, 2, . . .,M − 1
SOME ADDITIONAL AND RECENT RESEARCH TOPICS
Pm+M (z )
In this section, we want to briefly mention several other top-
ics on multirate filterbank theory and applications.
x (n)
M M
Linear-Phase Perfect Reconstruction Filterbanks
–1 –1
z z
In some applications, such as image/video processing, it is de-
sired that analysis and synthesis filters have no phase distor-

z–1 M 1 1 M z–1
tion. This has motivated the study of linear-phase FIR PR
WM W *
multirate filterbank factorization and construction. Linear-
M M M
⋅ ⋅ ⋅ ⋅
phase multirate filterbanks were first studied independently
⋅ ⋅ ⋅ ⋅
⋅ ⋅ ⋅ ⋅
by Nguyen and Vaidyanathan (49,50), and Vetterli and Le
Gall (51). For more details, see, for example Refs. 7, 9, 11, 12,
–1 –1
z z
45, 49–54.
M M +
^
x (n)
Lapped Orthogonal Transforms

Blocked discrete cosine transform (DCT) plays an important

Figure 22. DFT (analysis bank) and IDFT (synthesis bank).
role in the current image/video coding standards, such as
----------------------- Page 389-----------------------
(a) Prototype filter (b)

Analysis filter
0 0
–10 –10
) )
B –20 B –20
d d
( (
e –30 e –30
s s
n n
o –40 o –40
p p
s s
e e
r –50 r –50
e e
d –60 d –60
u u
t t
i i
n n
g –70 g –70
a a
M –80 M –80
–90 –90
–100 –100
0 0.1 0.2 0.3 0.4 0.5 0
0.1 0.2 0.3 0.4 0.5
Normalized frequency
Normalized frequency
Figure 23. Cosine modulated filterbank prototype

and analysis filters (12).
JPEG, MPEG1 and MPEG2, H.261 and H.263 in blocked DCT and 80.
For more properties on multidimensional basic build-
transform coding, one first decomposes an image on a rectan- ing
blocks, such as decimator and expander, delay chain sys-
gular region into 8 # 8 blocks, and then implement 8 # 8 tems,
and constructions, see for example, Refs. 67–85, 85a.
DCT on each block. Then, one quantizes the blocked DCT co- For
multidimensional cosine modulated filterbanks, see Refs.
efficients to get compression gain. Due to the truncation of an 46–48.
For multidimensional LOT, see Refs. 64, 65, 86, 86a.
image in the block decomposition, the blocking effects degrade
the compression performance at low bit rates. The blocking
effects basically come from a hard truncation between blocks, Optimal
Multirate Filterbanks in Quantization
which causes discontinuities of sinusoidal bases at the bound- One of

the most important applications of multirate filter
aries of these blocks. Based on this observation, overlaps be- banks is
data compression, such as speech, image, and video
tween adjacent blocks are used, such that a smooth trunca-
compression. In this application, an analysis filter bank is
tion between blocks is achieved while no more coefficients in used to
decompose a signal and then a quantization scheme
the transform domain are used and moreover the orthogonal- is used
to compress the signal. In the reconstruction, a de-
ity is still maintained. These overlapped blocked DCT are
quantization scheme and a corresponding synthesis filter
called lapped orthogonal transforms (LOT) (55–58). LOT can bank are
used. This implies that the decomposition and recon-
be thought of as special cases of cosine modulated filterbanks, struction
of a filter bank and the quantization should be con-
(9,12,54), and order 1 PR multirate filterbanks, (9,31,32). sidered
together to achieve the maximal compression gain.
Time-varying filterbanks and LOT are studied in Refs. 59 and Along
this direction, several research works can be found
60. LOT over finite fields are studied in Ref. 61, where a com- (87–93).
plete factorization is given. In the continuous-time domain,
LOT is called Malvar wavelets or local sinusoidal bases. For
more details, see Refs. 62 and 63. For multidimensional Multirate
Filterbanks over Finite Fields
Malvar wavelets, see Refs. 64 and 65. What we
have studied on multirate filterbanks are all over
the
complex field, i.e., all signals and filter coefficients are
Multidimensional Multirate Filterbanks
complex
values. These values can certainly be in any finite
Multirate filterbanks have been used in multidimensional sig- field. Due
to the incompleteness of a finite field, it is impossi-
nal processing, such as image/video processing. One way to ble to
factorize all FIR PR filterbanks over finite fields. A
use multirate filterbanks is to use one-dimensional fil- counter
example was given in Ref. 94. Although this is the
terbanks at each dimension of signals. Another way to use case, a
completely study on LOT (order one multirate fil-
multirate filterbanks is to use multidimensional filterbanks terbanks)
over finite fields was obtained in Ref. 61. It is found
directly to multidimensional signals, which has the potential that
nonmaximally decimated multirate filterbanks over fi-
to take advantage of multidimensional signals themselves, nite
fields are the same as convolutional codes for the error
such as nonhorizontal or nonvertical lines in image analysis. control
purpose (31,32,61,95). For filterbanks over finite
This motivates the study of nonseparable multidimensional fields,
see also Refs. 96, 97.
multirate filterbanks. It is known that all one-dimensional
paraunitary filterbanks can be decomposed into degree-one
Nonuniformally Decimated Multirate Filterbanks

building blocks, which are very useful in the construction of
optimal paraunitary filterbanks and also their implementa- In the
previous sections, we have studied uniformally deci-
tions. This property, however, does not hold for multidimen- mated
multirate filterbanks. There have been also some re-
sional paraunitary filterbanks. For more details, see Refs. 66 search
works on nonuniformally decimated multirate fil-
----------------------- Page 390-----------------------
terbanks (22–27). In particular, tree-structured multirate fil-

BIBLIOGRAPHY
terbanks that may lead to wavelets are nonuniformally deci-
mated multirate filterbanks (6–13). 1.
R. E. Crochiere and L. R. Rabiner, Multirate Digital Signal Pro-
cessing, Englewood Cliffs, NJ: Prentice-Hall, 1983.
2.
M. Vetterli, Perfect transmultiplexers, Proc. IEEE Int. Conf.
Nonmaximally Decimated Multirate Filterbanks
Acoust. Speech Signal Process., pp. 2567–2570, Tokyo, Japan,
1986.
Nonmaximally decimated multirate filterbanks have been re-
cently applied as precoders in communication systems, such 3.
R. D. Koilpillai, T. Q. Nguyen, and P. P. Vaidyanathan, Some
as intersymbol interference (ISI) cancellation (16–21,98,99)
results in the theory of cross-talk free transmultiplexers, IEEE
Trans. Signal Process., SP-39: 2174–2183, 1991.

and image quantization and coding (100,101). They are ap-
plied in the prefilter design for discrete multiwavelet trans- 4.
J. M. Shapiro, Embedded image coding using zerotree of wavelet
coefficients, IEEE Trans. Signal Process., SP-41: 3445–3462,
forms (98,99).
1993.
5.
A. Said and W. A. Pearlman, A new and efficient image codec
Time-Varying, Vector Filterbanks and Multirate
based on set partitioning in hierarchical trees, IEEE Trans. Cir-
Filterbanks with Block Sampling
cuits Syst. Video Tech., June 1996.
6.
O. Rioul and M. Vetterli, Wavelets and signal processing, IEEE
Because of the nonstationarity of signals, such as speech,
Signal Process. Magazine, 8: 14–38, 1991.
audio, and images, time-varying filter banks have been stud-
7.
M. Vetterli and C. Herley, Wavelets and filter banks: Theory
ied to exploit the nonstationarity, where filter banks change
and design, IEEE Trans. Signal Process., SP-40: 2207–2232,
dynamically with the nature of a signal. An important prob-
1992.
lem in time-varying filter bank theory is how to construct 8.
A. Akansu and R. Haddad, Multiresolution Signal Decomposi-
time-varying filter banks with PR property from local PR fil-
tion: Transforms, Subbands, and Wavelets, Boston: Academic
ter banks. For this problem, there have been several discus-
Press, 1992.
sions (60,102–114). 9.
P. P. Vaidyanathan, Multirate Systems and Filter Banks, Engle-
wood Cliffs, NJ: Prentice Hall, 1993.
Nonlinear Multirate Filterbanks 10.

N. J. Fliege, Multirate Digital Signal Processing, New York: Wi-
ley, 1994.
Conventional filter banks are all linear transformations. It is
˘ ´
11.
M. Vetterli and J. Kovacevic, Wavelets and Subband Coding,
well-known that all linear transformations may smear edges
Englewood Cliffs, NJ: Prentice-Hall, 1995.
that are important in image/video processing. Since nonlinear 12.
G. Strang and T. Q. Nguyen, Wavelets and Filter Banks, Welles-
transformations, such as median filtering, may preserve
ley-Cambridge Press, 1996.
edges better than linear transformations do, nonlinear filter
12a. P. P. Vaidyanathan et al., Improved technique for design of
banks have been recently studied (115–118).
perfect reconstruction FIR QMF banks with lossless polyphase
matrices, IEEE Trans. Acoust. Speech Signal Process., ASSP-37 :
1042–1056, 1989.
SUMMARY 13.
A. N. Akansu and M. J. T. Smith, Subband and Wavelet Trans-
forms: Design and Applications, Boston: Kluwer, 1996.
In this article, we introduced some fundamentals on 14.

M. Bellanger, G. Bonnerot, and M. Coudreuse, Digital filtering
multirate filterbank theory and applications. Beginning with
by polyphase network: Application to sample rate alteration and
filter banks, IEEE Trans. Acoust. Speech Signal Process., ASSP-

some application backgrounds, such as transmultiplexer in-
24: 109–114, 1976.

cluding TDMA, FDMA, and CDMA communications, and
15.
P. Vary, On the design of digital filter banks based on a modified
multiresolution image/video analysis and coding, we briefly
principle of polyphase, AEU, 33: 293–300, 1979.

mentioned the importance of the construction of PR multirate
filterbanks. For systematic construction of PR multirate fil- 16.
X.-G. Xia, New precoding for intersymbol interference cancella-
tion using nonmaximally decimated multirate filterbanks with

terbanks, in particular paraunitary multirate filterbanks,
ideal FIR equalizers, preprint, 1995, IEEE Trans. Signal Pro-

some basic building blocks, such as decimator and expander,
cess., 45, 2431–2441, 1997.
the Noble identities, and polyphase representations, and their
17.
H. Liu and X.-G. Xia, Precoding techniques for undersampled
properties were briefly described. These properties were used
multi-receiver communication systems, Tech. Report #97-3-1,
to convert a maximally decimated multirate filterbank into
Dept. Elect. Eng., Univ. of Delaware, 1997.
an MIMO system that has a system transfer matrix polyno-
18.
X.-G. Xia and H. Liu, Polynomial ambiguity resistant precoders:
mial with free variables (in mathematics, a free-module). This
Theory and applications in ISI/multipath cancelation, Tech. Re-
conversion makes the systematic study easier and the com-
port #97-5-1, Dept. Elect. Eng., Univ. of Delaware, 1997.
plete lattice factorization possible. After the study of general 19.
X.-G. Xia and G. Zhou, On optimal ambiguity resistant precod-
maximally decimated multirate filterbanks, we then briefly
ers in ISI/multipath cancellation, Tech. Report #97-5-2, Dept.
introduced the DFT and cosine modulated filterbanks, which
Elect. Eng., Univ. of Delaware, 1997.
are often used in practical applications. Finally, we briefly 20.
X.-G. Xia, Modulated coding and least square decoding via coded
mentioned some additional and recent research topics in
modulation and Viterbi decoding, Tech. Report #97-6-2, Dept.
multirate filterbank theory and applications. As we men-
Elect. Eng., Univ. of Delaware, 1997.
tioned in the introduction, this article covers only some basics 21.
G. B. Giannakis, Filterbanks for blind channel identification
of multirate filterbanks.
and equalization, IEEE Signal Process. Lett., 4: 184–187, 1997.
----------------------- Page 391-----------------------
22. P.-Q. Hoang and P. P. Vaidyanathan, Nonuniform multirate fil- 38.

H.-J. Park,A Computational Theory of Laurent Polynomial Rings
ter banks: Theory and design, Proc. IEEE Int. Symp. Circuits
and Multidimensional FIR Systems, Ph.D. dissertation, Dept. of
Syst., pp. 371–374, Portland, Oregon, 1989.
Mathematics, Univ. of Berkeley, 1995.
23. K. Nayebi, T. P. Barnwell III, and M. J. T. Smith, Nonuniform 39.

M. R. Portnoff, Time-frequency representation of digital signals
filter banks: A reconstruction and design theory, IEEE Trans.
and systems based on short-time Fourier analysis, IEEE Trans.
Signal Process., SP-41, 1993.
Acoust. Speech Signal Process., ASSP-28 : 55–69, 1980.
24. J. Li, T. Q. Nguyen, and S. Tantaratana, A simple design 40.

M. J. Bastiaans, Gabor’s expansion of a signal into Gaussian
method for nonuniform multirate filter banks, Proc. Asilomar
elementary signals, Proc. IEEE, 68: 594–598, 1980.
Conf. Signals Systems Computers, pp. 1015, Pacific Grove, CA, 41.
J. Wexler and S. Raz, Discrete Gabor expansions, Signal Pro-
1994.
cess., 21: 207–220, 1990.
25. J. Princen, The design of nonuniform modulated filterbanks, 42.

S. Qian and D. Chen, Discrete Gabor transform, IEEE Trans.
IEEE Trans. Signal Process., SP-43: 2550, 1995.
Signal Process., SP-41: 2429–2438, 1993.
26. J.-J. Lee and B. G. Lee, A design of nonuniform cosine modu- 42a.
S. Qian and D. Chen, Joint Time-Frequency Analysis, Prentice-
lated filter banks, IEEE Trans. Circuits Syst. II Analog Digit.
Hall, New Jersey, 1996.
Signal Process., 42: 732, 1995. 43.
R. D. Koilpillai and P. P. Vaidyanathan, Cosine-modulated FIR
27. T. Chen, Nonuniform multirate filter banks: analysis and design

filter banks satisfying perfect reconstruction, IEEE Trans. Sig-
with an H(#) performance measure, IEEE Trans. Signal Pro-
nal Process., SP-40: 770–783, 1992.
cess., SP-45: 572, 1997. 44.
T. Q. Nguyen and R. D. Koilpillai, The theory and design of
28. M. J. T. Smith and T. P. Barnwell III, Exact reconstruction tech-

arbitrary-length cosine-modulated filter banks and wavelets,
niques for tree-structured subband coders, IEEE Trans. Acoust.
satisfying perfect reconstruction, IEEE Trans. Signal Process.,
Speech Signal Process., ASSP-34 : 434–441, 1986.
SP-44: 473–483, 1996.
28a. M. J. T. Smith and T. P. Barnwell III, A new filter-bank theory 45.

Y.-P. Lin and P. P. Vaidyanathan, Linear phase cosine modu-
for time-frequency representation, IEEE Trans. Acoust. Speech
lated maximally decimated filter banks with perfect reconstruc-
tion, IEEE Trans. Signal Process., SP-43: 2525, 1995.
29. F. Mintzer, Filters for distortion-free two-band multirate filter 46.

Y.-P. Lin and P. P. Vaidyanathan, A review of two-dimensional
banks, IEEE Trans. Acoust. Speech Signal Process., ASSP-33 :
nonseparable filter bank design techniques, Multidim. Syst. Sig-
626–630, 1985.
nal Process., Special Issue on Multidimensional Filter Banks,
29a. H. J. Nussbaumer, Pseudo QMF filter bank, IBM Tech. Disclo-
1995.
sure Bulletin, 24: 3081–3087, 1981. 47.
Y.-P. Lin and P. P. Vaidyanathan, On the study of four-parallel-
29b. P. L. Chu, Quadrature mirror filter design for an arbitrary
ogram filter banks, IEEE Trans. Signal Process., SP-44: 2707,
number of equal bandwidth channels, IEEE Trans. Acoust.
1996.
Speech Signal Process., ASSP-33 : 203–218, 1985. 48.

Y.-P. Lin and P. P. Vaidyanathan, Theory and design of two-
29c. R. V. Cox, The design of uniformly and nonuniformly spaced
parallelogram filter banks, IEEE Trans. Signal Process., SP-44:
pseudo QMF, IEEE Trans. Acoust. Speech Signal Process., ASSP-
2688, 1996.
34: 1090–1096, 1986. 49.

T. Q. Nguyen and P. P. Vaidyanathan, Two-channel perfect re-
30. P. P. Vaidyanathan and S. K. Mitra, Polyphase networks, block
construction FIR QMF structures which yield linear phase FIR
digital filtering, LPTV systems, and alias-free QMF banks: A
analysis and synthesis filters, IEEE Trans. Acoust. Speech Sig-
unified approach based on pseudocirculants, IEEE Trans.
nal Process., ASSP-37 : 676–690, 1989.
Acoust. Speech Signal Process., ASSP-36 : 381–391, 1988. 50.

T. Q. Nguyen and P. P. Vaidyanathan, Structures for M-channel
31. P. P. Vaidyanathan and T. Chen, Role of anticausal inverses in
perfect reconstruction FIR QMF banks which yield linear phase
multirate filter-banks—Part I: System-theoretic fundamentals,
analysis filters, IEEE Trans. Acoust. Speech Signal Process.,
IEEE Trans. Signal Process., SP-43: 1090–1102, 1995.
ASSP-38 : 433–446, 1990.
32. P. P. Vaidyanathan and T. Chen, Role of anticausal inverses in 51.

M. Vetterli and D. Le Gall, Perfect reconstruction FIR filter
multirate filter-banks—Part II: The FIR case, factorizations,
banks: Some properties and factorizations, IEEE Trans. Acoust.
and biorthogonal lapped transforms, IEEE Trans. Signal Pro-
Speech Signal Process., ASSP-37 : 1057–1071, 1989.
cess., SP-43: 1103–1115, 1995. 52.
S.-M. Phoong, C. W. Kim, P. P. Vaidyanathan, and R. Ansari, A
33. P. P. Vaidyanathan, Theory and design of M-channel maximally
new class of two-channel biorthogonal filter banks and wavelet
decimated quadrature mirror filters with arbitrary M, having
bases, IEEE Trans. Signal Process., SP-43: 649, 1995.
perfect reconstruction property, IEEE Trans. Acoust. Speech Sig- 53.
L. Chen, T. Q. Nguyen, and K.-P. Chan, Symmetric extension
nal Process., ASSP: 35 : 476–492, 1987.
methods for M-channel linear-phase perfect-reconstruction filter
34. P. P. Vaidyanathan, Quadrature mirror filter banks, M-band ex-
banks, IEEE Trans. Signal Process., SP-43: 2505, 1995.
tensions and perfect-reconstruction techniques, IEEE ASSP 54.
R. L. de Queiroz, T. Q. Nguyen, and K. R. Rao, The GenLot:
Magazine, 4: 4–20, 1987.
Generalized linear-phase lapped orthogonal transform, IEEE
35. P. P. Vaidyanathan, Multirate digital filters, filter banks, poly-
phase networks, and applications: A tutorial, Proc. IEEE, 78: 55.
H. S. Malvar and D. H. Staelin, The LOT: Transform coding
56–93, 1990.
without blocking effects, IEEE Trans. Acoust. Speech Signal Pro-
˘
cess., ASSP-37 : 553–559, 1989.
36. Z. Doganata, P. P. Vaidyanathan, and T. Q. Nguyen, General
synthesis procedures for FIR lossless transfer matrices, for per- 56.
H. S. Malvar, Lapped transforms for efficient transform/sub-
fect-reconstruction multirate filter bank applications, IEEE
band coding, IEEE Trans. Acoust. Speech Signal Process., ASSP-
Trans. Acoust. Speech Signal Process., ASSP-36 : 1561–1574,
38: 969–978, 1990.
1988. 57.
H. S. Malvar, Signal Processing with Lapped Transforms, Bos-
37. X.-G. Xia and B. W. Suter, FIR paraunitary filter banks given
ton: Artech House, 1992.
several analysis filters: Factorization and construction, IEEE 58.

A. N. Akansu and F. E. Wadas, On lapped orthogonal trans-
forms, IEEE Trans. Signal Process., 40: 439–443, 1992.
----------------------- Page 392-----------------------
59. R. L. de Queiroz and K. R. Rao, Time-varying lapped transforms 81.

S. Venkataraman and B. C. Levy, A comparison of design meth-
and wavelet packets, IEEE Trans. Signal Process., SP-41: 3293–
ods for 2-D FIR orthogonal perfect reconstruction filter banks,
3305, 1993.
IEEE Trans. Circuits Syst. II, Analog Digit. Signal Process., 42:
60. X.-G. Xia and B. W. Suter, Construction of perfect reconstruc-
525, 1995.
tion time-varying FIR multirate filter banks with overlaps, Proc. 82.
M. Ikehara, K. Sawaaki, H. Isobe, A. Yamashita, and H. Kur-
IEEE Int. Symp. Time-Freq. Time-Scale Anal., Philadelphia,
oda, Design of 2-D perfect reconstruction filter banks for arbi-
Oct., 1994.
trary sampling lattices—Design of 2-D N-th-band digital filters,
61. S.-M. Phoong and P. P. Vaidyanathan, Paraunitary filter banks
Electron. and Commun. Japan. Part 3, Fundamental Electron.
over finite fields, IEEE Trans. Signal Process., SP-45: 1443–
Sci., 76: 85, 1993.
1457, 1997. 83.
B. L. Evans, R. H. Bamberger, and J. H. McClellan, Rules for
62. R. Coifman and Y. Meyer, Remarques sur l’analyse de Fourier
multidimensional multirate structures, IEEE Trans. Signal Pro-
` ˆ
cess., SP-42: 762–771, 1994.
a fenetre, C.R. Acad. Sci. Paris, 312, Serie I: 259–261, 1991.
63. P. Auscher, G. Weiss, and M. V. Wickerhauser, Local sine and 84.

R. A. Gopinath and C. S. Burrus, On upsampling, downsam-
cosine basis of Coifman and Meyer and the construction of
pling, and rational sampling rate filter banks, IEEE Trans. Sig-
smooth wavelets, in C. K. Chui (ed.), Wavelets: A Tutorial in
nal Process., SP-42: 812–824, 1994.
Theory and Applications, Boston: Academic Press, 1992. 85.
X.-G. Xia, B. W. Suter, and M. E. Oxley, On necessary and suf-
64. X.-G. Xia and B. W. Suter, A family of two dimensional nonsepa-

ficient conditions for perfect reconstruction multidimensional
rable Malvar wavelets, Appl. Comp. Harm. Anal., 2: 243–256,
delay chain systems, IEEE Trans. Signal Process., SP-43: 1515–
1995.
1519, 1995.
65. X.-G. Xia and B. W. Suter, Construction of Malvar wavelets on 85a.

T. A. C. M. Kalker and I. A. Shah, A group theoretic approach
hexagons, Appl. Comp. Harm. Anal., 3: 65–71, 1996.
to multidimensional filter banks: theory and applications, IEEE
66. S. Venkataraman and B. C. Levy, State space representations
Trans. Signal Process., SP-44:1392, 1996.
of 2-D FIR lossless transfer matrices, IEEE Trans. Circuits Syst. 86.
X.-G. Xia and B. W. Suter, A systematic construction method
II, Analog Digit. Signal Process., 41: 117–132, 1994.
for spatial-varying FIR filter banks with perfect reconstruction,
67. M. Vetterli, Multidimensional sub-band coding: Some theory

Proc. 1994 IEEE Int. Conf. Image Process., pp. 830–833, Austin,
and algorithms, Signal Processing, 6: 97– 112, 1984.
Texas, 1994.
68. D. E. Dudgeon and R. M. Mersereau, Multidimensional Digital

˘ ´
86a.
J. Kovacevic, Local cosine bases in two dimensions, Proc. Int.
Signal Processing, Englewood Cliffs, NJ: Prentice-Hall, 1984.
Conf. Acoust. Speech Signal Process., pp. 2125–2128, Detroit,
69. J. W. Woods, Subband Image Coding, Boston: Kluwer Academic
1995.
Publishers, 1991. 87.

R. A. Haddad and K. Park, Modeling, analysis, and optimum
70. V. C. Liu and P. P. Vaidyanathan, On the factorization of a sub-
design of quantized M-band filter banks, IEEE Trans. Signal
class of 2-D digital FIR lossless matrices for 2-D QMF bank ap-
Process., SP-43: 2540, 1995.
plications, IEEE Trans. Circuits Syst., CAS-37: 852–854, 1990. 88.
A. Benyassine and A. N. Akansu, Subspectral Modeling in Filter
71. G. Karlsson and M. Vetterli, Theory of two-dimensional
Banks, IEEE Trans. Signal Process., SP-43: 3050, 1995.
multirate filter banks, IEEE Trans. Acoust. Speech Signal Pro- 89.
S. Ohno and H. Sakai, Optimization of filter banks using cyclos-
cess., ASSP-38 : 925–937, 1990.
tationary spectral analysis, IEEE Trans. Signal Process., SP-44:
72. E. Viscito and J. P. Allebach, The analysis and design of multi-
2718, 1996.
dimensional FIR perfect reconstruction filter banks for arbitrary 90.
S. Wada, H. Yagi, and T. Saito, A flexible design of nonuni-
sampling lattices, IEEE Trans. Circuits and Syst., CAS-38: 29–
formly oversampled multirate filter banks, Proc. IEEE Int.
41, 1991.
Symp. Circuits Syst., 1996.
73. P. P. Vaidyanathan, Fundamentals of multidimensional 91.

B. Xuan and R. H. Bamberger, Complete FIR principal compo-
multirate digital signal processing, Sadhana, 15: 157–176,
nent filter banks, Proc. IEEE Int. Symp. Circuits Syst., 1996.
1990.
˘ ´ 92.
A. Delopoulos and S. Kollias, Optimal filter banks for signal re-
74. J. Kovacevic and M. Vetterli, The commutativity of up/down-
construction from noisy subband components, IEEE Trans. Sig-

sampling in two dimensions, IEEE Trans. Inf. Theory, IT-37:
nal Process., SP-44: 212, 1996.

695–698, 1991.
˘ ´ 93.
P. P. Vaidyanathan, Theory of optimal orthonormal filter banks,
75. J. Kovacevic and M. Vetterli, Nonseparable multidimensional
n
Proc. Int. Conf. Acoust. Speech Signal Process., pp. 1487–1490,
perfect reconstruction filter banks and wavelet bases for R ,
Atlanta, Georgia, 1996.

IEEE Trans. Inf. Theory, IT-38: Feb. 1992.
94.
P. P. Vaidyanathan, Unitary and paraunitary systems in finite
76. R. H. Bamberger and M. J. T. Smith, A filter bank for the direc-
tional decomposition of images: theory and design, IEEE Trans.
fields, Proc. IEEE ISCAS, New Orleans, LA., pp. 1189–1192,
1990.
77. A. A. C. M. Kalker, Commutativity of up/down sampling, Elec- 95.
X.-G. Xia, Applications of multirate filterbanks in error correc-
tron. Lett., 28: 567–569, 1992.
tion codes: Partial response channels, Proc. of SPIE: Wavelet Ap-
plications III, 2762: 478–489, Orlando, 1996.

78. T. Chen and P. P. Vaidyanathan, The role of integer matrices
in multidimensional multirate systems, IEEE Trans. Signal Pro- 96.
M. Vetterli, Filter banks allowing for perfect reconstruction,
cess., SP-41: 1035–1047, 1993.
Signal Process., 10: 219–244, 1986.
79. T. Chen and P. P. Vaidyanathan, Recent developments in multi- 97.

M. D. Swanson and A. H. Twefik, A binary wavelet decomposi-
dimensional multirate systems, IEEE Trans. Circuits Syst. Video
tion of binary images, IEEE Trans. Image Process., 5: 1637–
Technol., 3: 115–137, 1993.
1650, 1996.
80. S. Basu and C. Chiang, A complete parameterization of 2-D non- 98.
X.-G. Xia, J. S. Geronimo, D. P. Hardin, and B. W. Suter, Design
separable orthogonal wavelets, Proc. IEEE Signal Process. Symp.
of prefilters for discrete multiwavelet transforms, IEEE Trans.
Time-Freq. Time-Scale Anal., Victoria BC, Canada, 1992.
----------------------- Page 393-----------------------
MULTIVARIABLE SYSTEMS 51
99. X.-G. Xia, A new prefilter design for discrete multiwavelet

MULTISENSOR DATA FUSION. See DATA FUSION .
transforms, Tech. Report #96-11-2, Dept. of Elect. Eng., Univ. of
MULTISPECTRAL IMAGE ANALYSIS. See IMAGE CLAS-
Delaware, 1996.
SIFICATION.
100. Z. Cvetkovic and M. Vetterli, Oversampled Filter Banks, IEEE
MULTISPECTRAL PATTERN RECOGNITION. See IM-

Trans. Signal Process., to appear.
AGE CLASSIFICATION .
101. Z. Cvetkovic and M. Vetterli, Robustness of overcomplete ex-
pansions, in J. Zeevi and R. Coifman (eds.), Signal and Image
MULTISTAGE DECISION MAKING. See DYNAMIC PRO-
Representation in Combined Spaces, Boston: Academic Press,
GRAMMING.
1996.
MULTITARGET TRACKING. See SONAR TRACKING.
102. K. Nayebi, T. P. Barnwell, and M. J. T. Smith, Analysis-synthe-
MULTITHREADED ARCHITECTURES. See DATA-FLOW
sis systems with time-varying filter bank structures, Proc. IEEE
AND MULTITHREADED ARCHITECTURES .

Int. Conf. Acoust. Speech Signal Process., pp. 617–620, San Fran-
MULTIUSER SYSTEMS. See TIME-SHARING SYSTEMS.

cisco, 1992.
MULTIVARIABLE CONTROL SYSTEMS. See H #

103. I. Sodagar, K. Nayebi, T. P. Barnwell III, and M. J. T. Smith,
Time-varying analysis-synthesis systems based on filter banks
CONTROL.
and post filtering, IEEE Trans. Signal Process., SP-43: 2512,

1995.
104. C. Herley and M. Vetterli, Orthogonal time-varying filter banks

and wavelet packets, IEEE Trans. Signal Process., SP-42:
2650, 1994.
105. C. Herley, Boundary filters for finite-length signals and time-
varying filter banks, IEEE Trans. Circuits Syst. 2, Analog Digit.
Signal Process., 42: 102, 1995.
106. T. Chen and P. P. Vaidyanathan, Time reversed inversion for

time-varying filter banks, with applications in subband coding
of images, Proc. 27th Annual Asilomar Conf. Signals, Syst. Com-
puters, Pacific Grove, CA, 1993.
107. R. L. de Queiroz and K. R. Rao, Structure for time-varying para-
unitary filter banks with perfect reconstruction, Electronic Let-
ters, 29 (2): 217, 1993.
108. R. A. Gopinath and C. S. Burrus, Factorization approach to uni-

tary time-varying filter bank trees and wavelets, IEEE Trans.
109. S.-M. Phoong and P. P. Vaidyanathan, Time-varying filters and

filter banks: Some basic principles, IEEE Trans. Signal Process.,
SP-44: 2971, 1996.
110. M. R. K. Khansari and A. Leon-Garcia, Subband decomposition

of signals with generalized sampling, IEEE Trans. Signal Pro-
cess., SP-41: 3365–3376, 1993.
111. W. Li and Y.-Q. Zhang, A proposal for very low bit rate video
coding standard using vector subband coding (VSC), ISO/IEC
JTC1/SC29/WG11, MPEG-94-209, July 1994.
112. V. M. Gadre and R. K. Patney, Vector multirate filtering and
matrix filter banks, Proc. IEEE Int. Symp. Circuits Syst., pp.
1360–1363, 1994.
113. X.-G. Xia and B. M. Suter, Multirate filter banks with block
sampling, IEEE Trans. Signal Process., SP-44: 484–496, 1996.
114. X.-G. Xia and B. M. Suter, Vector-valued wavelets and vector

filter banks, IEEE Trans. Signal Process., SP-44: 508–518, 1996.
115. D. W. Redmill and D. R. Bull, Nonlinear perfect reconstruction
critically decimated filter banks, Electron. Lett., 32 (4): 310,
1996.
116. F. Hampson and J. Pesquet, A nonlinear subband decomposition

with perfect reconstruction, Proc. IEEE Int. Conf. Acoust. Speech
Signal Process., pp. 1523–1527, Atlanta, GA, 1996.
117. R. L. de Queiroz and D. Florencio, A pyramidal coder using a
non-linear filter bank, Proc. IEEE Int. Conf. Acoust. Speech Sig-
nal Process., pp. 1570–1573, Atlanta, GA, 1996.
118. D. Florencio and R. Schafer, Perfect reconstruction non-linear

filter banks, Proc. IEEE Int. Conf. Acoust. Speech Signal Process.,
pp. 1814–1817, Atlanta, GA, 1996.
XIANG-GEN XIA
University of Delaware
----------------------- Page 394-----------------------
} { { } }
●
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Ocean Thermal Energy Conversion
Standard Article

Stephen M. Masutani1 and Patrick K. Takahashi 1
1University of Hawaii, Honolulu, HI

reserved.
Search
DOI: 10.1002/047134608X.W3029
❍ Acronym Finder
Abstract
History of OTEC
Principles of Operation
Closed Cycle OTEC
Kalina Cycle
Open Cycle OTEC
Mist and Foam Lift OTEC Cycles
Hybrid Cycle OTEC
Comparison of OTEC Open and Closed Cycles
Open and Closed Cycle OTEC System Performance
Economics of OTEC
----------------------- Page 395-----------------------

----------------------- Page 396-----------------------
OCEAN THERMAL ENERGY CONVERSION 93
OCEAN
THERMAL ENERGY CONVERSION
Ocean
thermal energy conversion (OTEC) indirectly converts
solar energy
into electricity. From a thermodynamic perspec-
tive, OTEC
power cycles operate as continuous heat engines
driven by
the transfer of energy between a thermal source
and sink.
OTEC expends renewable solar energy; therefore,
recurring
costs to generate electrical power are minimal.
However, the
fixed or capital costs of OTEC systems per kilo-
watt of
installed generating capacity are very high. This is a
consequence
of the low theoretical efficiency of OTEC, which
demands
large components, such as heat exchangers and
pipelines,
to accommodate the thermal energy transfers nec-
essary to
produce small amounts of electricity. The high fixed
costs
dominate the economics of OTEC to the extent that it
currently
cannot compete with conventional power systems
except in
limited niche markets. Toward this end, consider-
able effort
has been expended over the past two decades to
develop OTEC
by-products, such as freshwater, air condition-

----------------------- Page 397-----------------------
94 OCEAN THERMAL ENERGY CONVERSION
Figure 1. Temperature difference between surface

seawater and ocean depths of 1000 m.
ing, and mariculture, that might offset the cost penalty of between
the surface and 1000 m, this requirement limits the
electricity generation. number of
candidate OTEC sites. One of the principal techni-
Warm surface waters of tropical oceans make up the ther- cal
challenges of recent years, therefore, has been to devise
mal resource used by OTEC. The oceans, which cover more
strategies to export the benefits of the renewable OTEC re-
than 70% of the earth’s surface, intercept solar radiation source
to locations outside the tropics, possibly through the
passing through the atmosphere. Even though a portion of
production of synthetic fuels, such as the generation of hydro-
this energy is reradiated directly back to space, a significant gen gas
by electrolysis.
fraction is retained by seawater at the lower latitudes, heat-
ing the upper mixed layer of tropical oceans to an average
year-round temperature of approximately 28#C. This energy HISTORY
OF OTEC
is subsequently advected to higher latitudes—where radiant
losses exceed gains—by oceanic and atmospheric circulation The OTEC
concept was first proposed in 1881 by the French
(the net transfer of energy between the earth and its sur- engineer
J. A. D’Arsonval. D’Arsonval recommended using
roundings must, of course, be zero in order to maintain equi-
pressurized, liquefied gases as working fluids in a heat engine
librium). The amount of solar radiation retained by the operating
across the temperature difference between the sur-
oceans is enormous. Each day, energy equivalent to approxi- face and
deep waters of the equatorial oceans. A half-century
mately 250 billion barrels of oil is absorbed over the 60 mil- later,
in 1930, G. Claude, D’Arsonval’s former student, field-
lion km2 of tropical seas (1). This is more than three orders of tested a
variation of the OTEC concept in northern Cuba.
magnitude greater than the current daily energy consumption Claude’s
power cycle used warm seawater evaporated in a
of the world’s population. partial
vacuum as a working fluid. His shore-based, 50 kW
Inasmuch as the second law of thermodynamics dictates plant
consumed more power than it generated because of an
that only a fraction of the energy extracted from warm sea- overly
conservative cold seawater system that pumped 10
water can be converted to usable work by a power cycle, a times
more water than required. The plant operated for 11
thermal sink must be available to accept waste heat. The days
before the cold water pipe failed in a storm. In spite of
oceans provide such a sink in the form of a bottom layer of its
shortcomings, the experiment was the first to demonstrate
cold water lying beneath the warm, well-mixed (by wind and clearly
the technical feasibility of OTEC. Encouraged by the
waves) surface zone. This reservoir of cold water forms in the Cuba
tests, Claude launched a second attempt in 1933 to de-
polar regions and descends to flow along the sea floor toward velop
OTEC, this time in the form of a plantship moored
the equator (2). The warm surface layer, which can extend to about 100
km off the Brazilian coast. The power system was
depths of 100 m, is separated from the deep cold water by a sized to
produce 2 MW of turbine shaft power. Rather than
thermocline. The vertical gradient in temperature below the attempt
to export this power to the coastline directly, Claude
mixed layer is usually substantial. The temperature differ- decided
to produce ice, which had a favorable market value.
ence between the surface and 1000 m depth ranges from 10
Unfortunately, the expensive vertical cold water pipeline was
to 25#C, with larger differences occurring in equatorial and lost
during deployment, resulting in cancellation of the proj-
tropical waters. ect.
Although Claude continued to champion OTEC, the goal
The performance of OTEC power cycles depends ultimately of an
operating commercial plant was to elude him.
on the available difference in the temperatures of the warm
Following the 1933 project, several OTEC design studies
and cold seawater
T. The rule of thumb is that a
T of about were conducted. In 1956, a French team was preparing to in-
20#C is necessary to sustain viable operation of an OTEC stall a
shore-based, 5 MW Claude cycle plant at a site on the
power station. As shown in Fig. 1, which maps average
T west coast of Africa. The project was abandoned when it was
----------------------- Page 398-----------------------
announced that a large hydroelectric station would be con-

Warm
structed nearby. During the 1960s, J. H. Anderson and his
Thermal sea water
son published plans for a 100 MW floating OTEC power plant
resource temp. = TH
that revived D’Arsonval’s original concept of using a pressur-
∆QH
ized, low boiling point gas (propane) as a working fluid in a
Rankine power cycle. Although the public and private sectors
∆W
remained indifferent, the Andersons’ advocacy of OTEC at-
Heat n
engine
tracted several key supporters in the early 1970s, notably W.
E. Heronemus of the University of Massachusetts and C. Ze-
∆QL
ner of Carnegie-Mellon University. In 1972, the National Sci-
ence Foundation awarded a grant to the University of Massa-
Cold
chusetts to assess the technical and economic feasibility of the
Thermal sea water
OTEC process. A second grant was awarded a year later to
sink temp. = TL
Carnegie-Mellon to investigate other elements of the OTEC

system.
Figure 2. Cyclic OTEC heat engine.
The oil embargo of 1973 to 1974, combined with growing
public concerns over the safety of nuclear power, forced the
world to reassess its dependence on fossil fuels and triggered portion
of the heat extracted from the warm seawater must
a frantic search for alternative energy resources. Interest in be
rejected to a colder thermal sink. The thermal sink em-
OTEC therefore enjoyed a sudden resurgence, and numerous ployed
by OTEC systems is seawater drawn from the ocean
development projects were initiated throughout the world. depths
by means of a submerged pipeline.
Several milestones in OTEC research and development The
essential features of a cyclic OTEC heat engine are
have been achieved since that time. In 1979, a consortium of
presented in Fig. 2. A steady state control volume energy
private businesses led by Lockheed Missiles & Space Com- analysis
yields the result that net electrical power produced
pany and the Dillingham Corporation, in partnership with by the
engine must equal the difference of the rates of heat
the Hawaii state government, began operation of the first transfer
from the warm surface water to the cold deep water
floating OTEC plant to produce net power. Named Mini- or,
using the notation in Fig. 2,
OTEC, the plant was anchored off the western coast of the
#Wn = #QH − #QL (1)

island of Hawaii. The ammonia Rankine cycle produced about
50 kW at the generator terminals and provided over 600
The
conversion efficiency # of a heat engine is the ratio of
hours of data over three months. Two years later, Japanese
usable
power generated to thermal energy received:
researchers began operation of a shore-based 100 kW Ran-
kine cycle plant on the island of Nauru in the south Pacific.
η = #W /#Q (2)
n H
Tests were conducted for about one year. A record power out-
put of 120 kW (gross) was attained before the cold water pipe- This
efficiency is maximized when entropy production is zero
line was damaged by a storm. A nominal 210 kW Claude cycle (i.e.,
when the engine operates reversibly). Such an ideal de-
OTEC plant has been producing electricity since 1993 at the vice is
called a Carnot heat engine, and its performance estab-
Natural Energy Laboratory of Hawaii, also on the island of lishes
an upper bound for real thermal power generation sys-
Hawaii. The facility, which is operated by the Pacific Interna- tems.
The theoretical Carnot efficiency is a function only of
tional Center for High Technology Research (PICHTR), is the
absolute temperatures of the thermal resource TH and
funded by the U.S. Department of Energy and Hawaii state thermal
sink TL :
government. Over the past three years, it has set new records
for both gross (255 kW) and net (103 kW) electrical power. It
η = 1 − (T /T ) = #T/T (3)
L H H
also has successfully produced potable water.
Although funding for recent OTEC programs has waxed Assuming
a maximum ocean surface temperature TH of about
and waned over the more than 20 years since the first oil 30#C
(303 K) and a typical deep seawater temperature TL of
embargo, research and development activities have signifi- 4#C (277
K), the limiting performance of OTEC power systems
cantly advanced OTEC technology. Unfortunately, none of the is
programs to date has taken the critical step of constructing
and operating the first commercial-scale (MW scale) pilot
η = 1 − (277/303) = 0.086 or 8.6% (4)
OTEC power station. Until this happens, the vast OTEC en-
ergy resource will continue to remain untapped. This
implies that more than 90% of the thermal energy ex-
tracted
from the ocean’s surface is ‘‘wasted’’ and must be re-
jected
to the cold, deep seawater. The low efficiency of OTEC
PRINCIPLES OF OPERATION
necessitates large heat exchangers and seawater flow rates to
sustain
the
Q’s needed to produce relatively small amounts
OTEC power systems operate as cyclic heat engines. They re- of
electricity. In contrast, the Carnot efficiency of a state-of-
ceive thermal energy through heat transfer from a resource, the-art
combustion steam power cycle, which taps a much
here, surface seawater warmed by the sun, and transform a higher
temperature energy source, may exceed 60%. The
portion of this energy into electrical power. The Kelvin-Planck lower
thermal quality of the OTEC resource, therefore, im-
statement of the second law of thermodynamics precludes poses a
significant penalty on the heat engine that ultimately
complete conversion of the thermal energy into electricity. A is
manifested in high capital costs.
----------------------- Page 399-----------------------

In spite of its inherent inefficiency, OTEC, unlike conven- Latent

heat is transferred from the vapor to the cold seawater
tional fossil energy systems, uses a renewable resource and in the
condenser, and the resulting liquid is pressurized with
poses a minimal threat to the environment. a pump to
begin the cycle again.
Carnot efficiency applies only to an ideal heat engine. In The
success of the Rankine cycle is a consequence of more
actual OTEC systems, irreversibilities will further degrade energy
being recovered when the vapor expands through the
cycle performance from the already low theoretical limit. Suc- turbine
than is consumed in repressurizing the liquid. In con-
cessful implementation of OTEC power generation thereby ventional
(e.g., combustion) Rankine systems, this yields net
demands careful engineering to minimize irreversibilities. Al-
electrical power. For OTEC, however, the remaining balance
though OTEC consumes what is essentially a free resource, may be
reduced substantially by an amount needed to pump
poor thermodynamic performance will reduce the quantity of large
volumes of seawater through the heat exchangers. Rep-
electricity available for sale and, hence, will negatively affect
resentative values of closed cycle OTEC back-work are pro-
the economic feasibility of an OTEC facility by increasing the vided in
an example that follows.
payback period for capital investments. One
misconception about OTEC is that tremendous energy
The OTEC heat engine may be configured following de- must be
expended to bring cold seawater up from depths ap-
signs proposed by D’Arsonval or Claude, known, respectively, proaching
1000 m against the force of gravity. In reality, the
as closed cycle and open cycle OTEC. The following sections natural
hydrostatic pressure gradient provides for most of the
provide additional technical information on these two power increase
in the gravitational potential energy of a fluid parti-
cycles and their variants, as well as a hybrid cycle that pro- cle
moving with the gradient from the ocean depths to the
duces electricity and potable water. surface.
This can be seen by writing the modified Bernoulli
equation
for a submerged pipeline, with intake at depth z1
discharging at sea level z . The work

W required to move a
CLOSED CYCLE OTEC
2
unit mass
of seawater through this simple pipe is
D’Arsonval’s original concept for OTEC proposed to use a

2 2
#W
≈ [(P − P )/ρ] + [(V − V )/2] + [g (z − z )]
pure working fluid that would evaporate at the temperature
2 1 2 1 2 1 (5)
of warm seawater. The vapor would subsequently expand and
+ #[(fLV 2 )/(2D )]
do work before being condensed by the cold seawater. This
series of steps would be repeated continuously using the same where V
is the local mean velocity; g is the gravitational ac-
working fluid, whose flow path and thermodynamic process
celeration; f , L , and D are, respectively, the pipe friction fac-
representation constituted closed loops—hence, the name tor,
length, and inside diameter; and # is the seawater den-
closed cycle. The specific process adopted for closed cycle sity,
assumed to be constant (in reality, small variations of #
OTEC is the Rankine, or vapor power, cycle. with
depth occur due to salinity and temperature gradients).
Figure 3 is a simplified schematic diagram of a closed cycle The
difference in the pressures at the pipe inlet and exit,
P # P ,
is equivalent to the hydrostatic head:
OTEC system. The principal components are the heat ex- 2 1
changers, turbogenerator, and seawater supply system,
(P − P ) = ρg (z − z ) (6)
which, although not shown, accounts for most of the parasitic
2 1 1 2
power consumption and a significant fraction of the capital

Hence,
the pressure and gravity terms cancel on the right-
expense. Also not included in the schematic are ancillary de-
hand side
of the Bernoulli equation. For a constant diameter
vices such as separators to remove residual liquid down-
pipeline,
V2 # V1 and work must be supplied only to overcome
stream of the evaporator and subsystems to hold and supply
frictional losses. In a real OTEC cold water pipeline, some

make-up working fluid lost through leaks or contamination.
additional work is needed to compensate for the nonuniform

In a closed cycle system, heat transfer from warm surface
seawater
density.
seawater occurs in the evaporator, producing a saturated va-
Irreversibilities in the turbomachinery and heat ex-

por from the liquid working fluid. Electricity is generated
changers
reduce cycle efficiency below the Carnot value. Irre-
when this gas expands to lower pressure through the turbine.
versibilities in the heat exchangers occur when energy is
transferred over a large temperature difference. It is impor-

tant,
therefore, to select a working fluid that will undergo the
Warm sea Cold
desired
phase changes at temperatures established by the
water in sea water
discharge surface
and deep seawater. Insofar as a large number of sub-
Working stances
can meet this requirement (because pressures and the
fluid pressure
ratio across the turbine and pump are design param-
vapor eters),
other factors must be considered in the selection of a
Evaporator Turbogenerator Condenser working
fluid including cost and availability, compatibility
with
system materials, toxicity, and environmental hazard.
Leading candidate working fluids for closed cycle OTEC

Warm Working fluid Cold sea
applications are ammonia and various fluorocarbon refriger-
sea water pressurizer water in ants
[i.e., chlorofluorocarbons (CFCs), hydrochlorofluorocar-
discharge (boiler feed pump)
bons
(HCFCs), hydrofluorocarbons (HFCs)]. Some of these
substances are also used in

geothermal power stations and
Working fluid Rankine
bottoming cycles. Their primary disadvantage is the
condensate
environmental hazard posed by leakage, because maximum

Figure 3. Schematic diagram of a closed cycle OTEC system. pressures
in OTEC systems using these fluids would lie be-
----------------------- Page 400-----------------------
tween 6 and 9 atmospheres (absolute). Ammonia is toxic in Even

though a 25% performance improvement is signifi-
moderate concentrations. CFCs have been banned by the cant,
the Kalina cycle needs additional capital equipment,
Montreal Protocol because they deplete stratospheric ozone. and may
impose severe demands on the evaporator and con-
HCFCs and HFCs are major greenhouse gases. denser.
The efficiency improvement arises largely as a result
of
reducing the temperature difference over which heat trans-
KALINA CYCLE fer
occurs in these devices; therefore, maintaining a given
Q will require some combination of higher heat transfer co-

The Kalina, or adjustable proportion fluid mixture (APFM),
efficients, more heat transfer surface area, and increased sea-
cycle is a variant of the OTEC closed cycle. Whereas simple water
flow rates. Each of these changes has an associated cost
closed cycle OTEC systems use a pure working fluid, the Kal- or power
penalty. Additional analysis and testing must be
ina cycle proposes to employ a binary mixture of ammonia
conducted to determine whether the Kalina cycle is appro-
and water with varying proportions at different points in the priate
for OTEC applications.
system. The advantage of a binary mixture is that, at a given

pressure, evaporation or condensation occurs over a range of
OPEN
CYCLE OTEC
temperatures; a pure fluid, on the other hand, changes phase
at constant temperature. This additional degree of freedom
allows heat transfer-related irreversibilities in the evaporator Claude’s
concern about the cost and potential biofouling of
and condenser to be reduced (3). The local temperature gap closed
cycle heat exchangers led him to propose using steam
between the working fluid and the seawater flowing through
generated directly from the warm seawater as the OTEC
the heat exchanger can, within limits, be manipulated, with working
fluid. The steps of the Claude, or open, cycle are (1)
a resulting improvement in cycle thermodynamic efficiency. flash
evaporation of warm seawater in a partial vacuum, (2)
A schematic diagram of an OTEC Kalina cycle is provided
expansion of the steam through a turbine to generate power,
in Fig. 4. As expected, even though it is similar to a simple (3)
condensation of the vapor by direct contact heat transfer
closed cycle OTEC system, several additional components are to cold
seawater, and (4) compression and discharge of the
added to accommodate the binary fluid and to provide means
condensate and any residual noncondensable gases. Unless
to adjust the proportions of the ammonia and water flowing
freshwater is a desired by-product, open cycle OTEC elimi-
through the different devices. Saturated two-phase flow en- nates
the need for surface heat exchangers. The name open
tering the separator is divided into an ammonia-rich vapor cycle
comes from the fact that the working fluid (steam) is
stream and water-rich liquid. The vapor expands through the
discharged after a single pass and has different initial and
turbine before mixing with and being absorbed into the de- final
thermodynamic states; hence, the flow path and process
pressurized liquid. Phase change begins and ends at higher are
open.
temperatures as the mass fraction of ammonia in the mix- The
essential features of an open cycle OTEC system are
ture decreases.
presented in Fig. 5. The entire system, from evaporator to
Uehara and Ikegami analyzed the performance of an opti-
condenser, operates at partial vacuum, typically at pressures
mized Kalina cycle and a simple ammonia closed cycle OTEC between
1% and 3% of atmospheric. Initial evacuation of the
system (4). For warm and cold seawater temperatures of 28 system
and removal of noncondensable gases during opera-
and 4#C, respectively, they concluded that the Kalina cycle tion
are performed by the vacuum compressor, which, along
would improve cycle efficiency (neglecting back-work ex- with the
seawater and discharge pumps, accounts for the bulk
pended to pump seawater) by about one percentage point, of the
open cycle OTEC parasitic power consumption (i.e.,
from approximately 4% to 5%. back-
work).
The
low system pressures of open cycle OTEC are neces-
sary to
induce boiling of the warm seawater. Flash evapora-
Warm tion is
accomplished by exposing the seawater to pressures
sea water below
the saturation pressure corresponding to its tempera-
discharge ture,
which, at 28#C is about 3780 Pa. This is usually accom-
plished
by pumping it into an evacuated chamber through
spouts
designed to maximize heat and mass transfer surface
area.
Removal of gases dissolved in the seawater, which will
Evaporator Separator Turbogenerator come out
of solution in the low-pressure evaporator and com-
promise
operation, may be performed at an intermediate
pressure
prior to evaporation.
Warm sea Regenerator Absorber Vapor
produced in the flash evaporator is relatively pure
water in steam.
The heat of vaporization is extracted from the liquid
Valve phase,
lowering its temperature and preventing any further
boiling.
Flash evaporation may be perceived, then, as a trans-
Cold sea
Condenser water fer of
thermal energy from the bulk of the warm seawater to
discharge the
small fraction of mass
that is vaporized. Less than 0.5%
Circulation of the
mass of warm seawater entering the evaporator is con-
pump verted
into steam.
Cold sea
water in The
pressure drop across the turbine is established by the
cold
seawater temperature. At 4#C, steam condenses at 813
Figure 4. Schematic diagram of a Kalina cycle OTEC system. Pa. The
turbine (or turbine diffuser) exit pressure cannot fall
----------------------- Page 401-----------------------
Warm sea
Cold
water in
sea water
discharge
Noncondensable
Desalinated
water
gases
Deaeration Vacuum chamber vapor
(optional) flash
Turbogenerator Condenser Vent compressor
evaporator
Desalinated
water
(optional)
Cold sea
Noncondensable Warm sea
water in
gases water
discharge
Figure 5. Schematic diagram of an

open cycle OTEC system.
below this value. Hence, the maximum turbine pressure drop

MIST AND FOAM LIFT OTEC CYCLES
is only about 3000 Pa, corresponding to about a 3 : 1 pressure
ratio. This will be further reduced to account for other pres- The
mist lift and foam lift OTEC systems are variants of the
sure drops along the steam path and differences in the tem-
OTEC open cycle. Both employ the seawater directly to pro-
peratures of the steam and seawater streams needed to facili-
duce power. Unlike Claude’s open cycle, lift cycles generate
tate heat transfer in the evaporator and condenser. For
electricity with a hydraulic turbine. The energy expended by
example, the nominal 210 kW PICHTR open cycle OTEC the
liquid to drive the turbine is recovered from the warm
plant in Hawaii operates with a turbine pressure drop of
seawater (7).
about 1200 Pa when the warm and cold seawater tempera-
Figure 6 outlines the concept of OTEC lift cycles. In the
lift process, warm seawater is flash evaporated to produce a

tures are 27.5#C and 6#C, respectively (5). The maximum the-
oretical pressure drop for these conditions is approximately
2700 Pa.
Condensation of the low-pressure steam leaving the tur-
Lift
bine may employ a direct contact condenser (DCC) in which
process
cold seawater is sprayed over the vapor, or a conventional
surface condenser that physically separates the coolant and
Warm sea
Hydraulic
the condensate. DCCs are inexpensive and have good heat
water in turbogenerator
transfer characteristics because they lack a solid thermal
Sea level
boundary between the warm and cool fluids. Surface condens-
ers are expensive and more difficult to maintain than DCCs;
Mixed water
however, they produce a marketable freshwater by-product.
discharge
Effluent from the condenser must be discharged to the en-
Cold sea
vironment. Liquids are pressurized to ambient levels at the
water in
point of release by means of a pump, or, if the elevation of the
(a)
condenser is suitably high, they can be compressed hydrostat-
ically. As noted previously, noncondensable gases, which in-
clude any residual water vapor, dissolved gases that have
come out of solution, and air that may have leaked into the
system are removed by the vacuum compressor.
Open cycle OTEC eliminates expensive heat exchangers at
Sea level Warm sea Mixed water
the cost of low system pressures. Partial vacuum operation
water in discharge
has the disadvantage of making the system vulnerable to air

Lift
Hydraulic
in-leakage and promotes the evolution of noncondensable
turbogenerator process
gases dissolved in seawater. Power must ultimately be ex-
pended to pressurize and remove these gases. Furthermore,
as a consequence of the low steam density, volumetric flow
rates are very high per unit of electricity generated. Large
components are needed to accommodate these flow rates. In
Cold sea
particular, only the largest conventional steam turbine stages
water in
have the potential for integration into open cycle OTEC sys-
(b)
tems of a few megawatts gross generating capacity. It is gen-
erally acknowledged that higher capacity plants will require
Figure 6. Schematic diagram of (a) mist lift and (b) foam lift OTEC
a major turbine development effort (6).
systems.
----------------------- Page 402-----------------------
two-phase, liquid-vapor mixture—either a mist consisting of

ammonia evaporator may be further cooled by heat transfer
liquid droplets suspended in a vapor, or a foam, where vapor to
either the liquid ammonia leaving the ammonia condenser
bubbles are contained in a continuous liquid phase. The mix- or
cold seawater. The noncondensables are then compressed
ture rises, doing work against gravity. Here, the thermal en- and
discharged to the atmosphere.
ergy of the vapor is expended to increase the potential energy
Steam is used as an intermediary heat transfer medium
of the fluid. The vapor is then condensed with cold seawater
between the warm seawater and the ammonia; consequently,
and discharged back into the ocean. Flow of the liquid the
potential for biofouling in the ammonia evaporator is re-
through the hydraulic turbine may occur before or after the duced
significantly. Another advantage of the hybrid cycle re-
lift process. lated
to freshwater production is that condensation occurs at
Advocates of the mist and foam lift cycles contend that
significantly higher pressures than in an open cycle OTEC
they are cheaper to implement than closed cycle OTEC be-
condenser, as a result of the elimination of the turbine from
cause they require no expensive heat exchangers and are su- the
steam flow path. This may, in turn, yield some savings in
perior to the Claude cycle because they use a hydraulic tur- the
amount of power consumed to compress and discharge the
bine rather than a low pressure steam turbine. These claims
noncondensable gases from the system. These savings (rela-
await verification. tive
to a simple Claude cycle producing electricity and water),
however, are offset by the additional back-work of the closed

HYBRID CYCLE OTEC cycle
ammonia pump.
One drawback of the hybrid cycle shown in Fig. 7 is that

The power generation capacity of the Claude open cycle is water
production and power generation are closely coupled.
limited to a few megawatts by existing low pressure steam
Changes or problems in either the water or power subsystem
turbine technology. This is not the case for closed cycle OTEC, will
compromise performance of the other. Furthermore,
which, because it uses a pressurized working fluid, does not there
is a risk that the potable water may be contaminated
need to sustain comparable volumetric flow rates of vapor. by an
ammonia leak. In response to these concerns, an alter-
Closed cycle design studies have identified commercial turbo- native
hybrid cycle, comprising decoupled power and water
machinery that can be used to generate up to 100 MW of elec-
production components, has been proposed (9). The basis for
trical power. On the other hand, open cycle OTEC has the this
concept lies in the fact that warm seawater leaving a
benefit of being able to produce potable water. Some market- closed
cycle evaporator is still sufficiently warm, and cold sea-
ing studies have suggested that OTEC systems that can pro- water
exiting the condenser is sufficiently cold, to sustain an
vide both electricity and water may be able to penetrate the
independent freshwater production process.
marketplace more readily than plants dedicated solely to
The alternative hybrid cycle consists of a conventional
power generation. Hybrid cycle OTEC was conceived as a re- closed
cycle OTEC system that produces electricity and a
sponse to these studies. Hybrid cycles combine the potable
downstream flash-evaporation-based desalination system.
water production capabilities of open cycle OTEC with the
Seawater from the closed cycle evaporator, which typically
potential for large electricity generation capacities offered by has
been cooled by about 3#C, is pumped into a flash evapora-
the closed cycle. tor
evacuated to below the water’s corresponding saturation
Figure 7 depicts the elements of one type of hybrid OTEC
pressure. The desalinated steam produced is condensed in a
cycle proposed by Panchal and Bell (8). As in the Claude cycle,
surface heat exchanger by heat transfer to the effluent cold
warm surface seawater is flash evaporated in a partial vac-
seawater from the closed cycle condenser. Depending on the
uum. This low-pressure steam flows into a heat exchanger
configuration of the closed cycle power system, the cold sea-
where it is employed to vaporize pressurized ammonia. Dur- water
may experience up to a 6#C temperature rise as it
ing this process, most of the steam condenses, yielding desali- passes
through the closed cycle condenser. Although the ini-
nated potable water. The ammonia vapor flows through a tial
power generation step can reduce the
T of the warm and
simple closed cycle power loop and is condensed using cold cold
seawater by about 9# to 10#C—from, say, 24# to 14#C—
seawater. The uncondensed steam and other gases exiting the this
is still adequate to accomplish desalination.
Evaporator Ammonia vapor

Ammonia
system
turbogenerator
Vacuum flash Ammonia

evaporator evaporator
Ammonia
Steam
condenser
Noncondensable
gases
Warm sea Sea water Cold

sea
water
water
Ammonia Pump
##yy##yy
liquid
Figure 7. Schematic diagram of hybrid cycle OTEC

Desalinated water
system.
----------------------- Page 403-----------------------
With the alternative hybrid cycle, water production and and

surface condensers, these savings may be offset partially
electricity generation can be adjusted independently, within by
the relatively high cost of the low-pressure steam turbine
limits, in response to demand. Should either subsystem fail and
larger ducting and structures needed to accommodate
or require servicing, the other can continue to operate. The
partial vacuum operation. Because the largest (L-0) existing
primary drawbacks of the alternative hybrid cycle are that (1)
conventional steam turbine stages will only produce about 2
the ammonia evaporator uses warm seawater directly and is MW
under open cycle OTEC conditions, multimegawatt open
therefore subject to biofouling, and (2) additional equipment,
cycle installations probably will need to be configured, in the
such as the potable water surface condenser, is required, in-
near-term, using multiple 2 MW to 4 MW (single or double
creasing capital expenses.
wheel turbine) power modules. Open cycle OTEC is therefore
believed to be best suited to rural sites with low power de-

mand
and a need for the potable water by-product. Heat ex-
COMPARISON OF OTEC OPEN AND CLOSED CYCLES
changers notwithstanding, pressurized closed cycle systems,
which are relatively compact, can easily generate tens of

It is worthwhile to examine the common elements and points
megawatts of electricity using existing technology. The cur-
of contrast of the open (Claude) and closed (Rankine) OTEC rent
consensus is that closed cycle OTEC may be successfully
cycles. Even though the means by which these two primary
integrated into urban energy strategies or employed to power
OTEC strategies exploit the low-quality thermal resource to OTEC
plantship operations.
produce power are quite different, the end results are similar.
The small temperature difference between surface and deep
seawater establishes an upper limit on the thermodynamic OPEN
AND CLOSED CYCLE OTEC SYSTEM PERFORMANCE
conversion efficiency of any heat engine to about 8%. Irrevers-
ibilities in devices such as heat exchangers and turbomachin-
Tables 1 and 2 present the results of design studies of open
ery, and back-work associated with pumping the large vol- and
closed cycle OTEC plants. Table 1 corresponds to a shore-
umes of seawater required to sustain operation, make it
based open cycle plant producing both electricity and fresh-
unlikely that OTEC plant efficiency (as a percentage of ther-
water (10). The system has been optimized for power genera-
mal energy extracted from the warm seawater that ultimately
tion. For a seawater
T of 22#C, the plant can provide about
is available for export and sale as electricity) will exceed 2% 1.2
MW of electricity to the local power grid and supply 2200
3
6
to 3%. Analyses have determined that, for comparable power- m
(2.2 # 10 L) of freshwater per day. Table 2 describes the
generating capacity, the thermal performance of the two cy-
simulated operation of a larger, ammonia closed cycle OTEC
cles does not differ significantly. Even though the direct con-
power system designed for installation on a floating platform
tact heat exchangers employed in an open cycle system have
the potential to use more fully the thermal seawater resource,
irreversibilities associated with the vacuum compressor sys-
tem and noncondensable gases reduce this advantage in
Table 1. Open Cycle OTEC System Performance Summary
practice. Warm
water pipe inner diameter 2.5 m
Both open and closed cycles require similar volumes of sea- Cold
water pipe outer diameter 1.6 m
water to produce a unit of electrical power. For megawatt-
Mixed discharge pipe outer diameter 3 m
scale plants operating over a seawater temperature difference Cold
water pipe intake depth 1000 m
of 20#C, about 3.5 m3/s of warm surface water must be sup-
Gross power 1838 kW
plied per megawatt of electricity generated by the turbine.

Parasitic power (back-work) 595 kW
Studies suggest that net power output is optimized when
Back-work ratio 32%
warm-to-cold water ratios lie between 1.8 : 1 and 2 : 1. Back- Net
power 1243 kW
work to run the seawater pumps and other parasitics ac-
Fresh (potable) water production 25.9 kg/s (1550 l/min)
counts for about 30% of the turbine output for larger plants Warm
seawater flow rate 5978 kg/s (5.85 m3/s)
and a higher percentage of the gross in smaller facilities. This Cold
seawater flow rate 3085 kg/s (3.0 m3/s)
suggests that 5 m3/s of warm seawater are needed for each Warm
seawater supply temperature 26 #C
megawatt of electricity available for sale. An OTEC plant pro- Cold
seawater supply temperature 4 #C
ducing 20 MW at the generator terminals will therefore be
Steam flow rate through turbine 26.1 kg/s
able to export 14 MW to the grid and will require 70 m3/s
Turbine inlet temperature 22.7 #C
(70000 L/s) of warm seawater and about 35 m3/s (35000 L/s)
Turbine inlet pressure 2.74 kPa
Turbine outlet temperature 11.1 #C

of cold seawater. Seawater velocities have to be maintained
Turbine outlet pressure 1.29 kPa

below 2 m/s to control pumping power losses; as such, the
Turbine rotor diameter 5.65 m
submerged intake pipeline diameter scales as 1.8 m/MW (net)
on the warm water side and 1.2 m/MW (net) on the cold water Warm
seawater pump power 183 kW
side. It is obvious that the huge pipelines or pipe arrays Cold
seawater pump power 324 kW
Vent
(vacuum) compressor power 80 kW
needed to supply seawater to a commercial OTEC plant rep-
Potable water storage pump power 8 kW

resent a serious technological challenge and major capital ex-
pense.
Carnot efficiency 0.0735 (7.35%)
Open and closed cycles differ primarily in their economics
Thermal energy from warm seawater 63531 kJ/s
and target markets. Although a Claude cycle dedicated to
Cycle efficiency based on gross turbine 0.029 (2.9%)
electricity generation offers potential savings over a closed cy-
power
Cycle efficiency based on net power 0.020 (2.0%)

cle plant by eliminating the need for expensive evaporators
----------------------- Page 404-----------------------
Table 2. Closed Cycle OTEC System Performance Summary face

temperature anomalies. Analyses of OTEC effluent
plumes suggest that discharge between the 50m and 100 m

Warm water pipe inner diameter 4.6 m
depths should be sufficient to ensure minimal impact on the
Cold water pipe outer diameter 2.74 m ocean
environment.
Mixed discharge pipe outer diameter 5.5 m
The extent of outgassing from seawater typically will be

Cold water pipe intake depth 1000 m
greater in open cycle plants than in closed cycle systems be-

Gross power 7920 kW cause
water-side pressures are low in the flash evaporator
Parasitic power (back-work) 2660 kW and
in the condenser, if a DCC is employed. Outgassing also
Back-work ratio 34% will
occur when cold seawater is brought up from the depths
Net power 5260 kW to
the ocean surface. Although the mass of gas released per
Warm seawater flow rate 27000 kg/s (26.4 m3/s) unit
of seawater is quite small, the massive volumes of water
Cold seawater flow rate 14240 kg/s (13.9 m3/s) used
in OTEC may result in nonnegligible emissions. Fortu-
Warm seawater supply temperature 26 #C
nately, these gas emissions generally comprise benign species
Cold seawater supply temperature 4.5 #C
(e.g., N , O , CO ); however, CO emissions from OTEC have
2 2 2 2
Warm water return temperature 22.9 #C been
investigated in consideration of its role as a greenhouse
Cold water return temperature 10.3 #C gas.
It is believed that open cycle plants will release between
Ammonia flow rate through turbine 274 kg/s 6 g
and 38 g of CO2/kWh, whereas closed cycle systems have
Turbine inlet temperature 21.1 #C
an
upper bound of 17 g/kWh. These values compare very fa-
Turbine inlet pressure 890 kPa
vorably with fossil fuel combustion power stations, which

Turbine outlet temperature 11.9 #C
Turbine outlet pressure 656 kPa emit
one to two orders of magnitude more CO2 .
Warm seawater pump power 640 kW
ECONOMICS OF OTEC
Cold seawater pump power 1150 kW
Mixed discharge seawater pump power 680 kW
Ammonia pump power 190 kW
Studies conducted to date on the economic feasibility of OTEC
systems suffer from the lack of reliable cost data. Commer-

Carnot efficiency 0.0719 (7.19%)
cialization of the technology is unlikely until a full-scale plant
Thermal energy from warm seawater 336176 kJ/s is
constructed and operated continuously over an extended
Cycle efficiency based on gross turbine 0.024 (2.4%)
period to provide these data on capital and recurring ex-

power
penses (10). Only this type of demonstration will be sufficient
Cycle efficiency based on net power 0.016 (1.6%)
to
allay the doubts of potential investors and funding
agencies.
Uncertainties in financial analyses notwithstanding, pro-

(9). A net power output of 5.3 MW is calculated for a slightly
jections suggest very high first costs for OTEC power system
smaller seawater
T of 21.5#C. components (11).
Small land-based or near-shore floating
In both systems, back-work, primarily to operate the sea-
plants in the 1 MW to 10 MW (net) range, which would proba-
water pumps, consumes slightly in excess of 30% of the power bly
be constructed in rural island communities, may require
generated by the turbine. Even though the Carnot efficiencies
expenditures of between $10,000 and $20,000 (in 1995 US
are about 7%, irreversibilities and system parasitics greatly
dollars) per kilowatt of installed generating capacity. Even
degrade performance. Less than 2% of the thermal energy ex-
though there appears to be favorable economies of scale,
tracted from the warm seawater is exported as electricity by
larger floating (closed cycle) plants in the 50 MW to 100 MW
the open cycle plant. The closed cycle plant only attains about range
are still anticipated to cost about $5000/kW. This is
80% of this low value—its net power efficiency is slightly well
in excess of the $1000 to $2000/kW capital expense of
more than 1.6%. The marginally better open cycle perfor-
fossil fuel combustion power stations. Although the OTEC en-
mance can be attributed to the superior heat transfer charac- ergy
resource is free, low recurring costs will only partially
teristics of the direct contact flash evaporator relative to the
offset the capital cost disadvantage of OTEC power systems.
ammonia evaporator.
To enhance the economics of OTEC power stations, various
initiatives have been proposed based on marketable OTEC

by-
or co-products, such as potable water, air conditioning,
ENVIRONMENTAL CONSIDERATIONS
refrigeration, mariculture, and high-value energy carriers
(11,12). OTEC proponents believe that the first commercial

OTEC systems are, for the most part, environmentally be- OTEC
plants will be shore-based systems designed for use in
nign. Even though accidental leakage of closed cycle working
developing Pacific island nations, where potable water is in
fluids can pose a hazard, under normal conditions, the only short
supply. Many of these sites would be receptive to oppor-
effluents are the mixed seawater discharges and dissolved
tunities for economic growth provided by OTEC-related in-
gases that come out of solution when seawater is de-
dustries. Even though some of the by- and co-product concepts
pressurized. have
been tested successfully on a small scale, development
OTEC mixed seawater discharges will be at lower temper- of
commercial-size operations has been hampered by the ab-
atures than seawater at the ocean surface. The discharges sence
of an full-scale operational OTEC plant. Several of the
also will contain high concentrations of nutrients brought up
product options are described below.
with the deep seawater and may have a different salinity. It
Freshwater
is important, therefore, that release back into the ocean be
conducted in a manner that minimizes disruptions to the The
condensate of the open and hybrid cycle OTEC systems
ocean mixed layer biota and avoids inducing long-term sur- is
desalinated water, suitable for human consumption and ag-
----------------------- Page 405-----------------------
ricultural uses. Analyses have suggested that first-generation tem

also drip irrigates the crop via condensation of moisture
OTEC plants, in the 1 MW to 10 MW range, would serve the in the
air on the cold water pipes. M. Vitousek of the Uni-
utility power needs of rural Pacific island communities, with versity
of Hawaii carried out demonstrations and determined
the desalinated water by-product helping to offset the high that
strawberries and other spring crops and flowers could be
cost of electricity produced by the system (11). grown
throughout the year in the tropics using this method
(see
Ref. 15).
Refrigeration and Air Conditioning
Energy
Carriers
The cold deep seawater can be used to maintain cold storage
spaces and to provide air conditioning. The Natural Energy Even
though the most common scenario is for OTEC energy
Laboratory of Hawaii (NELH), the site of Hawaii’s OTEC ex- to be
converted to electricity and delivered directly to consum-
periments, has air conditioned its buildings by passing the ers,
energy storage has been considered as an alternative,
cold seawater through heat exchangers. Similar small-scale
particularly in applications involving floating plants moored
operations would be viable in other locales. Economic studies far
offshore. Storage would also allow the export of OTEC en-
have been performed for larger metropolitan and resort appli- ergy
to industrialized regions outside of the tropics. Long-
cations. Air conditioning of new developments, such as resort term
proposals have included the production of hydrogen gas
complexes, with cold seawater may be economically attractive via
electrolysis, ammonia synthesis, and the development of
even if inexpensive utility-grid electricity is available (13). shore-
based mariculture systems or floating OTEC plant-
ships
as ocean-going farms. Such farms would cultivate ma-
Mariculture rine
biomass, for example, in the form of fast-growing kelp
that
could be converted thermochemically into fuel and chem-
The cold deep ocean waters are rich in nutrients and low in ical
co-products or burned directly to produce heat.
pathogens and, therefore, provide an excellent medium for the
cultivation of marine organisms. The 322-acre NELH has
Environmental Enhancement
been the base for successful mariculture research and devel-
opment enterprises. The site has an array of cold water pipes, In the
very long term, there are also prospects for environ-
originally installed for the early OTEC research but since mental
enhancement. Should a large number of floating
used for mariculture. The cold water is used to cultivate OTEC
plantships be placed at a few strategic sites around the
salmon, trout, opihi (limpet; a shellfish delicacy), oysters, lob-
equatorial belt, maintaining an effective surface ocean tem-
sters, sea urchins, abalone, kelp, nori (a popular edible sea-
perature below 26.2#C, there are prospects that hurricanes
weed used in sushi), and macro- and microalgae. The earliest and
typhoons may be prevented or ameliorated (16). A similar
experiments on the productivity of this water centered on
environmental benefit could accrue toward reversing global
nori. Phenomenal daily biomass growth rates of 40% to 45% warming
through intelligent management of the upwelled
were recorded, and this research has since developed into a water,
although early computer models have shown that iron
commercial enterprise. Even though the techniques used for
supplementation may be necessary (17).
mariculture remain largely proprietary to the individual en-

trepreneurs, the commercial success of these fledgling enter-
BIBLIOGRAPHY
prises has been one important indicator of the potential for
OTEC-related mariculture. 1. T.
R. Penney and D. Bharathan, Power from the sea, Sci. Amer.,
The cold seawater may have applications for open ocean
256 (1): 86–92, 1987.
mariculture as well. Natural upwelling of deep ocean water 2. H.
V. Sverdrup, M. W. Johnson, and P. H. Fleming, The Oceans:
occurs off the west coasts of North America, South America,
Their Physics, Chemistry, and General Biology, New York: Pren-
West Africa, and other coastal regions where along-shore
tice-Hall, 1942.
winds push surface seawater away from land, allowing the 3. C.
H. Marston, Development of the adjustable proportion fluid
deeper waters to upwell. Stimulated by the elevated nutrient
mixture cycle, Mech. Eng., 114 (9): 76–81, 1992.
content of this deep water, high fish production has been ob- 4. H.
Uehara and Y. Ikegami, Parametric performance analysis of
served in these areas. These natural upwellings account for
OTEC using Kalina cycle, Proc. ASME Joint Solar Eng. Conf.,
0.1% of the world’s oceans surface but yield roughly 44% of
pp. 203–207, 1993.
the world’s fish catch (14). Artificial upwelling, generated for 5. L.
A. Vega and D. E. Evans, Operation of a small open cycle ocean
OTEC or exclusively for mariculture, has been suggested as a
thermal energy conversion experimental facility, Proc. Oceanol-
method of creating new fisheries and marine biomass planta-
ogy, 94: 5, 1994.
tions. Should development proceed as anticipated, open ocean 6. T.
R. Penney, Composite turbine blade design options for Claude
cages can be eliminated, and natural feeding would replace
(open) cycle OTEC power systems. Proc. 5th Int. Symp. Exhibition
expensive feed with temperature and nutrient differentials
OMAE, pp. 13–22, 1986.
being used to keep the fish stock in the kept environment. 7. K.
J. Beck, Ocean thermal gradient hydraulic power plant, Sci-
ence, 189: 293, 1977.
Agriculture 8. C.
B. Panchal and K. J. Bell, Simultaneous production of desali-
nated water and power using a hybrid cycle OTEC plant, J. Solar
An idea initially proposed by S. Siegal of the University of
Energy Eng., 109: 156–160, 1987.
Hawaii involves the use of cold seawater for agriculture. This 9. L.
A. Vega and G. C. Nihous, Design of a 5 MW OTEC pre-com-
concept involves burying an array of cold water pipes in the
mercial plant, Proc. Oceanology, 94: 5, 1994.
ground to create cool weather growing conditions not found in 10. G.
C. Nihous, M. A. Syed, and L. A. Vega, Conceptual design of
tropical environments. In addition to cooling the soil, the sys- an
open-cycle OTEC plant for the production of electricity and
----------------------- Page 406-----------------------
OFFICE AUTOMATION 103
freshwater in a Pacific island, Proc. Int. Conf. Ocean Energy Re-

covery, 1989.
11. L. A. Vega, Economics of ocean thermal energy conversion. In R.

J. Seymour (ed.), Ocean Energy Recovery: The State of the Art,
New York: American Society of Civil Engineers, 1992.
12. P. K. Takahashi and A. Trenka, Ocean Thermal Energy Conver-

sions; UNESCO Energy Engineering Series, London: Wiley, 1996.
13. P. Helfrich et al., A Preliminary Assessment of the Feasibility of

Employing Ocean Thermal Energy Conversion Technology on Kiri-
timati (Christmas) Island in the Republic of Kiribati. Technical
Report for the Pacific International Center for High Technology
Research. Honolulu: PICHTR, 1990.
14. O. A. S. Roels, M. F. Laurence, and W. Van Hemelryck, Organic
production potential of artificial upwelling marine environment,
Limnol. Oceanogr., 13: 18–23, 1978.
15. J. P. Craven, Keahole Point as a precursor model of government-

industry-academic ocean enterprise, The Ocean Enterprise Con-
cept; Report on the Ocean Enterprise Workshop, Washington, DC:
National Science Foundation Office of Engineering Infrastructure
and Development, 1989.
16. P. K. Takahashi et al., Marine macrobiotechnology systems,

J. Marine Biotechnol., 1 (1): 9–15, 1993.
17. P. K. Takahashi, Project blue revolution, J. Energy Eng., 122

(3): 114–124, 1996.
STEPHEN M. MASUTANI
PATRICK K. TAKAHASHI
University of Hawaii
----------------------- Page 407-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Optical Waveguide Transitions
Standard Article

W. K. Burns1
1
Naval Research Laboratory, Washington, DC

reserved.
Search
DOI: 10.1002/047134608X.W4408
❍ Acronym Finder
Abstract
Separating Waveguides and Horns
Local Normal Mode Formalism
Adiabatic and Abrupt Transitions
Coupled-Mode Theory Representation of Normal Modes
Scattering Matrix Formalism
Coupled Amplitude Equations
Coupled Mode Theory Representation of C and

ij ij
Branches
Waveguide Horns
Direct Computational Methods
----------------------- Page 408-----------------------

----------------------- Page 409-----------------------
354 OPTICAL WAVEGUIDE TRANSITIONS
OPTICAL
WAVEGUIDE TRANSITIONS
Most
useful integrated optical waveguide devices use a series
of
waveguide transitions where the cross section of the wave-
guide
structure changes in the direction of propagation. In
the Mach
Zehnder interferometer and in the directional cou-
pler,
both of which are commonly used, waveguide branches
are used
on both sides of a phase shift region to form an elec-
trooptic
modulator or switch. The performance of these de-
vices
depends on how well the transitions perform their in-
tended
function. To understand and optimize device
performance it is therefore useful to be able to predict the
transmission characteristics of waveguide transitions.

These
characteristics are completely defined by Maxwell’s
equations.
By using a computer approach, such as the beam
propagation method (1,2), computer-generated solutions for

each
particular geometry of interest can be obtained. This ap-
proach
has the advantage of providing accurate results for
specific
cases. However, computer approaches can be tedious
and
require multiple iterations to provide design guidance.
An
alternative approach is the local normal mode formal-
ism (3,4),
an extremely powerful prediction tool. Coupling be-
tween
local normal modes is caused by variations of wave-
guide
parameters in the direction of propagation. With this
formalism,
the transmission properties of most waveguide
transitions can be shown to depend on a few aggregate pa-

rameters
which make most devices easy to understand and
design.
Under certain approximations, analytic solutions can
be
obtained to describe the amount of power transferred be-
tween
these modes. These solutions can then be used to pre-
dict the
performance of transition structures. The drawbacks
of the
local normal mode formalism include its inability to
handle
more than a limited number of modes (usually two),
its
inability to account for radiation, and its generally approx-
imate
representation.

----------------------- Page 410-----------------------
OPTICAL WAVEGUIDE TRANSITIONS 355
(a)
Ψ Ψ
i j
Figure 3. Electric field profiles for the first two normal modes of a
three-layer waveguide. Normal modes are used in the modeling of

(b)
transition structures.
Figure 1. Waveguide branching structures. In (a) the waveguide

branch is used to divide optical power or to separate normal modes.
In (b) the separating waveguide is used to separate the outputs of
yields electric-field profiles, shown in Fig. 3. These field solu-
two closely spaced channels.
tions are called the normal modes of the waveguide. In Fig. 3
the geometry and optical wavelength are assumed such that
the structure supports two normal modes. We will identify
the first or symmetric mode as #i and the second or antisym-

In general the local normal mode formalism is useful for
metric mode as # . These modes have propagation constants
j
understanding the behavior of waveguide transitions and for
# and # .
i j
giving approximate design information. Computer solutions
are useful for more precise results, looking at the effects of
Local Normal Modes
radiation, and in cases where a two-mode formalism cannot
be employed.
By definition the normal modes of a structure are orthogonal,
i.e., they satisfy the orthogonality condition
SEPARATING WAVEGUIDES AND HORNS

# ∞ # (x )# ∗ (x ) dx = 0 (1)
i j
−∞
The most common waveguide transition structures are sepa-

rating waveguides (also called branches) and horns. A sepa-
where x is the coordinate transverse to the waveguide. In
rating waveguide may be a waveguide branch, as shown in
waveguide transitions in which the waveguide layer thick-
Fig. 1(a), or the separating region of two closely coupled wave-
nesses or indices vary in the direction of propagation (z), Eq.
guides, as in Fig. 1(b). Waveguide branches are known as
(1) is no longer true and one cannot strictly speak of the nor-
power dividers or mode splitters, depending on whether input
mal modes of the structure. Instead, we introduce the concept
power is divided between output arms of the branch, or
of a locally normal mode (5) as suggested in Fig. 4. At a wave-
whether the output power is predominately in a single output
guide position z0, consider the waveguide parameters at that
arm of the branch. Waveguide horns are used to change the
position, and solve for the normal modes of the structure as
dimensions of channel waveguides and to couple from planar
if it did not vary with z . These normal mode solutions are
guides to channel guides, as shown in Fig. 2. To provide a low
then said to be the local normal modes of the actual varying
loss transition, the horn must operate such that the input
waveguide structure at the position z0 . The local normal mode
mode propagates through the structure without cumulative
representation will now become a function of z .
power transfer to higher order modes. Obviously any of these
In contrast to the normal modes of a fixed structure, the
structures may be operated in reverse.
amplitudes of the local normal modes in a varying waveguide
structure are not necessarily constant. Power transfer occurs
between the local normal modes of a varying structure, and

LOCAL NORMAL MODE FORMALISM
the magnitude of that power transfer will depend on the rate
Normal Modes
For a simple three-layer waveguide structure a solution of

Maxwell’s equations for the guided modes of the structure
#
x
#
z0
z
y
z
#
Figure 4. A local normal mode is defined by replacing the slowly
# x
varying waveguide structure z0 at (solid
line) with a structure con-
#
stant with z (dashed line), and obtaining the normal mode solutions.
Figure 2. A channel waveguide horn is used as a transition struc-
Local normal modes are used as approximations to normal modes in
ture to connect single-mode and multi-mode sections of waveguide.
----------------------- Page 411-----------------------
x
Φ
a
D D + ∆D
3 3 3
z
#y#y##yy##yy
0 1
D3 ∞ (a)
Figure 5. Abrupt transition at which the waveguide separation in-
creases at a plane. This model represents a section of a branching or
separating waveguide.
Φb
of change of the geometry of the structure. This rate of change

can be either slow or fast, yielding adiabatic or abrupt transi-
tions, respectively.
(b)
ADIABATIC AND ABRUPT TRANSITIONS
Figure 6. Uncoupled mode field profiles for waveguides at large sep-

A slow or adiabatic waveguide transition is defined as a tran-
aration. Linear combinations of these modes can be used to represent
the
normal modes of the coupled waveguides with finite D .
sition between two waveguide structures that takes place suf-
3
ficiently gradually with propagation distance z so that negligi-
ble power transfer occurs between the local normal modes, as
they propagate from one structure to the other. In an adia-
guides. It can also be used to provide an approximate repre-
batic transition, power injected in a given local normal mode
sentation for the normal modes of coupled waveguides (3,4).
will stay in that mode throughout the transition. The local
If D3 is the separation between the two coupled waveguides
normal mode may change its shape as the geometry changes, of
Fig. 7, the coupling coefficient between them is assumed to
but coupling to other local normal modes is assumed not to be of
the form
occur.
K = K e−γ3D 3 (2)
The opposite extreme is a fast or abrupt transition in
0
which a transition between two waveguide structures is made
where
K0 is a constant and #3 is the transverse component of
so abruptly that the maximum amount of power transfer be-
momentum in the region between the waveguides. The differ-

tween the local normal modes occurs, consistent with the ge-
ence
in the guide propagation constants, at large separation,
ometry of the two waveguide structures. The limiting case of
is ##
# # # # . We introduce the coupled waveguide parame-
an abrupt transition is one that occurs at a plane, perpendicu-
a b
ter X
:
lar to the direction of propagation of the modes. A plane or
step transition is illustrated in Fig. 5. Power transfer between
#β
local normal modes at a step transition can be readily calcu-
X = (3)
2K
lated from a consideration of boundary conditions at the tran-
sition. We
show below that, in this approximation, power transfer
between local normal modes at an abrupt transition between

COUPLED-MODE THEORY REPRESENTATION
OF NORMAL MODES
For separating or branching waveguides, we will be dealing

with two waveguides which can each independently support
a single mode but which are sufficiently close to each other so
(a)
that the optical fields of the two guides overlap in some re-
gion. The waveguides are then said to be coupled because
transfer of optical power between them can occur. We first
define the modes of the guides when they are uncoupled; that
D3
is, the separation between the guides approaches infinity so

that optical coupling between them does not occur (Fig. 6).
Then, if the waveguides are denoted a and b, the uncoupled
(b)
mode field profiles are denoted by # # # and # # # , and
a ia b ib
their propagation constants by # and # , respectively. These

a b
are the normal modes of each guide considered separately and
Ψ Ψ
are called the modes of the uncoupled waveguides.
i j
Coupled-mode theory (6–8) is a perturbation approach

Figure 7. Coupled waveguides at finite separation. The coupled
which uses the modes of the uncoupled waveguides to de- mode
theory representation of the normal mode field profiles is shown
scribe optical power transfer between two coupled wave- for
## # 0.
----------------------- Page 412-----------------------
separating waveguides is a function only of the change in X

where the element S denotes scattering from mode i inci-
j 1i0
across the transition.

dent on side 0 to mode j exiting from side 1, and so on. Equa-
For finite D3 the normal modes of the coupled waveguide
tion (7) defines power division from a normalized input mode
system of Fig. 7 can then be expressed as a linear combina-
to the normalized output modes at a transition. For example,
tion of the uncoupled modes #a and #b
for mode i incident on side 0(# # # ), the output fields on
0 i0
side 1 are S # in mode i, and S # in mode j . The total
i1i0 i1 j 1i0 j 1
#i ≈ d #a + e#b (4a)
output field is
#j ≈ −e#a + d #b (4b)
#1 = Si 1i0 #i 1 + Sj 1i0 #j 1 (8)
Normalization is provided by d2
e2 # 1. Since the normal
modes should be orthogonal, Eq. (1) applies. However, this

Reciprocity and energy conservation require S to be symmet-
becomes an approximation due to the overlap of the uncou-
ric and unitary (9). Then Sxy # Syx and the unique elements of
pled modes for finite D3 . The constants d and e are given by
Eq. (7) can be represented by
d = # 1 #1 + 2 X 1/2 ##1/2 (5a)

Si 1i0 Si 1j 0 =
(1 − P )1/2ej φi 1i 0 P 1/2ej φi 1j 0 (9a)
2 (X + 1)
Sj 1i0 Sj 1j 0 P 1/2ej φj 1i 0 (1 − P )1/2ej φj 1j 0
e = # 1 #1 − 2 X 1/2 ##1/2 (5b)
with
2 (X + 1)
φi 1j 0 + φj 1i0 − φi 1i0 − φj 1j 0 = ±π (9b)

The difference in normal mode propagation constant, ##ij #
# # # , is
i j
where P is the power

transferred between the local normal
#βij = 2K (X 2 + 1)1/2 (6)
modes in the waveguide transition. Note that P(j # i) # P(i
# j ). Equation (9b) provides the necessary relationship be-
tween the phases of the elements of the scattering matrix.

As an example of the coupled-mode theory representation
These equations then totally define the scattering matrix for

of the normal modes, the case of ## # 0 is plotted in Fig. 7.
a general waveguide transition that supports two normal

In this case X # 0, so e # d. The normal modes are made up
modes. For adiabatic transitions P # 0 and phases are imma-

of linear combinations of the uncoupled modes, with equal
terial. For abrupt transitions between separating waveguides

amplitude.
we will see subsequently that power transfer and phases can

The coupled-mode theory representation of the normal
be calculated analytically simply from a knowledge of the

modes outlined here is useful in that it provides an analytic
change in X at the transition. For intermediate transitions,

representation with a direct dependence on the waveguide pa-
between adiabatic and abrupt, power transfer and phases

rameters. However, it is an approximation, and assumes ei-
must be calculated numerically for each case.

ther that the coupling is weak, or that the waveguides are not
too dissimilar.
Power Transfer at Abrupt Transitions
SCATTERING MATRIX FORMALISM

Next we show how the scattering matrix approach can be
used to calculate P , the power transferred between local nor-

Four Port Model for a Waveguide Transition
mal modes at an abrupt transition (10,11), for separating
waveguides. Figure 5 shows a model abrupt transition, where

In treating waveguide devices with arbitrary transitions it is
two separating waveguides increase their separation by an
useful to develop the scattering matrix for a generalized four-
amount #D3 across the boundary. Propagation is assumed in
port junction (9). With this approach one can derive output
the direction of increasing z and the regions before and after
equations for any device involving two normal modes. Con-
the boundary are denoted by 0 and 1. Since the local normal
sider a four port model of a general junction where modes i
modes of a separating waveguide structure are characterized
and j may enter or exit on sides 0 or 1. The waveguide transi-
by the parameter X , the abrupt transition can be character-
tion may be any one or two guide structure, and may be adia-
ized in terms of the parameter X before (X ) and after (X ) the
0 1
batic or abrupt, or intermediate between these two cases. The
transition. An input mode i is assumed. The input mode

modes on a given side may coexist in the same waveguide,
# # # is given by Eq. (4a), with d,e given by d ,e . The out-
0 i0 0 0
may be in separate waveguides, or a mode may have its power
put modes are given by Eqs. (4a) (for # ) and (4b) (for # ),
distributed between two waveguides. Back reflections, radia-
i1 j 1
with d,e given by d ,e . The total output field is
1 1
tion, and losses are neglected, so that only forward scattering
is accounted for. The scattering matrix S can be defined by
(10):
#1 = (1 − P )1/2ej φi 1i 0 #i 1 + P 1/2ej φj 1i 0 #j 1 (10)
#i0  0 0 Si0i 1 Si0j 1  #i0 

For a plane boundary there is no accumulated phase across
    
    
#j 0 0 0 Sj 0i 1 Sj 0j 1 #j 0
the boundary. We can pick three phases to be zero and the
  =    (7)
    
# S S 0 0 #
fourth must be chosen to satisfy Eq. (9b). This choice is re-
i 1 i 1i0 i 1j 0 i 1
  S   
lated to the boundary conditions, and may be chosen consis-
# S 0 0 #
j 1 out j 1i0 j 1j 0 j 1 in
tent with the boundary conditions from a knowledge of
----------------------- Page 413-----------------------
Table 1. Sign Rules for Mode Coupling at Abrupt Transitions

Here a(z) is the amplitude at a point z , #(x ,z) is the x-de-
X increases i0 # i1
j 1 pendent field distribution, # is the
mode propagation con-
j 0 # j 1 # i1
stant, and # is a phase constant. Boundary conditions for TE
X decreases i0 # i1 #j 1
modes require Ey and Hx to be continuous across the step.
j 0 # j 1
i1 This requirement leads to a set of coupled
amplitude equa-
tions given by
A e−j αj 1 = c A e−j αi 0 + c A e−j αj 0 (14a)
j 1 ij i0 jj j 0
whether X is increasing or decreasing across the boundary by

A e−j αi 1 = c A e−j αi 0 + c A e−j αj 0 (14b)
i 1 ii i0 ji j 0
using the sign rules given in Table 1.

The desired result is obtained by imposing the require-
where
ment of continuous transverse electric fields across the
I
boundary, that is, # # # . This yields
i0,j 1
0 1
c ≈ (15)
ij √I I
i0,i0 j 1,j 1
(f 1 − f 0 )2
P = (11)
and I # ## # dx, for #,# # i,j . In Eq. (14), A is the mode am-
(f 1 − f 0 )2 + (1 + f 1 f 0 )2
#,# # # #
plitude a# normalized by a mode amplitude that corresponds
to unity power. Values of c and c are obtained by the appro-
jj ii
where f # e /d and f # e /d , and f (X ) is given by
0 0 0 1 1 1
priate substitution of i andj in Eq. (15), and c # #c , a conse-
ji ij
quence of power
conservation. These equations are appro-
e
f = = −X + (X 2 + 1)1/2 (12)
priate for a single step. By dividing a transition into many
d
small steps and iteratively applying these equations one can
determine the output mode amplitudes, and thus powers, for

For symmetric steps (about a centerline), Eq. (11) is inde-
the entire transition. The only requirement is that the fields
pendent of step size. For asymmetric steps, Eq. (11) is valid
throughout the structure are known. Later, we will show that
for small step sizes.
approximate analytic expressions for cij can be obtained, so
The approach developed in this section may be used to ob-
that specific knowledge of the fields can be avoided.
tain the output equations for any device (12) composed of di-
rectional couplers, interferometric sections, and combining
Differential Form
and separating branches, so long as the branches are as-
By transforming the coupled amplitude equations of Eq. (14)
sumed to be abrupt. It allows one to keep track of modal
into differential equations we can obtain coupled mode equa-
fields, and thus modal power, throughout the device, by seri-
tions for the local normal modes (15). These equations can be
ally treating each transition, phase shift, or coupling section.
used to show that transitions shaped in a particular way have
This formalism may be used to demonstrate how devices are
analytic solutions, within the approximations made. Alterna-
affected when transitions assumed abrupt may in fact not be.
tively they can be numerically solved themselves. The coupled
For devices with transitions intermediate between adiabatic
amplitude equations can be expressed
and abrupt this approach may still be used to develop output
equations, but power transfer and mode phase changes in the
Aj 1 = # cγ iA γ 0 , γ = i,j (16a)
intermediate transitions must be computed with a numerical

γ
technique.
A i 1 = # cγ j A γ 0 , γ = i,j (16b)
COUPLED AMPLITUDE EQUATIONS

which relate the transmitted local normal mode amplitudes
on side 1 of a small step to those incident on side 0. The am-
plitudes are now considered to be complex so that A # #A #
Single Step Model

i i
exp [ #j (# z
# )]. We generalize to a step in which an unspec-
i i
One commonly used numerical approach to treat waveguide
ified parameter p varies from p to p
#p across the step.
transitions is to divide the transition into a series of small
Thus, p may be the channel width in an expansion horn or
abrupt steps, connected by regions of constant waveguide pa-
the waveguide separation in a separating waveguide struc-
rameters, and calculate the mode propagation through this
ture. We expect that cij will be proportional to the change #p
idealized geometry (13,14). For the single step of Fig. 5, the
across the step, and remove this dependence by defining a
waveguide modes of interest are assumed to be incident on
new coefficient Cij
side 0, and the transmitted mode amplitudes are computed
by imposing boundary conditions at the interface, i.e., the
Cij = lim #cij # (17)
transverse field components must be continuous.

δp →0 δp
For two transverse electric (TE) guided modes, i andj , elec-
For a small step c # c # 1, c # #c # C #p , and Eq. (16) can

tric fields can be written as
jj ii ij ji ij
be written as
Ey = a (z )(x ,z )e−j α(z ) (13a)

dAj dp
= C A − j β A (18a)
dz i j dz i j j
where
dAi dp
= −Cij A − j β A (18b)
a (z ) = βz + φ (13b)
dz dz j i i
----------------------- Page 414-----------------------
which are the differential equations describing coupling be-

COUPLED MODE THEORY REPRESENTATION OF Cij AND ##ij
tween the local normal modes. These equations will be re-
ferred to as the coupled normal mode equations.
The coupled normal mode equations developed previously
show that coupling between local normal modes in a transi-
tion structure is governed by the coupling coefficient between

Analytic Solutions
them, Cij . Since we have analytic representations for the local

For specifically shaped transitions the coupled normal mode
normal modes from coupled mode theory, we can use these
equations have an analytic solution (15,16). If Cij dp/dz # 0,
representations to calculate an analytic expression for Cij
no power transfer between local normal modes occurs and the

(11,17). This is done here for the separating waveguide struc-
evolution of local normal mode i is described by a phase fac-
ture of Fig. 5, where the separation between guiding struc-
z
tor exp[ #j # # dz #]. Consider reduced mode amplitudes a #
tures is increasing across the step. From Eqs. (15a) and (17)
0 i i
z z
A exp[j # # dz #] and a # A exp[j # # dz #]. The variation of p
we have
i 0 i j j 0 j
with z is determined by the shape of the transition. For a

shape such that
c = C #D ≈ # ∞ # # dx (23)
ij ij 3 i0 j 1
−∞
dp = γ ##βij # (19)
dz Cij
where the parameter p is replaced by #D , the increase in
separation between the two guiding regions across the step.
Also, # refers to # before the step on side 0, and # refers

where ## # # # # and # is an arbitrary constant, Eq. (18)
i0 i j 1
ij i j
to # after the step on side 1. Equation (23) is solved by sub-

becomes
j
stituting the coupled mode representation for the normal

da
modes, yielding
j ju
= γ e ai (20a)
du
γ3X
dai −ju
Cij = 2 (24)
du = −γ e aj (20b)
2 (X + 1)
where u # z
which gives Cij as a function of X . This result is valid for sym-
# ## dz #. Equation (20) is a set of conventional
0 ij
metric branch shapes. For asymmetric separating wave-
coupled mode equations with constant coefficient # for the
guides, there will be a geometric contribution to Cij, which is

amplitudes a and a . These equations have the solution for
i j
not included in Eq. (24). For separating waveguides, ##ij is
a
j
given by Eq. (6).
aj = ai0 # 2γ eju /2 sin #1 (#4γ 2 + 1)u# (21)
(4γ 2 + 1) 2
BRANCHES
for the initial conditions a # a and a # 0 at z # 0.

The separating waveguide of Fig. 1(b) and the branching
i i0 j
The quantity # is a measure of the strength of the coupling

waveguide of Fig. 1(a) can be described in an equivalent man-
between the local normal modes. For small # the amount of
ner if the parallel waveguides in Fig. 1(b) are sufficiently close
power converted to mode j has a maximum value of
for the normal modes to have power in both guiding regions.
We will assume this

to be true and from now on will refer
Pmax (a ∗ 2
generically to both structures as branches. Branches fall into
j j a )max 4γ
= j ∗ = 2 (22)
two categories, depending on whether ## is or is not identi-
Pi0 ai0ai0 4γ + 1
cally zero. For ## # 0, it follows that X # 0 throughout the
structure and no power transfer occurs. The normal mode

Adiabatic structures should be designed with # # 1, while
shapes are defined by Eqs. (4) and (5) and each normal mode
abrupt structures need # # 1. The approach considered here
always has its power evenly divided between the two arms of
is valid if the principal coupling occurs between mode i and
the branch. Subsequently we will discuss the case ## # 0, for
mode j , and coupling to other modes can be neglected.
linear and shaped branches.
Minimum Length Structures

Linear Branches
Abrupt structures tend to be short and transition length is

Linear branches have been treated both by coupled amplitude
not a problem. However, adiabatic structures tend to be very
equations using the small step approximation (14) and by nu-
long and it is of great practical interest to design minimum
merical integration of the coupled normal mode equations
length structures. The shape described by Eq. (19) holds
(18). As noted previously branch behavior can be separated
(dp/dz)(C / ## ) constant throughout the length of the transi-
into two classes, power dividers and mode splitters. Power
ij ij
tion. This results in a constant pressure or tendency for mode

dividers tend to divide input power between the output arms
conversion between local normal modes everywhere (16,17). If
of the branch. In mode splitters output power is predomi-
we consider averages over mode interference effects, this
nately in a single output arm of the branch, and the output
should produce a transition which is as short as it can be for
arm selected depends on which input normal mode is excited.
a given amount of allowed mode conversion.
Power dividing and mode splitting behavior is illustrated in
----------------------- Page 415-----------------------
Φ
branch behavior depends only on the parameter ##/ ##3
and
a
the variation of X (z) in the branch. This simple result is a

Ψ Ψ
consequence of constant # in the linear branch. Eq. (26) can
i θ D3 j
be integrated to obtain output mode amplitudes and phases.
Φb
Shaped Branches
(a)
As discussed previously the analytic solution to the coupled
normal mode equations may be used to define a

shaped
branch with interesting properties (17). It has analytic solu-

–
Φa Φa
tions so that its behavior may be easily understood, and it
can provide design guidance for minimum length structures.

Ψ Ψ
As assumed, we take p # D3 in Eqs. (18) and (19). Coupled
i j
mode theory approximations for ##ij and Cij are assumed. For
branch angle # (no longer a constant or necessarily small), Eq.

Φb Φb
(19) becomes
(b)
tan #θ # = 1 dD3 = γ #β (X 2 + 1)3/2 (27)
Figure 8. The limiting behavior of (a) mode splitting and (b) power
2 2 dz γ3 X 2
dividing branches for single local normal mode input i or j. The mode
splitting branch separates the local normal modes; the power dividing
branch divides input optical power while maintaining relative phase
This result can be integrated to give the branch shape
between the branch arm outputs.
cos (tan−1 X0 ) − cos (tan−1 X ) = 2γ #βz (28)
Fig. 8 for input normal modes i and j . By using the approxi-
which is plotted in Fig. 9 for different values of X 0 . In order

mations of coupled mode theory, it has been shown (14) that
to satisfy Eq. (27) with a constant #, the branch angle # be-

operation as a mode splitter or as a power divider depends on
comes very large when the ratio Cij/ ##ij is small, at the begin-
whether
ning and ends of the

branch. In an actual device these large
#β #β
angles would have to be moderated to avoid excessive radia-
> 0.43 or < 0.43
(25) tion losses. The minimum value of # is
θγ3 θγ3
Operation will be as a mode splitter if the first limit holds,

θm = 33/2 γ #β (29)
and as a power divider if the second limit holds, where # is the
γ3
full branch angle. The validity of this dependence has been
demonstrated for calculated branch output using the coupled
amplitude equations (14), and by experimental branch results
for Ti : LiNbO3 channel waveguides (19).
5
This dependence on the quantity ##/ ##3 can also be dem-
onstrated by transforming the coupled normal mode

equa-
tions to a form appropriate for a linear branch (18). For a
4
branch with separation D3 between the arms, we have
# #
dD3/dz, for small angles. The coupling coefficient K is given
by Eq. (2), with D3 # 0 at z # 0. Then, from Eq. (3), X can be
3
expressed as X # X exp[# D ] where X # ##/2K is the
value
0 3 3 0 0
–
of X at z # 0. The coupled normal mode equations [Eq. (18)]
3 –– x0 = 0.01
D – 2
3 –
with p # D3, and with the reduced mode amplitudes intro-
γ –
2
duced previously, can be written for constant # as
x0 = 0.1
daj 1
= a exp[−ju (X ,X )]
(26a) 1
dX 2 (X 2 + 1) i 0
x0 = 0.5
da 1
i
= − 2 aj exp[ju (X ,X0 )]
(26b)
dX 2 (X + 1)
0
0 0.1 0.2 0.3 0.4 0.5

where
βγ
∆ z
u(X ,X0 ) = #β # X √X #2 + 1 dX #
(26c) Figure 9. Normalized plot of branch arm separation versus direction
θγ3 X 0 X #2
of propagation z for shaped structures with initial conditions X . The
shape is symmetric about a center line between the two guiding re-
In Eq. (26) we have used coupled mode theory representa-
gions. This shaped structure maintains a constant level of
power
tions for ## and C . It is clear from this representation that
transfer between normal modes throughout its length.
ij ij
----------------------- Page 416-----------------------
which occurs at X # #2 where the coupling effects are the

1 a1
strongest (C / ## is a maximum). The branch length (l ) is
ij ij b
given by taking D3 # 0
1
l = (30)
b √ 2
2γ #β X0 + 1
2 a2
Adiabatic operation with smaller values of #, therefore, corre-

sponds to longer branches. Cumulative power conversion from
1 (a)
mode i to mode j is given by Eq. (21)
a b
i j
|Aj |2 = 4γ 2 sin2 #u #4γ 2 + 1# (31a)
|A i0 |2 4γ 2 + 1 2
–b
with
2 a
1
−1 −1
u = (tan X − tan X0 ) (31b)
(b)
2γ
Figure 11. The use of superposition to solve a more complicated

which is a function of # and X . This result allows one to cal-
branching problem. In (a) unequal amplitudes a1 and a2 are incident
0
on an ideal power dividing branch in reverse. In (b) the original

culate power transfer for a shaped branch. The interference
modes are decomposed into a new set with in-phase amplitudes, a,

term approaches #1/2 for large values of # and X , that is, at
and out-of-phase amplitudes, b, which combine to form output normal
the branch output in the abrupt limit.
modes i and j .
In the shaped structure defined previously, # is a measure
of the strength of the coupling between local normal modes
and is held constant in the structure. In a linear branch, # #
already available (11). For example Fig. 11 shows the case of
#C / ## becomes z-dependent and a plot of # versus z shows
ij ij
power incident on an ideal power dividing branch with ## #
where mode coupling takes place in the branch. To demon-
0, from the branch arms. We assume unequal powers incident
strate the length advantages of shaped structures we compare
with amplitude a in guide 1 and a in guide 2 (a # a ), but
1 2 1 2
in Fig. 10 plots of # versus z for comparable shaped and linear
with equal phase [Fig. 11(a)]. We introduce a new set of mode
adiabatic branches. The linear branch has a constant # equal
amplitudes a and b defined by a1 # a
b and a2 # a # b, and
to #m [obtained from Eq. (29)] for the shaped branch. The
decompose the original set of modes into a new set with equal
branches are chosen so that they produce comparable mode
in-phase (a) and equal out-of-phase (b) amplitudes, as shown
conversion. The length advantage for the shaped device is
in Fig. 11(b). The in-phase mode amplitudes will combine to
clear. However, as noted before, the shaped devices may have
form local normal mode i in the output waveguide, and the
to be modified to maintain acceptable radiation loss, which
out-of-phase mode amplitudes will combine to form local nor-
may somewhat reduce their advantage.
mal mode j . Conservation of power will show that the output
amplitudes of modes i and j are related to the input ampli-

Superposition of Solutions
tudes by a # #2a and b # b. Problems of this kind can of
i j
course be worked out more formally using the scattering ma-

For a single normal mode incident on a branch, its behavior
can be understood in terms of the local normal modes of the
trix approach developed earlier.
structure, as in Fig. 8. For more complicated cases, with more

than one mode incident, or with propagation reversed in the
Current Branch Design
branch, outputs can be obtained by superposition of solutions

The overwhelming majority of branches and separating wave-
guides in current use in integrated optical devices are abrupt
power dividers. Because these branches are so short (a few
hundred microns in Ti : LiNbO3), there is no incentive to use

.10
anything other than a linear branch. Adiabatic mode splitting

.08 Shaped
branches are seldom used due to the long length required
(5–10 mm in Ti : LiNbO3) for good adiabatic operation. One

.06
important exception is the design of the active branching
γ
modulator (18,20) which is used in switch matrices (21,22)
.04
Linear
and in the digital switch (23). As shown in Fig. 12 the active
.02
branching modulator is a symmetric branch with electrodes
placed to symmetrically raise and lower the index in opposing

0
arms in a reversible, push–pull fashion. With no voltage ap-
0 0.4 0.8 1.2 1.6 2.0
plied, ## # 0, and the device

acts as a power divider. In one
z (mm)
of its switch states a positive voltage is applied and the asym-
Figure 10. The parameter # is plotted versus z for shaped and linear
metric index change makes ## large, so the device acts as an
branches designed to be mode splitters. The magnitude of # shows
adiabatic mode splitter. By reversing the sign of the applied
where power transfer between normal modes occurs in each branch.
voltage, the asymmetric index change reverses, and power ex-
----------------------- Page 417-----------------------
creasing
coupling problem between the modes as W increases
in a linear
horn with constant #. A solution to this problem is
to shape
the horn as described subsequently.
###yyy For an
analytic solution to the coupled normal mode equa-
yy
## tions
using a shaped horn (16), the required horn shape is
obtained by
substituting p # W in Eq. (19). Using the expres-
sions
obtained for ##ij and Cij in that result yields the local
θ
m horn angle
# (Fig. 13)
1 dW 4π λ
θ ≡ = −γ g (34)
###yyy
2 dz 3 W
yy
##yy##
where # is
a constant. The shape of this horn will be parabolic
with
W = (2αλ z + W2 )1/2 (35)

Figure 12. Active branching modulator with a shaped structure: #m
g 0
is the minimum angle in the branch. A voltage is applied to the center
yy
##yy##
electrode relative to the outside (grounded) electrodes, creating index where W0

is the horn width at z # 0 and # # #8##/3. The
changes of opposite polarity in the arms of the branch. The branch is solution
of power transfer from initial mode i to mode j is
driven to a mode splitter configuration, in which the input mode exits given by
Eq. (31a), with u given by (2#/ #)ln(W/ W ).
0
one output arm. Switching is achieved by changing the polarity of the
Arguments for the length efficiency of this shape are the
applied voltage. same as
those made earlier for the shaped branch. With a
parabolic
shape # # 2Cij#/ ##ij remains constant throughout
the horn,
and the strength of mode coupling between i and j
is evenly
distributed along the length. This is not true for
its the other arm, giving switching action. Minimizing the other horn
shapes where # will be a function of z . For the
length of this adiabatic device has generated considerable same
amount of mode conversion, in horns with the same
theoretical and experimental (24) interest, with the shaped expansion,
the parabolic horn should be shorter than other
branch solution (25,26) playing a significant role. shapes,
within the approximations made.
WAVEGUIDE HORNS
DIRECT
COMPUTATIONAL METHODS
Waveguide horns cannot make use of the coupled mode theory

An
alternative to the coupled mode theory approach is a di-
results for two coupled waveguides. However, the coupled
rect
computational approach which calculates the propaga-
normal mode equations do apply with p # W, the waveguide
tion of
light through a waveguide structure by applying an
width. In a horn we are generally interested in adiabatic
appropriate
computational scheme to the wave equation. One
propagation only, as typically one wants to expand the width
only needs
to assume the geometry of the index distribution,
of a channel waveguide while maintaining power in the low-
which may
vary with z , the input field, and any approxima-
est order mode. For symmetrical horns coupling between the
tions
inherent in the computational scheme employed. Entire
first and second local normal modes does not occur (as the
devices
including a variety of transition structures can be cal-
overlap integral is zero), and we will be concerned with cou-
culated,
given sufficient computer resources. Although a num-
pling between the first and third order modes. First we give
ber of
suitable computation schemes exist, only the more com-
expressions for ##ij and Cij appropriate to this problem (16).
monly
used beam propagation method (BPM) (1,2) will be
The value of ##ij can be approximated from the dispersion
discussed
here. The BPM is a split operator method used to
relationship for a channel waveguide. The propagation con-
solve the
scalar homogeneous Helmholtz equation. It effec-
stant difference for channel modes of different orders (first
and third) in the x direction, but of the same order in the y
direction (Fig. 2), is
2πλ
#βij = W2g (32)
θ
####
where # # 2#/# is the modal wavelength in the guide for the

g 0
first order mode. Also, Cij can be obtained from a consider- W0
W
ation of boundary conditions at a small step, and is given by
3
Cij = − 4W (33)
x
####
for a step increasing with z . These results are valid in a wide

z
channel where the modes are far from cut off. The ratio Figure 13.
Parabolic channel waveguide horn. Using the parabolic
#C / ## # increases as W increases. This will result in an in- geometry
allows one to analytically calculate horn transmission.
ij ij
----------------------- Page 418-----------------------
tively models an optical structure as a series of infinitely thin

4. W. H. Louisell, Analysis of the single tapered mode coupler, Bell
lenses separated by an incremental axial distance #z .
Syst. Tech. J., 34: 853-870, 1955.
For monochromatic light propagating in an optical wave-
5. D. Marcuse, Theory of Dielectric Optical Waveguides, New York:
guide, the homogeneous Helmholtz equation is (27)
Academic, 1974, Chapt. 3.
6. S. E. Miller, Coupled wave
theory and waveguide applications,
ω2
Bell Syst. Tech. J., 33: 661-719, 1954.
2 2
∇ E = 2 n (x ,y )E = 0 (36)
7. W. H. Louisell, Coupled Mode and Parametric Electronics, New
c
York: Wiley, 1960.
8. A. Yariv, Coupled-mode theory for guided wave optics, IEEE J.

where the refractive index n(x ,y ) is assumed to depend only
Quantum Electron., QE-9: 919-933, 1973.

on the x and y coordinates. For a forward propagating field in
9. D. S. Jones, The Theory of Electromagnetism, New York: Macmil-

the
z direction, E(#,x ,y ,z) may be expressed as E(x ,y )exp(#
lan, 1964, pp. 254-261.

jkz ), where k # n #/c, and n is the index of the substrate.
0 0
10. W. K. Burns, Normal mode analysis of waveguide devices. Part
With these assumptions one can obtain
I: Theory, J. Lightwave Tech. 6: 1051-1057, 1988.

# #
11. W. K. Burnsand A. F. Milton, Waveguide transitions and junc-

∂ E # + j ∇2 + k2 n2 − 1 E # = 0 (37a)
tions, in T. Tamir (ed.), Guided-Wave Optoelectronics, Berlin:
∂z 2 ⊥ n2
Springer-Verlag, 1988.
0
12. W. K. Burns, Normal mode analysis of waveguide devices. Part

where
II: Device output and crosstalk, J. Lightwave Tech. 6: 1058–1068,
∂2 ∂2
1988. (see also errata, J. Lightwave Tech. 7: 1425, 1989.)
∇2 = + (37b)
13. D. Marcuse, Radiation losses of tapered dielectric slab wave-
⊥ ∂x2 ∂y 2
guides, Bell Syst. Tech. J., 49: 273-290, 1970.
14. W. K. Burns and A. F. Milton, Mode conversion in planar-dielec-

This is the paraxial or Fresnel form of the wave equation
tric separating waveguides, IEEE J. Quantum Electron., QE-11:
where E # denotes a slowly varying approximation for the field
32-39, 1975.
in the z direction. This form has a solution which we write as
15. A. F. Milton and W. K. Burns, Mode coupling in tapered optical
(28)
waveguide structures and electrooptic switches, IEEE Trans. Cir-
cuits Syst., CAS-26: 1020-1028, 1979.

# # j #z ∇2 #
16. A. F. Milton and W. K. Burns, Mode coupling in optical wave-
E (x ,y, #z ) = exp − ⊥
2k
guide horns, IEEE J. Quantum Electron., QE-13: 828-835, 1977.
j #z 2 #n2 # # 17. W. K. Burns and A. F. Milton,
An analytic solution for mode cou-
exp − 2k k n2 − 1 E (x ,y, 0) (38)

pling in optical waveguide branches, IEEE J. Quantum Electron.,
0
QE-16: 446-454, 1980.
18. W. K. Burns, M. M. Howerton, and R. P. Moeller, Performance

where E #(x ,y ,0) is the initial condition of the field at z # 0.
and modeling of proton exchanged LiTaO3 branching modulators,
J. Lightwave Tech., 10: 1404-1408, 1992.

Equation (38) is written with two exponential factors which
is the basis for the BPM. The two factors represent a step of
19. W. K. Burns et al., Optical waveguide channel branches in Ti-
diffused LiNbO , Appl. Opt., 19: 2890-2896, 1980.

propagation over a length #z , and a phase or lens term. This
3
approach provides a method of visualizing the propagation
20. W. K. Burns, A. B. Lee, and A. F. Milton, Active branching wave-
guide modulator, Appl. Phys. Lett., 29: 790, 1976.

process by replacing the waveguide with a series of infinitely
21. P. Granestrand et al., Integrated optics 4x4 switch matrix with

thin lenses, separated by the axial distance #z . By using a
digital optic switches, Electron. Lett., 26: 4, 1990.

Fourier series representation of the initial field, one can use
22. A. C. O’Donnell, Polarization independent 1 16 and 132 lith-

fast Fourier transform algorithms to implement the propaga-
ium niobate optical switch matrices, Electron. Lett., 27: 2349,

tion terms in Eq. (38). This allows for algebraic implementa-
1991.
tion of the propagation term in the spatial frequency domain,
23. Y. Silberberg, P. Perlmutter, and J. E. Baran, Digital optical

while the phase or lens term is performed with simple multi-
switch, Appl. Phys. Lett., 51: 1230, 1987.
plication in the spatial domain. The BPM then consists of a
24. H. Okayama, T. Ushikubo, and M. Kawahara, Low drive voltage
set of Fourier transforms interspersed with complex multipli-
Y-branch digital optical switch, Electron. Lett., 27: 24, 1991.
cations in the spatial domain in an iterative algorithm, thus
25. W. K. Burns, Voltage-length product for modal evolution-type
advancing the solution in successive steps along the optic axis
digital switches. J. Lightwave Tech., 8: 990-997, 1990.
(29). The BPM approach has been widely used for a variety of
26. W. K. Burns, Shaping the digital switch, IEEE Photon. Tech. Lett.,
4: 861-863, 1992.
27. M. D. Feit and J. A. Fleck, Jr., Mode properties and dispersion

for two optical fiber-index profiles by the propagating beam
BIBLIOGRAPHY
method, Appl. Optics, 19: 3140-3150, 1980.
28. M. D. Feit and J. A. Fleck, Jr., Light propagation in graded-index

1. J. van Roey, J. van der Donk, and P. E. Lagasse, Beam-propaga-
optical fibers, Appl. Optics, 17: 3990-3999, 1978.
tion method: Analysis and assessment, J. Opt. Soc. Am., 71: 803-
29. P. E. Pace and C. C. Foster, Beam propagation analysis of a par-
810, 1981.
allel configuration of Mach-Zehnder interferometers, Optical
2. L. Thylen, The beam propagation method: An analysis of its ap-

Engr., 33: 2911-2921, 1994.
plicability, Opt. Quantum Electron., 15: 433-439, 1983.
3. A. G. Fox, Wave coupling by warped normal modes, Bell Syst.

W. K. BURNS
Tech. J., 34: 823-852, 1955.
Naval Research Laboratory
----------------------- Page 419-----------------------
364 OPTIMAL CONTROL
OPTICS, MATERIALS. See OPTICAL MATERIALS.

OPTICS, NONLINEAR. See NONLINEAR OPTICS.
----------------------- Page 420-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
●
HELP

Your Librarian
Profile
Power System Control
Standard Article

Roger L. King1 and Rogelio Luck2
1Mississippi State University, Mississippi State, MS
2Mississippi State University, Mississippi State, MS

Search
reserved.
DOI: 10.1002/047134608X.W6213
❍ Acronym Finder
Abstract
The Control Problem
Governor Control System
Control Areas and Supplementary Control
Constrained Dispatch of Generation
Power Pools and Energy Management Systems
Deregulation
Facts
Emerging Power Control Issues

%20Conversion/W6213.htm17.06.2008 16:44:01
----------------------- Page 421-----------------------

c
POWER SYSTEM CONTROL
An electric utility has the responsibility of generating electricity and delivering

it on demand to a remote
customer. If an electric utility consisted of one generating plant connected to one
customer by one transmis-
sion path, then the control of the system would be trivial. However, today’s modern
electric utility consists
of multiple, geographically spaced sources of electric power generation,
interconnected by a spider-web-like
system of transmission and distribution networks to ensure efficient delivery of
electricity to its customers.
For example, one large electric utility in the southeastern United States generated
156,000,000,000/kWh for
about 11,000,000 people living and working in over a 120,000 square mile area in
1996. This large-scale sys-
tem permits an electric utility to offer to its consumers a reliable product of
high quality and economically
priced. It also gives the utility the opportunity to be competitive with its
resources while still being environ-
mentally friendly. It is the role of power system control to ensure these benefits
to the electric consumer and
provider are realized. Therefore, power system control can be viewed as the problem
of controlling the power
output of each generator such that the combined power of the generators matches the
combined power demand
of the loads at the nominal frequency while meeting a variety of economic,
operational, and environmental
constraints.
The following explanation of power system control is based on the traditional
approach of the vertically
integrated utility. However, the electric supply industry is in a rapid period of
flux. Functions previously in
the exclusive domain of the vertically integrated utility are now being
performed by nonutility marketers
and independent system operators (ISOs). The traditional regulated monopoly
environment with exclusive
franchise territories and a mandated obligation to serve is shifting rapidly toward
an open access, market-
based system with flexible boundaries and only economic obligations. Today wholesale
customers have open
access. In the future, retail access, in some form or fashion, will probably be
available to most customers.
This will radically change (and complicate) the power system control process from
what is explained herein.
However, the fundamental premise of power system control will not change—generation
will still be required
to meet load.
The Control Problem

Control of any process requires there be variables that can be sensed to determine
the state of the process and
actuators that can be controlled to change the state of the system to a new state.
This normally happens within
the context of a closed loop, or feedback, control system. Figure 1 shows a typical
feedback control system. In
power system control, some of the variables of importance are system frequency,
voltage, and power flow (real
and reactive). For a thermal unit (e.g., coal-fired boiler) the actuators would be
the turbine governor and steam
admission valves that adjust output power. Another actuator could be the settings
on the generator’s exciter
which change the output voltage of the generator. In Fig. 1 the Plant represents
the interconnected power
system (generators, transmission lines, etc.). Information about system frequency,
power flows, or voltage may
be sensed and fed back to the referenced system input. An error is calculated
between the actual and expected
----------------------- Page 422-----------------------
2 POWER SYSTEM CONTROL
Fig. 1. Block diagram of a typical feedback control

system.
Fig. 2. Multiple generators supplying a system

load.
system variables and this is translated into a control action via the actuator.
With the control action initiated
by the actuator, the plant moves to a new state or operating point and the cycle
continues.
Figure 2 will help to illustrate the control problem in modern power systems.
The principal function of
the power system control unit is to adjust the output of the available generators
to meet the system load while
meeting all constraints imposed by engineering design, regulatory, and financial
considerations. This desire
can be expressed as
In general, a utility will have a large number of generating units of

various size (MW capacity) and type
(thermal, hydroelectric, etc.) available to meet its customers demand or
load. These units are operated in
synchronism which means they all operate at the same frequency and will all attempt
to respond to any change
in load.
One of the first control problems is the effective operation of machines in
parallel. Most people think of
electricity as being a constant frequency (50 Hz or 60 Hz) commodity. However, a
truly constant frequency
requires a generator to operate at a constant speed. This constant frequency
control is an impossible task
to perform on an interconnected power system such as that shown in Fig. 2. To
enable efficient operation
----------------------- Page 423-----------------------
POWER SYSTEM CONTROL 3
of parallel generators a speed governor must be used on the turbines of

the various units. This turbine
governor permits the units to depart from the nominal system frequency in order to
participate in effective load
sharing.
Governor Control System
If Eq. (1) were satisfied all of the time, then there would be no need for a
governor control system. However,
loads change according to the needs of the consumers and are strongly dependent on
weather patterns and
work day schedules. Since system load is a dynamic quantity that cannot be
precisely predicted, the governor
control systems are often reacting to correct for the mismatch of generation to
load. The generating units with
governors are acting to ensure the synchronous network can withstand these
imbalances without collapsing.
Therefore, the governor control system provides primary frequency control to the
synchronous network.
The result of this mismatch of generation to load is a frequency deviation
from the nominal value. Under-
generation or an increase in load will cause the system frequency to drop, while
over-generation or a decrease
in load will cause the frequency to rise. Equipment constraints limit the maximum
frequency deviation from
nominal values. Larger deviations lead to problems such as unacceptable vibration
levels in steam turbine
blades.
Governor control or primary frequency control is a short time frame (seconds)
mechanism for ensuring
generation meets load. It will not, however, reset frequency to nominal after a
disturbance. Automatic gen-
eration control (AGC ) operating in the tie line bias mode provides secondary
frequency control on the longer
time frame (minutes) and is responsible for resetting system frequency to nominal
after a sudden frequency
shift has been arrested by the primary frequency control system. The two systems
work together to provide
stable operations at or near setpoint frequency on the synchronous network. In
general, governor control can
be thought of as the high-frequency component of frequency control and AGC as the
low-frequency and/or bias
component. They should be thought of as two different control loops.
The governor control system can be illustrated from Fig. 1. Considering a
thermal unit, the system input
is steam flow, the actuator controls a steam admission valve, the plant block
represents the unit’s prime mover
(turbine), a device for measuring speed provides an input to the
feedback loop, and the system output is
mechanical power to the generator. The key to this control system is the speed-
droop characteristic illustrated
in Fig. 3. This characteristic shows that when there is a load increase on the
generator that it will slow down.
The slope of this characteristic, R , is referred to as the droop of the generator.
Then R is the steady-state
relationship between system frequency and generated power. Normally, all of the
individual units should have
their droop characteristics set to be about equal to minimize turbine governor
oscillation. Also, with the same
droop characteristic, each unit will share a change in system load in proportion to
the individual unit’s MW
rating. Values for droop ideally range from 3 to 5%. However, according to a recent
EPRI survey, the true
measured droop characteristic of generators actually varied from 1% to 21% with 5%
being the most common
(1).
The governor control system is an integral controller with a feedback loop
gain of R . Therefore, in operation
when the speed of the turbine is sensed to be slowing down (i.e., the load on the
machine has increased), the
feedback loop will provide a signal to open the steam admission valve. When more
steam is admitted into the
turbine this will result in more mechanical power out of the turbine and eventually
more electrical power out
of the generator to serve the load increase. When an increase in speed occurs (load
decreasing), the opposite
actions will be taken and the electrical output of the generator will be decreased.
More detailed analysis of the
governor control system can be found in Refs. 2,3,4.
----------------------- Page 424-----------------------
Fig. 3. Governor or speed-droop

characteristic.
Control Areas and Supplementary Control
In general, electric utilities have operated within a franchised area. In

other words, the local utility was
responsible for meeting the load within a given geographical area.
Although this is changing (see section
entitled Deregulation), the classic view of a franchised service area is still
useful in understanding the larger
picture of supplementary control. The North American Electric Reliability
Council (NERC ) was formed in
1968 as an industry group responsible for setting guidelines to ensure reliable
operation of the power system.
NERC took the interconnected utilities in the United States and Canada
and created four synchronously
operated regions—the Eastern interconnection, the ERCOT interconnection, the
Western interconnection, and
the Quebec interconnection. These four interconnections were further divided into
ten regional councils. And,
finally the regions were divided into smaller operating units known as control
areas. These control areas are
responsible for ensuring the load within their geographic area of responsibility is
reliably met.
A control area is defined by NERC as an electric system or systems, bounded by
interconnection metering,
and telemetry, capable of controlling generation to maintain its interchange
schedule with other control areas
and contributing to frequency regulation of the interconnection (5). A control area
meets its obligations of
meeting its interchange schedule and contributing to frequency regulation by either
generating power from
units owned by utilities within the control area, by purchasing power from other
sources, or by demand side
management. One control area is connected to another via transmission lines that
are specified as tie lines
(see Fig. 4). It is on these tie lines that the interchange schedules are
maintained.
Supplementary control is often referred to as either load frequency control
(LFC ) or automatic generation
control (AGC ). It differs from the governor control system in that it is on a time
scale of several seconds where
governor response is designed to be very nearly instantaneous. The primary
objectives of AGC are: (1) to keep
tie line interchanges as scheduled, (2) to maintain the system frequency at the
nominal frequency (60 Hz in
the USA), (3) to operate with security in mind (i.e., with sufficient energy reserve
to accommodate disturbances
in demand or generation), (4) to optimize economical operation, and (5) to maximize
compliance with NERC
guidelines.
Tie line bias control has typically been the control algorithm used for
ensuring a control area meets its
obligations to the grid. This control scheme attempts to maintain frequency support
and interchange schedules
by defining an area control error (ACE ) term as follows
----------------------- Page 425-----------------------
Fig. 4. Interconnected control areas.
where Ta is the measured tie line or interchange power flows in megawatts, Ts

is the scheduled flow, Fa is
the measured system frequency in hertz, Fs is the scheduled frequency, and B
is a term representative of the
control area’s obligation to frequency support in MW/0.1 Hz. A more detailed
explanation of tie line bias control
is given in Ref. 6.
Figure 5 shows a block diagram of a supplementary control system. Note that
it has metered tie line and
frequency data which it uses to calculate a value for ACE. This metered data is
normally filtered to provide a
smoother function for control. A close examination of Eq. (2) shows that a positive
value of ACE indicates that
the control area is exporting more power than scheduled, while a negative ACE
indicates the area is importing
more than it has scheduled. A judicious examination of Eq. (2) can show to a
control area whether the problem
on the grid causing inadvertent tie line power flows is being caused by their
control area or another one. This is
an important factor in determining the new dispatch levels for your control area’s
generation. A simple method
to help locate the source of the lost generation is to remember that if the signs
of the two components of ACE
are the same, then the problem is within the control area. If they differ, the
problem is outside the control area.
The supplementary control cycle typically occurs on a 2 s to 6 s time frame.

When the new dispatch
levels are calculated, this information is transmitted to the plants by either a
setpoint or pulse methodology.
It can also be seen from Fig. 5 that supplementary control is typically a
centralized control function; however,
some work is on-going to make it more of a distributed control function (7). In
fact, the new NERC Control
Performance Standard is fully decomposable. A more complete description of an
automatic generation control
system can be seen in Ref. 8. Currently, supplementary control systems are
primarily used to dispatch real
power. Reactive power flows on the grid are recognized; however, the dispatching of
units for reactive power
flow is not a generally accepted practice.
Constrained Dispatch of Generation
Figure 5 indicates that the units available for dispatch are selected by the
economic dispatch (ED ) program.
When units are dispatched to a new generation level there are a number of equality
and inequality constraints
that can be applied to the control problem. Equation (1) illustrates the simple
equality constraint that gener-
ation must equal load. However, this simple constraint is complicated in reality
when consideration is given
to the fact that impedance of the transmission and distribution network are between
the generation and load.
Therefore, there are real power losses associated with the transmission network
that must be added into the
----------------------- Page 426-----------------------
Fig. 5. Block diagram of a supplementary control

system.
load term as
or
There are also inequality terms such as minimum and maximum power outputs for the
individual plants.
Figure 2 also shows that each plant has a fuel cost associated with it (F i ).
Therefore, it is possible to apply the
principles of optimal control theory to this problem and develop an optimal
dispatch level for the units subject
to some defined constraint. Economic dispatch is the optimal control paradigm for
finding the power output
levels for multiple units while minimizing cost. This is described by two equations
known as the coordination
equations. One of them is Eq. (4) and the other is
There are obviously many constraints that could be used in finding the optimal
dispatch for a power system
(e.g., environmental and hydro-thermal). References 3, 4, 9, and 10 are good
sources.
----------------------- Page 427-----------------------
Power Pools and Energy Management Systems
Transactions between interconnected utilities are simplified and optimized

when the utilities belong to a
power pool. Typically, a power pool central dispatch office would be responsible for
setting up the interchanges
between the members and for other administrative tasks. Power pools range
from tight pools with strong
central authority to loose pools with very little central coordination. The benefits
and costs of pool operation
vary widely depending upon the specifics of how the pool is set up. Advantages of
belonging to a power pool
include minimizing operating costs, performing a system-wide unit
commitment, minimizing the reserves
throughout the system, coordinating maintenance scheduling to minimize costs and
maximize reliability, and
maximize the benefits of emergency procedures (4). The disadvantages of belonging to
a power pool include
having to relinquish the right to engage in independent transactions outside of the
pool, having to operate
under the increased complexity added by the pool agreement and by other regulatory
agencies, and paying
operating and investment costs associated with the central dispatch office.
Power pools can operate as a brokerage house that arranges transactions among
members or as a fully
staffed central office with extensive telemetry data and computational equipment to
calculate the optimum
economical dispatch. The simplest form of brokerage house arrangement is the
bulletin board scheme. In this
approach the utility members post offers to sell or buy power at regular intervals.
Members are free to contact
each other and make direct transactions between them. More elaborate schemes
require the members to sell
and buy power to and from the pool, respectively. For example, the pool can average
the sellers incremental
costs and the buyers decremental costs and then redistribute the energy based on
the average costs plus a
constant percentage fee used to compensate those systems that provided transmission
facilities.
Deregulation
As a regulated monopoly the electric utility industry has been the focus of a
variety of regulators at all levels
of government—the Federal Energy Regulatory Commission (FERC ), the Rural
Electrification Administration
(REA ), state public utility commissions, and even itself in the form of NERC.
However, over the past two decades,
in an attempt to introduce competitiveness into the electric utility industry, the
US Congress and FERC have
introduced a series of legislation and rules. This legislation and rule-making was
aimed at opening the customer
base within a utility’s franchise area to competition. The Congress, in the Public
Utility Regulatory Policies
Act of 1978 (PURPA ), began this attempt by confirming that the electric industry is
an interstate industry
subject to federal control in almost all dimensions of its regulated activities
(11). The next major legislation
was the Energy Policy Act of 1992. This act laid the legal groundwork for
competitive wholesale markets and
led to what has been referred to as the FERC’s Mega-NOPR (Notice of
Proposed Rule-Making). From the
Mega-NOPR, two final orders were issued in 1996.
Order Number 888 stated that it was the goal of FERC to remove impediments to
competition in the
wholesale bulk power marketplace and to bring more efficient, lower cost
power to the nation’s electricity
consumers. This was primarily to be accomplished by guaranteeing all
generating systems open access to
utility transmission networks. An important aspect of the order was that the
reliability of the interconnected
system was not to be compromised by the wholesale wheeling of power. To
ensure this did not occur, the
transmission provider was permitted to charge an access tariff to the
transmission network that included
ancillary services needed to maintain reliability and power quality at the
transmission level.
Order Number 889 contained rules for establishing and governing an Open
Access Same-time Informa-
tion System (OASIS). This order requires each public utility that owns, controls,
or operates facilities used for
the transmission of electric energy in interstate commerce to create or participate
in an OASIS. The OASIS
will provide open access transmission customers and potential open access
transmission customers with in-
formation, provided by electronic means, about available transmission capacity,
prices, and other information
----------------------- Page 428-----------------------
that will enable them to obtain open access nondiscriminatory transmission service.
The latest information on
FERC orders may be found at the NERC homepage (http://www.nerc.com/).
Facts
As the electric power transmission network grows in size and complexity the system
response is more difficult to
predict and control. The problem is aggravated as an increasing number of
independent power producers (IPP)
and non-utility generators (NUG) tie into the transmission system, and the
seemingly inevitable ruling of "open
access of transmission systems" by the Federal Government takes effect. A method of
accurate, dependable
control of the transmission system is needed if utilities are to provide reliable
service in the future. To answer
this challenge, the Electric Power Research Institute (EPRI) has launched a
research project called Flexible
AC Transmission Systems (FACTS).
First put forward in 1988, the FACTS program would utilize advances in
silicon science to develop FACTS
devices that can allow for compensation and control of system parameters much
faster and more accurately
than the mechanical versions in place today. These devices make use of thyristors
as solid state switches to
replace the slower, bulkier mechanical switches. FACTS devices would prove very
useful in controlling the three
parameters that determine the flow of power in a transmission system: the line
impedance, the magnitude of
the bus voltages, and the phase angle between the bus voltages.
The ideal result of the FACTS program is a power transmission system that
employs many different
FACTS compensators to eliminate problems such as loop flows, inefficient use of lines
to thermal limits, and
poor voltage regulation; all while improving power quality and system stability to
transients and disturbances.
Loop Flows. One problem that the FACTS program deals with is that of
loop flow on power systems.
Loop flow is defined as the difference between scheduled and actual power flows on the
interface between two
systems. It is the flow of scheduled power from a generation area to a load through
a transmission path where
it is not intended. In other words, one utility ends up using another utility’s
system to transmit power without
permission or payment. Loop flow occurs when there are many parallel paths from the
generation site to the
load center. The problem of loop flow from one utility to another is not currently
addressed except by after the
fact examination of scheduled transactions versus system conditions to see if in
fact a loop flow situation was
present during a given period. This is done because of the limited number of
transactions that occur between
utilities and the reasonably rare occurrence of serious loop flows that accompany
them. However, when open
access of transmission systems becomes a reality, the number of transactions on the
system, and therefore the
number of opportunities for loop flow to occur, will increase.
The problem of loop flows may be resolved by using FACTS devices to control
the power flow between
the possible paths to the load. The phase shifter appears to be the most effective
compensator to avoid loop
flows. The phase shifter is a device used to inject a series voltage into the
transmission line at a controllable
phase angle. This variable voltage source is used to inject a voltage of the
appropriate magnitude and phase
to improve the power transfer capability and the transient stability. Phase
shifters can be used to decrease
the flow of power in the nonscheduled path and thereby increase the flow in the
scheduled primary path.
Conventional phase shifters based on series transformers with mechanical tap
changers are already in use
in power systems but their operation is much slower in comparison to the expected
response of FACTS-based
phase shifters. The thyristor controlled phase shifter is expected to be able to
control the phase angle quickly
enough to increase the power transfer before the natural power swing occurs.
Although FACTS devices do have the potential to allow much finer control of
power flows on the network
than is currently available, they have not yet been widely utilized.
This is largely due to their cost and
unfamiliarity to system design and operations personnel. Another problem with FACTS
devices is that there
is no widely recognized methodology available at this time to determine
in real time how they should be
coordinated on a regional basis to optimize the network. In fact, there is no
consensus as to what such an
----------------------- Page 429-----------------------
POWER
SYSTEM CONTROL 9
overall optimization function should be for the network. Much additional work needs
to be done in this area
before widespread use of FACTS devices will become evident.
Line Loading to Thermal Limits. One application of series capacitors
is to use them to adjust line
impedance to make full use of a line’s thermal limits. For example, suppose there
are two parallel transmission
lines that compose the primary scheduled power path, or corridor, to a given load
area. One of the lines is
loaded to its thermal limits while the other has spare capacity. The power transfer
of the two lines cannot be
increased without exceeding the limit of the loaded line. Thyristor switched series
capacitors may be used to
decrease the impedance of the less loaded line, increasing the power flow on the
line. In this manner the total
power transfer over the two lines may be increased until both lines are at thermal
limits. Phase shifters could
also be used to control the power flow between the two lines by changing the phase
angle. However, series
capacitors appear to be better suited for this case because they serve the dual
purpose of controlling power
flow and generating substantial amounts of vars required by long transmission line
corridors operating near
thermal limits. Phase shifters, on the other hand, do not supply vars but consume
them (which in turn must
be supplied by additional shunt compensation). The thyristor switched series
capacitors, then, are the more
cost-effective choice. However, it must be remembered that NERC requires utilities
to operate their systems
such that all lines will be below their thermal limit after the first contingency.
The use of series capacitors to change the effective impedance of a
transmission line brings up a topic in
need of much further research—the effect of FACTS on protection schemes. The
present relaying and protection
schemes would certainly have to be modified to accommodate the presence of the
system changing FACTS
devices. One possible solution would make use of intelligent protection schemes
capable of receiving information
from the FACTS controllers about the status of the system and adapting the relay
settings accordingly.
Voltage Regulation on Long Lines. The static var compensator (SVC) is
a thyristor controlled shunt
device primarily used to raise the voltage on long transmission lines. The SVC
supports the voltage on these
lines and, therefore, increases the power transfer. Studies have been done to
determine the most effective
placement of SVCs on a long line. The studies have shown that midpoint compensation
has the most positive
effect on voltage regulation and power transfer with the least negative effects on
system dynamic performance
(12). Although industry practice for voltage control is still dominated by switched
capacitor banks, SVCs have
shown potential for more widespread use in the future.
Emerging Power Control Issues
Power systems are equipped with numerous voltage controlling devices to ensure the
system’s voltage profile
remains within predetermined limits for a variety of topological and loading
conditions. As system load con-
tinues to increase without the construction of new sources of generation, the fear
of a voltage collapse also
increases. To aid in controlling the voltage profile of systems it will become more
important in the future to not
only have AGC control active power flow, but also reactive power flow.
Federal legislation will continue to impact the control of generation units
primarily from an environmental
viewpoint. Environmental dispatch will take on more of a supervisory role to
classical AGC to ensure that the
proper mix of units are operating to maintain a utility within environmental
regulations (clean air and clean
water) and such that the most judicious use of allowances is made. However, some
federal legislation that
mandates the use of alternative fuels for US Government automobile fleets and
alternative fuel suppliers may
have the effect of load leveling (due to the use of chargers at night) that the
utility industry has sought for
many years.
Demand side management (DMS ) is another tool to assist in the balancing of
load and generation. A smart
DMS administered by AGC is an issue that needs to be further explored. This would
permit AGC the option
to drop load at the distribution level, in lieu of requesting additional expensive
(economic or environmentally)
generation, if it determines this is the best economical solution. Traditional DSM
programs can also be rapidly
----------------------- Page 430-----------------------
supplemented with the use of market-driven programs such as real time pricing
(RTP ). Incorporating RTP into
the unit commitment and AGC programs is another interesting problem.
The control of distributed generation is another emerging issue. As
additional nonutility generators are
added to the system (at both the transmission and distribution level) AGC will need
to be able to properly
dispatch utility owned units to meet load, provide voltage and frequency support,
and overall system security.
This will become a highly complex communication, control, and protection problem.
In today’s rapidly changing electric utility industry, mergers are becoming
commonplace, energy futures
are being traded on the commodity exchange, independent power producers (IPPs) are
coming on-line wherever
the market looks promising, and power marketers are acting as the go-between for
those with energy to sell and
with those looking to buy. As these things transpire the traditional utility is
looking to realign itself to remain
competitive in this new market. To accomplish this, most utilities are
restructuring. Separate generation
companies, transmission companies, and distribution companies will become more
prevalent in the future. The
distribution company (i.e., load) will drive the market by seeking low-cost sources
of energy from a variety of
sources. This energy may be delivered via one or more transmission companies or it
may be generated locally
at the distribution level (therefore, avoiding transmission charges). The idea of
building generation at the
distribution level is attracting several entrepreneurs. This dispersed or
distributed generation, as it is referred
to, is normally smaller capacity generation, but it has the potential of becoming
very important in system
reliability.
Even the control structures are presently being debated. One scenario is a
regional control area (RCA)
consisting of a regional transmission group (RTG ) and an associated
regional system operator (RSO ) in a
control area whose boundaries may be based on existing NERC regions. The RSO, which
may be nonutility
owned (an independent system operator or ISO ), would perform some or all of the
functions of the existing pool
control centers, depending on specific implementation. The primary function of the
RSO would be to ensure
comparable access and fair play among all transmission system stakeholders while
maintaining the integrity
and reliability of the network.
It is also important to consider the human side of power system operations.
In the past, system operators
were highly experienced and had intimate knowledge about the flow patterns and
operational characteristics of
their system. The operators of today, however, are confronted with significantly
more complex systems charac-
terized by rapid changes in stakeholders and flow patterns. Traditional
relationships between stakeholders are
changing and new stakeholders are being added at a breathtaking pace. For example,
pool control center oper-
ators have traditionally been generation and transaction coordinators. They match
power needs with available
generation. This matching has been based primarily on cost and often ignored the
use of security contingency
tools (or utilized precalculated limits from transmission or operational planning
studies) in decision making.
These operators typically handled 10 to 15 bulk power transactions per hour and
their primary concerns were
those of pricing mechanisms and peculiarities of the generation resources and
transactions.
However, now system operators must be prepared to operate in a more complex
system, highly impacted
(if not actively constrained) by the complexities of transmission system loading,
open transmission access, IPPs,
wholesale marketers, OASIS (real-time information networks), etc. In such a system
most, if not all, generation
control and transaction scheduling decisions will have to be screened for potential
system reliability impacts.
These will include both the traditional static constraints (thermal and
voltage limits) as well as dynamic
impacts (generation ramp rates and stability) and impacts on import and export
capabilities.
BIBLIOGRAPHY
1. Epic Engineering, Inc., Impacts of Governor Response Changes on the

Security of North American Interconnections ,
EPRI TR-101080, Project 2473-53, EPRI Final Report, October 1992.
2. A. R. Bergen Power System Analysis , Englewood Cliffs, NJ: Prentice-Hall,
1986.
----------------------- Page 431-----------------------
3. A. S. Debs Modern Power Systems Control and Operation , Boston: Kluwer, 1988.
4. A. J. Wood B. F. Wollenberg Power Generation, Operation, & Control , 2nd ed.,
New York: Wiley, 1996.
5. NERC, Glossary of Terms, http://www.nerc.com/glossary/index.html, August 1996.
6. N. Jaleeli et al. Understanding automatic generation control, IEEE Trans.

Power Syst. , 7: 1106–1122, 1992.
7. R. K. Green Transformed automatic generation control, IEEE Trans. Power
Syst. , 11: 1799–1804, 1996.
8. R. L. King R. Luck Intelligent Control Concepts for Automatic Generation
Control of Electric Power Systems , EPRI
Project 8030-12, EPRI Final Report, September 1996.
9. J. H. Talaq Ferial El-Hawary M. E. El-Hawary A summary of
environmental/economic dispatch algorithms, IEEE
Trans. Power Syst., 9: 1508–1516, 1994.
10. Application of Optimization Methods for Economy/Security Functions in
Power System Operations , IEEE Tutorial
Course 90EH0328-5-PWR, Piscataway, NJ: IEEE Press, 1990.
11. C. G. Stalon Increasing competition in the electric power industry, in
Regulating Regional Power Systems: Case Studies
and Perspectives on Emerging Competition, C. J. Edwards (ed.), Piscataway, NJ:
IEEE Press, 1995, p. 28.
12. Flexible ac Transmission Systems (FACTS): Scoping Study , EPRI EL-6943 Vol. 2
Part 1, Prepared by General Electric
Company, 1991.
ROGER L. KING
Mississippi State University
ROGELIO LUCK
Mississippi State University
----------------------- Page 432-----------------------

} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Primary Cells
Standard Article

Terrill B. Atwater1 and Alvin J. Salkind2
1US Army Communications Electronics Command,
2Rutgers, the State University of New Jersey,

Search
reserved.
DOI: 10.1002/047134608X.W3032
❍ Acronym Finder
Abstract
Electrochemical Thermodynamics
Electrochemical Kinetics
Impedance Spectroscopy and Electrochemical Modeling
Cell Design Characteristics
Performance Characteristics of Battery Systems

%20Conversion/W3032.htm17.06.2008 16:44:25
----------------------- Page 433-----------------------

c
PRIMARY CELLS
Batteries are reactors that convert chemical energy into electric energy
on demand. Battery reactions are
oxidation–reduction reactions similar to those that occur in ordinary chemical
reactors. However, in an electro-
chemical cell, the basic component of a battery, the charge-transfer aspect of the
oxidation–reduction reaction
occurs on a surface rather than in the bulk. The electron transfer is isolated and
is available as a direct current.
The reduction and oxidation intermediates are transported through electrolytically
conductive media. The elec-
trode referred to as an anode is the site of the oxidation reaction. The reduction
reaction of the electrochemical
couple occurs at the electrode referred to as the cathode. Oxidation is a chemical
process in which an electron
is given up by the reactant. Reduction is a chemical process in which an electron
is accepted by the reactant.
The oxidation–reduction reaction proceeds when the exchange electron from the
reaction is conducted from
one electrode to the other through an external wire (or electrically conducting
circuit) and intermediate ions
formed in the process are internally transported through the electrolyte to
complete the reaction. The current
of an electrochemical cell results from this chemical reaction in which electrons
are released at one electrode
and taken up at the other. Equations (1) and (2) show the oxidation half-cell
reaction and the reduction half-cell
reaction. Equation (3) shows the overall cell reaction:
where M is the species that will be oxidized, R is the species that will be
reduced, e − represents an electron,
and − and + denote the charge of the species.

Traditionally, the simplest electrochemical cells have a pair of electrodes
(an anode and a cathode), or
more, submerged in a common electrolyte separated by a porous mechanical
separator. The electrolyte is
required to be an ionically conducting (usually a solution of a solvent and a
solute), electrically nonconductive
and chemically stable. The separator must be an electrical insulator and
mechanically and chemically stable as
well as porous to the electrolyte. The porosity of the separator allows for the
continuum of the electrolyte and a
transport path for the ions of the reaction. An electrically conductive wire
provides the path for the exchanged
electrons. It is also possible to have solid electrolytes.
In ordinary bulk chemical reactions, an oxidizer and fuel react directly with
nonisolated electron transfer
to form the reaction products and release and absorb energy as heat. In
electrochemical reactions the reaction
----------------------- Page 434-----------------------
2 PRIMARY CELLS
Fig. 1. Schematic diagram of an electrochemical cell. In electrochemical

reactions, in which the oxidizer and fuel are
isolated, energy is manifested as a voltage and current flow through a wire
(conducting lead). Bulk chemical reactions, with
nonisolated electron transfer that forms the reaction products, release and absorb
energy as heat.
is manifested as a voltage and current flow through a wire (conducting lead). Figure
1 depicted schematically
the energy transfer in electrochemical reactions. In generating electric current,
the electrons involved in the
electrochemical reaction are transferred from the active materials undergoing
oxidation to the oxidizing agent
by means of an external circuit. The passage of electrons through the external
circuit provide a direct means for
utilizing energy without heat as an intermediate step. Figure 2 gives a graphical
depiction of a cell or battery
(1).
Electrochemical Thermodynamics
An electrochemical cell keeps the oxidation and reduction reactions

separate with the exchanged electron
traveling through a wire and ionic species transported through an ionically
conducting electrolyte. The sepa-
ration of the reaction allows for work to be performed by the electrons as they
travel through the wire. The
electromotive force (emf) is a measure of energy available to perform such work.
This energy is a function of
the equilibrium constant for the chemical reaction of the cell and the activities
of the reactants and products.
The electromotive force itself is measured by the difference in electric potential
between the electrode active
materials and can be related to fundamental thermodynamic properties (relating to
the energy of formation
and solvation of the materials involved).
The electric potential of an electrochemical cell reaction is a function of
an isothermal formation process
that does not involve pressure–volume work. Therefore the work performed in a
oxidation–reduction reaction
(the quantity of electric charge transported through a potential difference) can be
expressed as a change in
Gibbs energy.
----------------------- Page 435-----------------------
PRIMARY CELLS 3
Fig. 2. Graphical representation of an electrochemical cell. In generating

electric current, the electrons involved in the
electrochemical reaction are transferred from the active materials undergoing
oxidation to the oxidizing agent by means
of an external circuit. The passage of electrons through the external circuit
provides a direct means for utilizing energy
without heat as an intermediate step.
Equation (4) shows the relation between the change in Gibbs energy and the
electromotive force.
where #G is the change in Gibbs energy (J/mol), n is the stoichiometry of the

electron exchange, F is the
Faraday constant (96,487 C/equivale mol), and E is the electrochemical potential
(V).
The electrochemical potential for each group of chemicals comprising a
battery system is unique. However,
it is extremely important in the design and use of batteries to realize that the
reaction thermodynamics depends
on the surface chemistry and not the bulk chemistry of the reacting species.
In considering the free-energy relationship for bulk chemical reactions an
alternative representation of
the change in the Gibbs energy is shown as
where # G0 is the standard molar Gibbs energy, R is the gas constant (8.31433
J/mol·K), a is the activities of
the reactants (A, B) and products (C, D), a and b are the stoichiometry of the
reactants A and B, and c and d
are the stoichiometry of the products C and D.
Equation (6) is the expression of the relation between the standard molar
Gibbs energy and the standard
electrode potential:
where E 0 is the standard electrochemical potential of the reaction. Substituting

Eqs. (4) and (6) results in
Eq. (7), relating the electrochemical potential to the standard electrochemical

potential and the reactant and
products activities. The relation shown in Eq. (7) is called the Nernst equation:
----------------------- Page 436-----------------------
4 PRIMARY CELLS
The value of E 0 , the reversible potential, is determined by the atomic structure

of the reactants, especially
the energy levels related to the work function and the valence electrons. The value
of E 0 can be predicted
by summing the enthalpies of atomization, ionization, and solvation in a Born–Haber

cycle as discussed by
Sanderson 2. It is important to realize that only the surface composition of the
electrode in contact with the
electrolyte should be considered in calculating E 0 . Material buried deep in an
electrode completely covered
with active material, and without direct electronic contact to the current
collector, has no effect on E 0 and
measured values of cell voltage. When the reactants and product of an electrode
reaction share a single phase
or crystal habitat or form an intercalate, the voltage of that electrode depends on
the ratio of concentration
of the species. This results in a sloping discharge curve that is typical of cells
with MnO , TiS , MoS , etc.,
2 2 2
electrodes. When the reactants and product of an electrode reaction form separate
phases, the voltage of the
electrode is constant during discharge in the multiphase region. Typical
examples are cells with Hg/HgO,
Cd/Cd(OH) , and PbO /PbSO electrodes. These are often used as reference
electrodes in experimental studies
2 2 4
because of their invariant voltages.
Equations (8) and (9) are fundamental thermodynamic equations relating
entropy with the Gibbs energy
and enthalpy with entropy and Gibbs energy:
where T is the temperature (K), P denotes constant pressure, #S is the change in

entropy (J/mol·K), and #H
is the change in enthalpy (J/mol).
From Eqs. (4), (8), and (9) the change in entropy and the change in enthalpy
can be expressed in terms
of the electrochemical potential. Equations (10) and (11) show the introduction of
Eq. (4) into Eq. (8) and then
into Eq. (9). They define fundamental thermodynamic properties of the Gibbs energy,
entropy, and enthalpy in
terms of the electrochemical potential:
By measuring the voltage of a cell as a function of temperature, one can

derive the entropy and enthalpy
of the electrochemical couple by using Eqs. (10) and (11), respectively. Table 1 is
a listing for the standard
electrochemical reduction potentials for some common electrochemical half-cell
reactions. The standard elec-
+ − 1
trode potentials are determined with respect to the standard electrode potential of
hydrogen, H + e → 2
H2 at 1 atm, which has been assigned by international convention a zero value.
(The basis for the standard
hydrogen electrode potential is a weak acid.) An electrode that is more negative
than the hydrogen electrode
----------------------- Page 437-----------------------
PRIMARY CELLS 5
will have a negative standard potential, and an electrode that is more positive
than the hydrogen electrode will
have a positive standard potential. An electrochemical cell comprising two half-
cell reactions will result in one
electrode being reduced and the other oxidized. The electrode that has a more
negative standard potential will
be oxidized and the electrode with the more positive standard potential will be
reduced. This type of analysis
allows for the determination of the cell reaction.
The combination of standard electrode potentials for two half-cell reactions
can be used to determine the
cell voltage. Equation (12) shows the relation between the standard potential for
oxidation and the standard
potential for reduction, and Eq. (13) shows how the cell voltage can be determined
from the standard electrode
potential data.
Standard electrode potentials can be determined from the standard Gibbs

energy. Equation (14) shows
this relationship, which is a form of Eq. (10):
Equation (15) shows the thermodynamic equilibrium constant defined as a

function of the activities of the
reactants and products:
where K is the thermodynamic equilibrium constant. Equation (16) shows

the relation between chemical
potential and the thermodynamic equilibrium constant. For reactants and products in
their standard state,
the standard molar Gibbs energy is equal to the chemical potential. Equation (17)
shows the relation between
the Gibbs energy and the chemical potential.
----------------------- Page 438-----------------------
6 PRIMARY CELLS
where µ is the chemical potential. Combining Eqs. (16) and (17) gives Eq. (18):
Combining Eqs. (10) and (18) as shown in Eq. (10) gives the relation between
the standard electrochemical
potential and the thermodynamic equilibrium constant. The thermodynamic equilibrium
constant K is defined
in Eq. (20) by rearranging this equation:
This thermodynamic analysis of the electrochemical cell shows how

the cell potential and standard
electrochemical potential are related to fundamental thermodynamic properties. This
analysis also shows how
electrochemical measurements can be used to determine fundamental thermodynamic
properties.
Electrochemical Kinetics
The reaction rates of electrochemical reactions follow the usual kinetic laws. The
rate of reaction is expressed
as a current flow in coulombs per second (amperes) rather than a change
in concentration per unit time.
Operational battery systems exhibit a kinetic hindrance when producing current,
which manifests itself as a
loss in voltage. Thus the voltage during use (discharge) is lower than the voltage
measured under open-circuit
(reversible) conditions. This lowering of voltage is called polarization and is
usually represented by η. There are
three types of polarization: (1) activation or reaction rate polarization, (2)
concentration polarization, and (3)
resistance or ohmic polarization. Activation polarization is related to the charge-
transfer mechanism process,
concentration polarization is related to the availability of reacting species at
the electrode–solution interface
(often a diffusion- or mass-transfer-controlled process usually in the electrolyte
but occasionally in the solid
crystalline lattice of the active material), and resistance polarization
is related to the ohmic hindrance to
electron or ion flow. The same kinetic factors take place during the recharging of
secondary cells, and the
energy loss is reflected as higher voltage necessary to recharge.
In almost all cases, the positive electrode in a battery system is a metal
oxide, sulfide, or halide. In general,
these are relatively poor electric conductors and would give extremely high ohmic
polarization (impedance) if
not blended or reinforced with supporting conductors. An exception is pyrite
(FeS2 ), which is a semiconductor.
Conductors added to positive electrodes include graphite, lead, silver, copper,
nickel, and other materials in
rod, mesh, wire, grid, and other shaped configurations. In almost all cases the
negative electrode is a metallic
element of sufficient conductivity to require only minimal supporting conditions. An
exception is the hydrogen
electrode, which requires a substantial conductive surface support, which also
serves for electron transfer.
----------------------- Page 439-----------------------
PRIMARY CELLS 7
Fig. 3. Current–voltage curve of an electrochemical cell, showing polarization

regions. This plot shows cell voltage versus
discharge current and is known as a power curve. This curve shows the regions
controlled by activation, concentration, and
resistance polarization and the change as the battery is discharged.
The thermodynamics of electrochemical reactions does not provide information

regarding the cell when
work is being performed. The complex mechanisms of electrochemical cell reactions
require thorough analysis.
The voltage of the electrochemical cell, when current is flowing, can be represented
by an empirical equation.
At low reaction rates the polarization in battery reactions is almost always
controlled by activation energy
(charge transfer). This is illustrated in Fig. 3. Using reaction-rate theory it can
be shown that the activation-
energy barrier can bdisplaced by application of a potential for a net current flow.
Equation (21) shows the
relation among the current density, the exchange current density, and the
charge-transfer coefficient. The
exchange current density is the reversible current density when the electrochemical
reaction is in equilibrium.
2
2
where i is the current density (mA/cm ), i is the exchange current density
(mA/cm ), α is the charge-transfer
p 0
coefficient, and η is the overpotential (V). At equilibrium the net current i

= 0 and the overpotential η =
p
0 and the rate of the forward and reverse reactions are equal. The forward and
reverse action rates can be
expressed as a current flow i0 across the activation energy barrier at equilibrium
(reversible potential) and is
termed exchange current. The exchange current is directly related to the reaction-
rate constant, and the usual
kinetic expressions apply directly to battery reactions. A low rate constant is
associated with a high degree of
irreversibility, so that the smaller the i0 the more irreversible the reaction and
hence the greater the activation
polarization for a given net current flow. This implies that high exchange currents
are favored for battery
reactions. However, high exchange currents often lead to changes in electrode
morphology with a consequent
and sometimes deleterious change in surface structure.
When the electrode reaction occurs with polarization larger than
about 0.05 V, then Eq. (21) can be
simplified. Equation (21) reduces to Eq. (22) at large positive overpotentials and
reduces to Eq. (23) at large
negative overpotentials.
----------------------- Page 440-----------------------
8 PRIMARY CELLS
Equations (24) and (25) are the result of taking the logarithm of Eqs. (22)
and (23), respectively:
Equation (26) relates the voltage deviation of a cell from the rest potential
caused by a pulse of current
(this also depends on surface area of the electrode and the pulse should be defined
in terms of A/cm2 ). This
relation is referred to as the Tafel relation.
where a is an empirical constant and b is a constant referred to as the Tafel

slope.
From a comparison of Eqs. (26) and (24), the constants of the Tafel relation,
a and b, can be determined for
large positive overpotentials: a = −RT/α F ln (i0 ) and b = RT/αF . From
comparison of Eqs. (26) and (25), the
constants of the Tafel relation can be determined for large negative
overpotentials: a = [RT/(1 − α) F ] ln(i0 )
andb = −RT/(1 − α) F . Equation (27) shows the Tafel relation, which defines the
discharge current density of
a cell as a function of overpotential and exchange current density:
The Tafel relation holds true for most of the electrochemical reactions over a wide
range of overpotentials.
At low overpotentials the relationship fails to fit the data. By plotting the data
of overpotential against the
logarithm of current, a linear relation is clear at large overpotentials. The slope
of the line is the Tafel slope
and the x-axis intercept, after extrapolating the curve, is the exchange current
density. Figure 4 shows a Tafel
plot demonstrating the relationship (3).
Whenever the local concentration of a reacting component in a battery is low
or nears exhaustion, the rate
of reaction becomes controlled by the transport of that component to the reaction
surface site on the surface
or within the electrode. At this point diffusion equations, for example, Fick’s
laws, apply. Equation (28) shows
this relation:
where il is the limiting current density, D is the diffusion coefficient, [A ]B is

the concentration of the diffusing
species in the bulk solution, δ is the thickness of the diffusion layer, t± is
the transference number of the
reacting species. The relationship between concentration polarization and limiting
current density is
----------------------- Page 441-----------------------
PRIMARY CELLS 9
Fig. 4. Tafel plots for zinc and zinc–mercury alloys in

33% KOH.
Figure 3 shows the performance of a typical cell. This plot, showing cell
voltage versus discharge current,
is known as a power curve. This curve shows the regions controlled by activation,
concentration, and resistance
polarization and the change as the battery is discharged (1).
Equation (30) shows the relation between discharge capacity and the mass of
the electroactive species.
Equation (30) is a form of Faraday’s law, and Eqs. (31) and (32) define Q and Gew .
where Q is the discharge capacity (C), W is the mass of the electroactive species,
and Gew is the gram equivalent
mass of the electroactive species.
where MW is the molecular weight of the electroactive species and n is the

stoichiometry of the electrons
involved in the reaction.
where i is the discharge current (A) and t is time(s).
Rearranging Eq. (30), we arrive at Eq. (33). Equation (33) shows the reaction
rate in mass per unit time
as a function of discharge current and the gram equivalent mass of the reacting
species:
The activation energy and reduced diffusion are the major contributions to
the reduced voltage during
discharge at lower temperatures. Thus, the energy must be initially introduced to a
system in order for the
reaction to proceed. This energy is called the activation energy. Equation (34)
shows a chemical reaction and
----------------------- Page 442-----------------------
10 PRIMARY CELLS
its activated state:
where A are the reactants, B are the products, and X ∗ is the activated species.
The activated species is a molecule in which a bond is in the process of
being formed or broken. Equation
(35) shows the rate constant defined as a function of both the activation energy and
temperature:
where K is the rate constant, C is a constant, and Ea is the activation energy

(J/mol). In an electrochemical
reaction, Ea appears as a voltage drop due to cell reactions. Catalysts are often
employed to alter the speed of
a reaction. Catalysts provide an alternative path for the reaction but remain
unchanged after the completion
of the reaction. This alternative path typically has a lower activation energy and
thus proceeds at a faster
rate. The other source of voltage drop during discharge is diffusion. Diffusion is
a function of viscosity and
temperature, as shown in Eq. (36). It defines the diffusion coefficient of a
noncharged sphere as a function of
temperature and viscosity. Viscosity is an activated process describing the
tendency of a liquid to flow defined
by Eq. (37).
2
23
where D is the diffusion coefficient(cm /s), N 0 is Avogadro’s
number(6.02252 × 10 molecules/mol), r is the
radius of the sphere (cm), and n is the diffusing media viscosity (P).
where A is constant and Ev is the activation energy for viscosity.

During discharge, diffusion introduces a resistive component to the reaction,
lowering the voltage of the
cell due to IR losses. Another property of an electrochemical cell that adds a
resistive component is conductivity.
Conductivity for electric systems is the measure of the resistance that a material
exerts on the flow of current.
Ohm’s law gives us the relation shown in Eq. (38). Equation (39) is the identity
defining conductance as a
function of resistance. Equation (40) shows the relation among resistance,
conductance, and the dimensional
properties of the material. The resistivity of a material is the proportionality
constant for Eq. (40), and the
conductivity of a material is the reciprocal of the proportionality constant.
where E is the electromotive force (V), i is the current (A), and R is the
resistance(#).
where L is the conductance.
Techniques that are used for studying electrochemical reactions in non-

steady-state conditions include (1)
sweep voltammetry, (2) chronopotentiometry, (3) chronoamperometry, (4) pulse
voltammetry, and (5) impedance
----------------------- Page 443-----------------------
PRIMARY CELLS 11
spectroscopy. These techniques provide data regarding the processes governing and
limiting electrochemical
cells.
Sweep voltammetry is a technique in which a linear voltage sweep
is imposed across a cell and the
resultant current is observed. Cyclic voltammetry is a form of sweep voltammetry in
which the voltage scan
is reversed after the initial sweep. Qualitative and limited quantitative electrode
data can be obtained from
sweep voltammetry. Reversibility of an electrode, voltage stability of an
interface, charge transfer versus
diffusion limitations, rate constants, and reaction mechanisms are examples of
physical properties that can be
determined by sweep voltammetry. The data for evaluating the Tafel relation for a
cell can also be obtained
through sweep voltammetry.
Chronopotentiometry is a technique that studies voltage transients at
an electrode during constant
current discharge. Chronoamperometry is a technique that studies current transients
at an electrode when
a potential change is imposed across the electrode. Chronoamperometry can
aid in determining diffusion
coefficients and for investigating kinetic mechanisms. Pulse voltammetry is similar
to chronoamperometry,
but the voltage input to the electrode being studied is in the form of a square
wave. This allows for many
variations depending on the setup of the wave amplitude and sign. This includes
reverse current through
positive and negative amplitudes.
Impedance spectroscopy is a technique in which a cell is perturbed from
steady state by a frequency
sweep of alternating current of low magnitude. The potential of the cell due to the
signal gives insight to the
electrochemical processes at the interface. Impedance spectroscopy will be
discussed in more detail in the next
section.
Impedance Spectroscopy and Electrochemical Modeling
Impedance spectroscopy is a technique for studying electrode processes directly at

the interface. The interface
between electrode and electrolyte when subjected to an ac signal sweep provides an
output signal that can
be modeled as an electronic equivalent circuit. Impedance spectroscopy measures the
change in the electrical
impedance and phase of a system by analyzing the response to a sweep alternating
signal (ac) input. This
response to the frequency sweep for electrochemical systems relates to
the electrode–electrolyte interface.
In order to understand the significance of this response, it is necessary
to establish an equivalent circuit
with passive and reactive elements that represent the dynamic characteristics of
the interface. The elements
considered for this equivalent circuit is resistive, capacitive, and inductive.
Table 2 shows the ac impedance of
these elements.
When evaluating impedance spectroscopy data, it is important to note the sign
of the imaginary portion
of the elements. This allows for proper selection of elements to represent the
electrochemical system. Nearly
all electrochemical reactions exhibit negative imaginary impedance. Since
inductance has a positive imaginary
part, typically this element will not be considered. Inductance in its pure form
can appear in data, but this,
in most cases, can be traced to the wire leads and cell geometry in the setup.
Equipment setup can introduce
stray capacitance. This is typically due to pressure contacts. The last circuit
element in Table 2, a capacitor and
resistor in parallel, has both real and imaginary resistance as a function of
frequency. This circuit constitutes
the building blocks for equivalent-circuit analysis. Figure 5 is a simple
representation of an electrochemical
interface. This representation is simplified by not taking into account diffusion or
absorption phenomena.
The expression defining the various electrochemical parameters are relatively
straightforward to derive
but can be complex in format. This is especially true when absorption and capacity
effects are considered.
Typically, a network of resistive and capacitive elements can represent the
impedance behavior of electrochem-
ical processes. The building block of this equivalent circuit is represented by
Fig. 5. This network describes
a simple interface between two different phases. This equivalent circuit typically
describes the electrode and
electrolyte interface. The components represent the combined resistance of the
electrolyte and the electrode
----------------------- Page 444-----------------------
12 PRIMARY CELLS
Fig. 5. Equivalent circuit for a simple electrochemical interface. This network

describes a simple interface between two
different phases. This equivalent circuit typically describes the electrode and
electrolyte interface.
Fig. 6. Equivalent circuit for a two-electrode electrochemical cell. This network

describes a simple electrochemical cell
including the anode, cathode, and electrolyte ionic conductivity.
substrates, R , and double-layer capacitance of the interface, CDL , with its
associated polarization resistance
Rp . The resistance associated with double-layer capacitance represents charge
transfer. By linking a series of
these equivalent circuits, a complex multi-interface electrochemical system can be
described. Figure 6 shows
the basic equivalent circuit for a two-electrode electrochemical cell.
The elements for the equivalent circuit described in Fig. 6 are (1) Ra , the
resistance of the anode substrate,
(2) Cae , the double-layer capacitance of the anode–electrolyte interface, (3) Rae
, the polarization resistance of
the anode, (4) Re , the electrolyte resistance (ionic conductivity), (5) Cce ,
the double-layer capacitance of the
----------------------- Page 445-----------------------
PRIMARY CELLS 13
Fig. 7. Equivalent circuit for an electrode interface with Warburg impedance.

This network describes the equivalent
circuit representing an electrode with additional diffusion-controlled processes at
the interface.
Fig. 8. Equivalent circuit for an electrode interface with an adsorption

intermediate. This network describes the equivalent
circuit representing the addition of absorption process at an electrode.
Fig. 9. Equivalent circuit for a multilayer electrode with two interfaces. This
network describes the equivalent circuit for
a multilayer electrode.
cathode–electrolyte interface, and (6) R , the polarization resistance of

the anode, and (7)R , the resistance of
ce
c
the cathode substrate.

Elements also need to be added to the basic building block to represent
additional phenomena within the
electrochemical interfaces. Diffusion, represented by Warburg impedance, and
absorption are good examples of
phenomena that require additional elements to be added to the basic network. Figure
7 shows the equivalent
circuits representing an electrode with additional diffusion-controlled processes
at the interface. Here Ci is
the double-layer capacitance of the interface, Ri is the polarization
resistance of the interface, and W is the
Warburg impedance. The mass-transfer Warburg impedance follows the mathematical
relationship.
where σ is the Warburg diffusion coefficient.

Figure 8 shows the equivalent circuits representing the addition of
absorption process at an electrode.
Here Cc is the capacitance of the absorption layer and Rc is the
resistance of the absorption layer. Figure
9 shows the equivalent circuit for a multilayer electrode. In this equivalent
circuit, each resistor–capacitor
network represents an interface between phases. For Figs. 7 to 9, the electrolyte
resistance as well as the
electrode substrate resistance has been eliminated, thus providing an equivalent
circuit for the interface only.
----------------------- Page 446-----------------------
14 PRIMARY CELLS
Fig. 10. Prismatic cell. This cell design typically has a high power density
and low-to-moderate energy densities.
The value of the proposed equivalent circuit is in its ability to

simulate and represent the physical
properties of the electrochemical process being examined. Typically, the data
suggest an equivalent circuit that
can then be refined once the data are fitted to the model.
Cell Design Characteristics
A knowledge of current distribution and factors that affect current densities are
essential in battery design
considerations. The distribution of current on electrodes may not be uniform, even
when the electrodes are
vertically opposed and uniformly spaced. If the current distribution across the
surface and/or through the
electrode is nonuniform, undesirable effects in cell operation occur, including
nonuniform chemistry, excessive
heat evolution, and low utilization of active materials (4).
Most battery electrodes are porous structures, often with a mixture of solid
nonconducting active mass and
solid electronically conductive inert materials, filled with an electrolyte
permeating the structure. The porous
----------------------- Page 447-----------------------
PRIMARY CELLS 15
Fig. 11. Button cell. These are characterized as high energy density but
low power density, and they are simple to
manufacture.
electrode may be viewed as a macroscopically homogeneous volume of electrolyte,

inert electronic conductor,
and electroactive material. This electrode structure does t have a well-defined
planar surface but has a complex
surface extended throughout the porous electrode volume. Porous electrodes offer an
enhanced surface area for
the electrode reaction and increased mass-transfer rates because of shorter
diffusion paths. Current-producing
reactions can occur at the surface, as well as at a considerable distance inside
the surface of the electrode.
The effectiveness of a porous electrode over a plane surface electrode is given by
the product of the active
2 3
surface area A s (cm /cm ) and the penetration depth Lp of the reaction process
into the porous electrode. The
penetration depth is directly related to the diffusion coefficient and concentration
of the reactant and product,
to the exchange-current density, and to electrode thickness. When the LpA s
product is large, the performance
of the porous electrode is enhanced over that expected for a planar surface of the
same geometric surface area.
The reaction distribution in a porous electrode can be predicted, based on
Ohm’s law and the equation for
reaction kinetics and transport phenomena. Porosity, pore size, and tortuosity all
play a role. The equations
show that the ratio of the volume conductivity of the electrolyte and electrode
materials, the exchange-current
density, the diffusional characteristics of reactants and products, and
the total current, all determine the
current distribution. The nonuniform current distribution lowers efficiency and
increases heat evolution due
to lower operating voltages and to incomplete usage of the active materials inside
the porous electrode.
Cell geometry, such as terminal positioning and battery configuration, can
also influence current distri-
bution. Figures 10, 11, 12, 13 show the basic cell designs (5). The basic prismatic
cell designs are shown in
Figs. 10, 11 Figure 10 shows the prismatic plate cell. This cell design typically
has a high power density and
low-to-moderate energy densities. Figure 11 shows a variation of the prismatic
design, the button cell. Button
cells are characterized as high energy density but low power density, and they are
simple to manufacture. The
basic cylindrical cell designs, spiral wound and bobbin, are shown in Figs. 12 and
13, respectively. Spirally
wound cells characteristically provide moderate-to-high power and energy. Bobbin
cells, like button cells, are
simple to manufacture and provide high energy with low-to-moderate power
capability.
The current distribution across the face of an electrode in a battery system
is seldom uniform. In general,
the current tends to be higher on the edges of the electrode and lower in the
middle surface. This primary
current distribution is governed by geometric factors such as height (or length)
and the distance between the
electrodes, by the resistance of the anode and cathode structures, by the
resistance of the electrolyte and by the
polarization resistance or hindrance of the electrode reaction processes. More
uniform distribution is favored by
shorter, highly conductive electrode structures, by high solution resistance and
wide spacing between electrodes,
and by a high polarization resistance. These are the same considerations used for
predicting uniformity of
electroplating. The slope of the Tafel plot, b, is an indicator of the degree of
modification to primary current
----------------------- Page 448-----------------------
16 PRIMARY CELLS
Fig. 12. Spirally wound cell. These cells characteristically provide

moderate to high power and energy.
distribution by electrode polarization to produce a more uniform distribution. The

addition of polymers and
gelling agents to electrolytes is one common method of controlling current
distribution, as these increase the
slope of the Tafel plot. Most battery designers attempt to minimize high current
areas by masking surfaces
such as edges on the electrode.
Performance Characteristics of Battery Systems
The value of production of primary batteries was approximately $14

billion in 1998. The most common
portable battery systems use MnO2 as the positive electrode and
zinc as the negative in three different
´
electrolytes,ZnCl –NH Cl, ZnCl , and KOH. These electrolytes represent LeClanche,
heavy duty (zinc chlo-
2 4 2
ride), and alkaline batteries, respectively, and are used largely in flashlights,
portable electronic devices such
as radio, cassette players, and toys. They are available in a wide variety of sizes
and shapes. They have a
sloping discharge voltage, which is a function of the thermodynamics of the
manganese dioxide (MnO2 ) crystal
habitat.
The silver and mercury cells are used to power smaller electronic
devices that do not have moving
parts, such as hearing aids, cameras, and calculators. They are usually
manufactured in small, round, thin
----------------------- Page 449-----------------------
PRIMARY CELLS 17
Fig. 13. Bobbin Cell. Bobbin cells, like button cells, are simple to manufacture,
provide high energy with low-to-moderate
power capability.
configurations, often called button or miniature cells. They have a flat or constant
discharge voltage. The zinc
air cell is used for hearing aids. It has a short (one month) life after being
activated but can deliver twice the
energy of the same size silver and mercury cells. It also has a flat discharge
curve. Mercury cells have largely
been phased out, since 1990, for environmental reasons.
Lithium–copper oxide Li/CuO and lithium–iron sulfide Li/FeS systems are 1.5 V
replacements for the
mercury and silver button cells. The Li/CuO system also has potential as a long-
life, higher-energy replacement
for the Zn/MnO2 systems. They find application for which high-current pulse demands
are not required. They
have excellent shelf or storage life and operate over a very wide temperature
range. A higher-voltage cell
(∼3 V) can be made using FeS22 − with lithium.
----------------------- Page 450-----------------------
18 PRIMARY CELLS
Two lithium systems,Li/MnO and Li/CF , have 3.0 V cell

voltages and take advantage of the high
2 x
electronegativity of lithium on the EMF scale. These systems find

application in a large number of small
electronic devices such as cameras and calculators. They have exceptional shelf or
storage life and operate over
a wide temperature range. They do not have the good pulse power capability of the
zinc-based mercury, silver,
and alkaline systems.
The Li/l2 system is used primarily to power low-rate (current) cardiac
pacemakers, but some high-rate
pacemakers use other cathode materials, for example, CuS and Ag CrO . Over 95% of
the heart pacemakers use
2
4
a lithium battery to power the pacing systems. It has a 7 to 10 year life in these
applications. For higher-current
pulse devices, such as defibrillators, lithium vanadium oxide and lithium silver
vanadium oxide systems are
used.
Li/SO2 and Li/SOCl2 systems are used, almost exclusively, for military
applications. They have reasonably
good high-current capability, have a flat discharge profile, have outstanding long
shelf or storage life, and
◦ ◦
operate over a wide temperature range (−40 C to +70 C).
There are many other battery systems, not listed here, that have
been developed for specialty uses.
Thermally activated (“thermal”) batteries are one example. These cells employ an
electrolyte (typically, an alkali
halide mixture) that is solid at room temperature but that becomes very conductive
at operating temperatures
◦ ◦
of 400 C to 600 C. An internal pyrotechnic is used to bring the cell
to its operating temperature (6). The
conductivities of the molten salt electrolytes can be orders of magnitude
greater than aqueous or organic
electrolytes, which allows for large power and current densities to be attained.
Many battery chemistries are capable of a single discharge but are
incapable of having the original
chemistry restored by electric means internal to a cell. This lack of
reversibility choccurs from charges in
morphology, crystal habitat, conductivity, and inability to be chemically
formed in the battery electrolyte.
7 characterized the material irreversibility as arising from (1) irreversibility in
phase changes and chemical
interconversion, (2) isolation, electrolytically or electronically, of active
materials, and (3) local poor conductivity
of materials in the discharged state.
BIBLIOGRAPHY
1. A. J. Salkind R. J. Brodd Primary and secondary batteries,

Electrochem. Solid State Sci. Ed. , PV 87-3: 127–145,
1987.
2. R. T. Sanderson J. Chem. Education, 43: 584, 1966.
3. S. U. Falk A. J. Salkind Alkaline Storage Batteries , New York: Wiley, 1969,
p. 547.
4. R. J. Brodd Electrochem. Tech. , 6: 289–294, 1968.
5. Navy Primary and Secondary Batteries, Design and Manufacturing Guide , NAVSO
Publication P-3676, Philadelphia:
Navy Publications and Forms, 1991.
6. D. Linden (ed.) Handbook of Batteries , 2nd ed., New York: McGraw-Hill, 1995.
7. B. E. Conway Lecture notes in Rechargeable Battery Seminar, Florida Atlantic
University, March, 1986.
TERRILL B.
ATWATER
US Army
Communications Electronics Command
ALVIN J.
SALKIND
Rutgers,
the State University of New Jersey
----------------------- Page 451-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Quantum Storage Devices
Standard Article

Olin L. Hartin1 and Dean P. Neikirk2
1University of Texas at Austin, Austin, TX
2University of Texas at Austin, Austin, TX

Search
reserved.
DOI: 10.1002/047134608X.W6827
❍ Acronym Finder
Abstract
Definition of Quantum Storage
Summary

%20Conversion/W6827.htm17.06.2008 16:44:48
----------------------- Page 452-----------------------

c
QUANTUM STORAGE DEVICES
The semiconductor industry has used the ability to develop smaller and faster
devices to fuel the explosive
growth in productivity and functionality of electronic products. However, as
devices shrink below 0.1 µm in
size, physical phenomena must be identified that can produce devices that
work as well or better than the
larger devices utilized over the last 40 years. In particular, quantum devices that
work because of their small
size, rather than in spite of it, become more attractive. It is critical to explore
the potential of this new class of
ultra-small devices.
The continued trend in the integrated circuit industry toward smaller
individual devices (transistors) has
actually become part of the financial structure of the industry. There is a focus in
every area of device fabrication
toward next generation technologies that scale (shrink) in a favorable way.
Unfortunately, the resources and
technology required to decrease minimum feature size in each generation of
semiconductor devices continue
to increase. Industry is now wrestling with fundamental physical device size
limitations that threaten to limit
further scaling. This suggests that new device building blocks need to be developed
whose ideal size is in the
new operating regimes. Many quantum phenomena evidence themselves most strongly in
this sub-100 nm size
range. This gives quantum devices, which may have no intrinsic scaling
limitations, a potential role in the
future of the industry.
Definition of Quantum Storage
Quantum storage cells are composed of a device or of several devices which depend
on quantum confinement,
tunneling, or an interaction between these phenomena and space charge or other
effects. Preferably, quantum
storage cells should be composed of a single device that has multiple, electrically
distinguishable, stable states
accessed through one line and set through another. To be useful these
memory cells should have terminal
characteristics which allow them to drive variable-length interconnects with
acceptable noise margins and fan
out.
Quantum storage-based memory may be volatile or nonvolatile, static or
dynamic, as in other types of
memory. Nonvolatile memory retains its memory state when power is turned off.
Static random access memory
(SRAM) is an example of volatile static memory. Dynamic random access
memory (DRAM ) is volatile and
dynamic, requiring regular refresh cycles that consume significant amounts of power
that must be dissipated
as heat. DRAM is significantly slower than SRAM but it is often used
because of its low cost and smaller
memory cell size. There would be an enormous amount of interest in a device with
SRAM performance at cost
and densities of conventional DRAM. Quantum storage-based memory has the potential
to accomplish this.
There are two methods of achieving quantum storage. One is to use logic
devices whose operation is based
on quantum phenomena to make memory circuits. The other is to use a single device
that exhibits memory
characteristics using quantum phenomena. It is difficult to imagine a single device
that is robust enough to
perform all of the functionality of a memory cell so most cases are a hybrid of
these two methods. In either
----------------------- Page 453-----------------------
2 QUANTUM STORAGE DEVICES
case, there is a significant advantage in compatibility with established fabrication

methods and with existing
linear and digital electronics families.
Cellular Automaton. There are several examples of quantum
phenomena-based logic which may be
used to make memory. Lent and co-workers (1,2,3,4,5) have proposed a family of
logic based on coupled quantum
dots that can be used to make interconnect lines and perform basic logic functions.
These are known as quantum
cellular automata (QCA). It is tempting to design devices from low-dimensional
structures, such as quantum
dots whose charge is governed by quantization effects, particularly for digital
logic. In order to be small enough
to benefit from quantum confinement a quantum dot must be a few tens of nanometers in
diameter or smaller.
This makes it difficult to make physical contact between a quantum dot and a metal
interconnect. The QCA is
a suggested solution.
In this scheme each cell is composed of five quantum dots, as
shown in Fig. 1. Each dot is about 20
¨
nm across and 10 nm from its nearest neighbor. Interaction between cells is
governed by Schrodinger’s wave
equation
where H is the Hamiltonian, ψis the wave function solution, E is the energy
eigenvalue, E0 is the on-site energy
which is the same for all sites, ni,σ is the number operator generating an
electron at site i with spin σ, ti,j is
the tunneling energy from site i to site j , and EQ is the coulombic
energy required to have two electrons of
opposite spin on the same site. The interactions between cells are predominantly
controlled by the coulombic
perturbation between neighboring cells so that the Hamiltonian at cell 1 due to
cell 2 is given by
where
where ρ is some assumed fixed positive charge to achieve space charge neutrality.
The polarity of a cell may be
defined as
where ρn is the charge at location n. This is the numbering scheme for calculating
the polarization of each cell,
and it has values in the range of −1 to 1. The lowest energy and preferred solution
occurs when neighboring
cells align themselves with the same polarization. If an initial polarization is
supplied by a driver to a QCA
quantum interconnect, within a couple of cells alignment to a polarization of 1 or
−1 occurs and continues to
subsequent cells (1).
----------------------- Page 454-----------------------
QUANTUM STORAGE DEVICES 3
Fig. 1. Numbering scheme used to determine the polarity of a cell. In a QCA wire’s
lowest energy configuration, each cell
has the same polarity.
A range of logic circuits has been designed using QCA including fundamental
AND and OR gates as well
as more complicated circuits like a full adder. It may be dangerous to make
assumptions about the state of a
cell when the polarizing effects of the driver are removed, say to infinity. Since
work is done in reading the
state of a cell the very act of determining its state may change the polarity of a
cell or an entire QCA wire in
an unpredictable way. In order to reliably achieve memory functions, a memory cell
constructed of basic QCA
logic elements must function without making any assumptions about the volatility of
the data in each cell.
Figure 2 shows an SR latch composed of two NOR gates constructed from QCA
logic. One NOR gate is
outlined for reference. The cell size is about 0.3 µm by about 0.4 µm. This quantum
dot-based SRAM cell is
comparable in size to current conventional DRAM cells.
A more critical assessment of this logic family should, however, be
considered. For instance, the thought
experiment used to determine how the QCA operates is carried out at T = 0 K. The
devices could be tested
at 4 K but would probably have to work at least at 77 K (liquid nitrogen) to be
useful. In addition, for space
charge neutrality there must be a charge of +2q (where q is the magnitude of the
charge of an electron) on
each cell. This suggests precise placement of exactly two donor dopant
ions in each cell would be required.
Unfortunately, current doping processes always produces a distribution of dopant
atoms. Expected statistical
variations in doping would significantly change the electrical
characteristics of each cell. Finally, although
direct write methods of photolithography do have the resolution required to create
these patterns of dots, these
methods have not been used for production. At the time of this writing QCAs have
not been demonstrated in
the laboratory.
Resonant Tunneling Diode-Based Memory. Another example of a
novel family of logic that may be
used to create memory is based on resonant tunneling diodes (RTD s). RTDs are
ultrahigh frequency, generally
two terminal, majority carrier devices composed of a heterostructure quantum well
usually built from III/V
materials using MBE (molecular beam epitaxy). The quantum well is usually on the
order of 5 nm in width,
formed by sandwiching a low conduction/valence band offset material (e.g.,
GaAs or Inx Ga1 −x As) between
barriers made of high conduction/valence band offset materials (e.g., AlAs). These
devices are characterized
by an “N”-shaped current–voltage (I/V ) curve with a peak in current
when the bias aligns a large carrier
population at the contacts with a resonance in the well (6). A valley in the I/V
curve occurs when the applied
bias aligns a very small carrier population in the contacts with resonances in the
quantum well. Hence, the I/V
curve of these devices may have negative differential resistance (NDR ); an example
of the I/V curve of a typical
AlAs/InGaAs RTD is shown in Fig. 3. Since transport is ballistic, to the first
order, carriers do not scatter from
----------------------- Page 455-----------------------
Fig. 2. QCA-based SR latch (8) divided into functional units: A full

logic family may be constructed.
one energy to another. Tunneling through a quantum well is a quantum phenomenon

typically producing peak
currents in the hundreds of kiloamperes/cm2 . In contrast, if classical physics is
used to analyze these devices
essentially no current would flow at all. Operating by ballistic transport they may
be assumed to be, and are,
high-speed devices.
The switching characteristic used to achieve memory operation from
RTDs can be divided into three
categories. The first category is load line switching, where more than one
intersection may occur between the
I/V characteristics of a nonlinear device and the load line. The history of the
operation of the device governs
which of these intersections is the operating point. This may be used
to make logic. The second category
is intrinsic bistability which occurs in some devices where charge is
stored in the device during rising bias
and removed during decreasing bias. This induces a hysteresis loop in the I/V
curve, producing two possible
operating voltages at a given current. Both load line switching and intrinsic
bistability produce volatile memory.
A third category, memory switching, refers to a case in which there are two stable
states maintained with no
power dissipation and without input, and hence, is nonvolatile. Any of the
discussed approaches might be used
to create memory devices since they provide multiple operating voltages at the same
current dependent on
prior history.
Load Line Switching and Intrinsic Bistability Devices. A
number of attempts to make fast RTD-
based logic devices have been made (7) that are based on load line
switching. Load line switching RTDs
maintain their state only while under load. Unfortunately, as isolated devices they
generally have poor terminal
characteristics. Various schemes have been used to overcome these shortcomings by
adding conventional devices
to the memory cell. A representative approach is shown in Fig. 4 (8).
This design from Texas Instruments
employs “ultra-low” current density RTDs that dissipate 50 nW of standby power.
These devices are constructed
in the InGaAs/InAlAs and InP materials system. HFETs (heterostructure field effect
transistors) are used for
the switching transistors. This cell does not require a refresh cycle, giving the
functionality of an SRAM cell
----------------------- Page 456-----------------------
Fig. 3. RTD I/V curve. The peak current is about ±0.9 V and the valley is about
±1.2 V. An example load line is shown
with drive voltage V and resistance R . The intersection (large dot) between this
load line and the I/V curve of the RTD is
the operating point. Multiple intersections between the two curves occur for a
range of values of V (and R ).
while occupying an area comparable to a DRAM cell. Access times below 0.5 ns have
been demonstrated. This
architecture lends itself to vertical integration that aids in achieving high
density. Multiple valued logic can
also be achieved by using cascaded RTD structures. The device count to perform
basic logic functions is much
lower than a standard SRAM cell (typically six transistors). Variations on this
scheme use cascaded RTDs to
achieve multilevel logic and intrinsic bistability (9,10).
Memory Switching. Memory switching is less developed in RTD
research for making memory cells.
A number of nonquantum memory switching devices have been reported in
the literature. Since they are
nonvolatile and static the device count to create functional memory cells
is very low. Langmuir–Blodgett
film/metal sandwiches, for instance, show conductivity changes due to light
irradiation or applied voltages
(11). For electrically switchable metal/film/metal structures a preliminary high
voltage is applied as a forming
procedure; these devices then display memory switching characteristics. A film
thickness of 18 nm has been
used and the off-state resistance is 106 # while the on-state
resistance is 20 to 100 #. The structure of
the film may actually change during switching to cause this phenomenon. This device
is simple to fabricate
and is not necessarily incompatible with standard silicon fabrication procedures.
One negative possibility is
that switching is due to filamentary pathways. Sandwiches of
metal/SiO2/metal have also been observed to
show memory effects after a similar forming procedure, but these have been
attributed to highly conductive
filamentary pathways of heavily doped silicon which form and are destroyed during
switching (12).
Typically, observations of memory switching are difficult to explain. Charge
trapped in interface layers at a
metal/semiconductor boundary can result in memory switching. Such devices are
generally not useful because
they are difficult to reproduce. In any case, two terminal devices that
change resistance when a voltage is
applied are difficult to use without supporting active electronics. For instance,
on/off state resistances must not
be ambiguous when observed through an interconnect of varying length/resistance to
any particular memory
cell.
Flash memory is a conventional type of memory that attempts to use trapped
charge in a reproducible
way. In this device a floating gate accumulates charge that affects the source drain
current characteristics of
----------------------- Page 457-----------------------

Fig. 4. Proposed memory cell. The transistors are HFETs and the bidirectional
diodes are RTDs. Bit and word lines allow
random addressing of each memory cell.
the device. In the EEPROM (electrically erasable programmable read-only memory) hot
electrons are used to
reset the memory state of the device. They do have comparatively very long write
times and some long-term
deterioration with the number of write cycles. Even so, for a variety
of applications these are very useful
devices.
A flash memory cell which accumulates charge on a floating quantum dot above
the channel has been
built by IBM (13). This device is an excellent example of the concept
of a quantum device whose operating
principles become more favorable as it becomes smaller. The change in threshold
voltage (#Vt ) resulting from
charge trapped in the quantum dot increases as the quantum dot size
decreases. This device is interesting
because it has very small off currents in the 10 pA range, small operating
voltages, and potentially excellent
scaling characteristics. It is not clear whether write times, which have always
been the major drawback for
electrically erasable flash memory, will improve in this design.
Quantum Storage Device. Gullapalli and Neikirk have proposed a
method to engineer charge trap-
ping in an attempt to build a memory switching device. The QSD (quantum
storage device) is a modified
quantum well diode that relies on the interaction of a quantum well region with N −
/N +/N − doped layers to
achieve multiple conduction states (14,15,16). Unlike other multiple state quantum
structures, the QSD has
different current versus voltage curves corresponding to the different conduction
states. Preliminary experi-
mental findings indicate that these states remain stable even under short circuit
conditions and can only be
switched from one state to another with the application of bias in excess of
certain threshold voltages. Further-
more, calculations using a self-consistent coherent tunneling model indicate that
it is possible to design QSD
cells with more than two states, creating the possibility of multistate logic and
multibit storage (15,16,17).
----------------------- Page 458-----------------------
Fig. 5. I/V curve for a QSD showing four cycles of the memory switching operation.
The two curves are multiple solutions
for the same device. These solutions are accessed by applying a switching voltage
of about ±2.0 V to the terminals.
QSDs are functionally dissimilar to Shockley diodes or thyristors,

which also change resistance at a
break over voltage but return to the original resistance at low voltage.
The distinct differential resistance
corresponding to each state in a QSD is retained even at zero bias until another
switching voltage is applied,
at which point it changes to that associated with the other state, as shown in Fig.
5.
This multi-state behavior has been shown to occur in two terminal
devices with a thin heterobarrier
structure in close proximity to N − /N +/N − doped layers. For instance, a
double barrier structure consisting of
a thin, lightly doped n-type GaAs layer surrounded by thin AlAs barriers has been
used. The double barrier
quantum well structure is then placed within a few nanometers of an N
− /N +/N − doped GaAs structure
that itself forms a shallow quantum well. The doping sequence is crucial
to device operation. Only devices
with quantum interference between a barrier structure and a well with a highly
variable charge distribution
show evidence of multiple conduction curves. In particular, conventional DBRTDs
(double barrier RTDs) with
monotonically doped layers show no evidence of multistate behavior.
The device possesses memory, in that once the device is placed in one state
it will remain in that state
over a wide range of bias voltages, including zero bias (i.e., the multiple states
exist even at zero bias). Once
the device is placed on one branch of its I–V characteristic, it will
remain on that branch at zero bias. The
state of the device at zero bias (which can be sensed via the value of its
differential resistance), is determined
by whether the device was last switched to the high current curve or to the low
current curve. Even when the
device is completely disconnected from the bias supply, or its terminals are short
circuited, upon reconnection
the differential resistance is unchanged from its preset value. In this
sense, the device possesses memory
of its state, that can be retained without requiring any bias or
dissipating any power. Simulations suggest
that multistate behavior occurs in other structures including single and
triple barrier devices as long as a
N − /N +/N − doped layers are in near proximity to the heterobarrier.
The potential advantages of the QSD over existing technologies are

significant. First, the QSD can be
scaled down to the limit of the photolithography system. From initial findings, it
appears that the cell should
work at mesoscopic geometries creating the possibilities of extremely high-density
memory or logic. However,
unlike mesoscopic systems which require milli-Kelvin temperatures, the QSD has the
advantage of operating
----------------------- Page 459-----------------------
at room temperature. Furthermore, there is a possibility that the QSD can serve as
static, nonvolatile memory
element or logic device with zero holding power, since the multiple conduction
states are stable for extended
periods of time even when completely disconnected from any power supply. Finally,
since the QSD is a simple
two-dimensional structure, memory cells may be stacked on top of each
other. This fact, coupled with the
possibility of more than two conduction states per cell, offers other possibilities
of achieving very high densities.
There are several problems that must be overcome before the QSD can be a
viable alternative to con-
ventional memory. First the switching characteristics of the two terminal devices
built to date are poor: write
currents are high, and on/off resistances are at best differ by a factor of two.
Resistive memory elements require
supporting electronics to make viable memory cells, as in Fig. 4. The devices that
have been made so far have
been in the GaAs/AlAs materials system which is costly and difficult to
integrate with conventional silicon
technology. Attempts have been made to develop a three terminal version by making
direct contact to the N +
layer (18) that may overcome some of the deficiencies in present devices.
Summary
In each technology generation, scientists and engineers are called on to

help identify the directions of the
next generation. The technology to make the next generation of devices, however, by
definition does not exist.
Determining which developments will thrust any particular technology into the
forefront is obviously difficult.
At any particular instant it is important to be aware of the short comings of
proposed new technologies without
assuming these difficulties will necessarily be fatal ones. With this in mind, a
critical assessment of the new
quantum technologies discussed herein should be considered (19).
In some schemes new, novel devices are used in memory cells containing
conventional devices in order to
improve functionality. It is important to note that the fundamental size limitation
of a cell is determined by the
largest device in the cell. Although lower device counts would be extremely
important, the cell size decrease
possible without fundamental changes in the technology of all the components is
limited to a few generations.
Even given advances in device technology, interconnects are becoming an
increasing bottleneck in perfor-
mance. Although they have not been built, Lent’s QCA wires may pose a solution to
interconnect limitations.
Interconnect problems may also be mitigated by moving components closer together.
For instance, the main
advantage of smaller SRAM would be the inclusion on a central processing unit of
large amounts of fast mem-
ory, eliminating or reducing caching requirements and reducing interconnect
lengths. This would require a
compatible materials system with the CPU technology, which may in some cases be
difficult to achieve. In any
case, as we go to smaller devices, design assumptions must take into account
statistical variations inherent in
using nanostructures (20).
Regardless of the problems inherent in developing quantum devices
that work because of their small
size, their potential to provide new generations of technology still warrants
extensive study. The first steps
toward implementing real logic and memory circuits have already been taken, and in
the near future their
contributions to future electronic systems should become clear.
BIBLIOGRAPHY
1. C. S. Lent P. D. Tougaw Lines of interacting quantum-dot cells: A binary wire,

J. Appl. Phys., 74: 6227–6233, 1993.
2. C. S. Lent P. D. Tougaw W. Porod Bistable saturation in coupled quantum dots
for quantum cellular automata, Appl.
Phys. Lett. , 62: 714–716, 1993.
3. P. D. Tougaw C. S. Lent Effect of stray charge on quantum cellular automata:
Part 1, Jap. J. Appl. Phys., 34: 4373–4375,
1995.
----------------------- Page 460-----------------------
4. P. D. Tougaw C. S. Lent Logical devices implemented using quantum cellular

automata, J. Appl. Phys., 75: 1818–1825,
1994.
5. P. D. Tougaw C. S. Lent W. Porod Bistable saturation in coupled quantum-dot
cells, J. Appl. Phys., 74: 3558–3566, 1993.
6. B. G. Streetman Solid State Electron Devices, 3rd ed., Englewood Cliffs, NJ:
Prentice Hall, 1990.
7. C. H. Mikkelson et al. Coupled-quantum-well field-effect resonant tunneling
transistor for multi-valued logic/memory
applications, IEEE Trans. Electr. Dev. , 41: 132–137, 1994.
8. P. V. D. Wagt A. Seabaugh E. A. Beam III RTD/HFET low stand by
power SRAM gain cell, Corporate Research
Laboratories, Texas Instruments, Dallas, TX, 1997.
9. F. A. Bout A. K. Rajagopal Hysteresis of trapped charge in AlGaSb barrier as a
mechanism for the current bistability
in Al GaSb/InAs/AlGaSb double barrier structures, Appl. Phys. Lett. , 64: 2994–
2996, 1994.
10. Y. Hu S. Stapleton Self-consistent model of a double barrier resonant
tunneling diode: Dependence of intrinsic bistability
on structural parameters, J. Appl. Phys., 73: 5254–5263, 1993.
11. K. Sakai et al. Electrical memory switching in Langmuir–Blodgett films, Thin
Solid Films, 179: 137–142, 1989.
12. S. M. Gage et al. Anomalous high zero bias resistance in metal amorphous
silicon metal structures, J. Non-Crystaline
Solids, 115: 171–173, 1989.
13. J. J. Welser et al. Room temperature operation of a quantum-dot flash memory,
IEEE Electr. Dev. Lett. , 18: 278–280,
1997.
14. K. K. Gullapalli A. J. Tsao D. P. Neikirk Experimental observation of multiple
current-voltage curves and zero-bias
memory in quantum well diodes with N − –N+–N − spacer layers, Appl.
Phys. Lett. , 62: 2856–2858, 1993.
15. K. K. Gullapalli A. J. Tsao D. P. Neikirk Multiple self-consistent solutions

at zero bias and multiple conduction curves
in quantum tunneling diodes containing N − –N+–N − spacer layers, Appl.
Phys. Lett. , 62: 2971–2973, 1993.
16. K. K. Gullapalli D. R. Miller D. P. Neikirk Simulation of quantum
transport in memory-switching double-barrier
quantum-well diodes, Phys. Rev. B (Cond. Mat) , 49: 2622–2628, 1994.
17. K. K. Gullapalli D. R. Miller D. P. Neikirk Wigner–Poison simulation of memory
switching heterostructure tunneling
diodes, IEEE Int. Electron Devices Meeting , Washington, DC, 1993, pp. 109–112.
18. O. L. Hartin et al. Design and fabrication of a three terminal

quantum storage device, Proc. IEEE/Cornell Conf.
Advanced Concepts High Speed Semiconductor Devices Circuits , Ithaca, NY, 1995,
pp. 209–218.
19. R. Landauer Need for critical assessment, IEEE Trans. Electr. Dev. , 43: 1637–
1639, 1996.
20. G. J. Iafrate M. A. Stroscio Application of quantum-based devices: trends and
challenges, IEEE Trans. Electr. Dev. , 43:
1621–1625, 1996.
OLIN L. HARTIN
University of Texas at Austin
DEAN P. NEIKIRK
University of Texas at Austin
----------------------- Page 461-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP
Your Librarian
Profile
Short-Term Load Forecasting
Standard Article

X. Ma 1 2 3 4
, A. A. El-Keib , H. Ma , M. Kaddoura
1Cegelec ESCA Corporation, Bellevue, WA
2The University of Alabama, Tuscaloosa, AL

3ABB System Control, Cary, NC
Search
4The University of Alabama, Tuscaloosa, AL
❍ Acronym Finder
reserved.
DOI: 10.1002/047134608X.W6117
Abstract
Stlf Methods
Hybrid Statistical Methods for STLF: an Example
Prospective of the Development of the Stlf Methods
Keywords: load forecasting; general exponential smoothing; power spectrum analysis;

nonlinear temperature sensitive load model;
auto-regressive model; partial autocorrelation analysis; artificial neural networks

%20Conversion/W6117.htm17.06.2008 16:45:10
----------------------- Page 462-----------------------
SHORT-TERM LOAD FORECASTING 233

The
close and constant tracking of system load by the sys-
tem
generation is a basic requirement in the operation of
power
systems. For economically efficient operation and for
effective
control, this must be accomplished over a broad spec-
trum of
time intervals. STLF is generally defined as forecast-
ing the
system load from one hour ahead to one week (168 h)
ahead. The
principal objective of the STLF function is to pro-
vide
accurate load predictions for basic generation scheduling
tasks,
such as economic dispatch, unit commitment, and in-
terchange
evaluation, in addition to the important task of sys-
tem
security assessment.
The
timeliness and accuracy of STLF have significant ef-
fects on
power system operation and production costs (1).
Utilities
have to make up adequate levels of spinning reserve
and
standby reserve in order to have sufficient generation to
satisfy
the demand and to have the desired measure of system
security
and reliability. Underprediction of load results in a
failure to
provide the necessary reserve which, in turn, trans-
lates into
higher costs due to the need to use expensive peak-
ing units.
Overprediction of load, on the other hand, involves
the
start-up of too many units resulting in an unnecessary
increase
in reserve and, hence, operating costs. Thus, by re-
SHORT-TERM LOAD FORECASTING ducing
the forecasting error, reserve levels may be reduced
while
satisfying the demand and maintaining the system se-
This article is intended to provide an overview of the state-of- curity.
the-art methodologies for electric power system short-term The
behavior of the system load is directly governed by
load forecasting (STLF). Three categories of STLF methods the
customers’ activity pattern. These include the seasonal,
are described in the second section with technical details for weekly,
and daily periodicities, legal and religious holiday ef-
readers who are interested in advanced research in STLF. fects,
which form the inherent properties of the load of a sys-
Enhanced modeling techniques based on statistical methods tem.
Meanwhile, some external factors, such as weather, spe-
are presented with simulation results in the third section to cial
events, and so forth, greatly affect the individual
demonstrate how to employ the various existing modeling
consumption pattern and, in turn, the load behavior of the
techniques to build a realistic application. This article con- system.
Thus, the system load behavior is not only governed
cludes with the prospective of the development of STLF meth- by a self
regressive rule but is explicitly affected by a number
ods in the future. An introduction to the STLF problem of
external factors. This behavior is system and environment
follows. dependent,
a fact that makes the load forecasting problem
Power systems are large-scale, nonlinear, and geographi- quite
complicated. The search for a robust STLF algorithm
cally distributed systems. The objective of power systems op- which is
suitable for different systems and different cases has
erators is to provide high-quality services at a reasonable been a
challenging task for the power systems community.
price to the individual end users. This objective can be The
total load demand of a power system consists of energy
achieved only through efficient planning, scheduling, and
consumption of individual consumers. An individual con-
minute-to-minute operation. The availability of sufficient in- sumer may
be a residence, a factory, or a shopping mall. The
formation about a power system is critical for its efficient op- activities
of an individual consumer have to some extent cer-
eration. The future system demand is a crucial piece of infor- tain
regularity. For example, a residential user goes to work
mation. Much effort has been made in developing at around
8 AM and comes home at about 5 PM Monday
sophisticated methodologies to solve this problem. through
Friday, a factory operates 12 h daily seven days a
Power system planning deals with determining the opti- week, a
mall is open from 10 AM to 9 PM five days a week
mal mix and capacity of generation, capacity, and voltage except
weekends, and so on. This regularity causes the hourly
level of transmission and distribution system additions, and load
demands to follow some patterns from one day to the
the type of facilities required in transmission expansion next.
Figure 1 shows the plot of hourly loads (in megawatts)
plans, in the long or medium term. Scheduling focuses on on a
power system for two consecutive days in January of a
when to start up or shut down generating units, how to coor- given
year. It can be seen that there is a daily pattern for
dinate different energy sources (e.g., hydro and thermal en- hourly
loads of the two days, though differences in magnitude
ergy), and which equipment to maintain at what time, so that exist.
This can be generalized and is called daily periodicity.
consumers are served reliably and economically. Careful Similarly,
there are weekly, seasonal, yearly, and other peri-
scheduling is a prerequisite for high-quality service to con- odicities.
sumers. It is the minute-to-minute operation, however, that In

principle, one could determine the load patterns of a
makes reliable and economic services possible. Without system if
each of the individual consumption patterns were
knowledge of future load demand, none of these three facets known.
However, demand or usage pattern of an individual
can be achieved. Load forecasting, therefore, plays a very im- load
customer is quite random and highly unpredictable. In
portant role in the planning, scheduling, and operating of addition,
there are other factors affecting hourly loads, such
power systems. as
temperature, humidity, wind speed, holidays, economic fac-

----------------------- Page 463-----------------------
234 SHORT-TERM LOAD FORECASTING
1600 The
second application of STLF is for predictive assess-
1400 ment of
the power system security. System load forecasts are
1200 an
essential data requirement for the off-line transmission
system
analysis for the detection of future conditions under
d 1000
n which
the power system may be vulnerable. This information
a
m 800 permits
the operating personnel to prepare the corrective ac-
e
D 600 Jan. 9 tions so
that the system can be operated securely.
400 Jan. 10 The
third application of STLF is to provide system op-
200 erating
personnel with timely information, that is, the most
0 recent
load forecast. They need this information to operate
1 3 5 7 9 11 13 15 17 19 21 23 the
system economically and reliably.
Hour The
technical literature displays a wide range of methodol-
Figure 1. Hourly load demands for two consecutive days. ogies

and models for STLF. In general, there are three catego-
ries
for the existing approaches: (1) statistical methods, (2)
expert
systems, and (3) artificial neural networks.
tors, and other random factors. These factors must be taken
Statistical approaches generally employ two kinds of mod-
into account for accurate load forecasting. els:
static models (1–5) and dynamic models (6–9). Static
Different load forecasting methods were developed for vari- models
assume that the load is a linear combination of some
ous purposes. Based on the purpose of application, they can
functional elements which describe the variation of weather
be categorized into three categories:
variables or the basic characteristics of load behavior. The
model
parameters are estimated by using multiple linear re-
1. Very short-term load forecasting (minutes or hours
gression
or exponential smoothing techniques. These models
ahead)
are
structurally simple and require relatively low computa-
2. Short-term load forecasting (days to several weeks tional
effort. Dynamic models are based on a time series de-
ahead)
scription, or its equivalent state space description (10), of the
3. Medium- and long-term load forecasting (months to load
behavior. These models treat the load pattern as a time
years) series
signal with known seasonal, weekly, and daily periodic-
ities.
This gives a rough prediction of the load at the given
Very short-term load forecasting is mainly intended to pro-
season,
day of the week, and time of the day. The difference
vide load demand information in the very near future so that
between
the predicted and the actual load is considered as a
automatic generation control functions are appropriately ad-
stochastic process. The analysis of this random process leads

justed to follow load fluctuations economically. Short-term
to a
more accurate prediction. The techniques used for the
load forecasting is necessary for the operational planning of a
analysis
of this random process include Kalman filtering (6),
power system. Its main objective is to predict the hourly load
the
auto-regressive moving average (ARMA) models (9), Box–
demands, one day or one week ahead. Medium- or long-term
Jenkins
method (11), and the general exponential smoothing
load forecasting is essential for scheduling equipment mainte-
(12).
ARMA models are the most commonly used dynamic
nance, and planning generation, transmission, and distribu-
models
which model stationary processes with finite vari-
tion of a power system.
ances.
Nonstationary processes can be modeled by differenc-
The three categories are important. However, this article
ing the
original process. The differencing operation produces
will concentrate on the category of short-term load forecast-
ing; that is, daily to weekly load forecasting. an auto-
regressive integrated moving average (ARIMA) model
The primary application of the STLF function is to drive (11).
Some dynamic models explicitly include weather infor-
the scheduling functions that determine the most economic mation
as input variables (8). Others rely on a more heuristic
commitment of generation resources that comply with relia- approach
where the load process is preliminarily corrected for
bility requirements, operational constraints, and environmen- weather
influences (13).
tal constraints. For power systems with hydro energy being
Expert systems-based approaches have emerged as a re-
the major generation resources, load forecasts are required to sult of
advances in the field of artificial intelligence in the last
determine the optimal releases from the reservoirs and gener- two
decades. An expert system is a computer program
ation levels in the power house. For systems of dominant (though
not algorithmic) which has the ability to act as an
thermal generations, the load forecasts are used for generat- expert.
This program can reason, explain, and have its knowl-
ing unit commitment to determine the minimal cost hourly edge
base expanded as new information becomes available to
strategies for the start-up and shutdown of units to meet the it. In
the case of the STLF problem, the forecasting system
predicted loads. For hybrid hydro and thermal systems, the emulates
the knowledge, experience, and analogical thinking
load forecasts are used to coordinate hydro and thermal gen- of
experienced system operators. The objective is to identify
eration resources so that production costs are minimized.
variables and rules that are used by system operators in esti-
These scheduling applications require hourly system load mating
or forecasting the system load and the criteria for em-
forecasts for the next day or the next week for the determina- ploying
different rules in different situations. Examples dem-
tion of the least cost operating plans. A closely associated
onstrating the application of this approach to STLF problems
scheduling task is the scheduling and contracting of inter- can be
found in various references (14–16). In conjunction
changes between electric utilities. For this application, the with
expert systems, fuzzy set theory was proposed as a tool
STLF is useful for calculation of the economic levels of inter- to
handle the uncertainties in load models, weather variables,
change. and
operators’ heuristic rules (17). Expert system models
----------------------- Page 464-----------------------
seem to be robust and adaptable to changing conditions than

and a random noise #(t). The order of this process depends on
other methods.
the oldest previous value at which y (t) is regressed on. For an
In recent years, artificial neural network (ANN) tech- AR
process of order p , this model can be written as:
niques have been applied to perform STLF (18–23). The
ANNs can extract the implicit nonlinear relationship between
y (t) = φ y (t − 1) + φ y (t − 2) + · · · + φ y (t − p ) + #(t) (2)
1 2 p
past load or weather variables and forecasted load. They do

not rely on explicit function representation of input variables
With the introduction of the back-shift operator B that de-
and the load to be forecast. It also has the capability to adapt
fines y (t # 1) # By(t), and consequently y (t # m) # Bm
y (t), Eq.
to a changing forecasting environment through the concept of
(2) can be re-written in the following form:
self-learning. ANN application in STLF has been an ongoing
research area and promising results have been achieved.
φ(B )y (t) = #(t) (3)
2 p
where #(B) # 1 # # B # # B # . . . # B .
STLF METHODS
1 2 p
Moving Average Process. In the moving average (MA) pro-

Statistical Methods
cess, the current value of the time series y (t) is expressed lin-
early in terms of current and previous values of a white noise

There are many different variations of statistical methods for
series #(t), #(t # 1), . . .. This noise series is constructed from
STLF. They can be divided into the following four general
the forecast errors or residuals when load observations be-
methodologies:
come available. The order of this process depends on the old-
est noise value at which y (t) is regressed on. For a moving

1. Multiple linear regression
average model of order q, this model can be written as:
2. Stochastic time series
y (t) = #(t) − θ #(t − 1) − θ #(t − 2) − · · · − θ #(t − q) (4)

3. General exponential smoothing
1 t q
4. State space and Kalman filter

A
similar application of the back-shift operator on the white
noise series would allow Eq. (4) to be re-written as:

Multiple Linear Regression. In the multiple linear regression
method, the load is represented in terms of explanatory vari-
y (t) = θ(B )#(t) (5)
ables and weather and non-weather variables that influence
the electrical load. The multiple linear regression model
2 q
where #(B) # 1 # # B # # B # # B .
STLF can be described in the following form:
1 2 p
Auto-Regressive Moving-Average Process. In the auto-regres-

y (t) = a + a x (t) + · · · + a x + #(t) (1)
sive moving-average process, the current value of the time
0 1 1 n n
series y (t) is expressed linearly in terms of its values at pre-

vious periods (y (t # 1), y (t # 2), . . .) and in terms of
where y (t) is the electrical load demand to forecast, x1(t), . . .,
current and previous values of a white noise series (#(t),

xn(t) are explanatory variables correlated with y (t), #(t) is a
#(t # 1), . . .). The order of the ARMA process is deter-

random variable with zero mean and constant variance, and
mined by both the oldest previous value of the series and

a , a , . . ., a are regression coefficients.
0 1 n
the oldest white noise value at which y (t) is regressed on.

The explanatory variables of this model are identified on
For an ARMA process of order p and q, the model is written

the basis of correlation analysis on each of these independent
as:
variables with the load variable to forecast. The estimation of
the regression coefficients is usually computed with the least-
y (t) = φ y (t − 1) + · · · + φ y (t − p ) + #(t)
square estimate technique. Statistical tests, such as the F-
1 p (6)
+ θ #(t − 1) − θ #(t − q)
statistic test, are applied to determine the significance of
1 q
these regression coefficients. The t-ratios resulting from these
tests determine the significance of each of the coefficients, and By
using the back-shift operator previously defined, Eq. (6)
correspondingly the significance of the associated variables
can be re-written as follows:
with these coefficients.
φ(B )y (t) = θ(B )#(t) (7)

Stochastic Time Series. Stochastic time series based STLF
methods appear to be the most popular. This method has been
Auto-Regressive Integrated Moving Average Process. The time
widely utilized and is still being used for STLF in electric
series that can be defined as an AR, MA, or as an ARMA
power utilities. The theory of stochastic time series is dis-
process, is called a stationary process. This means that the
cussed in many text books, and there are many publications
mean of the series and the covariances among its observa-
on stochastic time series based STLF methods. With this
tions do not change with time. If the process is nonstationary,
method, the load series, y (t), is modeled as the output from a
transformation of the series to a stationary process needs to
linear filter that has a random series input, #(t), usually be
performed to obtain a stationary process. This can be
called a white noise. Depending on the characteristic of the
achieved by differencing operations on the nonstationary pro-
linear filter, different models can be classified as follows (11).
cess. By introducing the operator #, a differenced time series
Auto-Regressive Process. In the auto-regressive (AR) proof
order 1 can be written as #y (t) # (1 # B)y (t) using the
cess, the current value of the time series, y (t), is expressed
definition of the back-shift operator B . Therefore, an order d
differenced time series is written as #d d

linearly in terms of its previous values (y (t # 1), y (t # 2), . . .)
y (t) # (1 # B) y (t). The
----------------------- Page 465-----------------------
differenced stationary series can be modeled as an AR, MA, b

is response lag time of the independent variable x(t) with
or an ARMA process to yield an ARI, IMA, or ARIMA time
respect to y (t), and #(t) is a colored (nonwhite) noise series.
series model. For a series that needs to be differenced d times The
series #(t) can be modeled in terms of its past values and
and has orders p and q for the AR and the MA components, a
white noise by using any of the techniques discussed pre-
the model can be expressed as follows:
viously.
The identification of the time series models can be obtained

d
φ(B )∇ y (t) = θ(B )#(t) (8) by
analyzing the raw historical load data. This analysis may
include the use of the range-mean, autocorrelation function,

Seasonal Processes. As was pointed out in the first section, and
partial autocorrelation function plots. This usually leads
there are daily, weekly, yearly, or other periodicities in load to
initial guesses of the required data transformation and de-
demand time series. As a result, a different class of models
gree of differencing to obtain a stationary process. The de-
that can describe this property is designated as seasonal pro-
grees of the AR and the MA polynomials are initially deter-
cesses. Seasonal time series could be modeled as an AR, MA,
mined by means of autocorrelation function analysis and the
ARMA, or ARIMA seasonal models similar to the nonseasonal
partial autocorrelation function analysis. For the case of the
time series. The general multiplicative mode (p , d, q) #
transfer function time series model, the cross-correlation
(P , D , Q)S for a time series model can be written in the follow-
function plot between the historical load time series, y (t), and
ing form: the
independent variable, x(t), is also needed in order to eval-
uate the response lag time, b, along with the orders of r and
φ(B )#(BS )∇d ∇Dy (t) = θ(B )#(BS )#(t) (9) s
of the polynomials #(B) and #(B).
S
The estimation of parameters of the identified load fore-

where #d , #(B), and #(B) were defined previously. Similar
casting model is usually achieved through the use of an effi-
definitions for #D S S
cient estimation method. For a pure AR model, the applica-
S , #(B ), and #(B ) are given:
tion of the Yule–Walker equation solution gives the estimates
∇D = (1 − BS )D (10) of
the parameters of this process (11). Other methods such as
S
the
maximum likelihood techniques are capable of being ap-
#(BS ) = 1 − # BS − # B2S − · · · − # BPS (11)
1 2 P
plied to other models. In the third section, an adaptive
#(BS ) = 1 − # BS − # B2S − · · · − # BQS (12)
weighted recursive least-square estimate method is described.
1 2 Q
Along with the estimation of the load forecasting model, esti-
mation of the standard deviation and correlation of the pa-

The model described in Eq. (9) can be extended to the case
rameters of the model with the variances and covariances of
where two seasonalities are accounted. The order of the model
the
residuals are established for analysis.
is (p , d, q) # (P , D , Q)S # (P #, D #, Q#)S #, and is expressed as fol-
The load forecast model obtained can be assumed to be cor-
lows:
rect only if the model passes the diagnostic checking test.

# # #
This test can be performed simply by checking whether the
S # S d D D S S
φ(B )#(B )# (B )∇ ∇S ∇S# y (t) = θ(B )#(B )#(B )#(t) (13)
residual series is a white noise. If not, the inadequacy of the
model needs to be corrected in view of the autocorrelation

where definitions for #D #, ##(BS #), and ##(BS #) are similar to
S #
function and partial autocorrelation function.
Eqs. (10), (11), and (12).
An example demonstrating the seasonal time series models
is the model for an hourly load time series. If the daily period-
General Exponential Smoothing. With the general exponen-
icity is determined to be included in the model, Eq. (9) can be
tial smoothing (GES) method, the load at time t, y (t), can be
applied with D # 1 and S # 24. If data analysis shows that
modeled by using a fitting function as follows:
both daily and weekly periodicities should be reflected in the

time series model, Eq. (13) can be used to build an ARIMA
y (t) = c (t)T f (t) + #(t) (17)
model with D # 1, D # # 1, S # 24, and S # # 168.

Transfer Function Method. The previous models permit y (t)
where f (t) is the fitting function vector for the process, c(t) is
to be expressed in terms of its history and a white noise. If
the
coefficient vector, #(t) is a white noise, and T is vector
other variables affect the value of y (t), the effect of these vari-
transpose operator.
ables should be accounted for in the model. This can be
The estimates of the coefficients can be found by using

achieved by using the transfer function model. For the case of
weighted mean square error for the recent N sampled inter-

one independent variable x(t), such as the temperature, the
vals. This is achieved by minimizing the following function:

transfer function model can be written in the following form:
N −1
# # #
y (t) = ω(B )x (t − b) + #(t) (14)
wj y (N − j ) − f T (−j )c 2 0 < w < 1 (18)
σ (B )
j =0
where
This minimization gives the estimate of the coefficients in

2 r the
following form:
ω(B ) = ω − ω B − ω B − · · · − ω B (15)
0 1 2 r
σ (B ) = 1 − σ B − σ B2 − · · · − σ Bs (16)
cˆ(N ) = F −1 (N )h (N ) (19)
1 2 s
----------------------- Page 466-----------------------
where
At any time t , there will be an estimate for the process
based on knowledge of the process up to tk #1. This estimate is

N −1
# j T
called the apriori estimate and is expressed as y (k/k # 1). The
F (N ) = w f (−j )f (−j ) (20)
associated error between the actual and the previous esti-

j =0
mates of the process is given as follows:

N −1
# j
h (N ) = w j (−j )y (N − j ) (21)
e (k/k − 1) = y (k) −y (k/k − 1) (31)
j =0
The forecast of the series at lead time l is found as:

This error vector has an error covariance matrix expressed by
Eq. (32):
yˆ (N + l ) = f T (l )cˆ(N ) (22)
# T #
E e (k/k − 1)e (k/k − 1) = P (k/k − 1) (32)
The coefficient estimates and the forecasts can be updated,

as new observations are available, respectively as follows:
The posteriori estimate is obtained as a linear combination
from the apriori

estimate and the measurement noise as in
T −1 # #
cˆ(N + 1) = L cˆ (N ) + F f (0) y (N + 1) −yˆ(N ) (23)
Eq. (33).
# #
yˆ (N + 1 + l ) = f T (l )cˆ(N + 1) (24)
y (k/k) = y (k/k − 1) + K (k) y (k) − H (k)y (k/k − 1) (33)
where F #
lim F(N ). The L matrix is called the transition
N# #
where y (k/k) is the updated estimate and K(k) is the blend-
matrix and is constructed on the basis that the model will
ing factor.
have a fitting function satisfying the following relationship:
The error associated with the actual and the posteriori es-
timate of the process is:

f (t) = L f (t − 1) (25)
e (k/k) = y (k) −y (k/k) (34)

State Space and Kalman Filter. State space and Kalman filter
is a general forecasting approach. It can include the pre-
The covariance matrix of this error vector is expressed by:
viously described methods and more, such as time-varying co-
# T #
efficient models. In this method, the load is modeled as a state
E e (k/k)e (k/k) = P (k/k) (35)
variable by using state space formulation. The state space for-
mulation is designated by two sets of equations.
The blending factor K(k) is computed such that y (k/k) is
optimal in some sense such as the minimum mean squares

System state equations:
error criterion. This factor is known as Kalman gain and the
procedure for implementing Kalman filter for load prediction

y (k + 1) = #(k)y (k) + W (k) (26) is
as follows (24):
Measurement equations:
1. Find the processapriori estimate y (k/k # 1) and the

z (k) = H (k)y (k) + V (k) (27)
error covariance matrix associated with it, P(k/k # 1).
2. Compute the Kalman gain

where y (k) is an (n # 1) process state vector at time tk , #(k)
K (k) = P (k/k − 1)H (k)T # T − R (k)−1#

is an (n # n) state transition matrix relating y (k) to y (k # 1)
H (k)P (k/(k − 1))H (k)
when no forcing function exists, W(k) is an (n # 1) white noise
(36)
with a known white covariance Q(k), z(k) is an (m # 1) load
measurement vector at time tk , H(k) is an (m # n) matrix
3. Compute the updated estimate error covariance matrix
relating y (k) to z(k) without noise, and V(k) is an (m # 1) load
# #
measurement error that is a white noise with a know covari-
P (k/k) = 1 − K (k)H (k) P (k/(k − 1)) (37)
ance R(k).
The covariance matrices for vectors W(k) and V(k) have the
4. Calculate the apriori estimate y (k # 1/k) and the error
following properties:
covariance matrix P(k # 1/k) associated with it
# T #
Q(k) i = k y ((k + 1)/k) = #(k)y (k/k)
(38)
E W (k)W (i) = (28)

0 i =# k
p ((k + 1)/k) = #(k)P (k/k)Q(k)T + Q(k) (39)
# T # R (k) i = k
5. Go to Step 2 moving to the next time step.
E V (k)V (i) = (29)
0 i =# k
The state space method is very attractive for real-time ap-

The process noise, W(k), and the measurement noise,
plications as a result of the recursive nature of the Kalman
V(k), are also assumed uncorrelated. This assumption is de-
filter. The optimal forecast is based on the assumed model.
scribed in the following:
The model has to be known prior to using the Kalman filter.
#
T
Identifying the model is the main task of this method, in addi-
E W (k)V (k)] = 0, for all k and i (30)
tion to estimation of the noise covariances Q(k) and R(k).
----------------------- Page 467-----------------------
Expert System Based Method

nodes in the output layer provide the desired output signal.
The required number of hidden layers and the number of

An expert system based load forecasting model can be built
nodes for each layer are problem dependent. Usually, one or

by using the knowledge about the load forecast domain from
two hidden layers are sufficient for the STLF problem. The
a human expert in the field. The knowledge engineer extracts
number of nodes in the input layer depends on the number

this knowledge from load forecast domain expert by what is
of the identified primary variables that affect load demand

called the acquisition module component of the expert system.
significantly. The number of nodes in the output layer de-

This knowledge is represented as facts and IF-THEN produc-
pends on how the ANN is designed to generate outputs. For

tion rules. This representation is built in what is called the
instance, if an ANN is designed for each hour of a day and it

knowledge base component of the expert system. The search
generates the hourly load forecasting, then one output node

for solution or reasoning about the conclusion drawn by the
is needed. But, if the ANN is designed to provide hourly load

expert system is performed by the inference engine of the ex-
forecasts for each of the 24 h of a day with one put-through

pert system. An expert system has to have the capability to
of the input signal, 24 output neurons are needed. Other vari-

trace its reasoning if asked by the user. This facility can be
ants of the ANN architecture are also possible.

built through an explanatory interface module.
In the feedforward ANN, signals can only be propagated

The main efforts of building an expert system based model
from the input layer to the hidden layers and from the hidden
are acquiring expert knowledge in the STLF domain and or-
layers to the output nodes. Signal propagation between nodes

ganizing the acquired knowledge. Specifically for STLF, vari-
within the same layer or from the input layer directly to the
ables that affect load values need to be identified first. The
output layer is not permitted.

relationship between the identified variables and the load
For each neuron in the input layer, the neuron output is

need also to be constructed. Correlation analysis is useful for
the same as the neuron input. For each neuron in the hidden
variable identification. Relating the variables and the electric
layer or the output layer, the net input is given by:

load is essentially to build the IF-THEN production rules of
the expert system. It should be recognized that the relation-
netj = # wi j oi (40)
ship between the variables and the electric is not static, but
i
is dynamic. This means that the same magnitude of tempera-
ture changes affect load values to different degrees for differ-
where i is a neuron in the preceding layer, oi is the output of
ent day types and seasons. Numerical values are expected
node i, and w is the connection weight from neuron i to neu-
ji
outputs of an expert system based STLF. Therefore, some pro-
ron j . The neuron output is given by Eq. (41):
duction rules may be qualitative and others are quantitative.
As was mentioned previously, the expert system based
1
o = = f (net , θ ) (41)
model is used to emulate the knowledge, experience, and ana-
j 1 + e−(netj +θj ) j j j
logical thinking of an experienced human expert. An advan-
tage of a human expert is his/her capability to learn from
where net is the input signal to node j and # is a bias. In the
j j
new cases and to upgrade his/her knowledge and reasoning.
training process, # can be regarded as a connection weight
j
Therefore, self-learning capability is a desired feature of an
between node j and a fictitious node whose output always re-
experts system.
mains at unity.
The feedforward network is trained with an error back-

Multilayer Feedforward Artificial
propagation learning algorithm via selected sets of input–
Neural Network Based Methods
output training patterns. The algorithm begins with as-

A multilayer feed forward artificial neural network (ANN) as
signing a set of random numbers to the connection weights.
When a pattern p with target output vector t # [t , t , . . .,

a computing system consists of nodes or neurons connected
p p 1 p 2
t ]T is presented, the connection weights can be updated as

by links. The nodes are divided into several layers: (1) the
pM
input layer, (2) the output layer, and (3) some hidden layers
follows:
in between. A three-layer neural network is shown in Fig. 2.
#ω (p ) = ηδ o + α#ω (p − 1) (42)
The nodes in the input layer take the input signal, and the
ji pj pi ji
where, for the output nodes:
δ = (t − o )o (1 − o ) (43)
pj pj pj pj pj
⋅⋅⋅⋅⋅ Output layer
and for other nodes:
⋅⋅⋅⋅⋅ Hidden layer

δ = ## δ ω o (1 − o ) (44)
pj pk k j pj pj
These equations are derived

based on the criterion of min-
⋅⋅⋅⋅⋅ Input layer
imizing the following error function:
M
#
Ep = (t − o )2 (45)
2 pj pj
Figure 2. A multilayer feedforward artificial neural network.
j = 1
----------------------- Page 468-----------------------
where M is the number of output nodes. The method of

where y (t) is hourly load at hour t, yb (t) is base load compo-
steepest descent is used.
nent reflecting the daily or weekly regular changing patterns
It is noted that the learning rate # and the momentum of
the load and independent of weather conditions at hour t,
constant # in Eq. (42) affect the learning speed of an ANN yt
(t) is weather-sensitive load component at hour t, yr (t) is
significantly. Fine-tunings of them are usually needed when
random changing load component at hour t, and #(t) is a white
ANN training is performed, as their optimal values may be
noise with zero mean and constant variance.
problem dependent.
The base load component is dominated mainly by daily and
An ANN can make forecasts only on the basis of the way
weekly periodic load changes in normal conditions (except
it is trained. In power systems, numerous previous load and
holidays). The GES model is quite suitable for modeling this
temperature data can show divergent load patterns and dy-
component (26,27). Previously reported GES models are built
namic ranges. The selection of training cases from the data
through off-line spectral analysis of historical load data and
available can significantly affect the forecasting accuracy.
applied on-line without modification. As load patterns change
Methods to select appropriate training cases need to be de-
with time, especially during season-changing periods, the pri-
vised. For instance, the immediate previous two weeks data
mary harmonics, whose composition reflects the changing
may be used for training and the load is forecasted for the
load profile, also change. An adaptive GES modeling approach
present week. In this case, the ANN needs to be retrained is
able to adapt to the load pattern changes and is described
each week. The general criterion of selecting training cases is
subsequently.
that the ANN trained with cases of high similarity to the case
The random load component, yr (t), represents a stochastic
to be forecasted can produce better forecasting results.
process. Its analysis can always result in higher forecasting
accuracy. An adaptive AR modeling method, enhanced by par-
tial autocorrelation analysis, is used. This model is found to

HYBRID STATISTICAL METHODS FOR STLF: AN EXAMPLE be
highly effective in modeling the random load component.
Weather conditions have a strong influence on the behav-

Statistical STLF technology has been intensively studied over
ior of electric load demand. They are generally presented in a
the past years and is widely used by electric utilities. Physi-
nonlinear manner with considerable uncertainty (25). A gen-
cally meaningful models have been developed. They are capa-
eral nonlinear weather-sensitive load model is used here and
ble of describing influences due to weather pattern deviations
its description of this model is given.
from normal, and random correlation effects using a few ex-
In the proposed STLF algorithm, the GES model is identi-
planatory variables. Such models are usually set up via off-
fied weekly and the AR models are identified daily. With the
line simulations. This reduces their adaptability and ro-
proposed models, the weighted recursive least-square esti-
bustness.
mate (WRLSE) algorithm with a variable forgetting factor is
Improved forecasting performance of the statistical-based
employed for model parameter estimation. The utilization of
STLF techniques can be obtained by:
this algorithm improves model tracking capability and nu-
merical stability (26). This is important since in a real-time

1. Adaptive identification of explanatory variables
environment, STLF results are updated using the most recent
load and weather information when situations such as those

2. Advanced adaptive modeling methods
of
abnormal or abrupt weather changes, unexpected social
3. Adaptive parameter updating schemes
events, or even holidays occur. Also, for the purpose of on-line
economic dispatch, interchange transaction evaluations, and

Different adaptive model parameter estimation algorithms
network security assessment, updating is required for more
have been devised for STLF (25,26), but models with fixed
accurate system analysis. Details of the proposed techniques
structure were used. The models to be described in this sec-
follow.
tion focus on developing robust and adaptive modeling tech-
niques based on statistical methodologies (30).
Adaptive Modeling Techniques
The algorithm considers the load to be composed of base
Enhanced Adaptive GES Modeling

load, weather-sensitive load, and random load components.
Model Description. The base load component, yb (t), is mod-

Hybrid modeling techniques, which include an adaptive GES
eled with the GES model, which can be expressed as:

model, a nonlinear weather-sensitive load component model,
and an adaptive AR model, are adopted to model the three
yb (t) = c (t)T f (t) (47)
load components. Power spectrum analysis and partial auto-
correlation coefficients are applied to identify explanatory
where c(t) is a column vector of estimates of ‘‘n’’ locally con-
variables of the GES and AR models, respectively. The AIC
stant coefficients, f (t) is a column vector of ‘‘n’’ linearly inde-
criterion (29) is utilized to determine the optimal order of
pendent fitting functions.
both GES and AR models.
Equation (47) can be constructed by a finite Fourier series,
as
follows:
Description of the Hybrid Models
n
yb (t) = a + #(a sin ω t + b cos ω t) (48)

Based on the properties of the load process, the following sta-
0 i i i i
tistical-based hybrid models for the STLF are constructed.
i= 1
The load at hour t can be expressed as follows:

where # # 2#/ 168 k , k is an integer (#84, Nyquist limit).
i i i
Also, f (t) # [1 sin# t cos# t . . . sin# t cos# t]T and c(t) # [a
1 1 n n 0
y (t) = yb (t) + yt (t) + yr (t) + #(t) (46)
T
a
b . . . a b ] .
1 1 n n
----------------------- Page 469-----------------------
A major task in building an appropriate GES model for

where the model residual is #w(t) # y (t) # yb (t)#, and # is the
STLF is to identify the primary harmonics to be included in

estimated variance of the residuals of the kth order GES
Eq. (48). An adaptive modeling scheme to perform this task
model. According to AIC criterion, the optimal model order
was developed. A presentation of this scheme follows.
corresponds to the minimum AIC value computed in Eq. (51).
Adaptive Identification of Primary Harmonics. A systematic
The order of a GES model is usually high. This makes the
approach to adaptively determine the order of the GES model
previously defined modeling procedure computationally inef-
via power spectrum analysis is devised. Normalized power
ficient. To speed up the modeling process, the following gen-
spectrums are defined and utilized to build the GES models.
eral heuristic rules are designed: Include in the initial GES
model, those harmonics with P (# )
# (#, a prespecified
The power spectrum of the load series #y (t)#, is defined by:
n i
threshold value); and rank the remaining harmonics with

 
P (# ) # # in ascending order in terms of their P (# ) and in-
J
n i n i
l
#
1 2
clude them in the initial GES model one by one until Eq. (51)
 
P (ω ) = σ + 2 RYY (j ) cos j ω (49)
i Y i
π
reaches minimum.
j = 1
A threshold # # 0.5 is considered to be appropriate. This
choice of # is based on the observation that P (# )

0.5 indi-
n i
where RYY(k) is the auto-covariance of #y (t)#, #Y is the variance
cates that the relative significance of the harmonic in the
of #y (t)#, J is determined such that the R (J ) calculated us-
1 YY 1
GES model is above average compared to other harmonics.
ing the available sample data is statistically reliable. Usu-
ally, J1 # N/M with M greater than 10 (M # 13 is found to
Weather-Sensitive Load Component Model. The identified

be appropriate for the test case to be presented), and N is the
GES model does not take into account the effect of weather
number of data used for modeling.
deviations from normal patterns on electric load consumption.

The magnitudes of the coefficients (a and b ) of the GES
i i
Explicit inclusion of weather variables in the forecasting
model are directly proportional to those of power spectrums
model is necessary to improve model accuracy.

of the corresponding harmonics (28). As a result, a harmonic
Weather factors that are generally considered to signifi-

with a larger power spectrum magnitude has a more signifi-
cantly affect load demand are temperature, humidity, wind

cant effect on the GES model representing load change pat-
speed, and cloud cover. The relationship between changes in

terns. Inclusion of such harmonics improves the model ac-
load and weather is appropriately described by nonlinear

curacy.
polynomials (26,27). The following general model for weather-
The normalized power spectrums, instead of the values cal-
sensitive load component was applied:
culated using Eq. (49), were utilized because they reveal the
relative significance of the harmonics in the GES model. They
L M L
1 j 2
are defined by:
yt (t) = c0 + # # # cijk T (t − i)k (52)
j =l k =l i=0
P (ωi )
P (ω ) = (50)
n i Pav
where L 1 is the number of weather variables to be modeled,
Mj is a constant reflecting the nonlinearity of the relationship

where
between weather variable and the weather-sensitive load
component, cijk is the coefficient of the kth order of the j th

J
weather variable at the ith lagging hour, T (t) is the value of
2
j
#
1
Pav = |P (ω ) |
the j th weather variable at hour t, and L2 is a constant re-
J j
2 j = 1
flecting time delay effect of weather change.
Inclusion of a weather variable in the weather-sensitive

is the average power spectrum value. In general, the upper
load component model depends on the following factors:
bound for J2 is the Nyquist limit. Simulation tests are usually
strength of dependence of the weather-sensitive load compo-
needed to determine a proper value: J2 # 50 is used in the
nent on this variable; and accuracy of the forecasted value of
the weather variable. Use of a weather variable that has very

following forecasting results.
high unpredictability may deteriorate rather than improve

The proposed adaptive GES modeling method calls for add-
the load forecasting accuracy.

ing harmonics one by one into Eq. (48) in the ascending order
Based on these considerations, which variables to include

of P (# ) until the sum of the squares of the model residuals
n i
in the model may be system dependent and need to be tested
reaches minimum. The model thus obtained can be considered
on specific system data. In the following example case, tem-

to possess the optimal forecasting performance.
perature is considered to be the most prominent weather vari-

The amount of error reduction evaluated by the sum of the
able. Accordingly, a third-order polynomial weather-sensitive

squared residuals may be insignificant and does not justify
model is used in the form of Eq. (52).

the additional computational effort due to the increased
A hybrid model formed by combining the base load model
model order. In general, a model determined by this manner
component in Eq. (48) and the weather-sensitive load compo-
is not parsimonious. Model parsimony can be achieved by us-
nent model is constructed to estimate model parameters us-
ing the AIC criterion (29). The AIC for the GES model with k
ing the WRLSE algorithm.
harmonics is calculated by Eq. (51):
# #
2
Enhanced Adaptive AR Modeling. Once the task of building
AIC (k) = N · log σ˜W (k) + 2 (2k + 1) (51)
the base load and the weather-sensitive load components’
----------------------- Page 470-----------------------
models is completed, the random load component, yr (t), can

2. Add the lagging term at the top of the current ranking
be calculated by:
to the AR model.
n
3. Estimate parameters of the current AR model.
# ˜
4. Compute the sum of the squared model residuals.
yr (t) = y (t) − a˜ − (a˜ sin ω t + b cos ω t)
0 i i i i
i= 1
5. Evaluate the significance of the decrease of the sum of
L M L (53)
l j 2
the squared residuals using the AIC criterion. The AIC
− # # # c˜ T (t − i)k
for the p th order AR model is calculated by Eq. 56:
ijk j
j =l k =l i=l
AIC (p ) = N · log[σ˜ 2 ] + 2p (56)
#
where tilde parameters refer to the estimated values. The
random series, #yr (t)#, can be considered a stationary time se-
where # is the estimated variance of the residuals of
#
ries and can generally be modeled by an ARMA model.
the p th order AR model. If AIC (p ) # AIC (p # 1), go to
ARMA modeling methods have been used in STLF
step 2; otherwise, the optimal parsimonious AR model
(1,9,26,27). It has been shown in (9) that AR models have
is obtained.
forecasting quality equivalent to ARMA models. Promising
forecasting was obtained using AR models (26). In addition,
WRLSE Algorithm. For a model expressed by:
parameter estimation of AR models is computationally much
less demanding than that of ARMA models. This is an attrac-
y (t) = H (t)TX (t) + #(t) (57)
tive feature for the consideration of developing AR models
with an enhanced modeling algorithm for use in this example.
where x(t) is the vector of the known values of the variables,

Based on the assumption that the most recent lagging term
H(t) is the column vector of coefficients to be estimated, and
has a larger partial autocorrelation coefficient than the subse-
y
(t) is the output of the model. The WRLSE algorithm with a
quent term, AR models were identified in (9) by successively
variable forgetting factor # can be described by the following

adding the immediate lagging terms into the AR model one
set of equations:
by one. This process continues until the sum of squares of
# #
the model residuals reaches minimum. Fixed AR models were
T
R (t − 1) x (t)x (t) R (t − 1)
utilized in (26,27). An improved adaptive AR modeling algo-
R (t) = I − T (58)
λ(t − 1) λ(t − 1) + x (t) R (t − 1)x (t)

rithm enhanced with partial autocorrelation analysis is de-
scribed later. In addition to being adaptive, this algorithm
# T #
H (t) = H (t − 1) + R (t)x (t) y (t) − H (t − 1) x (t) (59)

recognizes that the most recent lagging term does not neces-
sarily have a larger partial-autocorrelation coefficient than
#(ˆ t) = y (t) − H (t)Tx (t) (60)
the subsequent one.

Q(t) = x (t)TR (t)x (t) (61)
According to the enhanced AR modeling algorithm, the
t
general form of an AR model of order p can be described by:
S (t) = # λ(j )t −j #(ˆ j )2 (62)
θ(B )yr (t) = #(t) (54)

j = 1
 ! 
 
p
1 #(ˆ t)2 #(ˆ t)2
qi
λ(t) = 1 − Q(t) − + 1 − Q(t) − + 4Q(t)
where #(B) # 1 # #i #1 # B , # is coefficient of the AR process,
qi qi
 
2 S (t) S (t)
B is back-shift operator, qi refers to the ith lagging term with
q hours preceding the current hour in the AR(p ) model. It is
(63)
determined by partial autocorrelation analysis.
An outline of the proposed adaptive AR modeling method
Test Results
follows.
The adaptive modeling techniques were implemented. A cubic
temperature-sensitive load model was used in order to evalu-

1. Compute the partial autocorrelation coefficients (PACs)
ate the proposed adaptive modeling techniques against the

of #yr (t)# to the order J3 as follows:
fixed structure approach. The WRLSE algorithm with the

N −j
variable forgetting factor was applied to estimate model pa-
# # ## #
1
yr (t) −yr yr (t + j ) −yr
rameters. In this
simulation, four-week historical data (N #
N − j − 1 t = 1
672) were used for building the GES and AR models. Identifi-
ρYR−YR (j ) = (55)
N
cation of the GES model was performed weekly; whereas AR
# # #
1 2
N yr (t) −yr
models were identified daily. The error statistics were calcu-
t = 1
lated with respect to daily peak loads. A threshold error value
of one percent was used to activate the updating algorithm.

where
Historical data from two utilities were used. Utility A has

N
a maximum and a minimum hourly load, respectively, of
# # #
1
yr = yr (t)
32,343 and 11,071 MW. The maximum and minimum hourly
N t = 1
average temperatures were 36.9#C (98.5#F) and 8.4#C (16.9#F).
Utility B has a maximum and a minimum hourly load of

J is determined in the same manner as J . The PACs
2,043 and 633 MW, respectively. The maximum and mini-
3 1
are ranked in ascending order #q , q , . . ., q # in terms

mum hourly average temperatures were 33.9#C (93#F) and
1 2 J
3
of their magnitudes.
30.6#C (#23#F). Utility A is a summer peak system with one
----------------------- Page 471-----------------------
yearly peak in July. Utility B is a summer and winter peak

Table 2. AME Error Statistics of 7 Day Types, Utility A
system with yearly peaks in August and February. The stan-
dard deviation of daily average temperature of Utility A is
Jan.–Mar. April–June July–Sept. Oct.–Dec.
smaller than that of Utility B. Actual recordings of tempera-

Day AME SDE AME SDE AME SDE AME SDE
ture from the two utilities were used in the following simula-
Mon. 1.275 1.629 1.266 1.612 1.042 1.305 1.344 1.753
tion studies.
Tue. 1.095 1.403 1.128 1.369 0.979 1.182 1.173 1.621

A fixed modeling technique, which is being used in several
Wed. 1.054 1.352 1.009 1.345 0.880 1.144 1.089 1.391
utility systems (26), is utilized as the base to evaluate the
Thu. 1.103 1.363 0.977 1.278 0.983 1.235 1.159 1.522
improvement that can be achieved by the enhanced modeling
Fri. 1.220 1.497 0.998 1.263 0.864 1.099 1.112 1.452
techniques. The fixed models are described as follows:
Sat. 0.894 1.180 1.009 1.293 0.826 1.044 1.144 1.476
Sun. 1.298 1.748 1.197 1.554 0.940 1.217 1.340 1.730
Harmonics in the GES model:
{1, 2, 3, 6, 7, 8, 13, 14, 15, 20, 21, 22, 28, 35, 49}
racy from the proposed method ranges from 10 to 30% for the
Nonlinear temperature sensitive model:
four time periods on average. This is a result of using the
proposed adaptive modeling techniques.

3
#
Table 2 presents the seasonal error statistics in terms of

j j j
yt (t ) = {c ljT (t ) + c T (t − 1) + c T (t − 2) }
1 a 2j a 3j a day
types. Results show that AME and SDE are less than
j = 1
1.5% and 2%, respectively, for all seven day types. The adap-
tive modeling approaches exhibit very stable performance for

where c , c , and c are model parameters and T (t) is
1j 2j 3j a all
day types. This verifies the designed adaptability and ro-
the average equivalent temperature for two adjacent
bustness.
hours.
The error statistics of peak load forecasting are given in

The AR model:
Table 3. It is noted that errors for the time period from Oct.
23 to
Dec. are larger than those for the other three periods. This
yr (t + 1) = (θ + θ B + θ B )yr (t)
0 1 23 is
because of holidays such as Thanksgiving Day, Christmas
Day, and the New Year Eve. When compared to the fixed
Table 1 shows absolute mean error (AME) and standard
modeling approach, the ratios of improvement in accuracy
deviation error (SDE) statistics of 24 h for Utility A by the
reach up to 34% on average. Considering the importance of
proposed method. Highly accurate forecasting was achieved
peak load forecasting, this improvement is quite significant.
with seasonal average errors less than 1.5%. Compared with
By using the proposed adaptive modeling techniques, the
the fixed modeling method, the ratios of improvement in accu-
identified models for Utility A in two consecutive weeks of
April follow.
The identified harmonics for the two GES models are

Table 1. AME Error Statistics of 24 Hours, Utility A
Jan.–Mar. April–June July–Sept. Oct.–Dec. {1,

3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 16, 18, 19, 21, 28, 35, 49}
Hour AME SDE AME SDE AME SDE AME SDE {1,
2, 3, 5, 6, 7, 8, 9, 10, 11, 12,
1 0.859 1.068 0.536 0.692 0.526 0.670 0.600 0.751

13, 14, 16, 18, 19, 21, 24, 26, 28, 35, 49}
2 0.805 1.104 0.763 0.966 0.642 0.710 0.819 1.015
3 0.909 1.154 0.860 1.039 0.684 0.861 0.874 1.072
The harmonics used in the fixed structure model are
4 1.346 1.562 0.888 1.121 0.588 0.779 1.042 1.290 #
1,2,3,6,7,8,13,14,15,20,21,22,28,35,49#. In addition to most of
5 1.053 1.098 0.795 1.065 0.581 0.706 0.864 1.066
these harmonics, other harmonics which vary from one week
6 2.187 2.610 1.531 1.950 1.388 1.532 2.270 2.676
to
another are also identified for the GES models by the mod-
7 1.356 1.692 1.298 1.632 1.117 1.406 1.431 1.819
eling method described previously. Different thresholds were

8 1.725 1.926 1.257 1.559 0.879 1.127 1.759 2.060
9 1.001 1.283 1.099 1.171 0.959 1.232 1.071 1.239
used to identify the primary harmonics in the GES model and
10 1.153 1.398 0.977 1.057 0.956 1.161 0.936 1.227

11 0.923 1.147 0.994 1.119 0.953 1.178 0.776 0.999
12 0.993 1.198 1.094 1.103 0.853 1.060 0.931 1.186
Table 3. AME Error Statistics of

13 0.923 1.199 1.089 1.172 1.036 1.244 1.115 1.403
Peak Load Forecast, Utility A

14 0.955 1.028 1.222 1.312 0.779 0.974 0.903 1.216
15 0.852 1.075 0.940 1.104 1.032 1.239 1.201 1.541
Jan.–Mar. April–June July–Sept. Oct.–Dec.
16 0.798 0.994 1.267 1.584 1.063 1.293 1.193 1.484
Day AME SDE AME SDE AME SDE AME SDE

17 1.398 1.093 1.076 1.293 0.949 1.263 1.227 1.564
18 1.453 1.744 1.330 1.828 1.109 1.366 1.914 1.947
Mon. 1.148 1.655 1.032 1.340 0.937 1.097 1.990 2.223
19 1.362 1.577 1.335 1.526 1.120 1.423 1.718 1.989
Tue. 1.041 1.235 1.635 1.677 0.811 0.863 2.044 1.874
20 1.467 1.452 1.543 1.949 1.176 1.412 1.434 1.826
Wed. 1.091 1.553 1.205 1.343 1.298 1.643 1.533 1.824
21 1.225 1.182 1.385 1.334 1.199 1.489 0.843 1.077
Thu. 1.413 1.470 0.884 1.191 1.055 1.228 1.579 1.536
22 0.708 0.904 0.967 1.169 0.746 0.890 0.821 1.019
Fri. 1.172 1.450 1.096 1.206 1.123 1.333 1.625 1.598
23 0.826 1.039 0.812 1.034 0.720 0.862 0.588 0.740
Sat. 1.145 1.610 0.709 0.877 1.088 1.185 1.342 1.771
24 0.908 1.094 0.651 0.831 0.755 0.967 0.788 1.041
Sun. 1.730 0.959 0.984 0.969 0.942 1.181 1.250 1.436
Avg. 1.133 1.468 1.071 1.391 0.909 1.179 1.130 1.532

Avg. 1.249 N/A 1.078 N/A 1.036 N/A 1.623 N/A
----------------------- Page 472-----------------------
a threshold of 0.5 was used as a better tradeoff between com-

tions of time series can improve the performance of the
putational effort and modeling accuracy. In particular, this
dynamic STLF models to some extent, but they will still have
threshold value generally guarantees that the primary har-
the bias of historical data. Static models, such as the multiple
monics are all included in the final GES model.
linear regression model, depend on identifying the so-called
The AR models of 7 successive days in the two weeks:
explanatory variables via complicated correlation analysis or
even empirical knowledge. The main source of errors for static

10 23
Mon./Tue.: yr (t) = # θ yr (t − i) + # θ yr (t − i)
models is that functional relationships between explanatory
i= 1 i i=21 i
and dependent variables are not stationary for load forecast-
9 24
ing. They change over time. Similarly, various updating tech-
Wed.: yr (t) = # θ yr (t − i) + # θ yr (t − i)
niques can be used to alleviate this error. While each of the
i i
i= 1 i=21
statistical STLF approaches demonstrates considerable suc-
8 24
Thu.: yr (t) = # θ yr (t − i) + # θ yr (t − i)
cess in forecasting accuracy, they are all subject to the risk
i i
i= 1 i=21
that the future cannot be fully reflected by history. This is
8 23
# #
the risk associated with the fundamental assumption of all

Fri.: yr (t) = θ yr (t − i) + θ yr (t − i)
i i
statistical modeling approaches.
i= 1 i=21
In recent years, AI-based approaches have received inten-
5 24
Sat.: yr (t) = # θ yr (t − i) + # θ yr (t − i)
sive attention as an alternative method to STLF problems.
i i
i= 1 i=23
Expert systems belong to this class of methods. Expert sys-
5 24
# #
tems can employ human experience to validate intuitive in-

Sun.: yr (t) = θ yr (t − i) + θ yr (t − i)
i i
sight about a process with a certain degree of success. This
i= 1 i=22
approach has a high reliance on the existence of an expert

Simulations were also performed using one year historical
capable of making accurate forecasts, from whom computer
hourly load data of Utility B. Reductions of daily average er-
software can be designed to emulate the expert. An inherent
rors range from 6 to 12%. The reductions of peak load fore-
difficulty with expert opinions, however, is that they may not
casting error range from 2 to 11%. This is achieved through
always be consistent, or the reliability of such opinions may
identifying the appropriate harmonics for GES models and be
in question. Similar to maintaining models for statistical
lagging terms for AR models, in terms of normalized power
methods, updating of knowledge base and production rules of
spectrum and partial autocorrelation analysis, respectively. an
expert system is very likely to happen relatively frequently
The developed AR modeling approach produced AR models to
reflect the latest changes in load data and expert opinions.
that are quite different from one day to the next and can bet-
The main challenges to applications of expert systems to
ter reflect the load change. This example demonstrates that
STLF are as follows: (1) The approach is not easily reproduc-
improved forecasting has resulted from the enhanced model-
ible, that is, an expert system based on selected experts from
ing techniques.
certain utilities cannot assume an easy portability to other
utilities, and (2) the lack of self-learning mechanisms, which
are fundamentally different from a human expert whose

PROSPECTIVE OF THE DEVELOPMENT
knowledge continuously evolves.
OF THE STLF METHODS
Another category of AI-based methods are ANNs. Neural
networks are a more promising area of artificial intelligence

Load forecasting is crucial for the expansion planning, relia-
for STLF since they do not rely on human experience, but

bility maintenance scheduling, and operational planning of
power systems. STLF is a particularly important function in
attempt to draw a link between a set of input data and ob-
power system operations as the principal driving element for
served outputs. This approach does not rely on an explicit
daily and weekly operations scheduling. In the last two de-
adoption of a functional relationship between past load or
cades, various modeling and forecasting techniques have been
weather variables and forecasted load. It has the capability
developed and have advanced the state-of-the-art of STLF to
adapt to a changing forecasting environment through self-
methods dramatically.
learning. In this sense, the ANN-based methods are very dif-
Of the many models published in the literature and ap-
ferent from the other methods and demonstrate obvious ad-
plied in power utilities, the time series based (or dynamic)
vantages. However, the way in which an ANN is trained in-
models, like the exemplar STLF modeling method presented
fluences its performance significantly. If the training data set
in the previous section, are the most popular. Such models is
not thoughtfully chosen, the resulting network is unlikely
are capable of describing time-correlated random phenomena to
hold up well. Massaging the set of training data probably
with relatively few explanatory variables and parameters.
demands most of the efforts in building a real-world ANN. A
This type of model is relatively easily developed and updated,
systematic method of preparing the set of training data helps
with modest computational requirements. However, it should
create a more robust ANN with less effort.
be realized that time series based approaches contain the fun-
As was discussed, any individual load forecasting method
damental assumption that from the behavior of the time se-
has its own merits and defects, and none of them has a deci-
ries in the past, its behaviors in the future can be determined.
sive advantage. Therefore, there is always room for improve-
This assumption may not always hold, because the system
ment. Hybrid forecasting techniques that blend various types
represented by the time series may not be self-explanatory, of
techniques for STLF, such as combining statistical and ex-
then this assumption will lead to erroneous predictions. Er-
pert system based methods incorporating fuzzy logic into the
rors will be greater if some abrupt changes in the system oc-
conventional ANNs and others, represent the trend of the fu-
cur, or if a long lead time is chosen. Some adaptive modifica-
ture research.
----------------------- Page 473-----------------------
244 SIGMA–DELTA MODULATION

BIBLIOGRAPHY 23.
C. N. Lu, N. T. Wu, and S. Vemuri, Neural network based short-
term load forecasting, IEEE Trans. Power Syst., 8: 1993, 334–342.

1. P. C. Gupta and K. Yamada, Adaptive short-term load forecasting 24.
G. G. Brown, Introduction to Random Signal Analysis and Kalman
of hourly loads using weather information, IEEE Trans. Power
Filtering, New York: Wiley, 1983.
Appar. Syst., 91: 1972, 2085–2094. 25.
Q. C. Lu et al., An adaptive nonlinear predictor with orthogonal
2. K. Srinivasan and R. Pronovost, Short-term load forecasting us-
escalator structure for short-term load forecasting, IEEE Trans.
ing multiple correlation models, IEEE Trans. Power Appar. Syst.,
Power Syst., 4: 1989, 158–164.
94: 1975, 854–858. 26.
J. Y. Fan and J. D. McDonald, A real-time implementation of
3. M. Abou-Hussain et al., An accurate model for short-term load
short-term load forecasting for distribution power systems, IEEE
forecasting, IEEE Trans. Power Appar. Syst., 100: 1981, 4158–
Trans. Power Syst., 9: 1994, 988–994.
4165. 27.
M. T. Hagan and S. M. Behr, The time series approach to short-
4. S. Or, A weather sensitive model for short-term load forecasting,
term load forecasting, IEEE Trans. Power Syst., 2: 1987, 785–791.
Proc. IEEE/CSEE Joint Conf. High Voltage Transmission Syst. 28.
R. G. Brown, Smoothing, forecasting and prediction of discrete
China, Beijing, Oct. 17–22, 1987, 245–251.
time series, Englewood Cliffs, NJ: Prentice-Hall, 1965.
5. A. D. Papalexopoulos and T. C. Hesterberg, A regression-based 29.
H. Akaike, A new look at the statistical model identification,
approach to short-term load forecasting, IEEE Trans. Power Syst.,
IEEE Trans. Automat. Contr., 19, 1974, 716–723.
5: 1990, 1535–1547. 30.
A. A. El-Keid and X. Ma, Advancement of statistical based model-
6. M. A. Abu-El-Magd and N. K. Sinha, Two new algorithms for on-
ing techniques for short-term load forecasting, Electric Power Sys-
line modeling and forecasting of load demand of a multi-node
tems Research, 35, 1995, 51–58.
power system, IEEE Trans. Power Appar. Syst., 100: 1981, 3246–
3253.
X. MA
7. B. Krogh, E. S. Llinas, and D. Lesser, Design and implementation
Cegelec ESCA Corporation
of on-line load forecasting algorithm, IEEE Trans. Power Appar.
A. A. EL-KEIB
Syst., 101: 1982, 3284–3289.
The University of Alabama
8. A. Keyhani and S. M. Miri, On-line weather-sensitive and indus-
H. MA
trial group bus load forecasting for micro-processor application,
ABB System Control

IEEE Trans. Power Appar. Syst., 102: 1983, 3868–3876.
M. KADDOURA
9. S. Vemuri, W. L. Huang, and D. J. Nelson, On-line algorithm for
The University of Alabama

forecasting hourly loads of an electric utility, IEEE Trans. Power
Appar. Syst., 100: 1981, 3775–3784.
10. L. Ljung and T. Soderstrom, Theory and Practice of Recursive
Identification, Cambridge, MA: MIT Press, 1983.
SHORTWAVE BROADCASTING. See HIGH-FREQUENCY
11. G. E. P. Box and G. M. Jenkins, Time Series Analysis—
BRAODCASTING.
Forecasting and Control, Oakland, CA: Holden-Day, 1976.
SHUNTS, CURRENTS. See CURRENT SHUNTS.
12. W. R. Christiaanse, Short-term load forecasting using general ex-
ponential smoothing, IEEE Trans. Power Appar. Syst., 90: 1971,
900–910.
13. M. Honig and D. Messerschmitt,Adaptive Filter, Structures, Algo-
rithms, and Applications, Hingham, MA: Klumer Academic Pub-
lishers, 1984.
14. S. Rahman and R. Bhatnagar, An expert system based algorithm
for short-term load forecast, IEEE Trans. Power Syst., 3: 1988,
392–399.
15. K. Jabbour et al., ALFA: Automated load forecasting assistant,
IEEE Trans. Power Syst., 3: 1988, 908–914.
16. S. Rahman and O. Hazim, Generalized knowledge based ap-
proach to short-term load forecasting, IEEE Trans. Power Syst.,
8: 1993, 508–514.
17. K. Ho et al., Short-term load forecasting of Taiwan power system
using a knowledge-based expert system, IEEE Trans. Power Syst.,
5: 1990, 1214–1221.
18. D. C. Park et al., Electric load forecasting using an artificial neu-
ral network, IEEE Trans. Power Syst., 6: 1991, 442–449.
19. K. Y. Lee, Y. T. Cha, and J. H. Park, Short-term load forecasting
using an artificial neural network, IEEE Trans. Power Syst., 8:
1992, 124–132.
20. K. Ho, Y. Hsu, and C. Yang, Short-term load forecasting using a
multi-layer neural network with an adaptive learning, IEEE
Trans. Power Syst., 7: 1992, 141–149.
21. S. Chen, D. Yu, and A. R. Moghhaddamjo, Weather-sensitive
short-term load forecasting using non-fully connected artificial
neural network, IEEE Trans. Power Syst., 7: 1992, 1098–1105.
22. T. Peng, N. Hubele, and G. Karady, Advancement in the applica-
tion of neural networks for short-term load forecasting, IEEE
Trans. Power Syst., 7 (1): 1992, 250–257.
----------------------- Page 474-----------------------
} { { } }
HOME
●
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Solar Energy Conversion
Standard Article

Gaoling Zhao1 1 2
, Hiromitsu Kozuka , Hong Lin ,
Toshinobu Yoko2
1Zhejiang University, Hangzhou, China

2Kyoto University, Kyoto-fu, Japan
Search
reserved.
❍ Acronym Finder
DOI: 10.1002/047134608X.W3036
Abstract
Molecular Photochemical Energy Storage and Photosynthesis
Heterogeneous Photocatalysis
Wet-Type Solar Cells

%20Conversion/W3036.htm17.06.2008 16:45:36
----------------------- Page 475-----------------------
638 SOLAR ENERGY CONVERSION
SOLAR ENERGY CONVERSION
The
fossil fuels such as coal, oil, and natural gas, which main-
tain
our industrialized world, will be surely running out
sometime in the twenty-first century. Moreover, burning such
fossil fuels causes the global air pollution problems due to

the
production of CO , SO , and NO , leading to global warm-
2 2 x
ing
and acid rain problems. The development of alternative
clean energy resources is, therefore, one of the most urgent
subjects with which contemporary scientists have to struggle.
Among them, utilization of solar energy seems to be the most
promising and potential, and an important subject that a
number of researchers in the world are now studying. A vari-

ety
of ways for utilizing solar energy are known, for example,
thermal energy by heat collectors, electrical energy by silicon
solar cells, and chemical energy by photosynthesis, where the
latter is referred to as the conversion of solar energy into
chemical energy. The significant problems common to all
these are the low cost performance, the low energy conversion
efficiency, and/or the lack of persistence. In order to replace
solar energy for fossil energy on an economical basis, it is
absolutely necessary to overcome these problems as soon as
possible.
Here, we focus on the solar energy conversion processes

which involve chemical reactions in any way. The best known
example of such a case is the plant photosynthesis; that is,
plants can produce chemical energy in the form of a carbohy-
drate from CO and H O with the help of light (1-5). Although
2 2
the
conversion efficiency is not high, the plant photosynthesis
contains many important processes which we have to mimic

in
order to develop an artificial photosynthesis with a higher
energy conversion efficiency. It is, therefore, indispensable to

----------------------- Page 476-----------------------
SOLAR ENERGY CONVERSION 639
know the mechanism of plant photosynthesis. Next, we give

avoided, and chemical energy storage is achieved. For
several examples of artificial photosynthesis using heteroge-
instance, nitrosyl chloride (NOCl) is photodissociated
neous catalytic systems and wet-type solar cells using semi-
into nitrogen monoxide (NO) and chlorine (Cl), and then
conductor photoelectrodes. This research field was boosted by
the Cl atom reacts with another NOCl molecule to form
Fujishima and Honda (6) who found that slightly reduced
a NO molecule and a Cl2 molecule.
TiO2, a semiconducting oxide, can decompose water into hy-
drogen and oxygen under illumination of near ultraviolet
NOCl + hν → NO + Cl (6)
light. Again, the problem was its low conversion efficiency.
NOCl + Cl → NO + Cl2 (7)
We overview the principle of the process and the recent prog-
ress in this field.
2NOCl + hν → 2NO + Cl2 (8)
Since Cl2 is soluble in liquid carbon tetrachloride

MOLECULAR PHOTOCHEMICAL ENERGY
(CCl4) while NO is not, the photochemical products Cl2

STORAGE AND PHOTOSYNTHESIS
and NO can be separated; that is, chemical energy stor-
age is achieved. It is also the case with photoionization

Molecular Photochemical Energy Conversion
and charge-transfer reactions. Photoionization reaction
When molecules or complexes are photoexcited and trans-
products M# and e# in Eq. (4) are easily recombined in
formed into products possessing higher energy than the ini-

the gas phase. When photoionization is carried out in a
tial state, a portion of the energy of photon is recognized to be
condensed phase matrix, however, the electron can be
converted into chemical energy (Fig. 1). Such photochemical
stably trapped in the matrix. For example, when aro-
reactions of molecules and complexes include (1):
matic compounds or amines are photoionized in a hy-
drocarbon matrix frozen at liquid nitrogen temperature,

1. photodissociation:
a portion of the photogenerated electrons are known to
AB + hν → A + B (1)
be trapped stably in the frozen matrix.
2. photoisomerization (cis-trans etc.): 2.
Specific paths should be present that convert the chemi-
Mcis + hν → Mtrans (2)
cal energy into nonthermal energies. When the photo-
chemical reaction products are back to the initial state

3. photoaddition reaction:
by reverse reaction, thermal energy is released; that is,
A=B + C-D + hν → AC-BD (3)
the chemical energy stored in the photochemical reac-
tion products is converted into thermal energy. If the

4. photoionization:
thermal energy is the final, aimed energy to be ob-
M + hν → M+ + e− (4)
tained, one can directly utilize the solar energy as the
heat source without employing such a lossy solar-to-

5. charge transfer:
A + B + hν → A− + B+ (5)
chemical-to-thermal energy conversion route. Utiliza-
tion of electrons obtained via photoionization (d) and
charge-transfer (e) reactions for a specific redox reaction

In order to utilize the chemical energy of the right side prod-
is an example of a fascinating molecular energy conver-

ucts as the energy source, the following three conditions
should be satisfied.
sion process. In this case, however, the reverse reaction
should be avoided, of course, and the photogenerated
electrons should be transferred immediately to the re-

1. The reverse reaction should be prevented. When a mole-
dox system of interest. In order to realize this situation,

cule AB in Eq. (1), for example, is photodissociated in
# #
again the photochemical reaction products [(M and e )

the gas phase, the reverse reaction A + B # AB can
# #
and (A and B )] should be separated, and in addition,

easily occur in general, and the chemical energy stored
there should be appropriate electron acceptors spacially

in A and B is released as thermal energy. However, if
close to the reaction products. As easily imagined, in

the photochemical reaction products A and B can be
homogeneous systems in general, the electrons obtained

separated and stabilized, the reverse reaction is
by photochemical reactions are difficult to be utilized
efficiently for specific redox reactions. Well designed,
heterogeneous systems should be employed for such en-

Excited state
ergy conversion processes.
3.
If solar energy conversion is aimed, the visible light
Product
y
should be
used for photochemical reactions. In most
g
r ν
cases, however, photodissociation (a) and photoioniza-
e h
n
tion (d) proceed only with ultraviolet light. In such a
E U
case, only a minor portion of solar energy can be utilized
for energyconversion, and the total energy conversion

Ground state
efficiency becomes quite low.
Figure 1. Energy diagram of molecular photochemical reaction. Mol-

Photosynthesis in natural plants is an ideal solar energy con-
ecule(s) and complex(es) are photoexcited from ground state to excited
version system that satisfies all the conditions mentioned
state with light energy hn and transformed into product possessing above.
Through learning and mimicking photosynthesis in
free energy U.
plants, a growing number of researchers are now designing
----------------------- Page 477-----------------------
and constructing artificial photosynthesis systems to develop

efficient solar energy conversion systems.
Grana Lamella
Photosynthesis
Photosynthesis and Solar Energy Conversion. Green plants
are converting CO from the atmosphere and H O from soils
2 2
into organic substances, utilizing solar energy. This is what
Stroma
we call photosynthesis of green plants (2). The energy stored
in organic substances produced by plants is about 0.1% of the
Chloroplast
total annual solar energy reaching the Earth and is over 10
times the total amount of energy consumed by human activi-
ties per year. Organic substances produced by plants serve as
Thylakoid
food for human beings and animals, which is the energy
membrane
source for activity and growth, organic raw materials for in-
dustrial production, and fuels. Fossil fuels can also be said to
have resulted from the photosynthesis activity of ancient
plants.
Thylakoid
The net overall chemical reaction occurring in photosyn-

Figure 2. Schematic illustration of chloroplast and thylakoid. A bio-
thesis of plants is represented by
molecular membrane separates the chloroplast from the other compo-
nent of theplant cell. Lamella and grana are composed of saclike

H O + CO → 1/6(C H O ) + O (9)
disks called thylakoids. Absorption of photons, electron transfer,
2 2 6 12 6 2 H
O oxidation, and ATP synthesis occur on the thylakoid membrane,
and CO2 fixation takes place in stroma, the substrate part of the chlo-
where C H O is hexose. Thermodynamically, the right side
roplast.
6 12 6
reaction products have an energy 475 kJ per mole of CO2

higher than the left side reactants, and this uphill reaction is
assisted by solar energy. In other words, plants convert solar
membrane (Fig. 2). All the electron transfer processes in pho-
energy into chemical energy in the form of hexose and oxygen
tosynthesis take place in chloroplasts. In chloroplasts, sack-
molecules. Since photoexcitation of eight moles of chlorophyll
like disks called thylakoid are present, forming lamella and
grana as shown in Fig. 2. Absorption of photons, electron

is utilized per mole of CO2, and the absorption band of chloro-
transfer reactions, H2O oxidation, and ATP synthesis (to be

phylls lies around 700 nm (#170 kJ), the photoexcitation en-
discussed later) all take place on the thylakoid membrane. On

ergy utilized for converting one mole of CO2 is calculated to
the other hand, CO2 fixation (reduction into hexose) occurs in
be 170 # 8 # 1370 kJ. If each reaction in the photosynthesis
the substrate part of chloroplasts called stroma.

is assumed to occur with a quantum yield of 1, the energy
Mechanism Details. The electron transfer processes in-

yield is calculated to be 475/1370 # 0.35, which is a much
volved in Eqs. (10) and (11) in chloroplasts are described be-

higher value than those obtained in photochemical reactions
low. Figure 3 shows a schematic illustration of thylakoid

of simple molecules. Furthermore, the product, hexose, is a
membrane along with electron and proton flows. Figure 4

quite stable compound, which is another advantage of photo-
shows electron flow and the chemical reaction scheme repre-

synthesis.
sented with a redox potential scale. As shown in Fig. 4, photo-
synthesis consists of two photoexcitation steps (uphill, light

Mechanism of Photosynthesis
reaction) in pigment systems called PSI (photosystem I) and
A Brief Review of Mechanism. The overall reaction in photo-
PSII (photosystem II), and a series of electron transfer steps
synthesis represented by Eq. (9) is a redox reaction and can
(downhill, dark reaction) in the other parts.
be divided into two reactions:
PSI and PSII in green plants’ chloroplasts contain a vari-
ety of pigments. In particular, chlorophyll a plays an impor-
tant role in photosynthesis. Chlorophyll a is a Mg-porphyrin

2H O→O + 4H+ + 4e− (oxidation of H O) (10)
2 2 2
compound shown in Fig. 5, and is distributed in a wide vari-
CO + 4H+ + 4e− → 1/6(C H O ) + H O
ety of states in thylakoid membranes, which gives rise to a
2 6 12 6 2
(reduction of CO2 ) (11)
broad optical absorption band of chloroplasts. P700 and P680
are chlorophyll a molecules of specific states called reaction
center chlorophylls in PSI and PSII, respectively. The fraction

indicating that photosynthesis is an event occurring in plants of
the reaction center chlorophyll molecules are about 0.3–
where electrons are transferred from an H2O molecule to a
0.5% of the whole chlorophyll molecules. Chlorophyll b and
CO2 molecule with the aid of solar energy. In other words,
other auxiliary pigments are also contained in these pho-
photosynthesis involves photoinduced electron transfer reac-
tosystems.
tions.
In PSII, the energy of light absorbed by pigments is trans-
H2O and CO2 molecules, which are absorbed through roots
ferred from pigments to pigments and is finally transferred to
and stomata (pores) in leaf surfaces, respectively, reach pig-
P680, which is thus photoexcited. The photoexcited electron
ment-bearing organells called choloroplasts, which are sepa- in
P680 is transferred successively to pheophytin (a phorph-
rated from other components of plant cells by a bimolecular
yrin compound, Pheo), plastoquinone (a quinone compound,
----------------------- Page 478-----------------------
+ +
NADP + H NADPH
ADP + P ATP
PSII
PSI FD
Outside
QA
QB
ν
h
Pheo
PQ
ν
Chl h Tylakoid
membrane
P680
Mn Cyt f
P700
Z
H+ H+
Inside
H O 1/2 O + 2H+ H+ PC
H+ H+
2 2
Figure 3. Schematic illustration of thylakoid

membrane and electron
flow and chemical reac-
tions. PSI and PSII contain P700 and P680, chlorophyll
molecules of specific states in thylakoid
membranes. In PSII, the energy of light absorbed by
pigments is transferred to P680. Then the
photoexcited electron in P680 is transferred to Pheo, Q ,
Q , P , Cytf, PC, and finally to P700 in
A B Q
PSI. P680 is recovered from photooxidized P680# by
accepting an electron from an H O molecule,
which is catalyzed by proteins containing Mn

ions, and the electron is transferred via Z to
P680#. In PSI, the energy of light absorbed by
pigments is transferred between pigments and
excites P700. The photoexcited electron is transferred to

Chl. The electron is then transferred to
FD via several acceptors, and then to NADP to reduce
NADP to NADPH, which is catalyzed by
FNR. Then NADPH# reduces CO into C H O ,
which consumes energy released in the reaction,
2 6 12 6
ATP # ADP # Pi .
PSI
– 1.0
Chl
– 0.8
PSII
Several
– 0.6 Pheo acceptors
ν
h
FD
– 0.4
QA ADP + Pi
– 0.2
QB NADP
NADPH+
ATP
V
/ 0
E PQ acceptor
ADP
ν
h
0.2 + Pi 1/12(C H O )
1/2 CO
6 12 6
2
0.4 ATP
Cyt f PC P700
0.6
0.8 H2O
Mn
Figure 4. Schematic
illustration of electron flow and
Z
1.0 1/2 O2 + 2H+ P680
chemical reactions in redox potential scale. See Fig. 3
legend.
----------------------- Page 479-----------------------
H2C = CH H CH3 The

reaction expressed by Eq. (15) proceeds via the so-called
C3
cycle (Calvin–Benson cycle) including several interme-
diates.
CH3 C2 H5 As
illustrated in Fig. 4, ATP is also synthesized in chloro-
plasts
by phosphorylation of ADP, namely, the reverse pro-
N N
cess
of Eq. (14). Since this ATP synthesis is achieved by utiliz-
H Mg H ing
energies released in downhill transfer of electrons that
have
photoexcited in PSI and PSII, this process is called pho-
N N
tophosphorylation. Solar energy is converted to chemical en-
ergy
here again and stored in ATP.
CH3 CH3
H
Artificial Photosynthesis. A reaction scheme of photosynthe-
H
sis
can be represented by the following equation:
CH2 H
hν
CH2 COOCH3
# ∗
COOC20 H39 O2
+ 4H+ C− T2 PS → PS
2
Figure 5. Molecular structure of chlorophyll a.
2H O C T− PS+
2 2 2 (16)
T1 C− CO2 + 4H+
T− C 1/6(C H O ) + H O
Q , Q , PQ), cytochrome f (a heme protein, Cyt f), plastocya-
1 1 6 12 6 2
A B
nin (a copper-protein, PC), and finally, to P700 in PSI. On the

other hand, P680 is recovered from photooxidized P680# by where
PS is the photoexcitation center (or sensitizer), T1 and
accepting an electron released from an H O molecule. The oxi- T2
are the electron transfer media (or relays), and C1 and C2
2
are
the catalysts for CO reduction and H O oxidation, respec-
dation reaction of H2O is now thought to be catalyzed by pro-
2 2
teins containing Mn ions, and the electron is transferred via
tively. In order to mimic the natural photosynthesis, appro-
a plastoquinol (Z) molecule to P680#. priate
sensitizers, relays, and catalysts, of course, should be
In PSI, the energy of light absorbed by pigments is trans-
employed. Efficient photoinduced charge separation and vec-
ferred between pigments and excites P700, the reaction cen- torial
electron transfer, however, are the essential parts of
ter chlorophyll. The photoexcited electron is transferred to an- the
natural photosynthesis. The simplest way to realize them
other chlorophyll molecule (Chl). The electron is then is to
use the surface potential of a semiconductor at the
transferred to ferredoxin (iron-copper-protein, FD) via several
semiconductor/liquid interface as the device for separating
the
charges, where semiconductors like TiO , SrTiO , Fe O ,
acceptors, and then to nicotinamideadenine dinucleotide
2 3 2 3
phosphate (NADP) to reduce NADP to NADPH (a reduced and
CdS work as the electron transfer media (and as the sen-
form of NADP), where ferredoxin-NADP reductase (FNR) cat-
sitizer), organic dyes or organometallic complexes as the sen-
alyzes the NADP reduction. Then, NADPH# reduces CO2 into
sitizer, and Pt, Pd, and Rh metals as the catalyst.
C H O , turning back to NADP.
Compartmentalization, that is, the arrangement of the
6 12 6
Thus, photosynthesis reaction is regarded to consist of oxi-
photosynthetic apparatus in the thylakoid membrane is re-
dation of H #
garded, however, as an essential feature of the natural photo-
2O by NADP and reduction of CO2 by NADPH ;
namely, Eqs. (10) and (11) can be rewritten as
synthesis in approaches mimicking more closely to the natu-
ral
process (3). So far, micelles, LB films, and vesicles have
2H O + 2NADP + hν → O + 2NADPH+ been
used as biomimetic membranes. Aqueous micelles are
2 2
(oxidation of H O by NADP) (12)
spherical aggregates (4 nm to 8 nm in diameter) of surfac-
2
tants,
which possess hydrophilic head groups and hydropho-
CO + 2NADPH+ → 1/6(C H O ) + 2NADP + H O bic
alkyl chains containing 6 to 20 carbon atoms. LB films are
2 6 12 6 2
(reduction of CO2 by NADPH+ ) (13) formed
by spreading surfactants over water, where the polar
headgroups of the amphiphiles are in contact with water,

The reaction represented by Eq. (13), however, does not pro- while
their hydrocarbon tails protrude above it. Vesicles are
ceed spontaneously because the right side products of this re-
spherical or ellipsoidal aggregates possessing aqueous inner
action are higher in free energy than the left side reactants. spaces
and composed of bilayer phosopholipids or surfactants,
In green plants, energy released by the following reaction is
typically having diameters on the order of 50 nm and a thick-
utilized to accomplish the reaction of Eq. (13) ness
of 5 nm. Vesicles composed of phosopholipids and syn-
thetic
surfactants are referred to as liposomes and surfactant
ATP (adenosinetriphosphate) →
vesicles, respectively.
ADP (adenosinediphosphate) + Pi (14)
For example, micelle-mediated charge separation and elec-
tron
transfer have been demonstrated in phenothiazine (PTH)
In other words, reduction of CO2 proceeds by
solubilized in sodium dodecyl sulfate (SDS) micelles. A larger
photoionization quantum
yield of PTH was obtained in the
+ SDS
micelles than in homogeneous methanolic solutions, and
CO + 3ATP + 2NADPH → 1/6(C H O ) + 2NADP
2 6 12 6
electron transfer from PTH in the micelles to Cu2# and Eu3#
+ 3ADP + 3P + H O (15)
i 2
adsorbed on the micellar surface occurred at higher rates
----------------------- Page 480-----------------------
than in homogeneous solutions. Vesicle-mediated electron

Reduction nH+ CO2 Loaded metal
transfer has also been recognized from ethylenediaminetetra-
acetate, EDTA, to methylviologen, MV2#, via tris(2,2-bipyri-
process
dine)ruthenium(II), Ru(bpy)2#, a sensitizer, where EDTA was
3
distributed in the bulk aqueous solution, and MV2# and
– – – – ν
h
Ru(bpy)2# were placed on the inner and outer surfaces of di-
CO
3
HCHO
hexadecyl (DHP) vesicles, respectively.
CH3OH Eg
Intramolecular photoinduced charge separation and elec-
CH4
tron transfer is another challenge. Electron transfer has been
+ + + + O2
recognized in a synthesized triad consisting of a tetraaryl-
porphyrin (P) covalently linked to both a quinone (Q) and a
+4H+
carotenoid polyene (C). Excitation of C-P-Q by a picosecond
Semiconductor 2H2O
Oxidation
laser pulse resulted in a rapid formation of charge-separated
process
species, C# # -P-Q# # , with a quantum yield of 0.25 and a life-
time of several microseconds. Very recently, photoinduced
Figure 6. Schematic diagram of a photocatalysis process for reduc-
electron transfer with a quantum yield of 0.83 and a lifetime
tion of CO2 to methane and others by a semiconductor photocatalyst
of 55 ms has been achieved in a molecular pentad consisting
loaded with a metal or metal oxide.
of a porphyrin dyad covalently linked to a carotenoid polyene

and a diquinone moiety (4).
There is another challenge combining the membrane and
from reductants to oxidants. In the overall process, the semi-
intramolecular approaches (5). A molecular triad consisting of
conductor photocatalyst remains unchanged since both the
carotenoid polyene (C), an electron donor, and a naphthoqui-
electrons and the holes are simultaneously consumed.
none moiety fused to a norbornene system bearing carboxylic
acid group (Q), an acceptor, linked to a photosensitive tetraar-
Photocatalysis by Semiconductor Particles
ylporphyrin (P). This triad was incorporated into the bilayer
Small size semiconductor particles are effective for photocata-

of a spherical liposome. When the triad incorporating the lipo-
lytic reaction from the following view points: (1) increase in

some was photoexcited, reduction and oxidation potentials
the
number of adsorption sites based on the increase in the
were established near the outer and inner surfaces of the bi-
surface area, (2) increase in the reduction and oxidation abili-

layer, respectively. In response to this redox potential across
ties by the shifts of the energy levels due to the quantum

the bilayer, a freely diffusing quinone molecule alternates be-
size effect, and (3) decrease in the duration for photoinduced

tween its oxidized and reduced forms to convey protons across
charges to reach the surface, which correlates reciprocally

the bilayer with an overall quantum yield of 0.004, creating a
with the surface density of the charge calculated with the fi-

pH gradient between the inside and outside of the liposome.
nite depth well model. Small particles, however, decrease in

the
space required for the photoinduced electron-hole pairs to
HETEROGENEOUS PHOTOCATALYSIS be
separated, resulting in their easy recombination.
Semiconductors loaded with metals effectively separate

Heterogeneous photocatalysis is a photochemical process in
electron-hole pairs. Metal particles on the semiconductor col-
which the catalyst is present in a separate phase. In a general
lect electrons which are excited by irradiation and pass them
case, the catalyst is a solid, the reactants and products are in to
reactant. The loaded small metal particles which are dis-
gaseous or liquid phases, and the catalytic reaction occurs on
persed on the surface of the semiconductor thus improve the
the surface of the solid.
catalytic activity of the semiconductor in the consequent
chemical reactions. The activity and reaction selectivity are

Heterogeneous Photocatalysis Process
found to strongly depend on the metal and method of prepara-
tion (metal loading). For example, the Pt/TiO2 photocatalyst

In analogy with chlorophyll which acts as photocatalysts for
is
highly active for dehydrogenation of organic compounds
the photosynthesis of plants, dyes or semiconductors, which
compared with plain TiO2 .

are stable and relatively easy to handle, are used as photocat-
alysts in the artificial photosynthesis. The photocatalytic
Photosplitting of Water
mechanism in the photosynthesis involving water oxidation
and CO reduction is explained on the basis of electron trans-
Figure 7 shows the positions of a redox potential of H O and
2
2
fer mainly consisting of the following steps: (1) water oxida-

flatband potentials of various semiconductor catalysts. For
tion catalytic reaction (water as electron source), (2) electron
complete splitting of water, the conduction band edge must
transfer in photoexcited state (separation of photoexcited elec- be
more negative of the reduction potential of H# for the gen-
tron and hole), and (3) CO reduction catalytic reaction (syn-

eration of H , and the valence band edge more positive of the
2
2
thesis of carboxyl materials). Figure 6 shows the simplified

oxidation potential of H O to generate O . Semiconductors
2 2
mechanism of the photocatalysis by a semiconductor photo-

with a larger bandgap, however, have serious problems that
catalyst whose surface is loaded with a metal or metal oxide.
they cannot efficiently utilize solar energy, and photocorro-
As the semiconductor is irradiated by light with an energy
sion is liable to occur in aqueous solution as found for CdS.
larger than its bandgap Eg, electrons and holes are generated
Stoichiometric water photosplitting was first confirmed us-
in the conduction band and the valence band, respectively. ing
Pt/TiO2 covered with a very thin layer of water. The
The photoinduced electrons react with oxidants and the pho-
loaded metal is, however, also so active for the back reaction
toinduced holes with reductants, resulting in an electron flow of
hydrogen with oxygen that sustained evolution of hydrogen
----------------------- Page 481-----------------------
2
O
P
r
a e
Z
G S
–1.0
e d i
S T C S
3
d d
O
C C 5 3
i
2 3 2
T
O 3 O
r O
2 O 2 O O
S i
b O 2
T
V V V V n e i n
e e e e N Z W F B S
0.0
H /H O
2 4 7 1 2 2
. . . .
2 1 1 1
V V V
V Conduction band
e e e
e
V V V V V V
0 2 0
5 #e e e e e e
V . . .
.
e +1.0 5 3 3
2 4. 2. 5. 2. 6. 5.
/
3 3 2 2 2 3
E
H
O /H O
2 2
N
s
v
V ####
+2.0 ##
####
+3.0
Figure 7. Relationship between the redox potential

of H O and flatband potentials; bandgaps of various
2
semiconductor photocatalysts. +4.0

Valence band
##
###
and oxygen hardly occurs in aqueous suspensions of met-

possibility of photocatalytic fixation of CO2 has been shown by
allized semiconductors. Covering the surface of Pt/TiO2 with using
metal complexes such as [Ru(bpy)3]2#, semiconductors,
a layer of NaOH increases the quantum efficiency up to some

of which are listed in Fig. 7, and dye-doped semiconduc-
17% (7). tors
or metal hybrids which can adsorb visible parts of solar
Some layered niobates such as Ni-loaded K Nb O decom-
light. In addition, uses of photocatalysts as deodorant, bacte-
4 6 17
pose water into H2 and O2 in the stoichiometric ratio since the ria
killer, and their applications in organic synthesis and even
products of H2 and O2 are separated between different layers in
cancer treatment are currently very attracting.
to restrain their recombination. Ion-exchangeable layered
Perovskite type niobates show interesting photocatalytic ac-
tivity with visible response (8). Niobates and titanates with WET-
TYPE SOLAR CELLS
characteristic tunnel structures have been shown to be poten-

tial candidates as efficient photocatalysts for the splitting of Wet-
type solar cells convert solar energy into chemical and/
water. Pb K Nb O , M Ti O (M # alkaline metal atom or or
electrical energy by utilizing light to drive or promote the
1#x 2x 2 6 2 n 2n #1
proton, n # 3,4,6,8) and BaTi4O9 are the examples (9).
photoelectrochemical (PEC) redox reactions in the electrolyte
An Na2CO3 addition to Pt-TiO2 suspension in water pro- and
at the metal or semiconductor electrodes. The cell at-
motes the stoichiometric decomposition of water with the

highest H2 evolution rate of 568 #mol/h known so far (10).
The Na2CO3 addition has proved to be very useful to acceler-
R
ate water splitting for various oxide semiconductor catalysts.
A
The formed carbonate radicals in the system produce peroxy-
carbonates by self-coupling or reacting with OH # radicals, re-
V
sulting in the irreversible evolution of O2 gas and the easy
detachment from the surface of the catalysts. Decomposition
3
of water by an AM1 solar UV light irradiation has been tried
+ 1 2 –
for the first time and found to take place efficiently as well as
stoichiometrically in the presence of NiOx -TiO2 catalyst (11).
Dye 4
Dye
The present problems in photocatalysts for the complete wa-
h
ter splitting are the low efficiency and its difficulty in utilizing
– Diffusion
the visible light.
e Ox
Dye* Dye e–
Other Applications
+
Photocatalysts, especially TiO2 due to its oxidizing power and
Diffusion
photochemical stability, have been utilized for various pur-
Ox Red Red
poses such as air purification, CO2 reduction, and NOx fixation
[Ox] < [Red] [Ox] > [Red]

which all concern the environmental problem. The increase
of CO2 in the atmosphere and the accompanied global warm-
Figure 8. Photogalvanic cell. (1) and (2): metal electrodes, (3): porous
ing problem have been highlighted in the last decade. The
diaphragm, and (4): electrolyte.
----------------------- Page 482-----------------------
tracts much attention because of its low cost and promising

R
utilization.
A
Photogalvanic Cells
V
The cell (Fig. 8) consists of two metal electrodes separated by

a membrane, an electrolyte containing a dye as sensitizer,
– 2 1 CB e– +
ν
and a redox system either as an electron acceptor or as an
3 h
–
electron donor. In such a cell, light is absorbed by the dye.
e
The excited dye sensitizes the redox reaction, and homoge-

Red
neous photochemical reaction occurs at the electrodes, form-
Ox
+
ing products. The formed products then give or accept elec-
VB p
trons to or from electrodes, generating photovoltage. Using
this cell, one can only convert solar energy into electrical
energy.
Taking an electrolyte containing a redox couple Fe3#/Fe2#
(a)
and dye Ru(bpy)2# as an example, the following equilibrium
3
is established in the dark:
R
A
Fe3+ + Ru(bpy)2+ ⇔ Fe2+ + Ru(bpy)3+ (17)
3 3
V
When the system is illuminated by light, the dye is excited,

3#
and an electron transfer occurs from the excited dye to Fe :
+ 2 1 CB e– –
2+ ∗ 2+
3
Ru(bpy)3 + hν → Ru(bpy)3 (18)
e– Ox
h
Fe3+ + ∗Ru(bpy)2+ → Fe2+ + Ru(bpy)3+ (19)
3 3
Red
+
The equilibrium in the dark is then broken. The resultant
VB p
increase of the concentration of Fe2# and Ru(bpy)3# changes
3
the redox potential: that of Fe2#/Fe3# is shifted negatively,
while that of Ru(bpy)2#/Ru(bpy)3# is shifted positively. As a
3 3
2#
(b)
result, Fe gives an electron to one of the metal electrodes,
and Ru(bpy)3# accepts an electron from the rest metal elec-
Figure 9. (a) n-type and (b) p -type semiconductor photoelectrochemi-
3
trode, giving rise to a photovoltage between the two metal

cal cell. (1): semiconductor electrodes, (2): metal counter electrodes,
electrodes. Besides Ru(bpy)2#, thionine, methylene blue, and
and (3): electrolyte. See Fig. 10 legend.
3
acridine orange are often used as sensitizers.

It should be, however, pointed out that a photovoltage can
be expected only when the incident light is absorbed near the
obtain a band model of semiconductor PEC cells in the dark
one electrode because generation of a photovoltage depends
as shown in Fig. 10(a) for n-type and Fig. 10(c) for p -type. An
on a competition for redox reaction and electron exchange be-
electrical field arises in the semiconductor across the space
tween the electrode and the redox species. It is necessary to
charge layer because of the low carrier density. An electron
make a difference in the concentration of an electrode-active
is excited from the valence band to the conduction band
component between two identical electrodes so that the elec-
of a semiconductor by a light whose energy is larger than the
tron can exchange between the electrode and the redox spe-
bandgap of the semiconductor, leaving a hole in the valence
cies. For this reason, the conversion efficiency of this type cell
band. The electrical field in the semiconductor separates the
cannot be in substance very high. Its maximum power is cal-
electron-hole pair as shown in Fig. 10(b) and (d), producing
culated to be 0.7 Wm#2 when the incident light has an average
chemicals and/or generating electrical power accompanied by
terrestrial solar radiation of 30 Wm#2, and quantum efficiency
the change or lack of change of the electrolyte composition, re-
of unity is assumed for the photochemical process (11).

spectively.
Semiconductor Photoelectrochemical (PEC) Cells
Regenerative or Photovoltaic Cells. In this cell, light which

The cell (Fig. 9) consists of a semiconductor electrode as a
irradiates the semiconductor/electrolyte interface is converted
working electrode and a metal electrode as a counter elec-
into electricity with no change in the composition of the elec-
trode which both are immersed in an electrolyte containing
trolyte or the semiconductor. The band model of this cell is
redox couples (6). In such a cell, light is absorbed by a semi-
also presented by Fig. 10. In the dark, at equilibrium [Fig.
conductor electrode or by a dye molecule adsorbed on the
10(a) and (c)], the Fermi level is equal throughout the system,
semiconductor electrode. Using this type of cell, the solar en-
including the back side ohmic metal contact of the semicon-
ergy can be converted into either electrical energy (regenera-
ductor and the metal counter electrode. When such a system
tive or photovoltaic cells) or both chemical and electrical en-
is illuminated, electron-hole pairs are generated and sepa-
ergy (photoelectrosynthetic or photocatalytic cells).
rated by the electric field, leading to a decrease of the band
One can assume the semiconductor/electrolyte interface to
bending and producing a photopotential, U , as indicated in
ph
be similar to the semiconductor/metal contact. Therefore, we

Fig. 10(b) and (d). Connecting the ohmic contact to the
----------------------- Page 483-----------------------
E ' E C
F, m
eU E
E C
ph F, S Ox e–
E E E E
E E
F, m F, S F, red ox F, m
F, red ox F, m
EV
Red
p
EV
Metal Semiconductor Electrolyte Metal

E = E = E
eU = E ' – E = E – E
F, m F, S F, red ox
ph F, m F, m F, S F, red ox
(a)
(b)
E C
E C e–
Ox Red h+
E E
E F, m E F, S E F, red ox E F, m
F, red ox F, m
EV
eU
ph
(b)
E
F, S
E 'F, m EV p+

E = E = E
eU = E – E ' = E – E
F, m F, S F, red ox
ph F, m F, m F, red ox F, S
(c)
(d)
Figure 10. Energy scheme of a semiconductor

photoelectrochemical cell. (a) n-type, in the dark,
(b) n-type, under illumination, (c) p -type, in the dark,
and (d) p -type, under illumination. When
the cell is illuminated by light with energy greater than
the bandgap of the semiconductor, the
semiconductor electrodes absorb the photons. Electrons
are excited from the valence band to the
conductor band. The electric field in the space
charge layer separates the electrons and holes
thus generated, forming a photovoltage, U .
ph
counter electrode by a wire or a small load resistance, pho-

The photocurrent iph depends linearly on the light inten-
tocurrent flows through the external circuit, and redox reac-
sity, I :
tions occur at the electrodes while the electric field remains
unchanged upon illumination. In the case of n-type electrodes
iph ∝ I (22)
as shown in Fig. 10(b), the electrons excited by light reach
the counter electrode, through the external circuit, and are
and the photopotential U is given by
ph
transferred to the oxidized species there:
Uph = (kT /e) ln(1 + #p /p 0 ) ∝ (kT /e) ln I (23)

Ox + e− → Red (20)
for #p # n , where p is the minority carrier density in the
0 0
On the other hand, the holes created by light are transferred
bulk, #p is the increment of p 0 on illumination, and n0 is the
across the semiconductor/electrolyte interface to the reduced
majority carrier density. The highest possible photopotential
species:
depends on the band bending below the semiconductor sur-
face at equilibrium in the dark. Theoretically, for a heavily

Red + p+ → Ox (21)
doped semiconductor, U # E /e. Unfortunately, experi-
ph. max g
mental data (12) show that the photopotential is always con-

A similar situation occurs for p -type electrodes as shown in
siderably lower than the bandgap. One reason is that at equi-
Fig. 10(d).
librium, the Fermi level of the redox system is far above the
----------------------- Page 484-----------------------
valence band edge. Another possible reason is the recombina-

tioned above, a problem is that photoelectrochemically stable
tion at the surface states as shown in Fig. 10(b) and (d).
semiconductors have wide bandgaps and hence, visible light
Another topic of this cell is the stabilization of the semicon-
cannot be utilized, whereas narrow bandgap semiconductors
ductor surface. The Anodic dissolution process competes with
are often photoelectrochemically unstable. An enormous effort
the redox reaction, very often making many semiconductor
has, however, been made so far in order to enhance the con-
electrodes unstable. From the thermodynamic viewpoint, the
version efficiency.
semiconductor will be stable in water when the decomposition
Adsorption of dye on the semiconductor electrode surface
potential is much more positive of the O /H O redox potential.
and doping of transition metal in the semiconductor electrode
2 2
However, the thermodynamic data show that only wide band-

are the best known methods to extend the photoresponse of
gap semiconductors such as TiO2 and SrTiO3 meet the above
those photochemically stable, wide bandgap semiconductor
condition. Therefore, redox couples should be selected for the
electrodes to a visible region. A detailed description can be
PEC cell so that the redox potential is less than the decompo-
found in Ref. 13, and their simple mechanism is as follows.
sition potential of the semiconductor.
Adsorption of Dye on the Semiconductor Electrode Sur-
face. One can obtain a visible photoresponse for wide band-

Photoelectrosynthetic Cells and Photocatalytic Cells. In pho-
gap semiconductor electrodes by adsorbing a dye, such as ru-
toelectrosynthetic cells, light is used to drive the overall cell
thenium complexes, whose energy levels lie within the
reaction in a nonspontaneous direction (#G # 0), so that radi-
bandgap of semiconductor. Visible-light-induced current oc-
ant energy is stored as chemical energy (e.g., fuels). On the
curs when a visible-photon absorbed by a dye molecule gives
other hand, in photocatalytic cells, a reaction is driven in a
rise to electron injection into the conduction band of the semi-
spontaneous (#G # 0) direction by the light energy. Here, ra-
conductor. However, the conversion efficiency of the dye-sen-
diant energy is not converted to chemical energy but is used
sitization PEC cell is as low as about 1%, mainly because of
to overcome the activation energy of the process. In both
the poor light harvesting. Improvement of the light harvest-
cases, the composition of the electrolyte changes.
ing may be achieved by employing sol-gel-derived colloidal
Take photoelectrolysis of water as an example for the pho-
semiconductor electrodes, which will be described in the next
toelectrosynthetic cells. The band model of the open circuit
paragraph.
case for this cell can also be described, as in Fig. 10. It should
Sol-Gel-Derived Semiconductor Film Electrodes
be noted that in such a system, the position of the Fermi level
(14,15). These electrodes are characterized by many surface
of the electrolyte is not very well defined. It would be rather
states which mediate the electron transfer between the elec-
close to E0 and to the valence band at the surface if the
trode and redox species. Our recent research showed that a
O2/H2O
electrolyte is saturated with oxygen. On the other hand, it

large number of surface states shift the photoresponse to visi-
may be close to E0 and to the conduction band if the hy-
ble region. The same result was also obtained by Hagfeldt et
H2/H2O
drogen concentration is large. When the PEC cell is short-

al. (16). The sol-gel method provides an easy way to prepare
circuited, electrons in the conduction band are transferred to
film electrodes in a metastable phase such as anatase TiO2
the counter electrode, and H2 is generated there; whereas
that cannot be obtained by the traditional method. The elec-
holes in the valence band are transferred to the surface of the
tron mobility is known to be larger in anatase than in rutile
semiconductor electrode to generate O . When the Fermi lev-
TiO . The colloidal semiconductor film electrodes can be easily
2
2
els of metal electrode are far below E0 , the generation of
prepared by the sol-gel method (17). They have been found to
H2/H2O
H2 does not occur. Since no current flows, the band bending

exhibit a higher light harvesting ability even for diffuse day-
decreases, and a photopotential U is developed. This photo-
light because of the large surface area. The reported highest
ph
potential increases with increasing light intensity until the

overall light-to-electric energy conversion is 12% in such a
Fermi level of the metal is moved above E0 . Then, the
condition. The PEC properties of the colloidal electrodes are
H2/H2O
photocurrent starts to flow, and H2 and O2 are generated. It

sensitive to the size of the colloidal particles because of the
should be noted that the Fermi level can only be moved suffi-
barrier formed between the particles.
ciently upwards if the conduction band edge at the surface
Doping of Transition Metal in the Semiconductor Electrode. To
lies above E0 ; otherwise an externally applied voltage,
endow a wide bandgap semiconductor with a visible photore-
H2/H2O
Uappl, is necessary to shift the Fermi level. In this situation,

sponse, one can incorporate transition metals such as Cr, Co,
one still has an energy gain as long as U # E0 #
Mn, Fe, and Ni into the semiconductor. The mechanism is to
appl O2/H2O
E0 # 1.23 eV.
create energy levels within the bandgap to apparently ‘‘nar-
H2/H2O
For this cell, the problem of semiconductor stability also

row’’ the bandgap. This will, however, increase the recombi-
exists.
nation of the photogenerated electron-hole pairs at the same
time.
Improvement of the Conversion Efficiency. In order to
Doping of Noble Metal Nanoparticles in the Semiconductor
achieve a high conversion efficiency for the semiconductor
Electrode (18,19). One can obtain visible photoresponse by in-
PEC solar cells, the semiconductor electrode should meet the
corporation of noble metal nanoparticles or co-incorporation
following four requirements: (1) an optimum bandgap of 1.5 of
metal nanoparticles and dye molecules in the wide bandgap
to 2.0 eV, which responses to the solar spectrum, (2) a conduc-
semiconductor electrodes (18,19). The mechanism is based on
tion band edge more negative of the reduction potential of
the enhancement of the dye sensitization or visible-light-exci-
redox species in an electrolyte combined with a valence band
tation of electrons at the surface states within the bandgap of
edge more positive of the oxidation potential, (3) rapid charge
semiconductor by the surface plasma resonance of the metal
transfer kinetics to the electrolyte, and (4) strong resistance
nanoparticles. For example, visible photoresponse was found
to corrosion of the semiconductor surface. No semiconductors
for Au or Ag nanoparticles doped TiO2 electrodes, which was
investigated so far satisfy all of the requirements. As men-
ascribed to the electron excition from the surface states en-
----------------------- Page 485-----------------------
hanced by the surface plasma resonance of the metal nano-

methane from carbon dioxide. Solar irradiation excites the
particles.
light absorber immersed in water, such as [Ru(bpy)3]2# and
Doping of Semiconductor Particles in the Semiconductor Elec-

TiO2 ; then, electron transfer occurs between the light ab-
trode. Visible photo response is induced by a narrow bandgap
sorber and the redox species of water, producing H2 as fuel
semiconductor doped in the wide bandgap semiconductor elec-
and O2 . Redox catalysts are required for attaining high con-
trodes; for example, TiO2 electrodes can be sensitized by
version efficiency. Such a system can be regarded as a selfcell.
doped CdS particles. Besides extending the photoresponse of
For example, the system contains TiO2 particles as light ab-
the wide bandgap semiconductor electrodes, other attempts to
sorbers, and Pt and RuO2 adsorbed on TiO2 particles as redox
improve the conversion efficiency have been made.
catalysts, where Pt works as the cathode, and RuO2 works as
A protective coating film on the narrow bandgap semicon-
the anode.
ductor electrode is to cover an unstable narrow bandgap semi-
conductor electrode with a protective coating film, for exam-
ple, the CdSe electrode covered with a TiO2 coating film. The
BIBLIOGRAPHY
problem is that the Schottky barriers form at the interface
between the protective coating and the narrow bandgap semi-
1. H. Tsubomura, Chemical conversion of light energy, Kagaku So-
conductor, reducing the mobility of the photogenerated elec-
setsu, 12: 29–44, 1976 (in Japanese).
trons and holes.
2. M. Pessarakli (ed.), Handbook of photosynthesis, New York: Mar-
Mixed Semiconductor Electrode. It is possible to alter the
cel Dekker, 1997.
bandgap of the semiconductors by mixing two semiconductors
3. J. H. Fendler, Photochemical solar energy conversion. An assess-
with different bandgaps. A solid solution of wide and narrow
ment of scientific accomplishments, J. Phys. Chem., 89: 2730–
bandgap semiconductors, which is called a mixed semiconduc-
40, 1985.
tor, may simultaneously have a resistance against the anodic
4. D. Gust et al., Photoinduced electron and energy transfer in mo-
decomposition and a visible photoresponse. For example, our
lecular pentads, J. Am. Chem. Soc., 115: 11141–52, 1993.
recent research has found a conspicuous visible photore-
5. G. Steinberg-Yfrach et al., Conversion of light energy to proton

sponse for the sol-gel-derived Ti V O solid solution film
1#x x 2
potential in liposomes by artificial photosynthetic reaction cen-
electrodes.
tres, Nature, 385: 239–241, 1997.

Semiconductor Electrode with Gradient Bandgap. Our re-
6. A. Fujishima and K. Honda, Electrochemical photolysis of water
search also has shown that it is possible to improve the con-
at a semiconductor electrode, Nature, 238: 37–38, 1972.
version efficiency by preparing composite electrodes con-
7. K. Yamaguti and S. Sato, Photolysis of water over metallized

sisting of a wide bandgap semiconductor at the surface to
powdered titanium dioxide. J. Chem. Soc., Faraday Trans. 1, 81:

absorb ultraviolet light and a narrow bandgap at the interior
1237–1246, 1985.
to absorb visible light. In order to eliminate the emergence of
8. J. Yoshimura et al., Visible light induced photocatalytic behavior

the Schottky barrier at the interfaces, the formation of a gra-
of a layered perovskite type niobate, RbPb Nb O , J. Phys. Chem.,

dient bandgap semiconductor electrode is required (Fig. 11).
2 3 10
97: 1970–1973, 1993.
Fuel-Producing Photochemical Redox Reaction

9. Y. Inoue, Y. Asai, and K. Sato, Photocatalysts with tunnel struc-
tures for decomposition of water, J. Chem. Soc., Faraday Trans.,
Fuel-producing photochemical redox reactions use solar en-
90: 797–802, 1994.
ergy to generate a fuel, for example, hydrogen from water or
10. K. Sayama and H. Arakawa, Significant effect of carbonate addi-
tion on stoichiometric photodecomposition of liquid water into hy-

0.25
drogen and oxygen from platinum-titanium (IV) oxide suspen-
sion, J. Chem. Soc.: Chem. Commun., 150–152, 1992.
TiO film Gradient bandgap Ti V O

11. H. Arakawa, Creation of new energy by photocatalyst: Technol-
2 1–x x 2
2 0.20 electrode film electrode
ogy of clean hydrogen fuel production from solar light and water,
–m
Laser Res., 25 (6): 425–430, 1997 (in Japanese).
c 0.1
1–
12. R. Memming, Solar energy conversion by photoelectrochemical
W 0.15
process, Electrochim. Acta, 25: 77–88, 1980.
A x
/
t
13. H. O. Finklea, Semiconductor Electrodes (Studies in Physical and
n
e
r 0
Theoretical Chemical), New York: Elsevier, 1988.
r 0.10
u Substrate Thickness Surface
c
14. T. Yoko, K. Kamiya, and S. Sakka, Photoelectrochemical proper-
o
t Depth profile of atomic fraction
ties of TiO film prepared by the sol-gel method, Yogyo-Kyokai-
o
2
h
P 0.05
Shi (J. Ceram. Soc. Jpn.), 95: 150–155, 1987.
15. T. Yoko et al., Sol-gel derived TiO film semiconductor electrode
for photocleavage of water: preparation and effects of post-heat-

0.00
ing treatment on the photoelectrochemical behavior, J. Electro-
300 350 400 450 500 550 600
chem. Soc., 138: 2279–2285, 1991.
Wavelength/nm
16. A. Hagfeldt, U. Bjorksten, and S.-E. Lindquist, Photoelectro-
chemical studies of colloidal TiO2 Film, Solar Energy Mater. Solar

Figure 11. An action spectrum of a gradient bandgap Ti V O film
1#x x 2
Cells, 27: 293–304, 1992.
electrode that is compared with one of a pure TiO2 film electrode. The
thicknesses of Ti V O and TiO film electrodes are 145 nm and 110
17. B. O’Regan and M. Gratzel, A low-cost, high-efficiency solar cell
1#x x 2 2
nm, respectively. The composition of Ti V O film changes gradually
based on dye-sensitized colloidal TiO2 films, Nature, 353: 737–
1#x x 2
as shown in the inset. Compared with the TiO2 film electrode, the
740, 1991.
photoresponse of gradient bandgap Ti V O film electrode extends to

18. G. Zhao, H. Kozuka, and T. Yoko, Effects of the incorporation of
1#x x 2
a considerably long wavelength.

silver and gold nanoparticles on the photoanodic properties of
----------------------- Page 486-----------------------
SOLAR HEATING 649
rose bengal sensitized TiO2 film electrodes prepared by sol-gel

method, Thin Solid Films, 277: 147–154, 1996.
19. G. Zhao, H. Kozuka, and T. Yoko, Effects of the incorporation of

silver and gold nanoparticles of rose bengal sensitized TiO2 film
electrodes prepared by sol-gel method, Solar Energy Mater. Solar
Cells, 46: 219–231, 1997.
GAOLING ZHAO
Zhejiang University
HIROMITSU KOZUKA
HONG LIN
TOSHINOBU YOKO
Kyoto University
----------------------- Page 487-----------------------
} { { } }
HOME
●
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Solar Heating
Standard Article

Sanjay K. Roy1 and Hongtan Liu 1
1PhD Research Group, Inc., Coral Gables, FL

reserved.
Search
DOI: 10.1002/047134608X.W3037
❍ Acronym Finder
Abstract
Solar Radiation and Its Distribution
Energy Collection and Storage
Energy Conversion and Utilization
Modeling and Simulation

Resources

%20Conversion/W3037.htm17.06.2008 16:46:07
----------------------- Page 488-----------------------
SOLAR HEATING 649
Issues
relevant to each of these are discussed separately in
the
following sections.
SOLAR
RADIATION AND ITS DISTRIBUTION
Background
The
solar constant G , defined as the energy flux (or irradi-
ance)
incident on a surface perpendicular to the incoming so-
lar
radiation outside the atmosphere of the earth at the mean
earth–
sun distance, has been measured by numerous re-
searchers. Based on a range of satellite and other measure-

ments,
the estimated value of the solar constant is 1366
W/m2
3 W/m2 (2), corresponding to an equivalent blackbody
temperature of about 5780 K. Spectral analyses show that

about
98% of this energy lies between the wavelengths of 0.3
SOLAR HEATING #m and
4.0 #m. Seasonal variations in the irradiance due to
the
elliptical orbit of the earth are on the order of 3%. Addi-
The world’s supply of fossil fuels is finite and is expected to tional
fluctuations caused by sunspot activities over its ap-
be depleted in the foreseeable future. Though nuclear energy
proximately 11 year cycle are believed to be less than 1.5%.
systems are not similarly constrained, particularly since the
For engineering applications, the solar constant is as-
development of breeder reactors, nuclear waste problems still sumed
to be fixed even though small changes have been re-
remain unresolved. The future of fusion technology, where the corded
in addition to the normal variations described. This
fuel is essentially unlimited, is also uncertain at the present
assumption has no impact on solar system design because the
time. As a result, there is a great potential for alternative energy
incident on the surface of the earth differs signifi-
renewable energy sources, such as solar energy. cantly
from the solar constant. Scattering effects are caused
Humankind has a very long history of solar energy utiliza- by air
molecules (Rayleigh scattering, mostly at wavelengths
tion dating back to prehistoric times. The total solar radiation below
0.6 #m) and by dust/water particles (Mie scattering).
incident on the surface of the earth is on the order of 80 tril- In
addition, absorption by ozone, water vapor, and carbon di-
lion kW, which is more than 10,000 times the annual global oxide
has a strong impact on the overall transmission of radi-
energy consumption for all human activities at present (1). ation
through the atmosphere. As a result, the wavelengths
Although energy availability is not an issue, its direct use is of
interest for terrestrial applications are limited to the region
constrained by three factors: (1) the low energy flux (a maxi-
between 0.29 #m and 2.5 #m (Fig. 1, see Ref. 3).
mum of about 1 kW/m2); (2) the large variations and unpre-
Terrestrial solar radiation consists of beam (i.e., direct)
dictability of the incident terrestrial radiation due to localized and

diffuse components. Limited experimental studies sug-
weather conditions; and (3) the need to convert the energy to gest
that the spectral distributions of both components of ra-
forms that are more easily transported/stored (e.g., electric-
diation are very similar, though the diffuse component is
ity, chemical fuels). These increase the overall cost of energy
shifted slightly toward the shorter wavelengths because of
obtained from solar radiation, so that solar energy systems higher
scattering in this region. The actual spectral distribu-
are uneconomical for most large-scale industrial and commer- tion
and energy content of radiation at any given location are
cial applications.
ultimately determined by three factors related to the scatter-
In spite of these limitations, direct solar heating systems ing
and absorption processes. These include the ‘‘air mass,’’ a
have become more acceptable, and in many cases, commer-
cially viable because of the increasing economic and environ-
mental costs of energy. This is particularly true for many low
2250
e
temperature applications (#100
C to 150
C), such as water c 2000
n
a
and space (i.e., building) heating and industrial processes, i
d
1750
a
Extraterrestrial
such as distillation, salt/brine production, and drying. A num- r
r
i
ber of other applications, such as cooking, power generation, l )

1500 Rayleigh attenuation
a
r m
t
cooling/air conditioning, and detoxification/disinfection have c µ
1250 O O = 0.35 cm (NTP)
e 2
3 3
also been investigated over the years. Most of these are still p m
w = 2 cm
s /
1000
l
H O β
in the developmental phase though a few designs have been a W
2 Clean atmosphere, = 0
m (
750 Air mass = 1
commercialized to a limited extent. r
O
o
3 H O,CO
n
500 2 2
The applications of direct solar heating applications are t
H O
c
2 H2O
e
250 H O,CO
quite diverse. However, the basic design in all cases is deter- r
2 2
i
O3 H O
mined by three factors: D
0 2
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Wavelength ( m)
• the availability of solar energy at a given location
Figure
1. Extraterrestrial solar radiation and the effects of Rayleigh
• the energy collection and storage requirements
scattering and atmospheric absorption on the spectral distribution of

• the end use of the energy as related to the actual appli- beam
irradiance (From J. A. Duffie and W. A. Beckman, in Solar
cation
Engineering of Thermal Processes, Wiley, 1991. With permission.)

----------------------- Page 489-----------------------
650 SOLAR HEATING
measure of the length of the radiation path in the atmo- bright

sunshine averaged over the duration under consider-
sphere, the precipitable water content, and the dust content ation, and N
the average length of the day.
of the atmosphere. The
clear sky irradiation Hc required in this correlation
can be
estimated from equations relating the terrestrial nor-
Measurements of Solar Radiation mal
irradiance Gc (including the effects of latitude, longitude,
altitude,
and climate type) to the extraterrestrial value Go (6).
A knowledge of the amount of solar radiation received at a The average
length of the day N can be easily evaluated be-
given location surface is essential for the design, evaluation, cause it is
a direct function of the solar constant, the orbital
and optimization of solar thermal systems. Thus, instruments position of
the earth around the sun during the time period
based on photovoltaic detectors have been developed to mea- of
interest, and the actual geographic location (latitude and
sure the duration of bright sunshine. Similarly, instruments longitude)
under consideration. The parameter n and the em-
to measure the direct solar radiation flux (pyrheliometers or pirical
constants a, b have been obtained from regression
actinometers) and the diffuse and total solar radiation fluxes analyses of
solar radiation data for various locations around
(pyranometers or solarimeters) are also available. More re- the world
and are available in literature. For areas where
cently, extensive work has been done to develop satellite- these values
are unavailable, suitable estimates are obtained
based techniques for measuring solar insolation over large by using
data from locations with similar geographic and cli-
areas (4). matic
conditions.
Standard and secondary standards pyrheliometers use a In
recent years, a number of other models/equations for
comparator-type approach where the incident radiant energy estimating
the solar radiation have been developed. Many of
is compared with a known heat input. These devices are dif- these
involve the clearness or cloudiness index H/Ho (Ho being
ficult to use in practice, and operational devices are usually the
extraterrestrial irradiance), which helps eliminate the pa-
based on direct temperature measurements using multijunc- rameter Hc
from the calculations. In the simplest cases, these
tion thermopiles. Before use, these field instruments must be equations
may be based on simple linear regression analyses,
calibrated, and when carefully used, errors with well-cali- as
described previously. Other equations are based on an-
brated instruments of this type are quite small (on the order other
parameter, the average cloud cover C, which is derived
of 5% or less), though thin clouds or haze may increase errors from
empirical data or from more complex models. Consider-
to as much as 10% or more. able
research in this area is continuing at present (7).
Pyranometers also use thermopile detectors and are cali-
brated against standard pyrheliometers, though they may be
ENERGY
COLLECTION AND STORAGE
occasionally calibrated against secondary standard pyrano-
meters. Alternative pyranometer designs are based on bime-
tallic strip transducers or photovoltaic detectors. Though the A critical
part of any solar thermal system is the collector
spectral response of a silicon cell differs from the typical solar which is
designed to absorb solar energy incident on it. These
spectrum, errors introduced by this difference are believed to are often
used with energy storage systems because the en-
be less than 5%. However, it is important to note that the ergy demands
for most applications are time-varying and are
errors may be much larger if reflected radiation from the sur- rarely
correlated with the available solar energy. Various
roundings is a significant fraction of the total incident radi- types of
collectors and storage devices have been developed
ation. for
different applications. These devices and practical issues
The diffuse component of the radiation is measured with surrounding
their use are discussed in the following sections.
pyranometers modified to include a shading ring. In this case,
Preliminaries
correction terms on the order of 5% to 20% are usually neces-
sary because the ring shades the detector from some of the Collectors
are classified into two broad categories, nonconcen-
diffuse radiation (in addition to the direct component). Mea- trating or
flat-plate collectors and concentrator-receiver-type
surements of radiation on inclined surfaces are also made collectors.
In either case, it is first necessary to estimate the
with conventional pyranometers. However, the measure- solar
radiation incident on their surfaces before designing a
ments may need to be corrected by as much as 10%, de- solar
thermal system. For a horizontal surface, methods for
pending on the instrument used. estimating
the incident energy have already been discussed
in the
previous section. For inclined surfaces, the problem is
Estimates of Solar Radiation more
complicated, and the beam and diffuse components of
the incident
radiation must be considered separately.
Solar radiation data required for systems design are often un-
available. Thus, the average incident radiation is often esti- Radiation
on Inclined Surfaces. The beam component of radi-
mated from empirical equations developed for this purpose. ation is
obtained directly from the sun. Thus, it can be evalu-
These were originally of the following form (5): ated from
geometric considerations once the angles relative to
the
horizontal plane are known. Any specular reflection from
H n
= a + b (1) the ground
or the surroundings can also be similarly calcu-
Hc N lated as
long as the reflecting surface locations/configurations
are well
defined. The diffuse component is more difficult to
where H is the irradiation on a horizontal surface, (i.e., the evaluate
because a nonhorizontal surface receives radiation
energy incident per unit area integrated over a specific time by
scattering from the sky and by (diffuse) reflection from the
period e.g., year, month) where the subscript c refers to a ground.
Further complications arise because the diffuse com-
‘‘clear sky’’ terrestrial value, n is the daily number of hours of ponent from
the sky consists of three separate components,
----------------------- Page 490-----------------------
SOLAR HEATING 651
isotropic diffusion which is independent of the direction, cir- Design

Considerations. In an actual collector, the radiation
cumsolar diffusion which is concentrated in the solid angle incident
on the collector surface is not entirely absorbed by
region surrounding the sun, and the horizontal brightening the
absorber plate because of some absorption/reflection at
effect which is seen near the horizon. the glass
surface(s) and some reflection at the plate itself.
Various models have been developed to account for differ- Though
this can be calculated from the transmittance of the
ent components of diffuse radiation incident on inclined sur- glass
cover and the absorptance of the absorber plate, it is
faces (8). Simple approximations assume that the overall dif- convenient
to define an effective transmittance-absorption
fuse radiation, including both sky and ground reflected product
(##)eff combining the effects of the absorber plate-
radiation, is isotropic. In such a model, the diffuse radiation glass
cover combination. In most collectors this is approxi-
is independent of direction, and only the beam radiation geo- mated by
the expression
metric effects are important in evaluating the total incident
radiation. An improved model, the diffuse isotropic model, ne-
(τα)eff ≈ 1.01 − 1.02(τα) (2)
glects anisotropic scattering effects (i.e., the circumsolar diffu-
sion and the horizon brightening) but considers the ground where #
and # are the absorptance of the absorber plate and
reflection separately and assumes that it is isotropic. Thus, the
transmittance of the glass covers, respectively. Though
corrections based on radiation shape factors are used to ac- these must
include both spectral and directional effects, hemi-
count separately for the energy received from the sky and the spherical,
total values are typically used because of lack of
ground. Many more advanced models considering the aniso- adequate
information, particularly related to directional prop-
tropic scattering effects have also been developed in recent erties.
years. Nevertheless, the diffuse isotropic model, which gives Three
parameters are of importance in collector design.
conservative results, is still widely used because it is rela- The
collector loss coefficient UL is the overall heat transfer
tively easy to implement. coefficient
summing up the effects of convective and radiative
heat
losses from the absorber plate:
Flat-Plate Collectors
Q = A [S − U (T − T )] (3)
A solar collector is a heat exchanger that collects thermal en-
c L p a
ergy in the form of radiation and transfers it to a working

fluid. Flat-plate collectors are usually designed to be station- where Q is
the energy gain of the collector, S is the incident
ary (i.e., nontracking), and their orientation is selected to op- radiation
absorbed by the absorber per unit area, Tp is the
timize the average system performance during the year (or mean
temperature of the absorber plate, and Ta is the ambi-
portion of the year). In general, they are classified into two ent
temperature. The collector loss coefficient can be calcu-
broad categories, air heaters or liquid-plate collectors, de- lated from
a 1-D (or network type) analysis including the indi-
pending on the heat transfer fluid. Though the physical con- vidual
(conductive, convective, and radiative) heat transfer
figuration of the two types of collectors is somewhat different,
resistances.
the design principles are similar. Because
Tp is not known a priori and is somewhat difficult
A simple liquid collector shown in Fig. 2 consists of a to
evaluate, it is convenient to define a second parameter,
‘‘black’’ solar energy absorbing surface to which are attached namely,
the heat removal factor FR, a measure of the ‘‘heat
a number of fluid conduits. The conduits are mounted in an exchanger
effectiveness’’ of the collector. FR is the ratio of the
enclosure with transparent cover(s) which transmit incoming actual
heat transfer to the maximum heat transfer (that
solar radiation while reducing heat losses from the upper sur- would
occur if the fluid temperature remained constant):
face. Liquid collectors are typically designed to have selective
surfaces with high absorption at wavelengths below 2.5 #m
FR = Q (4)
A [S − U (T − T )]
and low emittances above 3 #m (for a surface at 200
C, 99% c L i a
of the emitted blackbody radiation is at wavelengths greater
than 3 #m). This ensures that a very high fraction of the inci- where Ti
is the inlet fluid temperature. Because the tempera-
dent radiation is absorbed and the energy emitted is rela- ture of
the fluid is always lower than the temperature of the
tively low. The overall enclosure is also well insulated at the absorber
plate, FR is less than (or equal to) the collector effi-
rear to further minimize any heat losses to the surroundings. ciency
factor F
. F
is defined as the ratio of the fluid-to-ambi-
ent heat
transfer coefficient Uo to the plate-to-ambient (i.e.,
the
collector loss) heat transfer coefficient UL :
Absorber plate Glass covers

Fluid tubes
F # = Uo (5)
Then the
mean plate and fluid temperatures are related to
the
previous three parameters as follows:
Q/A c # FR #
T = T + 1 − (6)
f i F U F #
R L
Insulation Collector structure

Q/A c
T = T + (1 − F ) (7)
P i F U R
Figure 2. Schematic of a simple liquid solar collector.
R L
----------------------- Page 491-----------------------
652 SOLAR HEATING
Practical Considerations. In a practical system, the collector the

absorber is equivalent to a heat exchanger that transfers
performance is affected by a number of factors that are diffi- the
redirected radiant energy to the working fluid flowing
cult to quantify. Limited experiments suggest that dust depo-
through it. Because of the concentration effects, however, the
sition on the collector cover plate reduces the energy absorbed
temperature of the working fluid can be raised to much higher
by 1% to 2%. Shading effects are also significant, but these
values than possible with flat-plate collectors. Thus, most ab-
can be evaluated only on a case-by-case basis. Shading by the
sorbers include vacuum barriers of various forms (Fig. 3) to
structure of the collector itself can reduce the incident energy.
minimize convective heat losses which are proportional to the
In most cases these are expected to be on the order of 1%
temperature difference between the absorber plate and the
to 3%.
surroundings. With a smaller absorber area relative to the
Most analyses related to solar collectors assume that a
collector (i.e., concentrator) area, these are therefore quite low
steady-state condition exists as has been done in the previous in
well-designed concentrator-receiver systems.
equations. In reality, a solar thermal system, and particularly
the collector, is a transiently operated device. Two effects are
Design Considerations. As in flat-plate collectors, the pa-
important here, the early morning heating of the system and
rameters of importance for a concentrator–receiver collector
the transient (and often intermittent) nature of the daily solar
include the loss coefficient, the efficiency factor, and the heat
radiation. Thus, time-dependent models for solar collectors
removal factor. However, two additional parameters, the con-
have also been developed and are sometimes used for detailed
centration ratio and intercept factor, are also important in
analysis. In most cases, however, the effects of variations in this
case. The first is usually based on area and is defined as
ambient temperature, wind speed, and solar insolation are
relatively small and are neglected in system design.
A a
C = (10)
A r
Performance Testing. Performance standards have been de-
veloped for solar collectors as their commercial market has
where A a is the aperture area (i.e., the projected area of the
increased in recent years. Three parameters are used to char-
concentrator) and A r is the receiver area. This is a measure of
acterize the collector performance: the instantaneous collector the
energy concentration whose maximum values are given
efficiency, incident angle modifier, and the time constant. The by
first is a measure of the energy absorbed by the collector rela-
tive to the energy incident on it:
1
Ccirc,max = 2 ≈ 45, 000

Q
sin θs (11)
η = (8)
1
A G
c
C = ≈ 212
lin ,max sin θs
where is the instantaneous efficiency. The second parame-

ter, the incident angle modifier K## gives the effect of the
where the subscripts circ and lin refer to circular and linear
orientation of the collector with respect to the normal direc-
concentrators, respectively, and #s is the half-angle subtended
tion: by
the sun as viewed from the earth (#0.27
). The intercept
factor # of a concentrator is the fraction of the radiation inci-

(τα) # 1 # dent
on the concentrator that is ultimately reflected onto the
Kτ α = = 1 + bo − 1 (9)
(τα)n cos θ
where b is the incident angle modifier coefficient. This empir-

o
Outlet tube Dewar flask Delivery tube Absorber tube
ically determined constant is obtained from tests with the col-
lector at different orientations with respect to the incoming
radiation. The final parameter, the time constant, is a mea-
sure of the heat capacity of the collector and is defined as the
time required for the temperature of the fluid to change by a
Inlet tube
factor of 0.632 ( # 1 # 1/e) relative to the ultimate steady-
state value.
(a)
Concentrator-Receiver Collectors
Inlet tube Dewar flask Absorber surface Heat transfer tube

Concentrator-receiver collectors consist of two components, a
reflecting or refractive surface known as the concentrator (9),
which is used to direct incident solar radiation to the second
component, the receiver. Because of the higher costs of lenses,
particularly for larger sizes, most large systems use mirror-
Outlet tube
type concentrators. These range from simple diffuse back-re-

flectors (a slight modification of conventional flat-plate collec-
tors), to systems using minimal tracking, nonimaging concen-
(b)
trators, to fully tracking, imaging-type parabolic or Fresnel

Figure 3. Two types of evacuated Dewar tubes with cylindrical ab-
concentrators or large arrays of heliostats.
sorbers. (a) Dewar with delivery tube. (b) Dewar with inserted fin and
The receiver itself includes the absorber, related structural tube
(From J. A. Duffie and W. A. Beckman, in Solar Engineering of
hardware, and thermal insulation. Like a flat-plate collector,
Thermal Processes, Wiley, 1991. With permission.)
----------------------- Page 492-----------------------
SOLAR HEATING 653
In
phase-change energy storage systems, the storage me-
dium
undergoes a change of phase during the heat
absorption/release process. Because of latent heat effects, en-

ergy
absorption and release occur at a more uniform tempera-
ture and
the size/mass of the energy storage unit can be re-
duced
significantly.
Figure 4. Compound parabolic concentrators with cylindrical ab-
sorbers. Chemical
energy storage systems are based on reversible
chemical
reactions. Common approaches are based on decom-
position of
metal oxides, dehydration of metal hydroxides, and
photochemical decomposition. As in phase-change type sys-

receiver. In most concentrator–receiver collectors, this is
tems, the
energy storage density can be increased signifi-
greater than 0.9.
cantly.
Both phase-change and chemical energy systems are
Practical Considerations. Concentrator–receiver designs are limited by
the cost of the storage material(s) and often by
driven primarily by the beam component of the radiation, ex- material
compatibility and related issues.
cept for systems with low concentration ratios (#10 approxi-

mately, e.g., diffuse back-reflector-type collectors). To reduce Combined
Collector-Storage Systems
or minimize tracking requirements, however, nonimaging
Solar
Ponds. The most widely used combined collector-stor-
concentrators have been designed to reflect radiation to the
age system
is a solar pond. Solar ponds are 1 m to 3 m deep,
receiver over a wide range of incident angles. For flat receiv-
large
bodies of water in which convection is suppressed by
ers, compound parabolic concentrators are commonly used,
increasing
the salinity at the bottom of the pond or by using
but involute reflectors provide better performance for cylindri-
honeycomb
membranes or polymer gels. By minimizing con-
cal absorbers (Fig. 4). The large acceptance angle of nonim-
vection,
energy absorbed by the bottom layers of water can
aging concentrators means that a fraction of the diffuse radia-
be stored
for extended periods of time because heat losses by
tion is also collected by such reflectors.
conduction
at the pond edges and through the upper layers of
Parabolic or paraboloidal mirrors (or lenses) are used as
concentrators in imaging systems. For a perfect mirror in per- water are
relatively small. As a result, a strong temperature
fect alignment, linear or circular images are obtained at the inversion
(with bottom temperatures as high as 70
C to 80
C)
focal point. Imperfect reflectors and/or alignment results in exists
within a well-designed solar pond.
an image larger than those expected in the perfect configura-
Experimental studies show that as much as 95% of the
tion. Thus, the concentration ratios of practical systems must solar
energy in the wavelength region of 0.3 #m to 0.6 #m is
be lower than those of theoretically perfect designs. These ef- absorbed in
the lower layers of a solar pond when the incident
fects are particularly evident in linear concentrator-receiver radiation
is approximately normal to the pond surface. Radia-
collectors where tracking is typically possible only in one tion at
higher wavelengths is absorbed in the upper regions
plane. As a result, the image size can vary significantly and may not
provide usable energy. Heat is removed from the
through the day, depending on the system orientation (E–W, pond by
heat exchangers submerged within the pond or by
N–S, etc.). extracting
hot water from the lower layers and returning it
Conventional mirror-based approaches are inadequate for after use.
In the second approach, special care must be taken
large solar thermal systems designed for power generation or to avoid
destratification and to maintain the stability of the
similar processes. For these applications, the system configu- pond.
ration typically includes a central tower-mounted receiver to- Saline
ponds with salt concentrations increasing with
gether with a field of independently controlled plane mirrors depth are
the most common type of solar ponds. The concen-
(heliostats) which direct the solar radiation onto the absorber. tration
gradients in these solar ponds result in continuous
Such a heliostat configuration is equivalent to a large Fresnel diffusion
of salt from the lower layers to the upper layers. To
mirror and provides enormous flexibility in system design. A maintain
the temperature and salinity gradients in the pond,
number of such systems have been tested worldwide. it is
therefore essential to add salt to the lower layers and
dilute the
upper layers as required. Water evaporating from
Energy Storage Systems the pond
surface must also be replaced continuously. To mini-
mize wind-
induced mixing, wave barriers in the form of nets
The availability of solar energy rarely coincides with energy
or pipes
are often necessary, particularly for larger ponds. To
demand for a given application because of the obvious tran-
improve
long-term performance, the ponds are lined with
sient nature of solar radiation and the time constant associ-
membrane
liners and/or layers of clay so that leakage is kept
ated with all solar thermal systems. To increase the applica-
to a
minimum.
bility of solar thermal systems, these are often used with
energy storage subsystems which store thermal energy when
excess energy is available.
Collectors With Integral Storage. On a smaller scale, conven-
Three basic types of thermal energy storage systems have tional
solar collectors have also been designed with integral
been considered (10). In sensible-heat, thermal energy storage storage to
simplify the overall system. Such integrated sys-
systems, heat is transferred to an energy storage medium tems have
been developed for flat-plate and concentrator-re-
with a consequent rise in its temperature. Typical storage me- ceiver
configurations. These may be entirely passive devices,
dia include water in an insulated tank and rocks in a packed- in which
water, the heat transfer fluid, may be stored in the
bed configuration. Special designs include storage walls for collector
itself. In other configurations, the storage subsystem
buildings or aquifier storage for seasonal applications. may be
incorporated in the lower section of the collector/re-
----------------------- Page 493-----------------------
654 SOLAR HEATING
ceiver, below the absorber plate, so that the operating temper- with
risers and headers made of special materials so that they
ature of the system is more uniform. can
withstand occasional freezing.
Auxiliary heating is almost inevitably required in solar

water
heaters. This is true even in the most ideal geographic
ENERGY CONVERSION AND UTILIZATION
locations because of the mismatch between available energy

and
demand. Auxiliary heat is usually supplied by a conven-
Solar thermal energy has been used for numerous applica-
tional heater that provides the energy required either in the

tions over the years. The most common are residential/com-
storage tank, to the water leaving the tank, or to a separate

mercial water and space (building) heating systems. In addi-
bypass, as required.
tion, solar energy has also been used for cooking and a range
A
number of specialized solar water heating systems have
of industrial and process applications, such as distillation,
been
proposed over the years. One example is the hybrid solar
salt/brine production, drying, detoxification, power genera-
collector, which converts incident solar energy to both ther-

tion, and absorption cooling. These are discussed further
mal
and electrical energy. The electrical energy generated can
here.
be
used to drive the pump or fan to circulate the fluid in the
collector system or for other purposes. Another example is the

Residential Water Heating low
flow rate, forced-convection system that has been devel-
Water heating is probably the most common solar heating ap- oped
to obtain increased stratification in the storage tank.
plication. A typical configuration of a residential water heat- Such
a design provides higher performance at significantly
ing system is shown in Fig. 5. Residential water heating sys- lower
cost.
tems are of two basic types, natural circulation and forced A
second large residential application of solar energy is
circulation. In natural circulation systems, density differences
swimming pool heating. The general requirements of this ap-
due to the temperature rise of the fluid in the collector(s) in-
plication are similar to those of water heating, though there
duces a flow in the system. Because this driving force depends are a
few significant differences. The most important of these
on the heat absorbed, natural circulation systems are self- is
the set-point temperature which is relatively low for a
adjusting to a certain extent. The temperature rise in the col-
swimming pool. On the other hand, the total thermal energy
lector is about 10
C over a wide range of operating conditions. required is usually
much larger than that for a residential
In forced convection systems, a separate collector-side heat water
heating system because the mass of water is much
transfer loop is often employed, and a separate pump trans-
greater. As a result, the simplest approach is to use the pool
ports the heat transfer fluid in this loop. With this approach,
itself as the collector by covering it with a solar transparent
it is possible to use a heat transfer fluid other than water. (and
IR opaque) plastic cover. Alternatively, separate collec-
This is particularly advantageous in regions subject to freez- tors
are used in series with the pool pump to heat the water
ing because then antifreeze mixtures (ethylene glycol–water, as
required.
propylene glycol–water) or air can be used in the collector
loop. Furthermore, because the boiling point of antifreeze
Space
(Building) Heating
mixtures is higher than that of water, such systems can also
be designed to operate at higher temperatures without in-
Building (residential or commercial) heating systems are of
creasing the system pressure. three
general types: passive, active, and hybrid systems. In
Freeze protection is an important issue for solar water
passive ‘‘solar houses,’’ building elements are designed so that
heaters in many parts of the world. If antifreeze mixtures
individual components also absorb, store, and release thermal
cannot be used because of cost or other constraints, systems
energy as required. In one approach, the building may include
may be designed so that the water can be drained back to the
collector-storage walls that absorb and store incident radia-
tank or out of the system below a set temperature. Alterna- tion.
In such a design, glazing is put on part of the south wall
tively warm water can be circulated through the collectors, if and
the wall itself is painted black (i.e., with a solar radiation
freezing temperatures are rare. Systems can also be designed
absorbent paint). To maximize energy storage, the wall is
Cold water
supply
Vent
T
c
DHW
Collector
Doublewall
heat
exchanger
Antifreeze
T
s
solution
Pump
Figure 5. A typical residential water Pumps on when
Pump Storage Auxiliary
T is 10 F above T
heater
heating system. c ° s
tank
----------------------- Page 494-----------------------
SOLAR HEATING 655
the
heat pump is then used for conventional space heating.
Specialized phase-change storage techniques (e.g., wax im-
pregnated wallboards) have also been developed in recent

years,
and these are particularly useful for solar houses. In
Collector some
cases, entire housing complexes have been integrated
with
large, seasonal, energy storage facilities which provide
hot
air as required. For best performance, both passive and
active
methods may be combined in hybrid building heating
systems.
Solar
Cooking
Living space
Solar
cookers have been investigated for many decades, and
some
low cost commercial models are available. These cookers
Air return Auxiliary can be
classified into three basic types. Hot-box designs con-
Distribution
heater sist of
an insulated box with double glazing in which the cook-
Fan ing
vessel is placed. Therefore the basic configuration is
equivalent to that of a solar collector using the energy within

the
‘‘collector’’ itself. Such cookers are oriented manually, and
are
limited to relatively low operating temperatures. The heat
input
and temperatures can be increased by incorporating re-
Domestic Cold water flectors
within the box to increase the degree of concentration
hot water by a
small amount.
Cookers that provide higher temperatures use concentra-

tors
(usually mirrors), and the vessel is mounted at the focal
point.
In this configuration, the vessel corresponds to the ab-
Figure 6. A schematic of an active building heating system. sorber
in a concentrator-receiver collector. This type of design
requires bright sunshine for successful operation, and care

must
be taken to minimize heat losses from the vessel at
higher
wind speeds. Folding-umbrella concentrators have
made of masonry or includes a water tank. Absorbed/stored
been
developed for such systems to enhance portability.
energy is transferred to the living space by forced or natural
The
hot-box and concentrator cookers cannot be used in-
convection depending on the building design. If necessary,
doors
and require a high level of solar insolation. Therefore
movable insulations, such as screens, shutters, or special
their
use is limited to rural and/or portable, camping applica-
plastic foam insulations, are used to prevent heat losses dur-
tions
where the cost of fuels may be excessive. To overcome
ing periods of low solar insolation.
these
problems, the third approach is to use a collector-based
A second approach is to use direct-gain south windows in
system
which transports energy with a suitable heat transfer
combination with overhangs and movable insulations. In this
fluid to
an indoor ‘‘cooking range.’’ High costs and other con-
case, the room itself is the collector, and the window acts as
straints, however, limit the use of such cookers at present.

the glazing material. Alternatively, a greenhouse (or conser-
vatory), which is a glaze-enclosed extension of the house, can
Industrial and Process Applications

be used both as a living space and as a collector. Energy is
stored by selecting appropriate materials for the floor and se- Solar
heat is extensively used for a number of industrial pro-
lect (interior) walls. Various types of overhangs are used to- cesses,
such as distillation, salt/brine production,and drying
gether with passive systems to control the incident radiation of
agricultural products. Over the years, considerable work
on the ‘‘collector’’ surface. Heat distribution is a particularly has
been done to develop other solar thermal systems, such
important issue in passively heated buildings, and therefore as
power plants, cooling/air-conditioning units, and detox-
architectural design is very important to ensure that proper
ification/disinfection reactors. These are discussed further
comfort conditions are maintained. later
on.
Active building heating systems are very similar in concept
to water heating systems (Fig. 6). Air or water (or antifreeze
Distillation. Solar stills for producing drinking water from
mixtures) are the typical heat transfer fluids. Liquid-based
brackish or salt water have been used for more than a cen-
systems include a closed liquid loop from which heat is trans- tury.
Though they may have various physical configurations
ferred to the air via a suitable heat exchanger. Systems using (e.g.,
evaporation from wicks, flow over nonhorizontal sur-
air collectors may be direct or include a separate loop similar faces),
they are all thermally equivalent to the basin-type still
to liquid-based systems. If necessary, an energy storage sys- (Fig.
7). In this design, solar energy is absorbed by water (ei-
tem may be included in either case. As in water heating sys- ther
directly or via the dark lining of the basin) which evapo-
tems, auxiliary heating is typically required as are freeze/ rates
and rises to the top of the enclosure. Because of the
boiling protection systems and suitable controls to maintain lower
temperature of the cover, the vapor condenses, drains
the required set points. by
gravity to the troughs on each side of the still, and is col-
An alternative approach to direct heating with solar en- lected
for future use.
ergy is to use the heated air from the collector as the cold-
Basin-type solar stills are either shallow with depths less
side thermal reservoir for a heat pump. In this configuration, than
0.02 m or deep with depths on the order of 0.1 m or
----------------------- Page 495-----------------------
656 SOLAR HEATING
Cover–condenser
Power Generation. Solar thermal power generation sys-
tems have been investigated for more than a hundred years.
Early work was related to low-power (#100 kW) systems,
mostly for irrigation.

In recent decades, however, large sys-
Condensate collection
trough
tems generating up to 80 MWe (peak) have been operated ex-
perimentally. Typical systems have been based on concentra-
tor-receiver collectors (e.g., parabolic troughs, heliostats)
because heat engine efficiencies are a strong function of the

Black liner
temperature of the heat source. Alternatively, large solar
Basin with salt
water Insulation
ponds have been used to power low temperature power plants
####
generating as much as 5 MWe.

Figure 7. A schematic of a solar still.
Most solar thermal power plants have been based on the
Rankine cycle, though Brayton and Stirling cycles have also
been considered to increase the efficiency, particularly for
high temperature systems. Depending on the power output

more. Shallow designs are typically more expensive but pro-
and
application, the overall configuration of the system can
duce more water even though they are operational only dur-
vary significantly. Typically, the solar energy collection sys-

ing the day. Deep stills continue to operate at night because
tem
is a separate loop where the working fluid (organic/syn-
of the thermal inertial of the larger body of water in the ba-
thetic heat transfer medium) remains in the liquid phase

sin. From the viewpoint of heat transfer, a solar still is equiv-
throughout. Energy is transferred via a heat exchanger to a

alent to a ‘‘no-flow’’ solar collector with an additional evapora-
secondary medium (e.g., water/steam, fluorocarbons, organic

tion-condensation process within the still itself. Solar stills
fluids) which must satisfy the requirements of the power gen-

are typically characterized by their efficiency which is equiva-
lent to the ratio of the useful heat transfer to the water to the
eration cycle. Energy storage units have also been incorpo-
solar radiation incident on the still surface.
rated in large power generation systems to provide better en-
ergy management.
Salt/Brine Production. Salt/brine production by solar heat-

Cooling and Air Conditioning. Solar heat has also been used
ing and evaporation of sea or brackish water is also an indus- in
cooling applications. The most common approach is to use
trial process used throughout the world. Typically, the salt,
solar energy as the heat source for a conventional vapor ab-
usually sodium chloride, is obtained in three stages in shallow
sorption refrigeration system. Instead of using steam, hot wa-
ponds or pans. In the first step undertaken in the largest ter
from solar collectors is used to supply energy to the gener-
ponds, the solution is concentrated to a level at which it is
ator in such systems. Continuously operating units for
fully saturated at about 23% salts. The concentrate is trans-
building cooling may require an auxiliary heat source to pro-
ferred to the second-stage pans, where calcium, magnesium,
vide the necessary system performance. Simpler intermittent
and iron sulfates and carbonates precipitate. Sodium chloride
absorption systems can be used for food preservation. Proto-
and magnesium and potassium chlorides and bromides are
type solar absorption cooling systems have been tested in var-
obtained in the final crystallizing ponds.
ious locations but have not yet reached the commercial stage.
The heat transfer process in a solar evaporation process,
Solar thermal energy has also been used in desiccant cool-
whether for salt production or brine concentration, is similar ing
systems which are based on a dehumidification-humidifi-
to that in a solar still. However, no condensation is associated
cation process. Preconditioned air is dehumidified by a desic-
with this process, and in the absence of a cover, heat and
cant, an exothermic process which increases the temperature
mass transfer rates are significantly enhanced at high wind of
the air. Then the air is cooled initially using exhaust (or
speeds. At the same time, rainfall reduces the production rate
other low temperature) air from the conditioned space, after
by diluting the concentrates in the ponds.
which its temperature is brought down further by evaporative
cooling. The desiccant itself is regenerated by solar heat. Ex-

Drying. Drying of agricultural and forest products is an-
haust air may also be used for this purpose if it is available
other common use of solar thermal energy. Either a direct or at
a suitable temperature. Though various types of solar des-
an indirect approach may be used for this application. In di-
iccant cooling systems have been developed using both solid
rect systems, the product (e.g., grains, fruit) is spread out in and
liquid desiccants, these are mostly experimental at
a thin layer and exposed to solar radiation and wind. Mass is
present.
transferred from the grain surface by convective effects of the
wind and the latent heat of evaporation is provided by solar
Detoxification/Disinfection. Solar energy has been used for
radiation. Then moisture continues to diffuse outward from
photocatalytic decomposition of hazardous wastes, such as
the interior because of the resulting concentration gradients,
PCBs and other organic compounds, and for killing bacteria
and drying occurs over a period of time. and
other pathogens. Systems are typically based on concen-
Indirect systems have also been developed for drying wood,
trator or flat-plate collectors, though shallow pond reactors
crops, etc. These are similar to conventional dryers, but part
have been used in some cases. Collectors are specially de-
of the energy is provided by solar heat absorbed in a collector
signed for this application because they must be highly trans-
system. Another alternative is to use a greenhouse design in
parent to ultraviolet radiation. This is essential because the
which the kiln itself is equivalent to a solar collector with
oxidation-reduction reactions are driven only by UV radiation
internal circulation. In both cases, either a recirculation sys-
(wavelengths #388 nm based on the activation energy of tita-
tem or an open circuit configuration is possible.
nium dioxide, the most commonly used catalyst), which com-
----------------------- Page 496-----------------------
SOLAR HEATING 657
prises only about 4% to 6% of the total terrestrial solar radia- the useful
energy based on available weather data and an
tion. The overall collector is otherwise relatively simple and estimate of
the system losses (which must be less than the
requires no thermal insulation because the reaction rates are absorbed
radiation if there is to be any useful energy). For a
not a significant function of temperature. given system
design, these are typically quantified by a pa-
rameter known
as the utilizability, which is the useable frac-
tion of the
total incident energy available over any given pe-
MODELING AND SIMULATION riod (hour,
day, etc.).
Utilizability methods are usually based on the monthly av-

Modeling of solar thermal processes is particularly important eraged values
and are widely used, particularly for swimming
because of the high initial cost of many solar thermal sys- pool heating,
passive and hybrid heating, and air heating sys-
tems. As a result, considerable work has been done to develop tems. For
many liquid collectors, however, there may be large
simulation tools for these systems. These include highly spe- intramonth
variations in the critical radiation level (i.e., the
cialized models for individual or specific types of systems and radiation
level at which the incident radiation exactly equals
more general transient or time-averaged models that can be the absorber
losses, and the utilizability is zero). As a result
used for a range of applications. Only the second group of these methods
are not applicable for such systems.
models is discussed here because the others are not directly An
alternative approach is to correlate the results of a
practical for most applications. large number
of simulations and use the correlation to predict
the
performance of similar systems. This is the basis of the f -
Transient Models chart method
which is used to estimate the fraction f of the
A number of generalized transient models have been devel- total energy
required for a specific application that can be
oped by different research groups around the world [e.g., supplied by
solar radiation. This parameter is usually ob-
tained as a
function of the heat loss (X ) from the collector and
Commission of European Communities (EMPG2/EURSOL),
TRNSYS (United States), WATSUN (Canada)]. These are the energy
(Y) absorbed by the collector, both of which are
built on individual numerical simulation modules describing normalized
with respect to the heating load. Separate correla-
the transient performance of components that make up a typi- tions for f
are available for air- and water-based systems, and
cal (solar) thermal system. The modules are usually based on correction
factors which account for energy storage have been
a 1-D approach because the uncertainties associated with typ- developed.
Because the functions are obtained using a curve
ical solar thermal systems do not justify the increased compu- fit, the f
-chart method can be used only for the specified range
tational cost of more complex 2-D or 3-D models. Limitations of parameters
and where the minimum load temperatures are
of current multidimensional computer models for complex about 20
C. More advanced methods combining the utilizabil-
thermal processes also make the 2-D and 3-D models imprac- ity and f
-chart concepts have also been developed to overcome
tical in most cases. this minimum
temperature limitation but are not commonly
used.
The generalized transient models are implemented in the
form of large computer programs which include subroutines The
utilizability, f -chart, and related methods are directly
for the numerical solution of differential equations and utili- applicable to
active collector-based systems. In contrast, the
ties for mathematical analysis of various thermodynamic and heat
transfer process in passive ‘‘solar buildings’’ is related
transport processes. To simulate a given process/system, the closely to
its architectural design. Nevertheless, models based
modules describing the individual components are linked to- on the
utilizability concept and correlation-based methods
gether in series and/or parallel as required. Then the system have been
developed for such applications. Specialized charts
performance is studied by numerically integrating the set of and methods
based on past experience with passive heating
time-dependent differential equations with a given set of pa- are also used
in system design. These may be strictly empiri-
rameters and initial conditions. The results are output in var- cal or may be
based on more complete analysis of the building
ious forms, and with some programs, it is possible to conduct heat transfer
process. Similar detailed analyses may also be
an economic analysis of a given system/process using the re- required for
large active systems where the simple utilizabil-
sults of the simulation. ity and f
-chart methods are not adequate.
The performance of solar thermal systems is a strong func-

tion of local climatic conditions. Thus, some programs gener-
RESOURCES
ate a synthetic climate that is used as part of the simulation
process. Various transient forcing functions are also included
in the simulation packages for the same purpose. Meteorologi- Numerous
books covering theoretical and practical aspects of
cal data, averaged over suitable periods of time (e.g., weeks, solar energy
and its utilization have been published during
month, years, depending on the application), are used directly the past 25
years. Probably the most complete single volume
in many cases. on solar
thermal systems is by Duffie and Beckman (12). A
more
exhaustive 10 volume compilation of work on solar heat-
ing
technologies has also been published recently by MIT
Time-Averaged Models
Press (13).
Most handbooks on energy systems engineering
Transient models are not typically used for evaluating small- provide some
coverage of solar heating, but the most useful
scale, residential type solar thermal systems because of time one, limited
to solar energy systems, is by Kreider and
and cost constraints. Instead, time-averaged methods that are Kreith (14).
used to obtain quick results have been developed for such ap- The
primary professional organization associated with so-
plications (11). If the temperature of the fluid at the inlet of lar heating
is the International Solar Energy Society that has
the collector is known a priori, one approach is to evaluate affiliated
national organizations worldwide. Publications of
----------------------- Page 497-----------------------
658 SOLID MODELING
these organizations provide an useful perspective on progress 13.

C. A. Bankston (ed.), Solar Heat Technologies: Fundamentals and
in solar energy. The results of the latest research on solar
Applications. Vol. 1–10 (with individual volume titles and edi-
thermal systems are published in a number of periodicals re-
tors), Cambridge, MA: MIT Press, 1993.
lated to heat transfer and thermal engineering and in a num- 14.
F. Kreider and F. Kreith, Solar Energy Handbook, New York:
ber of specialized journals related to solar and alternative
McGraw-Hill, 1981.
energy.
SANJAY K. ROY
Solar radiation data are available from various sources,
particularly the national meteorological offices of individual
PhD Research Group, Inc.
countries [e.g., National Weather Service (NWS), National
HONGTAN LIU
Oceanic and Atmospheric Administration (NOAA), and Na-
University of Miami
tional Climatic Data Center in the US]. Extensive compila-

tions of data obtained from sources worldwide can also be ob-
tained from the World Radiation Data Center in St.
SOLAR HEATING STORAGE. See THERMAL ENERGY
Petersburg, Russia.
STORAGE.
Finally, it is also important to note that a significant frac-
SOLENOIDS. See SUPERCONDUCTING ELECTROMAGNETS.

tion of research on solar thermal systems has been funded by
SOLID ELECTROLYTES. See ELECTROLYTES.

government agencies throughout the world. Reports pub-
lished through the relevant organizations (e.g., DoE, NASA,
NREL in the US) are extremely important sources of detailed
information on many of the original research projects. Techni-
cal standards have also been developed by professional orga-
nizations (e.g., ASHRAE, SRCC, and FSEC in the US) for
components and systems related to active water and space
heating. Industry associations (e.g., SEIA in the US) are also
an important source of information regarding commercial as-
pects of solar energy.
BIBLIOGRAPHY
1. W. C. Dickinson and P. N. Cheremisinoff (eds.), Solar Energy

Technology Handbook. Part A. Engineering Fundamentals, New
York: Dekker, 1980, p. 12.
2. J. Lean, The Sun’s variable radiation and its relevance for Earth,
Ann. Rev. Astron. Astrophys., 35: 33–67, 1997.
3. M. Iqbal,An Introduction to Solar Radiation, New York: Academic

Press, 1983.
4. M. Noia, C. F. Ratto, and R. Festa, Solar irradiance estimation

from geostationary satellite data: 1. Statistical models, and 2.
Physical models, Solar Energy, 51: 449–465, 1993.
5. C. Gueymard, P. Jindra, and V. Estrada-Cajigal, A critical look

at recent interpretations of the Angstrom approach and its future
in global solar radiation predictions, Solar Energy, 54: 357–363,
1995.
6. C. Hottel, A simple model for estimating the transmittance of

direct solar radiation through clear atmospheres, Solar Energy,
18: 128, 1976.
7. J. Davies and D. C. McKay, Evaluation of selected models for

estimating solar radiation on horizontal surfaces, Solar Energy,
43: 153–168, 1989.
8. D. T. Reindl, W. A. Beckman, and J. A. Duffie, Evaluation of

hourly tilted surface radiation models, Solar Energy, 45: 9– 17,
1990.
9. W. T. Welford and R. Winston, The Optics of Nonimaging Concen-

trators, New York: Academic Press, 1978.
10. G. Beckmann and P. V. Gilli, Thermal Energy Storage, Vienna,

Austria: Springer-Verlag, 1984.
11. T. A. Reddy, The Design and Sizing of Active Solar Thermal Sys-
tems, Oxford, UK: Clarendon Press, 1987.
12. J. A. Duffie and W. A. Beckman, Solar Engineering of Thermal

Processes, 2nd ed., New York: Wiley, 1991.
----------------------- Page 498-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Steam Turbines

Standard Article

William P. Sanders1
1Turbo-Technic Services, Inc., Aurora, Ontario, Canada

reserved.
Search
DOI: 10.1002/047134608X.W3041
❍ Acronym Finder
Abstract
Turbine Types
Turbine Performance and Efficiency
The Turbine Stage
Turbine Control
Bearings and Lubrication System
Operating Problems
Starting and Stopping Steam Turbines
Materials for Turbine Construction

%20Conversion/W3041.htm17.06.2008 16:46:32
----------------------- Page 499-----------------------
STEAM TURBINES 509
STEAM
TURBINES
The
steam turbine is an engine designed to convert the ther-
mal
potential energy of steam to rotational kinetic energy on
a rotor
which is free to rotate when supported by a series of
axially
aligned bearings. This rotational energy is then used
to drive
some mechanism or machine, such as an electric gen-
erator,
pump, compressor, or ship propeller. Energy conver-
sion is
achieved by allowing for the controlled expansion of
the
high-energy (pressure and temperature) steam in a num-
ber of
series arranged stages, each stage allowing a portion of
the
total thermal energy to be released and then converted to
high
kinetic energy in a fixed blade row. The resulting jet of
high-
velocity steam is then directed into a following row of
blades
which are attached to a central spindle and caused to
rotate,
thereby driving an attached mechanical or electrical
device.
There can then also be a further energy release or
reaction
in the rotating row.
The
concept of expanding steam through some form of noz-
zle and
employing the resulting kinetic energy to produce
shaft
rotational energy has been recognized for centuries, by
Hero in
120 B.C. and later by Giovanni de Brance in 1629.
However,
it was not until the late nineteenth century that the
principle was applied to some practical application—first by

Sir
Charles A. Parsons, who designed a 10 horsepower unit
in 1894
and who in 1890 built four 75 kW units for electrical
lighting. Later other designs followed: in 1889 by de Laval

and 1896
by Curtis.
The
units resulting from these early designs formed the
basis of
the modern high-capacity units, which provide much
of the
energy conversion capability throughout the world.
Unit
sizes have increased so that today there are operating
units
rated at 850,000 kW in a single line (tandem com-
pound),
driving a single electrical generator, and units rated
up to
1,300,000 kW in a two-line (cross compound) application
driving
two separate but electrically interconnected genera-
tors.
While this increase in unit size has continued to grow,
so have
the conditions of the steam which is admitted to the
units.
There have also been advances in the manner in which
the
equipment of the total cycle is configured to increase the
efficiency of the conversion of the thermal energy to electric

or
mechanical output from the turbine units.
The
physical properties of steam, pressure, absolute tem-
perature, and specific volume are related by the ‘‘General Gas
Equation.’’ This equation provides a relationship capable of

----------------------- Page 500-----------------------
510 STEAM TURBINES
predicting the steam properties throughout the expansion tain

sections, this is normally done only to help balance the
passage or steam path. In examining these properties, possi- axial
thrust developed within the unit. There are units with
bly the most significant characteristic in terms of affecting a steam
path designed for steam to flow radially outwards
the arrangement and design of the steam turbine is that as from the
inlet, to flow radially in and then to exhaust axially,
the steam expands from inlet to exhaust conditions, most of- and even
to flow tangentially. However, the majority of units
ten in a subatmospheric condenser, there is a dramatic in- are of
the axial flow type.
crease in the volumetric flow. This increase in volumetric
flow, along with the limitation placed on axial velocities TURBINE
PERFORMANCE AND EFFICIENCY
throughout the unit, requires that in the lower-pressure sec-
tions either multiflow or parallel-flow stages are required. There
are a number of losses which occur within the steam
This multiflowing of stages allows increases in the quantity turbine,
and these losses are associated with a number of the
of steam which can be admitted to a unit and therefore allows
mechanical components comprising the unit. These losses in-

an increase in the output which can be achieved. clude,
in addition to the more obvious steam path, those in
Since the introduction of the steam turbine as the primary
bearings, couplings, and valves. There are also losses associ-
power generating unit, there have been advances in unit out- ated
with steam leakage, heat radiation and conduction, and
put. With advancing steam conditions and improvement in pressure
drops in lines from one section of the unit to another.
cycle arrangement there have also been improvements in unit These
losses are predictable, and in terms of mechanical
and cycle efficiency. losses
they remain relatively constant throughout the op-
erating
life of the unit. However, those losses experienced
within
the steam path are a function of a number of complex
TURBINE TYPES
factors
both mechanical and aerodynamic, and the efficiency
within
the blade rows tends to decrease with time. A portion
Despite their apparent similarities, there are a variety of of
these steam path losses are recoverable when the unit is
types of turbine, and they can be classified in a number of opened
for inspection and refurbishment, but it is normally
ways, the most apparent being their type of application—that not
possible to recover the full losses which have been in-
is, electrical power generation, mechanical drive, or marine. curred
and associated with operation, except at very high
They can also be classified by size, arrangement, and output. costs,
and such improvement can normally only be sustained
A factor having considerable influence on the details of any for
relatively short periods.
turbine is the type of thermal cycle in which it is to be em-
Steam path efficiency, #sl, is defined as the ratio of useful
ployed, the energy source, and the steam conditions used in energy
converted to power in the unit or section to energy
that cycle. which
is available for conversion (1). This expansion in the
A significant advance in cycle arrangement was the use of steam
path is shown as in Fig. 1. The steam enters the stage
‘‘regenerative feed heating.’’ Feed heating is an arrangement
by which a portion of the steam entering the unit is removed
from the expansion, at a number of suitable steam conditions,
P
a
and used to preheat the condensate returning from the con-
A Ta
denser to the boiler. This feed heating is applied to raise the
Ha
temperature of the condensate to close to the saturation tem-
perature of the water entering the boiler. The boiler has then
only to supply the latent heat and superheat to the working
fluid.
y
A straight condensing non-reheat turbine is one in which
g
e
the steam enters the unit and expands in a series of stages to
y n
g l
a defined exhaust pressure. As discussed above, the lower-
r u
e f
pressure stages can be paralleled, but this has no effect on
n e
e s
the non-reheat character of the unit. There is also the reheat-
e U
b
type arrangement, in which the steam, after expanding to a
a
i
suitable pressure, is removed from the turbine, returned to
a
A
the boiler or other reheating device, has its temperature level
C
raised, and is then returned to the turbine to continue its
T
expansion. Some advanced cycles have been arranged to in-
Tc c
corporate two stages of reheating.
t y
s r
The reheat cycle provides considerable advantage in terms
P b o e
L n
of cycle efficiency, but suffers to a small degree from the addi-
e
tional technical complexity associated with the addition of

H b
more equipment which can be subject to failure. However,
B
with modern design and manufacturing practices, these po-

tential problems are capable of being overcome.
A represents the inlet conditions
There are a variety of turbine design configurations in
B represents the isothermal discharge conditions
terms of the steam flow. The most common form is for the
C represents the actual discharge conditions
steam to enter the unit and flow axially from inlet to exhaust, Figure
1. The steam expansion line of the turbine stage, as repre-
and while the direction of axial flow may be reversed in cer- sented
on the Mollier enthalpy–entropy diagram.
----------------------- Page 501-----------------------
STEAM TURBINES 511
with conditions H , P , and T . The steam discharges from the

Steam Leakage Losses. The steam path is comprised of a se-
a a a
stage with steam conditions H , P , and T . Had the expansion ries
of stationary and rotating blade rows which have pres-
c c c
been isothermal, the conditions would have been H , P , and sure

drops produced across them and which operate with rela-
b b
Tb . Therefore, the state line efficiency can be defined as tively

small clearances, even at the highest blade tangential
velocities. Steam which bypasses the blade elements does no

Useful energy converted to power per pound of steam work.
Also, this leakage steam can reenter the steam path,
ηsl =
Available energy per pound of steam
disturbing the orderly flow from one blade row to the follow-
Ha − Hc ing,
thereby inducing further losses.
= (1)
Ha − Hb In
the case of shaft end packing, where the rotor passes
through the stationary outer casings, the steam leaks from

where the
internal portions of the unit to the lower pressure sur-
rounding the casing, or to some lower pressure, where it can

Ha is the enthalpy at inlet to the turbine section or stage at be
returned to the steam path or cycle, but at a lower thermal
pressure Pa energy
level. Again this leakage represents a loss to the
Hb is the enthalpy at discharge for an isothermal expan- steam
path.
sion from Pb The
most common seal used in the steam turbine is the
Hc is the actual discharge enthalpy at pressure Pb
labyrinth type which consists of a series of ‘‘fins,’’ forming a
number
of chambers into which the steam expands converting
If account is taken of the velocity (kinetic energy) of the a
portion of its thermal energy to kinetic in the chambers,
steam leaving the final stage in a turbine section, then this is this
kinetic being successively destroyed by the tortuous path
considered irrecoverable and is known as the ‘‘leaving loss.’’ of
expansion. The quantity of steam leaking past any series
In other than the low-pressure section, a portion of this en- of
labyrinth seals can be determined by the use of ‘‘Martin’s
ergy can be recovered, as the kinetic energy of the steam en-
equation’’ (3). A simple multitooth labyrinth seal is shown in
tering the section. This leaving loss is of little significance, Fig.
2.
except in the final or exhaust stage in low-pressure sections,
where the leaving velocity is carried to the condenser and rep-

2
2
resents energy lost to the unit.
# P1 1 −
# P
#
The steam path losses can be considered to be influenced
1
m = k ·A e (2)
# P
#
by several diverse factors including (a) the design philosophy
V N + ln 1
used to define the various components comprising the steam
s1 P2
path, (b) the manner and quality with which these compo-
nents are manufactured to meet the design specification and where
then installed in the unit, and (c) the manner in which the
unit is operated and maintained. m is
the quantity of steam leaking in unit time
The actual losses can be divided into four principal groups, k is
a constant, including a flow coefficient
which are as follows:
A e
is the circumferential leakage area, # D # C
Aerodynamic Losses. The steam expands through the steam D is
the mean diameter of the seal
path at high velocities; and it flows through the blade ele- C is
the running clearance
ments, causing a frictional interaction with the surfaces with P1
is the upstream steam pressure
which it is in contact. As the unit ages, there is a deteriora- Vs1
is the upstream steam specific volume
tion of this surface due to a number of phenomena which can- P2
is the downstream steam pressure
not be entirely avoided. These include the passage of solid N is
the number of series strips
particles of magnetite (Fe O ) through the steam path, these
3 4
particles being carried into the unit from the boiler and other
As
a unit operates there are a number of phenomena, in-
portions of the steam cycle. Also the steam carried into the
cluding starting and stopping, load changes, and other tran-
steam path can have chemical compounds suspended in the
sients
which cause the rotor to vibrate to the extent the knife
steam, which, as the pressure decays, come out of solution
edge
of the seal strips rub and therefore wear. This action
and attach themselves to the steam path surfaces, thereby
causes
the clearance to open, with a resulting increase in the
increasing roughness and therefore increasing frictional
leakage quantities. Therefore as a unit ages, there will be an

losses. Another influence of these chemical ‘‘carry-ins’’ is that
increase in leakage flow and a resulting reduction in output.

they may form corrosive compounds when deposited on the
steam path component surfaces. Under certain circum-
stances, these compounds can cause corrosive pits to form,
which cause flow disruption and therefore losses.
P V
1, s1
In addition to these surface losses, there can be losses due
Inlet
to high velocities above Mach 1 where supersonic losses are
Discharge
induced. There are also losses associated with components in-
P2
cluded in the steam path from structural considerations. Such

####y#y#y#y##yy#y#y
###yyy#y
components include tie or lacing wires, as well as notch blocks

C
D
or missing blades at row closing windows. These components
cannot be avoided, but modify the flow and cause distur- Figure
2. Parameters of steam and geometry defining the laby-
bances in the steam path. rinth
seal.
###yyy#y
----------------------- Page 502-----------------------
512 STEAM TURBINES
Moisture Losses. As the steam expands from its inlet condi-

Power Cycle Efficiency. The expression Eq. (1) for the steam
tions, there is a gradual reduction of its energy level, associ-
path efficiency does not take account of the mechanical losses
ated with the conversion to rotational energy in the rotating
which occur within the various components of the unit such
portion of the unit of energy. This reduction of steam energy
as bearings, couplings, and so on. Therefore, a quantifiable
continues until ultimately moisture is formed in the steam,
characteristic was required which would allow the total effi-
which is then required to transport these moisture particles
ciency of energy conversion of the system to be defined. Two
through the steam path.
such characteristics are available to indicate the performance
Associated with this moisture, there are two forms of en-
level or potential of the turbine generator or turbine-driven
ergy loss within the steam path. First is the kinetic effect.
unit. These are (a) the ‘‘heat rate,’’ which is a measure of the
This consists of the energy the steam particles must expend
quantity of thermal energy required to generate a specific
on changing the direction of the water as it passes through
electrical output for a given period of time (5), and (b) the
the blade rows. Second, are the losses associated with the col-
‘‘steam rate,’’ which provides a measure of the pounds of
lision of these water particles on the rotating blade rows and
steam operating between specific thermal conditions required
their effect of retarding or ‘‘braking’’ the rotational energy of
to generate a defined amount of electric or mechanical output
the blades.
for a specified time.
While these losses are not high, they do increase with the
These are defined as
increase in the moisture level of the steam (4).
Leaving Losses. As the steam exhausts from a turbine sec-
Thermal input
Heat rate (HR) =

tion, it contains velocity energy. In stages other than the last,
Electrical output
in any section this velocity energy is carried into the following
(3)
stationary blade row where a portion of it can be utilized. In
Steam rate = Quantity of steam entering the unit
any section of a unit, other than the low-pressure sections,
Electrical or mechanical output
this does not represent a significant loss. However, in the last
or exhaust stages, velocities are high and these losses are sig-
In the basic reheat cycle shown in Fig. 3, the heat rate can
nificant.
be defined by the following expression:
In fact the turbine exhaust loss comprises three distinct
components which contribute to the total loss. The first is the
m (H − h ) + m (H − H )
Heat rate (HR) = 1 1 c r hr cr (4)

component velocity, which is proportional to the axial leaving
kW
velocity squared. The second is the frictional loss within the
exhaust hood, and the third is a loss which occurs only at
In this expression, the term m (H ( # h ) is the heat added
1 1 c
low exhaust velocities and is associated with the steam being

to the feed water in the boiler, and the term mr(Hhr # Hcr) is
unable to clear the exhaust annulus and causing a pressure
the heat added to the steam in the boiler reheater section.
buildup just downstream of the last stage blades.
The output is given as kW, being the electric generator out-
Once an exhaust configuration is selected by the design
put. This term can be adjusted to reflect the power consumed
process, the magnitude of the losses cannot be changed except
by the boiler feed pump and other auxiliaries carried by the
by replacing the exhaust blades with one providing a greater
plant. There are therefore a number of definitions of heat rate
annulus area and therefore lowering the axial exhaust veloc-
in use. This term also takes account of the generator effi-
ity. This is obviously not an easy process, but in some cases
ciency because it includes the actual output from the gen-
it is justified by the savings in fuel cost.
erator.
From boiler
reheater
H m
hr, r
To boiler
reheater
H m
cr, r
From boiler H 1,m1

superheater V
V
High
Reheat Low Generator
pressure
pressure
kW
Figure 3. The steam and flow conditions

Condenser
of a reheat steam turbine, showing steam
Regenerative feed heaters
h
H enthalpies, water h enthalpies, and
c
Return to
flow quantities ‘‘m’’ at various locations
boiler
throughout the cycle. The unit electric
Boiler
output ‘‘kW’’ is also shown.
feed pump
----------------------- Page 503-----------------------

STEAM TURBINES 513
in
the stationary row and has a further drop in the rotating.
m (H )
1 1
Because a pressure differential exists across row blade rows,
there is a tendency for the steam to leak or bypass the blade
rows if an alternate path of less resistance can be found. The
occurrence of leaking steam therefore represents a condition

in
which the steam gives up its thermal energy and does no
work in its expansion (a throttling or nonexpansive expan-

Steam Driven
sion). The extent to which this leakage occurs must be con-
turbine equipment
trolled to maintain stage efficiency. This individual turbine
blade rows are therefore normally arranged

provide to sys-
kW or HP
tems which minimize this leakage and thereby maximize the
work done in the stage.

m (H )
The thermal energy of the steam must be released in dis-
1 e
crete steps, selected to maximize the efficiency of expansion.
This control of pressure and enthalpy drop is achieved by en-
suring that the discharge area from each blade row is con-
Figure 4. A smaller industrial or marine type unit, showing steam
enthalpies H and flow quantities m together with the output in kilo-
trolled within close tolerances and is a function of the quan-
watts ‘‘kW’’ or horsepower ‘‘HP.’’
tity of steam flow and the steam properties at inlet to the
stage. The discharge area required on any blade row is deter-
mined from the equation

A typical smaller installation, as required for industrial or
marine use, is shown in Fig. 4. Here the steam rate would be
m = AC d#2g γ P1Vs1 [R2/γ − R (γ + 1)/γ ] (6)
defined as
γ − 1
m
Steam rate (SR) = (5)
where
kW or HP
In this definition, the input, in thermal energy, is mea- m
is the quantity of steam flowing in unit row
sured as the steam flow m required for a specified time be- A
is the row discharge area
tween steam properties H 1 and He to generate an electric out- Cd
is the discharge coefficient
put kW or mechanical output HP for a specified time period. g
is the gravitational constant
P1
is the row inlet pressure
P2
is the row discharge pressure
THE TURBINE STAGE
Vs1 is the row inlet specific volume
#
is the ratio of specific heats of steam
The basic turbine stage consists of two blade rows. The first
is stationary, consisting of a number of vanes which are ar-
In the practical turbine this area is achieved by defining
ranged to produce passages between adjacent elements. the
form of the individual blade vanes forming the expansion
These vanes are attached to the stationary portions of the
passage and then controlling their spatial relationship rela-
unit; they both expand the steam by allowing for a controlled
tive to each other in the final assembly.
pressure drop, and they are oriented so as to direct the re-
sulting jet into a following row of rotating blade elements.
Impulse and Reaction Designs

These rotating elements provide a passage which deflects the
steam through a turning angle, and they discharge it to enter
There are two distinct philosophies concerning the manner of
the stationary blade row of the following stage, or to the con-
releasing the thermal energy within the turbine stages. The
denser in the case of the final or exhaust stage. In changing
first can be defined as the pure impulse stage, in which the
the steam direction from inlet to near axial discharge, these
total available thermal energy assigned to the stage is re-
rotating blade elements have a thrust developed on them,
leased in the stationary blade row. This energy conversion
which is transmitted to the central spindle or shaft, causing
produces a high-velocity jet which discharges from the row
it to rotate. Therefore, in its simplest form, the stationary and
is directed into the rotating elements. The rotating blade
blades convert the steam’s thermal potential to kinetic en- row
is then designed to convert this kinetic energy to thrust
ergy, and the rotating blades convert the jet kinetic energy on
the individual blade elements by changing the momentum
to rotational kinetic energy, which is available to drive any of
the jet. This change of momentum produces a tangential
machine or mechanism connected to the shaft.
thrust on the individual blades, which in turn drive the rotor
The description of the blade rows represents a simplified to
which they are attached. In the pure impulse stage the
explanation of the function of the two blade rows, and it ne-
pressure at discharge from the rotating blade row is the same
glects the effect of reaction or pressure drop which is designed as
at entry.
to occur in the rotating blade element. In the practical turbine
The second form of energy release is within the reaction
stage, there is in fact a pressure drop of some magnitude in
stage. In this design a portion of the total thermal energy is
the majority of rotating rows.
released in the stationary row, in precisely the same manner
The turbine stage consists of a series of stationary and ro- as
in the impulse stage; the remaining energy is released in
tating components, through which the working fluid expands. the
rotating row, as a reaction on the rotating elements. A
This steam, at entry to the stage, has a pressure which drops
stage in which half the thermal energy is converted to kinetic
----------------------- Page 504-----------------------
514 STEAM TURBINES
in the stationary row, with the remainder converted in the

blade tangential velocity, whereas in the 50% reaction stage
rotating row, is called a 50% reaction stage.
the velocity of the steam for maximum efficiency will be
In practice the reaction stage is normally designed so that
smaller. In fact the 50% reaction stage will require twice the
50% of the total energy is released in each of the stationary
number of impulse stages to accommodate the same thermal
and rotating blade rows. Also, in actual designs of modern
energy range. There are, however, few pure impulse stages
units the true impulse stage is only rarely used, and the ma-
built in modern units, and a realistic reaction for the ‘‘impulse
jority of stages have some small degree of reaction sufficient
stage’’ provides for a velocity ratio above 0.5; as a result, the
to ensure a positive pressure gradient throughout the steam
number of impulse stages are somewhat larger, and the ratio
path. The degree of reaction used is dependent upon the phi-
of reaction to impulse stages is about 60 to 65%.
losophy of the individual designer of the turbine.
The geometry of the individual blade rows is influenced by
Another stage design, often used at inlet to high-pressure
the magnitude of the pressure drop in the stage. The manner
turbine sections, are those in which a single pressure drop
in which this drop is divided between the stationary and ro-
occurs in a stationary blade row, followed by a rotating blade
tating blade rows, together with a consideration of the num-
row. Then an exhaust from this first rotating row there is a
ber of stages between any bearing span, has a direct effect on
direction changing stationary blade row, which is followed by
the efficiency.
a second rotating blade row. Because these rows tend to have
Consider the two stages in Fig. 6. Figure 6(a) shows the
poor efficiency, they are not used extensively in modern tur-
basic arrangement of the impulse stage which comprises the
bines.
wheel and diaphragm construction; in this design the station-
ary blade row is carried in a diaphragm, which forms its sta-

Turbine Stage Velocity Ratio and Efficiency
tionary stage seal on the central portion of the rotor. There
are also seals above the rotating blade row, even though the
There is a well-defined relationship between individual stage
pressure drop is relatively small. In Fig. 6(b) the typical ar-

efficiency and the ratio of blade tangential velocity to steam
rangement of the reaction stage is shown. Here, because of

jet theoretical or isentropic velocity (5). This relationship is
the larger number of stages, it is necessary to seal the station-

shown in Fig. 5 for a pure impulse, a 50% reaction, and the
ary blade row at a considerably larger diameter; but because

two-rotating-blade row design of stage. From these curves it
the pressure drop is smaller, there is a smaller leakage quan-

can be seen that for the pure impulse stage, that a ratio of
tity which bypasses the blade row. Again, a seal is used above
0.50 provides the optimum efficiency, whereas for a stage hav-
the rotating blade, which in this case has a somewhat higher

ing a 50% reaction design, that a ratio of 0.707 is optimum.
pressure drop. The total leakage losses in the impulse and

The two-stage design is optimum at 0.25.
reaction designs for any specified energy range are at about

The implication of these curves is clear. It can be seen that
the same level.

for any blade tangential velocity, which is a function of the
In both stage designs, it can be seen that above the rotat-

stage diameter and rotational velocity, the pure impulse stage
ing blade row, there is a circumferential band of material

will require a larger steam isentropic enthalpy for any speci-
attached to the outer surface of the blade vane. These cover

fied blade tangential velocity. Therefore, the reaction stage
or shroud bands are included in the design to limit any flow

will utilize a smaller enthalpy drop than the pure impulse
which would occur through the tip section of the vane. These
stage, which will require a steam velocity which is twice the
bands are also used to provide seals or a platform upon which

a
radial seal can be produced.
A stage can also include one, two, or, in certain older de-
100
signs, three lacing or tie wires. These wires provide a continu-
ous or semicontinuous coupling between adjacent blade ele-

90
ments, and they are included to transmit vibratory forces
% 80 50%
developed in one element to others to which it is connected.
η
Reaction
-
These ties tend to be limited to the blade rows with longer

e stage
g 70
vanes, because these blade elements have a series of lower
a
t
s
natural frequencies and are more easily excited by alternat-
l 60 Pure impulse
a stage
ing loads which are produced in the steam path. These wires
e
d
i
e 50
are included in the design to achieve structural integrity in
h
t
those stages where the possibility of vibratory loads could in-
f
o 40
duce some form of fatigue failure. These wires also include
y
c
efficiency losses, due to drag, into those stages which employ
n
e 30
i
them.
c
i
f
f
The individual blades are attached to the rotor by the use
E 20
Two-row
of specially designed fastening devices which are produced in-
10 wheel
tegral with the blade vane. There are a variety of forms for
these fastenings. Each turbine manufacturer has a series of

0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
designs available which can include geometries which allow
the blades to be assembled to the rotor entering it in the

Blade tangential velocity U
Velocity ratio " "=ρ
axial, radial, or tangential directions. There are advantages
Steam isentropic velocity C0
of security, assembly, and cost associated with each design,
Figure 5. The variation of ideal stage efficiency as a function of the
and the turbine manufacturer selects the best suited design
stage velocity ratio # for various stage configurations.
for each row, after careful evaluation.
----------------------- Page 505-----------------------
STEAM TURBINES 515
Vector
Diagrams for the Turbine Stage
C
Stationary and rotating blade vanes are
designed to produce
C a
ro a
profile which will expand the steam by providing a passage
### through
which it can flow, change direction, and exhaust at
the
correct pressure and velocity to enter the following blade
L row.
The design of profiles is one of the more complex aspects
no
L bo of
turbine engineering. This is because it requires that pro-
D D
ot et files be
defined which can be produced at realistic costs with
D it
existing machining capabilities and which will induce a mini-
mum of
energy losses into the stage. These blades must also
L have
sufficient mechanical strength so they will last for the
###
bi
L ni
predicted life of the turbine unit.
As
a preliminary step to defining the requirements of a
blade
vane, it is necessary to establish the required profile
inlet
and discharge angles (8). These angles are normally de-
termined during the design process by the construction of

D
D ir ## er
‘‘vector or velocity diagrams.’’ These diagrams establish the
angles
which are required of the profile to accept the in-
flowing
steam with a minimum of incidence loss, and they
also
provide a controllable exhaust which will ensure that the
steam
discharge area and direction from the row are correct.
Cir
Shown in Fig. 7 are velocity diagrams for both a pure im-
pulse
and a 50% reaction stage. In these diagrams the blade
##
tangential velocity U is the same on both designs as are the
D
or
stationary and rotating blade discharge angles and # . Fig-
1 2
(a) ure
7(b) shows the vane profiles required by these two velocity
diagrams to allow for incidence free entry and expansion

through
the resulting blade expansion passage.
Caulking
Form of
the Blade Vanes
C The
two blade rows comprising each stage are designed to
ro
accept
the inflowing steam, deflect it through the desired
## angle
sufficient to allow the required degree of energy conver-
sion,
and then deflect it into the following blade row, either
rotating or stationary. These functions are achieved by ensur-

L bo ing
that the profiles form the correct shaped passage between
D it Dot C D nt Det the
elements, and also form the correct area at discharge from
ao each
row.
In
an actual unit, the blades must be manufactured to
## achieve
both the required vane profile and also ensure that as
they
are mounted to the major components which give them
Cai
location within the unit, so their spatial relationship both to
elements within their own row and rows both up- and down-
stream
is correct.
## TURBINE
CONTROL
Cr L b D nr Der
D ir
The
turbine control system is designed to control both unit
rotational speed and output. For large utility units, connected

to
large power grids, it is necessary for the governing system
Dor Caulking to
control speed to ensure that system frequency is main-
# tained.
The control system must also react in response to sys-
(b) tem
load demands, to adjust control valve settings and admit
Figure 6. (a) The basic impulse stage showing component arrange- the
required quantity of steam.
ment and major dimensions. (b) The basic reaction stage showing In
non-reheat units, there are valves located at the inlet to
component arrangement and major dimensions. the
steam turbine unit; these valves open and close to control
the
admission of steam to the high-pressure section. In reheat
units,
there are large quantities of steam contained between
the
cold outlet and the hot reheat return to the unit. There-
fore,
it becomes necessary to incorporate valves between the
----------------------- Page 506-----------------------
516 STEAM TURBINES
Impulse stage (i)

U
α
1 β β 1
δ
W2 C1
W1
C2
50% Reaction stage (ii)

U
α 1 β2 β 1
W1
C2
W2 C1
δ
U is the tangential
velocity of the rotating blade.
C1 is the absolute
velocity of the steam discharging from the stationary blade row.
α 1 is the discharge
angle from the stationary blade row,
W1 is the relative
velocity of the steam entering the rotating blade row.
β 1 is the rotating
blade vane inlet angle.
β 2 is the rotating
blade vane discharge angle.
W2 is the relative
velocity of the steam discharging from the rotating blade row.
C2 is the absolute
velocity of the steam discharging from the rotating blade row.
δ is the steam
absolute discharge angle.
(a)
Figure 7. (a) The velocity vector dia-

grams for the basic impulse stage (i) and
50% reaction stage (ii). (b) Vane profiles
for (i) the pure impulse stage with a con-
stant expansion passage width and (ii) the
Pure impulse
vane profile (i) 50% Reaction profile (ii)
50% reaction stage with a converging
passage.
(b)
boiler superheater and reheat section inlet, so that in the

There are two basic control systems in use, which are de-
event of the unit being disconnected from the power grid or
scribed in the following section.
other driven device, this quantity of steam cannot reenter the
Mechanical Hydraulic System. The mechanical hydraulic
turbine and contribute to its overspeed. These intermediate
control (MHC) system consists basically of a governor system,
valves incorporate both stop and control functions.
connected to and driven from the main turbine rotor, and is
A second major function of the control system is the protec-
therefore sensitive to its rotational speed. In large utility
tion of the turbine-generator unit against both internal dis-
units connected to power grids, the turbine rotational speed
ruptions and system disturbances. Therefore, the control sys- is
governed by system frequency, and therefore the governor
tem is designed to react to disturbances, by disconnecting the
becomes in effect a load control device, which actuates control
unit from the grid. When disconnection occurs,the control sys-
valve position through an oil-pressure-sensitive system, in
tem is designed to limit overspeed transients (to levels which
which the oil pressure varies in response to unit speed. This
will allow the unit to continue to carry station auxiliaries)
governing system controls the positioning of the valves
and then to resynchronized and connected to the grid as the
(through servomotors), which open and close the control
disturbance is cleared. Alternatively, if the overspeed is suf-
valves to vary the quantity of steam admitted to the unit and
ficiently high, the control system must regulate the overspeed
thus vary the load which can be generated.
and bring the unit to a low turning speed, so the unit can then
The control system oil, which is controlled by rotor speed,
be restarted and re-connected to the grid after the disturbing
will, upon loss of electric load and in response to speed in-
condition has been corrected.
creases, cause the governor system to close the main and in-
----------------------- Page 507-----------------------
STEAM TURBINES 517
termediate control valves. This is a fast-acting system and As

load demands reduce further, the full arc admission
will, on a well-designed unit, cause the valves to go rapidly to unit
again closes all valves by some amount, to limit the
a ‘‘closed’’ position. steam
admission. In the partial arc design, the closure occurs
Electro-Hydraulic System. The electrohydraulic control on
only one valve (or two, if sufficient area reduction cannot
(EHC) system employs solid-state technology together with a be
achieved with the closure of just one).
high-pressure hydraulic system. This faster-acting system is
Therefore, sequential closing of the valves offers advan-
more responsive than the MHC system and provides not only tages
in terms of the unit performance at part loads. How-
a faster response, but a more reliable system. The EHC sys- ever,
such a partial admission unit will suffer problems asso-
tem is also more flexible, allowing interconnection with other ciated
with partial arc admission where the control (first)
external systems and functions such as ‘‘overspeed anticipa- stage
blades must pass alternately through arcs of high-
tory devices,’’ which cause main and intermediate stop and energy
steam jets and dead bands. This effect causes high-
control valves to close if the initial rate of rotor acceleration, impulse
loads to be placed on these blades and can result in
on loss of load, indicates an emergency overspeed will result. damage.
This EHC system also allows remote control and remote diag-
nostic monitoring and is capable of being computer-controlled Sliding
Pressure Control
from the control room.
An
alternate method of steam admission into the control
The high-pressure oil system used on the EHC design has
stage
is ‘‘sliding pressure operation’’ in which the initial pres-
strict requirements for cleanliness and fluid stability. Because
sure
of the steam produced in the boiler is reduced with re-
of the possible leakage problems associated with a high-pres-
ducing
load to a condition that the control valves can be left
sure oil, it is necessary to use a fire-resistant fluid in this
wide
open at partial loads. This method of steam admission
system.
has
the effect of requiring less energy to be expended in the
A well-designed EHC system is capable of taking the main
boiler,
to raise the conditions of the steam to one suitable for
and intermediate stop and control valves from a fully open to
admission to the turbine unit, and does not suffer from the
a fully closed position in about 0.2 s.
disadvantage of having to throttle the steam at inlet and

thereby
sacrifice a considerable energy penalty.
Stop and Control Valves
There is a requirement for close control of the boiler and

The steam turbine is furnished with a number of valves, each turbine
so that both respond to the demands of each other,
with various functions and operating modes. The largest and both
accurately and in a timely manner.
most significant valves in terms of unit operation are the
main and intermediate stop and control elements.
BEARINGS AND LUBRICATION SYSTEM

Under emergency conditions, when load is rejected, if the
unit overspeed does not exceed 8 to 10% normal operating
The
turbine, irrespective of its driven load, must be supported
speed, the stop valves will remain open. However, in the
on a
system of fluid film bearings, which in multishaft ar-
event the unit goes into an emergency overspeed, in excess of
rangements must be aligned both vertically and transversely

about 110% normal, then these valves will close and remain
to
minimize stresses introduced in the shaft. This will also
closed until the unit is restarted. These main stop valves nor-
help
eliminate the possibility of introducing vibrations into
mally have a small bypass function, a valve which opens to
the
rotor. These bearings must maintain alignment during
admit prewarming steam as the unit is started. This bypass
transient conditions, at startup and shutdown, when the unit

valve normally has the capability of bringing the unit to speed
passes
through critical speeds and will exhibit higher than
and synchronizing on full arc admission before the main stop
normal
levels of vibration.
valves are opened.
The control valves are located behind the main stop valves,
The
Axial Bearings
often within the same valve body, to reduce frictional losses.
These valves have a function of controlling unit speed and The
axial bearings must carry and support the rotor weights
also open and close in response to load demands. without
excessive deflection. The bearings are sized so that
an oil
film developed between the journal and bearing pads
Full and Partial Arc Admission prevent
metal-to-metal contact, thereby causing no wear on
the
journal or bearing pads themselves.
There are two modes for admitting steam to the turbine unit
Bearings can be subject to various forms of damage; but

at part load; these are full and partial arc admission. At full
provided that the oil is maintained at its design defined level

load these two admission systems operate in essentially the
of
cleanliness, the bearings are aligned, and oil supply is
same manner, with the valves in their ‘‘wide open’’ position.
maintained, they will operate with a minimum of mainte-

At this valve setting, steam is flowing through a complete in-
nance.
let arc. However, at part load the entire valve flow area is not
required, and the inlet area of the valves must be reduced to
The
Thrust Bearings
admit less steam. In the unit designed for full arc admission,
each of the control valves partially close, thereby throttling The
thrust bearings are used to carry the axial thrust devel-
the steam and thus destroying a portion of the available ther- oped in
the unit steam path and are also used to provide axial
mal energy. With the partial arc admission system, only the
location of the rotor, which is necessary to allow axial clear-
top (or last to open) valve closes, leaving the remaining ele- ances
to be maintained at design values. A thrust bearing is
ments to operate wide open and therefore not suffering from
supplied for each rotor; or, in the case of solidly coupled ele-
the energy degrading effects of throttling. ments,
one bearing which is sized to carry the total axial
----------------------- Page 508-----------------------
518 STEAM TURBINES
thrust is supplied. The thrust collar produced on the rotor is must be

of sufficient strength so they can withstand the pres-
aligned to a series of pads carried in a separate housing which sure
differentials which are produced across their walls. In
is located in the stationary portion of the unit, and therefore the case
of the high and intermediate or reheat sections, there
provides positive position of the rotating and stationary por- is a high
pressure within the unit. In the case of low-pressure
tions of the unit. sections,
the inner sections of the casing can be subjected to
both a
positive and a negative pressure gradient from inside
The Lubrication System to
outside the unit. For double-casing low-pressure designs,
the outer
casings must be able to resist the vacuum force de-
A reliable lubricating oil system is required to ensure that the veloped
across their outer walls.
bearings are able to operate in a manner which maintains The
high- and intermediate-pressure casings are designed
alignment and minimizes the frictional losses which can oc- to be
able to contain a rotor burst and therefore prevent the
cur. This system supplies oil to each of the axial bearings, the rotor
entering the power house. In the case of a low-pressure
thrust bearing, and any other location or piece of equipment section
the probability of rotor containment is low.
which requires it.
The major component of the lubricating system is the oil
The
Diaphragms
tank, which is a reservoir holding sufficient oil to charge the
system at all loads and under conditions of emergency opera- In the
case of wheel and diaphragm stages (impulse type) and
tion. In many applications there are oil pumps which are in many
low-pressure sections, the stationary blade rows are
driven directly from the main rotor. In addition, there are located
in and are carried by a diaphragm, which is itself lo-
backup systems: Normally, there is one which operates on an cated in
the casings, either outer or inner. These diaphragm
ac supply driven by power produced in the turbine, and there elements
can have large pressure drops developed across
is also a final backup system which utilizes batteries. This dc them and
can have an inner web which extends down from
backup system is only used in emergencies when the ac sup- the blade
row to the rotor, at which interface point it is neces-
ply is cut off or when the unit is totally disconnected. sary for
the diaphragm to carry seals to prevent excess steam
The oil tank contains a filtration system which receives the leakage
from one stage to the next.
oil returning from the unit, removes any particulate matter,
and returns the oil to the tank where it is reused. It is normal
practice for the condition of the oil to be monitored, both for
OPERATING PROBLEMS
purity and temperature rise within the various bearings.

Higher oil temperature rises can be an indicator of pending During
operation, the steam turbine can be subjected to a
bearing problems. number of
phenomena which have the potential to introduce
The steam turbine comprises two major groups of ele-
conditions and situations resulting in damage. These condi-
ments. These are the stationary and rotating portions of the tions, if
not corrected, could require that the unit be removed
unit. Both sets of components are precision engineered and from
service for extensive maintenance, or could even cause
are designed for operation up to 200,000 h (30 years) without
mechanical failure to the extent that components will require
the need for major replacement. However, it is recognized by expensive
replacement to correct the condition. These condi-
both designers and operators that some components are con- tions
include those associated with the steam environment,
sumable and will require replacement during the life of the the high
stress levels which are developed within the various
unit.
components themselves, and the possible ingress of chemical
ions
which have the potential to form aggressive compounds
capable
of causing corrosive-type damage. In addition to these
The Rotor
influences
which are associated with the turbine-generator
The main rotating component of the steam turbine is the ro- unit, its
design, and its operating environment, there can also
tor. The rotor comprises a central shaft, along with blades be
system disturbances which can introduce a number of
which are attached to it, and these blades may carry stage problems.
During the design phase, the design process at-
hardware such as cover bands and tie wires. Rotors are con- tempts to
anticipate and allow for these phenomena in select-
structed from monobloc forgings (a central spindle with ing and
dimensioning the individual parts of the unit, but it
wheels shrunk onto this spindle), or they can be built from is
always difficult to predict both the severity and frequency
individual forgings which are weld connected to form a single of these
conditions.
central shaft onto which the various components are The
following factors are among the more common condi-
mounted. tions
causing unit damage which can require component re-
The rotor is subjected to a number of stresses; these are placement
or repair, possibly resulting in a forced outage situ-
either (a) steady, due to rotation, or (b) alternating, due to ation or
the extension of a normal maintenance outage.
steam forces developed with the steam flow path. There are
also thermal stresses set up in the unit, which are most se- Steam
Environment Effects
vere during thermal transients and can ultimately initiate
cracks at regions of high stress concentration. The rotating The
steam environment in many units introduces conditions
components of a unit tend to be more susceptible to damage in which
the temperature reduces the mechanical properties
than the stationary portions. of the
materials of construction, and the pressure results in
high-
pressure drop loads which must be carried by the compo-
nents
themselves, resulting in high stresses.
The Casings
At the
lower-pressure conditions, moisture forms in, and is
The casings are the most important stationary components
transported by, the steam and is deposited on the various
and are designed to achieve various functions. First they steam
path components. This deposited moisture has the po-
----------------------- Page 509-----------------------

STEAM TURBINES 519
tential to cause various forms of erosion such as ‘‘worming,’’ Water

which leaks into the condenser, normally termed
‘‘washing,’’ and ‘‘impact.’’ Each of these mechanisms is capa- ‘‘raw
water,’’ contains a number of impurities, particularly if
ble of removing material from the surface of the various com- the
condenser is cooled by ‘‘brackish or salt’’ water.
ponents of the steam path and casing, introducing losses in Once
a corrodent has entered the turbine steam path, it
efficiency and also introducing the possibility of mechanical will
eventually come out of solution in the steam path, possi-
failure when sufficient material has been removed. bly
forming aggressive compounds which will concentrate at
various
locations and then has the potential to cause corrosive
Stress Level Effects pitting,
or will induce stress corrosion cracking or contribute
to
corrosion fatigue.
Stresses are induced in the various components of the steam Water
ingestion can be another cause for damage within
turbine due to a number of different phenomena. The most the
unit. The many lines carrying steam, water, and a steam–
obvious are those induced by the effects of rotation. There are, water
mix which are connected to the turbine unit can, under
however, other phenomena, particularly in the steam path, certain
conditions, cause water to flow back into the steam
where high loads are induced by the flowing steam and its path.
This water has both ‘‘quenching’’ effects, which cause
change of momentum in flowing through the blade rows. high
thermal stresses in both stationary and rotating compo-
There are also temperature change considerations, where the nents,
the rotating being the most sensitive. Reentering water
rapid increase or decrease of temperature will induce thermal
can also
produce excessive impact loads on the rotating blade
stresses into many components which cannot change temper-
elements, capable of their total destruction.

ature throughout their total volume to accommodate the
Many
lines which are connected to the turbine body proper
changes in expansion rates.
and have
access to the steam path are equipped with ‘‘nonre-
These various stresses will combine and are often ampli-
turn’’
valves. These valves are included to prevent the reentry
fied by vibratory effects to induce levels of stress in many
of water
which could collect in these lines under certain load
components which can ultimately fail. In the higher-tempera-
conditions or on unit ‘‘shutdown,’’ when condensate will drain

ture sections of any unit, the presence of both high tempera-
to the
lowest points to which they have access.
ture and high stresses has the potential to induce creep into
various components, ultimately leading to mechanical
rupture.
Vibratory Loading
The
turbine-generator unit is a high-speed machine, op-
Speed and Load Transients erating
at high direct stress levels and subject to a number of
vibratory loads introduced into the unit. These loads can be a

During normal operation there are often conditions intro-
consequence of variable flow in the steam path and can be

duced into the total system which will cause the unit to trip
caused
by rotating with unbalanced loads or by misalignment
and enter an ‘‘overspeed’’ transient. While the main and inter-
mediate stop and control valves will respond to this condition, of the
various portions. These vibratory loads can transmit
themselves back into the unit and induce failure.

it is normal for the unit to reach an overspeed condition,
which induces centrifugal stresses into various components, Many
of these phenomena are predictable and are taken
while they are still at operating temperature. This condition into
account during the design of the unit. However, it is en-
can consume a portion of their remaining life. The initial de- tirely
possible for stray and unpredictable exciting forces to
sign of these units anticipates the occurrence of, and allows exist,
which even at low amplitude have the capability of in-
for, these emergency conditions. However, there is a limit to ducing
failure if they are coincident with the fundamental or
the number of such incidents which can be tolerated. lower
harmonics of the various components of the unit.
Load changes will normally introduce a change of tempera-

ture distribution throughout the steam path, resulting in Unit
Alignment
temperature variation within the various components. There
A
multisection turbine unit comprises a number of individual
are limits set by the manufacturer for the rate of change
sections, each comprising an outer and possibly an inner cas-

which is acceptable, because these affect the thermal stresses
within both stationary and rotating components. ing
containing the rotor. These casings are subjected to an
internal
pressure which in other than the low-pressure sec-
tions is
above atmospheric.
Ingestion of External Matter In
the event that there is a vertical or transverse shift of
The steam power cycle is a complex arrangement of a number the

casing, this will not affect the rotor alignment, unless the
of mechanical components, many operating above, and a num- bearings
are located within the casing, a design feature of
ber operating below, atmospheric pressures. These pressures many
low-pressure sections. However, such casing movement
above and below atmospheric cause a continuous leakage, can
introduce ‘‘rubs’’ between rotating and stationary por-
both into and out of the working fluid cycle. tions,
causing uneven heating, and therefore it has the poten-
The leakage out of the cycle must be made up with replace- tial to
introduce problems associated with alignment.
ment fluid, which must be purged of any contaminants it may In
the case of bearing movement, whether this bearing is
contain. This purging or washing operation does have the po- located
in the casings or more likely is carried by a separate
tential, if not continuously controlled, of introducing com-
pedestal, there will be misalignment between the rotor sec-
pounds into the working fluid with a potential to form or pro- tions.
This movement will result in additional bending
mote the formation of corrosive compounds. Similarly, those stresses
being set up in the central portion of the rotor. This
elements which leak into the system are most commonly air can
introduce vibratory problems and will result in additional
(which contains oxygen) and water at condenser tube leaks. stresses
and shear forces introduced into the coupling flanges.
----------------------- Page 510-----------------------
520 STEREO IMAGE PROCESSING
External System Disturbances 4.

K. Baumann, Some recent developments in large steam turbine
practice, J. Inst. Electr. Eng., 59: 581, 1921.

External system disturbances are not a common cause of unit
5.
J. K. Salisbury, Steam Turbines and the Cycles, Huntington, NY:
failure or damage. However, in the event of a generator short
Krieger, reprinted 1974.
circuit, causing high-amplitude oscillations, the turbine rotor
6.
K. Fenton, Thermal Efficiency and Power Production, London: Pit-
could be damaged. Similarly, a distribution line disturbance,
man, 1966.
grounding, or lightning strike remote from the station can be
7.
Modern Power Station Practice, 2nd ed., vol. 3, prepared by the
transmitted back to the turbine which will possibly reject load
staff of the Central Electricity Generating Board, New York: Per-
and enter into an overspeed transient. This can induce high
gamon, 1971.
levels of shaft torque which has the potential to damage the 8.
W. J. Kearton, Steam Turbine Theory and Practice, 7th ed., Lon-
turbine unit.
don: Pitman, 1958.
9.
W. J. Kearton, Steam Turbine Operation, 7th ed., London: Pit-
STARTING AND STOPPING STEAM TURBINES
man, 1964.
The starting and stopping of steam turbines can be a complex
WILLIAM P. SANDERS
operation. The complexity is introduced by the effect of tem-
Turbo-Technic Services, Inc.
perature and the design requirement of not exceeding temper-

ature ramp and cooling rates which introduce excessive ther-
mal stresses into a number of components. For this reason
STEEL INDUSTRY. See METALS INDUSTRY.
the manufacturer will normally specify both heating and cool-
STEEPEST DESCENT ALGORITHM. See ADAPTIVE
ing rates, which, if exceeded, will consume a portion of the
FILTERS.
life of these major components.
In addition to the concerns for thermal stress, the turbine
unit will normally pass through a number of critical speeds
both during ‘‘startup’’ and ‘‘shutdown.’’ It is normal practice
for the operator to approach these speed levels, then to accel-
erate through them as rapidly as possible.
MATERIALS FOR TURBINE CONSTRUCTION
The steam turbine is designed to operate for periods of years

between being opened for inspection and the undertaking of
any maintenance refurbishment which may be required. Dur-
ing these operating periods, the components are subject to a
number of complex stresses, which are induced by rotational
speed, high temperatures, and pressure differentials. There-
fore, the material from which the various components are pro-
duced must be of high quality and must meet design require-
ments in every detail.
The design engineer specifies the materials to be used for
each particular application. For the major components, the
specification will indicate chemical composition, mechanical
properties, the method of manufacture, and the cycles and
form of heat treatment to which the material is to be sub-
jected to meet both short- and long-term requirements.
The most common and consistently used materials are the
alloy steels, the majority of which are 2.25% to 12 to 13%
chromium, and are used for all major stationary and rotating
components. While many of these steels conform very closely
to national standards, they are usually customized by the tur-
bine manufacturers to suit their individual requirements, the
actual materials being based on extensive research into the
long-term characteristics when operating in environments
similar to those encountered in the turbine unit.
BIBLIOGRAPHY
1. J. H. Horlock, Axial Flow Turbines, London: Butterworths, 1966.
2. L. M. Milne-Thomson, Theoretical Aerodynamics, New York: Mac-

millan, 1958.
3. H. M. Martin, Steam turbines, The Engineer (London), 167, 1913.
----------------------- Page 511-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Synchronous Generators
Standard Article

Ahmed M. El-Serafi 1
1University of Saskatchewan, Saskatoon, Canada

reserved.
Search
DOI: 10.1002/047134608X.W3006
❍ Acronym Finder
Abstract
Generated Voltage Waveshape
Steady-State Operation
Sudden Three-Phase Short Circuit
Dynamic Performance
Determination of Synchronous-Generator Parameters
Methods for Determining the Parameters Representing the Cross-Magnetizing

Phenomenon

%20Conversion/W3006.htm17.06.2008 16:46:57
----------------------- Page 512-----------------------

c
SYNCHRONOUS GENERATORS
Synchronous generators have long been used for converting mechanical to electrical
energy in ac power systems.
The great majority of these generators are three-phase machines operating
at 50 Hz or 60 Hz. In some
other applications, other frequencies are also used, such as 400 Hz in
aerospace and marine electric power
applications. For other special applications, single-phase generators and
generators with more than three
phases are built. In this article, three-phase synchronous generators operating at
power frequency, (i.e. 60 Hz
or 50 Hz) will be mainly considered.
A three-phase synchronous generator has essentially two elements: a three-
phase armature winding and
a magnetic field excitation winding or structure. For constructional
convenience and economic reasons, the
armature winding is commonly located in the stator (nonrotating part of the
machine), and the rotor (rotating
part of the machine) contains the magnetic field excitation winding or structure. In
nearly all synchronous gen-
erators, a field winding carrying a dc current is used to establish the magnetic
field. In some small synchronous
generators, permanent magnets are utilized to establish this magnetic field.
A three-phase armature winding consists essentially of sets of coils
equal to the number of pairs of
poles symmetrically distributed around the stator, and each set consists
of three groups of coils, which are
symmetrically distributed under every pair of poles. Each one of these groups
belongs to one of the phases,
and all the groups belonging to one phase are connected together, usually in series
but occasionally in parallel,
to form the phase winding. For the availability of a neutral and because of
insulation requirements, the three
phase windings are connected in wye (star) for most synchronous generators.
When the rotor of a three-phase synchronous generator is driven by a prime
mover, the magnetic flux
linking the armature winding coils alternates and an alternating voltage is
generated in them. The generated
voltages in the three phase windings are similar, but are always 120◦
apart in phase. The frequency of these
voltages will depend on the speed of the prime mover and the number of poles of the
synchronous generator
and is given by the following relationship (1,2):
where f is the frequency of the generated voltage in hertz,p is the number of

poles of the synchronous generator,
and n is the speed of the rotor in revolutions per minute. For operation at a
desired frequency, the choice of the
number of poles of a synchronous generator will depend on the speed of the prime
mover. For optimal operation,
the speeds of the various prime movers depend on their types, and thus the number
of poles and the form of
construction of synchronous generators will essentially depend on their prime
movers. The magnitude of the
generated voltage depends on the rate of change of the magnetic flux
linking the coils and, for the no-load
situation, is given by the following relationship (1,2):
----------------------- Page 513-----------------------
2 SYNCHRONOUS GENERATORS
Fig. 1. Sketch showing the rotor construction of a two-pole

cylindrical-rotor synchronous generator.
where Ep is the root-mean-square generated phase (line-to-neutral) voltage in

volts and is usually called the
open-circuit voltage, φ is the magnetic flux per pole in webers, f is the
frequency in hertz, N is the number of
turns in series per phase, and kw is a winding factor that depends on the winding
design (1,2). For the commonly
used wye-connected armature windings, Eq. (2) gives the line-to-neutral generated
voltage, and the generated
voltage between the machine terminals (the line-to-line voltage) is given by the
following relationship:
The construction of synchronous generators may vary in several

aspects. However, they are usually
divided into two different basic types related to the form of their rotor
construction:
Cylindrical-Rotor Generators. These generators have cylindrical rotors in

which the field winding is em-
bedded in axial slots (Fig. 1). The rotors usually have a length that is
several times their diameter, and
are made from a single steel forging or from several forgings. These types of
synchronous generators are
suitable when steam or gas turbines are used as prime movers. Such turbines
operate best at high speeds,
usually 3600 (3000) or 1800 (1500) revolutions per minute. At these
speeds, the number of poles of a
synchronous generator has to be either two or four for 60 (50) Hz operation.
Salient-Pole Generators. These types of synchronous generators have rotors of
salient-pole construction (Fig.
2). The rotor has usually a diameter several times its length, and the field
windings are wound around the
poles and connected in series. These generators are used when the prime
movers are hydraulic turbines
or internal combustion engines. Such types of prime movers operate at best at
relatively low speeds, in
the range of 50 to 1000 revolutions per minute. At these speeds, the number
of poles of a synchronous
generator has to be larger than four for 60 or 50 Hz operation. In this case,
it is more economical and
convenient to construct the field structure in the salient-pole form.
GENERATED VOLTAGE WAVESHAPE
As mentioned before, the voltages are generated in the armature phase windings of a
synchronous generator
as a result of the relative motion between the magnetic field and the armature
winding. The waveshape of
these generated voltages will only be sinusoidal if the magnetic-flux-
density waveshape is also sinusoidal.
----------------------- Page 514-----------------------
SYNCHRONOUS GENERATORS 3
Fig. 2. Sketch showing the rotor construction of a six-pole salient-

pole synchronous generator.
In practice, the magnetic-flux-density waveshape usually contains odd harmonic

components, and as a result,
corresponding odd harmonic components, though not of corresponding magnitudes, will
appear in the generated
phase voltages.
Usually, the most prominent harmonic components in the magnetic-flux-density
waveshape are the third,
fifth, and seventh. In general, the presence of such harmonic components can be
treated in Eqs. (1) and (2)
as if the synchronous machine, besides having the given number of poles,
also had a number of harmonic
poles equal to the harmonic order times the number of (fundamental) poles.
With regard to the line-to-line
generated voltage for the case of a wye-connected armature winding, Eq.
(3) is also applicable to all odd
harmonic components except the third harmonic and its multiples. In this case,
these harmonic components
do not appear between the line terminals of the synchronous machine, because they
are in phase in all the
three phase windings and hence will cancel each other in any line-to-line voltage
calculation. However, some
problems may still be caused by such harmonics in the generated line-to-neutral
voltages.
With the proper design of the synchronous machine and particularly its
armature winding, the magnitudes
of the most prominent harmonic components of the generated voltages can
be reduced and, in some cases,
eliminated. Through this, the harmonic content of the generated voltage can be kept
within the values specified
by standards (345–6). It is worthy of mention here that the generated voltage
waveshape specifications are
usually limited to the open-circuit conditions, since the loading of synchronous
generators affects the harmonic
contents of their generated voltage in different ways that depend on the levels of
saturation in the generator’s
magnetic circuit under load.
STEADY-STATE OPERATION
When the active power demand from a synchronous generator changes, small changes of
speed cause a governor
to adjust the input power of the prime mover accordingly. In this case, the prime-
mover speed and the system
frequency will be maintained as nearly as possible at the specified
values. On the other hand, the effect of
changing the field current of the synchronous generator will depend on the nature of
the system connected at
its terminals.
----------------------- Page 515-----------------------
Fig. 3. The relationships between the air-gap flux and the magnetizing ampere-turns
(saturation curves) along the direct
and quadrature axes of a salient-pole synchronous generator.
If the synchronous generator runs alone to supply power to a load, the

terminal voltage will depend on
both the field current of the generator and the nature of the load. Thus, to keep
the terminal voltage constant
at a specified value, the field current has to be adjusted if the load changes.
If, however, the synchronous generator is connected to a large
power system, as is usually the case,
changing its field current does not affect the system voltage and only the reactive
power supplied from the
synchronous generator will change. In other words, it changes only the power factor
of the power supplied by
the generator. A synchronous generator is called overexcited if its field current is
more than is needed to make
it operate at unity power factor. In this case, the generator will operate at
lagging power factor and is said to
be supplying reactive power. If the field current of the generator is less than is
needed to make it operate at
unity power factor, the synchronous generator is called underexcited. In this case,
the generator will operate
at leading power factor and is said to be absorbing reactive power.
Phasor diagrams. For the analysis of a synchronous generator
connected to a large power system,
it is commonly assumed that the generator is connected to an infinite bus system. An
infinite bus system is
defined as a system that can absorb power without its voltage and frequency
exhibiting any change. To carry
out such an analysis, phasor diagrams are usually used and certain machine
information is needed. These are
the direct- and quadrature-axis open-circuit saturation curves (Fig. 3), the
armature winding resistance Ra ,
the unsaturated direct- and quadrature-axis synchronous reactances X d and X q ,
and the leakage reactance X l
or Potier reactance X p (1,2,7).
For a cylindrical-rotor synchronous generator, it can be assumed that the
magnetic path is homogeneous
around the rotor periphery, that is, all the saturation curves representing the
magnetic relationship along all
axes around the rotor periphery are similar. With this assumption, the direct- and
quadrature-axis unsaturated
synchronous reactances will be equal, and both are simply called the synchronous
reactance. In this case, the
equivalent circuit of Fig. 4 could represent the synchronous generator.
In this figure, Ea is the generator
terminal voltage per phase, EG is the generated voltage per phase, I a is
the line current, Ra is the armature
winding resistance per phase, and X s is the synchronous reactance per phase.
Usually, the effect of the armature
winding resistance can be neglected. Figure 5 shows the phasor diagram for this
equivalent circuit for the case
of an overexcited generator supplying power at a lagging power factor. In this
phasor diagram, φ is the power
factor angle and δ is the angle between the generator terminal and generated
voltages, which is usually called
the load or power angle.
For a salient-pole synchronous generator, or when saliency is not neglected
in the case of a cylindrical-
rotor generator, account must be taken of the differing direct- and quadrature-axis
synchronous reactances.
----------------------- Page 516-----------------------
Fig. 4. The representation of a nonsalient-pole synchronous

generator by an equivalent circuit.
Fig. 5. Phasor diagram illustrating the relationship between the electrical

quantities of an overexcited nonsalient-pole
synchronous generator connected to an infinite bus system.
Figure 6 shows the phasor diagram for this case when the generator is
overexcited and supplying power
at a lagging power factor. In this phasor diagram, I adX d and I aqX q
are the voltage drops in the direct- and
quadrature-axis synchronous reactances calculated by using the direct- and
quadrature-axis components of
the armature current, respectively. To obtain the values of these two components, a
phasor equal to I aX q is
added to the terminal voltage to obtain the phasor Eq . The direction of Eq gives
the direction of the quadrature
axis, and the direction of the direct axis will, as a result, be known. Knowing the
directions of these two axes,
the components of the armature current along the direct and quadrature axes will be
known, and the voltage
drops I adX d and I aqX q can be calculated. It is clear from the phasor diagram
that the phasor of the generated
voltage EG will be in the same direction as Eq , that is, it will be in the
direction of the quadrature axis.
Effect of saturation on the steady-state performance. In
the analysis of the steady-state perfor-
mance of synchronous generators, the accurate calculation of their field excitation
and load angle depends to
a large extent on their saturation condition. In general, the effect of saturation
will depend not only on the
saturation curve in the axis (direction) of the resultant machine ampere-turns, but
also on the phase angle
between the resultant ampere-turns and the resultant magnetic flux. This phase angle
depends on both the
machine saliency and the different saturation levels along the different
axes of the synchronous generator
(8,9).
To take account of saturation in the synchronous-generator phasor diagrams,
empirical methods, which
in most of the cases seem to give fairly closer agreement to the measured values,
are used. In these methods,
the custom in the past was to use only one saturation factor or
increment corresponding to the saturation
along the direct axis. This has been replaced in present practice by the use of two
different saturation factors
corresponding to the saturation conditions along the direct and quadrature axes,
since the saturation levels
are very different along the two axes even in the case of the cylindrical-rotor
synchronous generators. These
saturation factors (or increments) are defined as the ratio of (or the difference
between) the actual excitation
and the excitation on the air-gap line of the saturation curves at the
operating point. The internal voltage
behind the leakage (or Potier) reactance is usually used to locate the operating
point on the saturation curves.
In using these two saturation factors, the magnetic coupling between the direct and
the quadrature axes (the
----------------------- Page 517-----------------------
Fig. 6. Phasor diagram illustrating the relationship between the

electrical quantities of an overexcited salient-pole
synchronous generator connected to an infinite bus system.
cross-magnetizing phenomenon) is ignored. The importance of allowing for the cross-

magnetizing phenomenon
has, however, been recognized recently, and an accurate two-axis model of saturated
synchronous generators
should include it (8,9).
Use of one saturation factor (or increment). The use of one
saturation factor or increment has been
practiced both for salient-pole and for cylindrical-rotor synchronous machines,
under different assumptions. In
the case of salient-pole machines, saturation is assumed to occur only along the
direct axis, and a saturation
increment corresponding to the voltage behind the leakage (or Potier) reactance, El
, or to the projection of this
voltage on the quadrature axis, Elq , is obtained from the direct-axis saturation
curve (Fig. 3). The excitation,
including the effect of saturation, will be equal to the sum of this
saturation increment and the excitation
from the air-gap line of the direct-axis saturation curve corresponding to the
voltage EG calculated using the
unsaturated direct- and quadrature-axis synchronous reactances (Fig. 6).
In the case of cylindrical-rotor synchronous generators, saturation can be
significant in both the direct and
quadrature axes. However, the use of one saturation factor (or increment) is
applied assuming the magnetic
path is homogeneous around the rotor periphery. Thus, the direct-axis
saturation curve will represent the
magnetic relationship along any axis around the rotor periphery and can
be used to find the saturation
correction corresponding to the total air-gap flux (i.e., corresponding to the
voltage behind the leakage or Potier
reactance). Adding this saturation increment in phase with the voltage behind the
leakage or Potier reactance to
the voltage behind the unsaturated synchronous reactance, the total excitation
including saturation is obtained
(Fig. 7). In a second approach, the unsaturated synchronous reactance, X SU , is
broken into two components,
and the larger component (X SU–X l or X SU–X p ) is adjusted by the
saturation factor to give, together with X l or
X p , the saturated synchronous reactances. Thus,
An internal voltage EQD is then obtained using this saturated

reactance, as shown in Fig. 8. The total
excitation is then obtained by multiplying this internal voltage by the saturation
factor. On comparing Figs. 7
and 8, it is clear that the two approaches are similar.
----------------------- Page 518-----------------------
Fig. 7. The inclusion of the effect of saturation in the phasor diagram of a

nonsalient-pole synchronous generator.
Fig. 8. The application of the saturated-synchronous-reactance method to include

the effect of saturation in the phasor
diagram of a nonsalient-pole synchronous generator.
Use of two saturation factors. In this approach, both the direct- and
quadrature-axis saturation effects
are represented by specifically adjusting the unsaturated direct- and quadrature-
axis synchronous reactances
by corresponding saturation factors K d and K q to obtain the saturated direct-
and quadrature-axis synchronous
reactances as discussed in the previous sub-subsection. In this case, K d is
calculated from the direct-axis sat-
uration curve to give X dsat , while K q is calculated from the quadrature-
axis saturation curve to give X qsat .
The internal voltage behind the leakage or Potier reactance is used to determine
the operating point on both
saturation curves. Using these saturated reactances X dsat and X qsat in
Fig. 6, an internal voltage (≡EG in
Fig. 6) is obtained. The total excitation is then obtained by multiplying this
internal voltage by the saturation
factor K d . Empirically, the use of the two saturation factors seems to give, in
some cases, an accurate represen-
tation of the effect of saturation along the intermediate axis of the total
resultant ampere-turns, that is, the
effect of both the saturation factor due to the total resultant ampere-turns using
the saturation curve along
the intermediate axis of the total resultant ampere-turns, and the phase angle
between the resultant flux and
the resultant ampere-turns.
Cross-magnetizing phenomenon. It has been recognized recently
that the magnetic coupling between
the direct and quadrature axes of saturated synchronous machines (cross-magnetizing
phenomenon) plays an
important role in their analysis (8,9). When the flux linkages along the direct and
quadrature axes due to the
components of the ampere-turns along these axes are handled separately, this
magnetic coupling causes changes
in these flux linkages. This can be represented in the phasor diagram of saturated
synchronous generators
----------------------- Page 519-----------------------
Fig. 9. The inclusion of the cross-magnetizing effect in the phasor diagram of

a saturated synchronous generator.
in the form of separate direct- and quadrature-axis voltage drops, Edq

and Eqd , which are proportional to
the changes in the quadrature- and direct-axis flux linkages, respectively, and
which can be called the cross-
magnetizing voltages (Fig. 9). The values of these cross-magnetizing voltages are
functions of the components
of the ampere-turns along both the direct and quadrature axes (8,9). In this case,
the saturated synchronous
reactances are calculated using saturation factors obtained from the direct- and
quadrature-axis saturation
curves using the projections of the voltage behind the leakage or Potier reactance
on the direct and quadrature
axes, respectively, as the operating points.
In a second approach, the effects of both the direct and
quadrature saturation factors and the cross-
magnetizing phenomenon are combined to give the values of saturated direct- and
quadrature-axis synchronous
reactances (10). In this case, the values of these saturated synchronous reactances
are functions of the compo-
nents of the ampere-turns along both the direct and quadrature axes and can be used
as shown in the phasor
diagram of Fig. 6.
Power–load-angle relationship. From the phasor diagram of Fig. 9,
the power delivered by a syn-
chronous generator connected to an infinite bus system, losses being neglected, can
be derived as follows:
----------------------- Page 520-----------------------

Fig. 10. A power capability diagram that shows the loading limitations imposed by
heating, prime mover, excitation, and
stability on a nonsalient-pole synchronous generator.
If the cross-magnetizing effect [the last two terms of Eq. (4)] and saliency
[the second term of Eq. (4)] are
neglected as represented in the phasor diagram of Fig. 8, the power–load-angle
relationship becomes
The maximum power that can be delivered by the synchronous

generator is then E E /X when δ
a QD sat
equals 90◦ . This is the stability limit under steady-state conditions. However, if
saliency is not neglected, the
angle at which a maximum power can be delivered by the synchronous generator is
less than 90◦ , and its value
is dependent on the operating condition. In this case, the maximum power will be
larger than EaEQD/X sat .
Loading limitations. The loading limitations imposed by heating, prime
mover, excitation, and stability
on a synchronous generator connected to an infinite bus system can be evaluated by
using power capability
diagrams (11121314–15). Neglecting losses and saturation, Fig. 10 shows
such a power capability diagram
for the case of a nonsalient-pole synchronous generator connected to an infinite bus
system. In this diagram,
each point represents an operating condition. For example, point e represents the
rated loading condition, the
line bo represents the constant-phase voltage of the infinite bus system Ea , the
line oe represents the rated
stator current to a scale or the rated MVA to another scale at power-factor angle
φ, the lines og and oh are the
corresponding active and reactive powers, and the line be represents the phase
generated voltage EG .
The operation of this machine must be so controlled that the operating power
remains within the bound-
aries set by:
(1) An arc ef of center b and radius be representing the rated field current, that
is, the rotor heating limit
(2) An arc of center o and radius oe representing the rated stator current, that
is, the stator heating limit
(3) The line dge representing the rated power output of the prime mover
(4) An arc ac that represents the minimum acceptable field current
(5) The curve ci, which represents the practical stability boundary beyond which
instability at all loads occurs
when the synchronous generator absorbs excessive reactive power, (i.e., is
underexcited)
----------------------- Page 521-----------------------
Fig. 11. A power capability diagram that shows the loading limitations imposed by
heating, prime mover, excitation, and
stability on a salient-pole synchronous generator.
For the case of salient-pole synchronous generators, the power capability

diagrams can be drawn from
the phasor diagram (Fig. 6) as illustrated in Fig. 11 when losses and saturation
are ignored. As in the case
of nonsalient-pole synchronous generators (Fig. 10), the lines oe, og,
and oh are the rated MVA and the
corresponding active and reactive powers, respectively. In Fig. 11, the
ratio om/oe = X q/X d , the line op =
Ea2/X d , the line on = Ea2/X q , the line ej is perpendicular to the
line pm , and the line pj represents the
generated voltage. At any loading condition, the length be on the line

through point n represents also the
generated voltage, since the angle pbn is always 90◦ and the point b moves on a
semicircle constructed on the
diameter pn . This semicircle represents the locus of the operating point e with
zero excitation. For any other
constant excitation, the locus of point e is not quite a circular arc; however,
this becomes evident only with low
excitation. In this figure, the active power represented by the line og is divided
into two components: ok, which
is the power contributed by the excitation, and kg (= bl), which is the reluctance
power due to the saliency.
The theoretical steady-state stability boundary nqr can be obtained by applying the
expression dP/dδ = 0. It
should be noted that salient-pole synchronous generators can be operated with
negative excitation (12,16).
SUDDEN THREE-PHASE SHORT CIRCUIT
If the three phases of a synchronous generator are suddenly short-circuited

together, a high current of up to
several times the full load value will flow in the stator. As shown in Fig. 12, this
current have basically two
components: a balanced three-phase alternating current component and an offset
direct current.
----------------------- Page 522-----------------------
Fig. 12. Typical oscillograms of the stator and field currents of a synchronous
generator on a sudden three-phase short
circuit at its terminal.
In the case that the synchronous generator is initially unloaded, the

balanced three-phase component of
the short-circuit current can be expressed by the following equation (1, 2, 17–19):
where I is the root-mean-square value of this balanced alternating current

component, E is the root-mean-
square value of the phase voltage prior to the short circuit, t is the time after
the instant of the short circuit,
X d## is the direct-axis subtransient reactance, X d# is the direct-axis
transient reactance, X d is the direct-axis
synchronous reactance, τd## is the direct-axis short-circuit subtransient
time constant, and τd# is the direct-
axis short-circuit transient time constant. The magnitude of the second component,
namely the offset direct
----------------------- Page 523-----------------------
current, depends on how near the instantaneous value of the phase voltage is to
zero at the instant of the short
circuit. For the instant corresponding to zero instantaneous phase voltage, the
offset direct current will be at
its maximum magnitude, and it will be zero for the instant corresponding to the
peak of the phase voltage. The
magnitude of the offset direct current in each phase will be equal and opposite to
the instantaneous value of
the balanced alternating current component at the instant of the short circuit, and
thus it can be equal to zero
in only one of the phases. In comparison with the alternating component, the offset
direct current components
decay at a rapid rate, with the armature short-circuit time constant τa .
The balanced three-phase component of the short-circuit stator
current gives rise to a direct current
component in the rotor circuits, namely the rotor damper circuits and the field
winding. This is because the
balanced three-phase component produces a magnetomotive force rotating
synchronously with the rotor. In the
case of the initially unloaded synchronous generator, the axis of this
magnetomotive force lies along the pole
axis and the magnetomotive force tends to reduce the flux linkages with the rotor
circuits. Such changes of the
flux linkages with the rotor circuits result in induced currents in the rotor damper
circuits and field winding.
The induced currents in the damper circuits decay with the direct-axis short-
circuit subtransient time constant
τd## , while the induced current in the field winding decays after the subtransient
period with the direct-axis
#
short-circuit transient time constant τd .
On the other hand, the stator offset direct current components produce a
stationary magnetomotive force
that will induce a rotational-frequency current in the rotating rotor
damper and field circuits. These rotor
currents, like the stator offset direct current components, will decay
with the armature short-circuit time
constant τa .
Effect of sudden three-phase short-circuit on synchronous generators.
As a result of the large
currents flowing in the stator windings of a synchronous generator on a
sudden three-phase short-circuit,
electromagnetic forces much larger than normal will result on the end
windings of the stator coils (18,19).
These forces can produce severe mechanical strains on the end windings and cause
insulation failure. These
end windings must, therefore, be adequately braced.
Besides the forces on the end windings of the stator coils, a sudden three-
phase short-circuit of a syn-
chronous generator results also in pulsating torques, which can be many
times the magnitude of the rated
torque (17,19). The casing, the shaft, and the foundation of the synchronous
generator must be designed to
withstand the stress developed as a result of these torques.
DYNAMIC PERFORMANCE
Although a synchronous generator normally operates in parallel with others at a

constant speed corresponding
to the system frequency and number of poles, variation in the rotor speed about its
average value can occur
because of variations in the terminal voltage, field current, load, or driving
torque. These changes in angular
velocity are associated with changes in the load angle δ and usually occur in an
oscillatory form. For the proper
operation of the system, such oscillations should damp as quickly as possible.
Synchronizing procedure. The operation of synchronous
generators in parallel with others raises
the problem of switching a generator into service (synchronizing) and ensuring that
it subsequently remains
in synchronism. In connecting a synchronous generator in parallel with another
machine or to an electrical
system, certain conditions must be satisfied by the incoming generator with respect
to the bus bars of the other
machine or of the system. These conditions can be summarized as follows:
(1) The phase sequence of the incoming synchronous generator should be the same as
that of the bus bars.
(2) The frequency of the incoming synchronous generator should be close to that of
the bus bars.
----------------------- Page 524-----------------------
Fig. 13. The application of the equal-area criterion to illustrate the dynamic
performance of a synchronous generator.
(3) The root-mean-square voltage of the incoming synchronous generator

should be close to that of the bus
bars.
(4) The incoming synchronous generator and the bus-bars’ voltages must be
momentarily in phase at the
instant of connecting the incoming generator to the system.
In the cases of synchronizing synchronous generators manually, a synchroscope

is usually used to satisfy
condition 4 and indicate the right instant to close the switch between the incoming
synchronous generator and
the bus bars. The synchronizing process can also be carried out by automatic means
that monitor the incoming
synchronous generator and bus bars’ frequencies and voltages and the phase angle
between these voltages and
initiate the switch closure at the right instant.
Synchronizing power. If a synchronous generator connected to an
electrical system is disturbed from
its steady state, for example by a change in the system voltage, the generator
electrical output power will no
longer balance the mechanical input power (presumed constant, and the losses are
neglected). In this case, the
generator rotor swings from load angle δ1 to a new load angle δ2 at
which the balance will be restored. The
change from δ1 to δ2 occurs usually in an oscillatory form, as illustrated for
the case of a single synchronous
generator connected to an infinite bus by the equal-area criterion in Fig. 13. In
this figure, the initial operating
point A moves to B as the infinite bus voltage drops from Ea1 to Ea2 ,
reducing the electrical output power of
the synchronous generator from P 1 to P2 . The difference between the mechanical
power input Pm (equal to P 1
and presumed to be kept constant) and the new electrical output power P2
will accelerate the rotor, and the
operating point moves from B to C on the power–load-angle curve of Ea2 . At point
C, the rotor speed is higher
than the synchronous speed, and thus the operating point continues to move till
point D . From C to D , the
electric output power is larger than the mechanical input power, and thus the rotor
will decelerate till its speed
reaches the synchronous speed at point D . Since the rotor continues to decelerate,
the operating point moves
back to point C and then to point B to complete one swing. The areas
ABC and CDE represent the energy
interchanged between the rotor kinetic energy and the electric energy, and if there
were no damping in the
system, these two areas would be equal and the generator rotor would continue to
oscillate between δ and δ .
1 3
However, due to the existence of damping, these oscillations will damp and the
operating point will settle at
the point C at which the electric output power is again equal to the mechanical
input power. In the case that
----------------------- Page 525-----------------------
the mechanical input power is too close to the peak of the power–load-angle curve
Ea2 , the area CDF will be
less than the area ABC and the rotor can never be slowed to the synchronous speed
before reaching point F . In
this case, the generator rotor starts to accelerate beyond point F and the
generator falls out of synchronism.
If the change in the load angle (#δ) around an operating point is small, the
accelerating or decelerating
power (synchronizing power) can be expressed as follows:
where Ps is the synchronizing-power coefficient. If the rotor oscillations were

very slow, the appropriate power–
load-angle curve for calculating this coefficient would be the steady-state curve as
in Fig. 13, and the synchro-
nizing power coefficient could be expressed (for the case of neglecting losses and
the cross-magnetizing effect)
as follows:
If the assumption that the rotor oscillations are very slow is

invalid, the use of Eq. (7) will result in
inaccurate values for the synchronizing-power coefficients. References 2021–22 give
some of the methods for
calculating the synchronizing power coefficients for such cases.
Damping power. As mentioned above, the change of the operating
condition due to a disturbance occurs
usually in an oscillatory form that damps to a final new operating condition. The
damping of these oscillations
depends mainly on the damping power, which is often considered to vary linearly
with the deviation of speed
from synchronous speed. Thus the damping power can be expressed as follows:
where Pd is the damping-power coefficient. This coefficient depends greatly

on the rotor construction, the
frequency of oscillations, the field excitation system, and the load type.
References 2021–22 give some of the
methods for calculating these coefficients.
The natural frequency of oscillations. The oscillations of the
synchronous generator rotor due to
a disturbance occur with a frequency equal approximately to its natural
frequency. This natural frequency
of oscillations can be calculated by assuming that only the synchronizing power is
acting on the oscillatory
system, which is constituted by the inertia of the coupled rotors of the
synchronous generator and its prime
mover. In this case, the natural frequency of oscillations is expressed as follows:
where H is the inertia constant (23).
----------------------- Page 526-----------------------
DETERMINATION OF SYNCHRONOUS-GENERATOR PARAMETERS
Experimental methods. Most of the parameters of synchronous

generators can be determined exper-
imentally. Reference 7 describes in detail the various methods for doing this. In
the following, a summary of
the commonly applied methods for determining the parameters used in this article is
given.
Synchronous reactances. The unsaturated direct-axis synchronous
reactance can be obtained from
the open-circuit and short-circuit tests (7). On the other hand, the unsaturated
quadrature-axis synchronous
reactance can be measured by the slip test (7) or by using the maximum-lagging-
current method (7).
In the slip test, the synchronous generator is rotated at slightly below or
above the synchronous speed
while its armature windings are connected to a three-phase supply with no field
current. On measuring the
maximum and minimum values of the current and the corresponding values
of the terminal voltage, the
quadrature-axis synchronous reactance is found as the ratio of the minimum voltage
to the maximum current.
The direct-axis synchronous reactance can also be calculated from this test as the
ratio of the maximum voltage
to the minimum current.
In the maximum-lagging-current method, the synchronous generator is
connected to an infinite bus
system through a tap-changing transformer. For a certain terminal voltage, the
synchronous generator’s field
current is reduced from its approximately rated no-load excitation to zero,
reversed in polarity, and gradually
increased in the opposite polarity. This will cause the rotor of the synchronous
generator to slip one pole. If
the negative excitation increases in small increments until instability
occurs, the armature current at the
stability limit will equal approximately its quadrature-axis component while the
direct-axis component equals
approximately zero. Neglecting the effect of the armature resistance, the terminal
voltage Ea will thus be equal
to the reactance voltage drop I a X q , and thus the quadrature-axis synchronous
reactance X q can be determined.
The quadrature-axis open-circuit characteristic curve, which is needed to
calculate the quadrature-axis
saturation factors, can also be obtained by applying this method. At the stability
limit determined as explained
above, the terminal voltage can thus be written as follows:
where EG is the internal electromotive force that represents the air-gap flux along
the quadrature axis. Mea-
suring the terminal voltage Ea and the armature current I a at the stability
limit and subtracting the armature
leakage voltage drop X lI a from Ea , the relationship between the
electromotive force EG and the armature
current I a can be obtained and can be used to represent the relationship
between the air-gap flux and the
excitation current along the quadrature axis, that is, the quadrature-axis open-
circuit characteristic curve.
Leakage reactance. For determining the value of the leakage reactance,
the Potier reactance X p mea-
sured at rated terminal voltage is usually used as an approximation. This reactance
can be determined with
the help of the open-circuit and zero-power-factor characteristic curves (7).
However, it has been found that
the value of the Potier reactance measured at rated terminal voltage may
not be a good approximation of
the value of the armature leakage reactance. To obtain a more accurate value, the
Potier reactance has to be
determined in the highly saturated region of the open-circuit characteristic curve
(24). This may however be
difficult, since the synchronous generator may not be able to stand the values of
the field current needed to
obtain the open-circuit and zero-power-factor characteristics in the highly
saturated region. Reference 24 has
proposed an alternative testing method to obtain accurate values of the armature
leakage reactance without
any risk to the generator under test.
Subtransient and transient reactances and time constants. The
direct-axis subtransient and transient
reactances and their respective time constants can be obtained from the three-phase
sudden short-circuit test.
----------------------- Page 527-----------------------
In this test, with the synchronous generator rotating at synchronous

speed and open-circuited, the three
phases are suddenly short-circuited and the stator and field currents are recorded.
Analysis of the envelope
of the balanced alternating component of the short-circuit stator current
leads to the determination of the
subtransient and transient reactances and time constants for the direct
axis. Nowadays, a computerized
implementation of this test and the analysis of the short-circuit currents is used
(25,26).
Recently, a test procedure for obtaining the direct-axis subtransient and
transient reactances and time
constants by frequency-response testing of synchronous generators at standstill
has been proposed (7). Un-
like the three-phase sudden short-circuit test, the standstill frequency-
response test can also be applied to
determine the quadrature-axis subtransient and transient reactances and time
constants.
Analytical methods. With the advances in computer technology and
the progress made in the area
of numerical analysis, most of the synchronous-generator parameters can be
calculated analytically with
acceptable accuracy. At present, the most commonly used technique for obtaining the
parameters numerically is
the finite-element method (2728293031–32). In this method, a detailed magnetostatic
field solution of Maxwell’s
equations, which can accurately reflect the effects of rotor saliency, slotting,
saturation, winding layouts, and
other effects, is used. The application of these methods is of great importance
when the values of the synchronous
generator parameters are to be determined in the design of synchronous generators.
METHODS FOR DETERMINING THE PARAMETERS REPRESENTING THE

CROSS-MAGNETIZING PHENOMENON
If a synchronous machine has an auxiliary excitation winding in the quadrature axis

and can thus be excited
from both the direct and quadrature axes simultaneously, the parameters
representing the cross-magnetizing
effect in it can be determined experimentally as explained in details in Ref. 8.
However, for industrial syn-
chronous machines that have no auxiliary excitation winding along their quadrature
axis, that cannot be done.
For such machines, Ref. 9 has proposed an analytical approach to determine these
parameters from their mea-
surable direct- and quadrature-axis saturation curves. An indirect experimental
method, which allows their
determination from measurements of the output active and reactive powers, the
excitation field current, the
terminal voltage, and the load angle under various loading conditions, has also
been proposed (33).
BIBLIOGRAPHY
1. M. S. Sarma Synchronous Machines: Their Theory, Stability and Excitation

Systems, New York: Gordon and Breach,
1979.
2. A. E. Fitzgerald C. Kingsley, Jr. S. P. Umans Electric Machinery , New York:
McGraw-Hill, 1990.
3. American National Standard Requirements for Salient Pole Synchronous
Generators and Generator/Motors for Hy-
draulic Turbine Applications, ANSI C50.12-1982 [R1989].
4. American National Standard Requirements for Rotating Electrical Machinery—
Cylindrical Rotor Synchronous Gener-
ators, ANSI C50.13-1989.
5. American National Standard Requirements for Combustion Gas Turbine-Driven
Cylindrical Rotor Synchronous Gen-
erators, ANSI C50.14-1977.
6. Rotating Electrical Machines, Part 1:Rating and Performance . International
Electrotechnical Commission Publication
34-1, 1969.
7. IEEE Guide: Test Procedures for Synchronous Machines , IEEE Standard 115-1995.
8. A. M. El-Serafi et al. Experimental study of the saturation and the

cross-magnetizing phenomenon in saturated
synchronous machines, IEEE Trans. Energy Convers. , 3: 815–823, 1988.
9. A. M. El-Serafi J. Wu Determination of the parameters representing
the cross-magnetizing effect in saturated syn-
chronous machines, IEEE Trans. Energy Convers. , 8: 333–342, 1993.
----------------------- Page 528-----------------------
10. A. M. El-Serafi A. S. Abdallah Saturated synchronous reactances of

synchronous machines, IEEE Trans. Energy
Convers., 7: 570–579, 1992.
11. W. Szwander Fundamental characteristics of synchronous turbo-generators, J.
IEE, 91, Part II: 185–194, 1944.
12. J. H. Walker Operating characteristics of salient-pole machines, Proc. IEE ,
100, Part II: 13–24, 1953.
13. R. W. Bruck H. K. Messerle The capability of alternators, Proc. IEE , 102,
Part A: 611–618, 1955.
14. IEEE Guide for Operation and Maintenance of Turbine Generators , IEEE Standard
67-1990 (R1995).
15. IEEE Guide for Operation and Maintenance of Hydro-Generators , IEEE Standard
492-1999.
16. F. Peneder H. J. Herzog R. Bertschi Static systems for positive and negative
excitation current, Brown Boveri Rev. , 78:
598–601, 1991.
17. C. Concordia Synchronous Machines, Bombay: New York: Wiley, 1951.
18. G. C. Jain Design, Operation and Testing of Synchronous Machines ,Bombay: Asia
19. M. G. Say The Performance and Design of Alternating Current Machines, New
York: Pitman Paperbacks, 1970.
¨
20. A. M. El-Serafi Synchronisierungs- und Dampfungskoeffizienten der
Synchronmaschine und ihre Anwendung auf
¨
Stabilitatsuntersuchung, Sci. Electrica, 16: 1–13, 1970.
21. R. T. H. Alden A. A. Shaltout Analysis of damping and synchronizing torques,
part I—a general calculation method,
IEEE Trans. Power Appar. Syst. , PAS-98: 1696–1700, 1979.
22. A. A. Shaltout R. T. H. Alden Analysis of damping and synchronizing torques,
part II—effect of operating conditions
and machine parameters, IEEE Trans. Power Appar. Syst. , PAS-98: 1701–1708,
1979.
23. P. Kundur Power System Stability and Control , New York: McGraw-Hill, 1994.
24. A. M. El-Serafi J. Wu A new method for determining the armature leakage
reactance of synchronous machines, IEEE
Trans. Energy Convers., 6: 120–125, 1991.
25. I. Kamwa et al. Computer software to automate the graphical analysis of
sudden-short-circuit oscillograms of large
synchronous machines, IEEE Trans. Energy Convers. , 10: 399–406, 1995.
26. I. Kamwa et al. Experience with computer-aided graphical analysis of sudden-
short-circuit oscillograms of large syn-
chronous machines, IEEE Trans. Energy Convers. , 10: 407–414, 1995.
27. N. A. Demerdash H. B. Hamilton A simplified approach to determination of
saturated synchronous reactances of large
turbogenerators under load, IEEE Trans. Power Appar. Syst. , 95: 560–569,
1976.
28. S. H. Minnich M. V. K. Chari J. F. Berkery Operational inductance of turbine-
generator by the finite element method,
IEEE Trans. Power Appar. Syst. , 102: 20–27, 1983.
29. General Electric Company, Improvement in Accuracy of Prediction of Electric
Machine Constants and Generator Models
for Subsynchronous Resonance Conditions, Projects 1288-1 and 1513-1,
EPRI EL-3359, Vol. 2: Application of Two-
Dimensional Finite-Element Techniques for Synchronous Machine Modeling, Final
Report , 1984.
30. S. R. Chaudhry S. Ahmed-Zaid N. A. Demerdash A coupled finite-element/state-
space modeling of turbogenerators in
the ABC frame of reference, part I—the no load case; part II—the
short circuit and load cases including saturated
parameters, IEEE Trans. Energy Convers. , 10: 56–62, 63–70, 1995.
31. S. R. Chaudhry S. Ahmed-Zaid N. A. Demerdash An artificial-neural-network
method for identification of saturated
turbogenerator parameters based on a coupled finite-element/state-space
computational algorithm, IEEE Trans. Energy
Convers., 10: 625–633, 1995.
32. F. Deng N. A. O. Demerdash A coupled finite-element state-space approach for
synchronous generators, part I: model
development; part II: applications, IEEE Trans. Aerosp. Electron. Syst. , 32:
775–784, 785–794, 1996.
33. A. M. El-Serafi A. S. Abdallah Methods for determining the parameters
representing the cross-magnetizing phenomenon
in saturated synchronous machines, Proc. Int. Conf. on Electrical Machines,
Paris, France, Vol. 3, pp. 468–473, 1994.
AHMED M. EL-SERAFI
University of Saskatchewan
----------------------- Page 529-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Synchronous Motor Drives
Standard Article

Keith A. Corzine1 and Scott D. Sudhoff2
1University of Wisconsin, Milwaukee, Milwaukee, WI
2Purdue University, West Layfayette, IN

Search
reserved.
DOI: 10.1002/047134608X.W3007
❍ Acronym Finder
Abstract
Machine Models
Power Converters
General Control Philosophy
Sensor Requirements
General Relationships for Analyzing Drive Systems
Voltage-Source, Inverter-Fed, Permanent-Magnetic, Synchronous Machine

Drives
Current-Regulated, Inverter-Fed, Permanent-Magnet, Synchronous Machine

Drives
Voltage-Source Inverter-Fed, Synchronous-Reluctance, Machine Drives
Current-Regulated, Inverter-Fed, Synchronous-Reluctance, Machine Drives
Semicontrolled, Inverter-Fed, Wound-Rotor, Synchronous Machine Drives
----------------------- Page 530-----------------------
Keywords: brushless dc motors; current regulation; current-regulated inverters;

hysteresis current control; inverter control; load-
commutated inverter; normal load commutation; PWM power converters; permanent
magnet; permanent-magnet synchronous
motor drives; revolving field; rotor speed; round rotor; salient pole; sine-
triangular modulation; synchronous reluctance; torque
angle; voltage source inverters; wound field

----------------------- Page 531-----------------------
246 SYNCHRONOUS MOTOR DRIVES
SYNCHRONOUS MOTOR DRIVES
Synchronous motor drives are used to convert between elec-

trical
and mechanical energy in applications which include
disk
drives in computers, compressors and fans, high-perfor-
mance
servo systems for robotics, and propulsion systems in
some
electric vehicles. This type of variable-speed drive sys-
tem offers
high efficiency and excellent torque and speed reg-
ulation.
Figure 1 is a functional representation of a generic
synchronous motor drive. A power converter, consisting of a

number of
power semiconductor switches, converts electrical
energy
from the power source to a form suitable for supplying
the
synchronous-machine stator windings. A low-power field
exciter
supplies the field winding if one is present. The power
source is
often the electric utility grid, but may also be a bat-
tery or
another rotating machine. The conversion between
electrical
and mechanical energy actually takes place in the
synchronous machine which may be a wound-rotor machine,

a
synchronous-reluctance machine, or a permanent-magnet

----------------------- Page 532-----------------------
SYNCHRONOUS MOTOR DRIVES 247
ans
apart are associated with each of these three phases. The
Field main
rotor winding is the field winding (fd-fd #). In addition,
exciter
the
rotor also includes damper ‘windings’ denoted (kd-kd #)
and
(kq-kq #). These windings may not actually be windings at
all;
instead they often represent currents flowing in alumi-
num
bars inthe rotor structure which provide damping
Power Power
source converter SM
torque during electromechanical transients. The q and d axes
of
the machine denote the magnetic axes of the q-axis damper
and
d-axis field and damping windings, respectively. Figure 2
Switching Sensors also
defines the mechanical rotor position #rm of the machine,
logic
which is measured from the as axis to the q axis. The mechan-
ical
rotor speed #rm is measured in the counterclockwise direc-
Command Control
tion. For the purpose of machine analysis, it is convenient to
define the electrical rotor position #r and the electrical rotor
speed, #r as P/2 times the corresponding mechanical quanti-

Figure 1. Generic synchronous motor drive. The power converter
ties, where P is the number of poles.
converts energy from the power source (a utility grid, a battery pack,
The steady-state operation of a synchronous machine is as
or another machine) to a form suitable for the synchronous machine
follows. First, a dc source is applied to the field winding. This
(SM).
makes the rotor iron in the area where the d-axis leaves the
rotor a North magnetic pole and the rotor iron on the opposite
side
a South magnetic pole. Next, a three-phase ac source ex-
machine. Another important aspect of this system is a sensor
cites the stator windings. The a-, b-, and c-phase applied volt-
array which often includes a rotor positional sensor, current
ages
have the form
sensors, and voltage sensors. Based on the sensor output and
a torque, speed, or other input command, the drive controller
v = √2 V cos(θ ) (1)
determines the on/off status of each of the power converter
as s e
semiconductors so that the desired performance is achieved.
√ # 2π # (2)
v = 2 V cos θ −
The operation of synchronous motor drives is dictated by
bs s e 3
the operating principle of a synchronous machine. Figure 2 is

a cross-sectional view of a wound-rotor synchronous machine. and
# #
The interior of the machine (the rotating member) is called
2π
the rotor and the exterior portion (the stationary member) is
v = 2 V cos θ + (3)
cs s e
the stator. Three sets of windings (as-as #, bs-bs#, and cs-cs #)
3
located in the interior of the stator are called phases. Physi-
where
cally, these windings are sinusoidally distributed in slots on
the inside of the stator although they are shown in lumped
θ = ω t (4)
positions in Fig. 2. Three magnetic axes, which are 2#/3 radi-
e e
In
Eq. (1–4), Vs is the rms amplitude of the applied stator
voltage, and #e is the radian frequency. In the steady state,

bs axis
the
resulting currents are expressed by
ω
rm q axis
as'
ias = √2 Is cos(θe + φ ) (5)
√ # 2π # (6)
ibs = 2 Is cos θe + φi −
cs θrm
3
bs
kg'
fd' and
kd kd as axis
fd
√ # 2π # (7)
ics = 2Is cos θe + φi +

bs' kq Stator iron
3
cs'
Air
where Is is the rms stator current. Currents of the form Eq.

(5–
7) result in a rotating magnetomotive force (MMF) which
as
travels counterclockwise around the interior of the stator sur-
d axis face
at an angular velocity of #e . The interaction of the rotor
cs axis Rotor iron
poles with the stator poles causes the rotor to rotate synchro-
nously with the stator magnetic field. Thus, the electrical ro-
Figure 2. Cross-sectional view of a wound-rotor synchronous ma-
tor
speed is # , and the mechanical (physical) rotor speed is
chine. The stationary member, labeled ‘‘stator,’’ contains three sets of
e
2# /P . The fact that the rotor speed is directly tied to the ra-
windings which are sinusoidally distributed (although they are shown e
here as lumped coils). The rotating member, labeled ‘‘rotor,’’ is a sa- dian
frequency of the stator excitation is the defining charac-
lient structure containing a field winding to produce the rotor MMF
teristic of synchronous machines.
and damper windings to produce damping torque during electrome-
To make this observation more quantitative, if the resis-
chanical transients.
tance of the stator windings is negligible, the electromagnetic
----------------------- Page 533-----------------------
torque produced by a synchronous machine under the stated

bs axis
conditions is given by (1)
ωrm q axis
as'
3 P Lmd i# √2 Vs
fd
T = − sin δ
e 2 2 ω L
e d (8)
# #
cs bs rm
√
3 P 1 1 1
S
− − ( 2 V )2 sin 2δ
N
4 2 ω2 L L s
e q d
as axis
In Eq. (8), L , L , and L are the d-axis magnetizing induc-
md d q
tance, the d-axis stator self-inductance, and the q-axis self-

S Stator iron
bs' N cs'
inductance, respectively, i# is the referred field current, and
fd
Air
# is the torque angle of the synchronous machine defined as
δ = θr − θe (9)
as
d axis
Rotor iron
Physically, the torque angle represents the difference be-
cs axis Permanent
tween the electrical rotor position and the instantaneous
magnet
angle of the applied voltages. In Eq. (8), the first term repre-
sents the torque produced by the interaction of the field wind-
Figure 4. Cross-sectional view of a surface-mounted, permanent-
ing with the stator magnetic field (field torque), and the sec-
magnet, synchronous machine. This machine differs from the config-
uration in Fig. 2 in that the rotor iron is round and a permanent

ond term is from the interaction of the rotor iron with the
magnet replaces the wound field.

stator magnetic field (reluctance torque). Figure 3 illustrates
the field, reluctance, and total torque as a function of torque
angle for a 3.7 kW synchronous machine.
are less expensive to manufacture than the wound field ma-

Although the wound-rotor synchronous machine just con-
chines because the field winding and the complication of mak-

sidered is a standard configuration, several important varia-
ing electrical connection to the rotating field circuit are elimi-

tions are used in electric drive systems. First, from the form
nated.
of the torque-delta characteristic, specifically the second term
In another variation of the synchronous machine, the field

in Eq. (8), torque is produced without a field winding. Such
winding is replaced with a permanent magnet. There are two

machines are known as synchronous-reluctance devices be-
variations of this system, the surface-mounted permanent

cause the torque is produced by reluctance. These machines
magnet and interior permanent-magnet machines, as de-
picted in Figs. 4 and 5, respectively. In the surface-mounted
case, the magnetic material is mounted on the surface of the

Total
rotor. Because of the relatively low permeability of most ma-
50 torque
terials used as permanent magnets, such a machine offers rel-
atively low inductances and thereby good bandwidth for con-

Field
torque
Reluctance
torque
bs axis
rm q axis
as'
)
m
⋅
N 0
cs θrm
(
bs
e
T
S as axis
bs' cs' Stator iron
Air
as
–50
d axis
–2π – 3π –π – π 0 π π 3π 2π
Rotor iron
2 2 2 2
cs axis
Permanent
δ (rad)
magnet
material
Figure 3. Field, reluctance, and total torque as a function of delta.
The field torque is produced by the interaction of the stator magnetic
Figure 5. Cross-sectional view of a buried, permanent-magnet, syn-
field with that of the rotor field winding. The reluctance torque is
chronous machine. This permanent-magnet machine has magnetic
produced by the interaction of the salient rotor iron and the stator
saliency due to the low relative permeability of the magnetic material
magnetic field.
buried in the rotor iron.
----------------------- Page 534-----------------------

trol applications. Furthermore, such a machine is relatively
In terms of the transformed and referred variables, the
easy to construct. Although more difficult to construct, the
stator and rotor voltage equations are expressed by
interior or buried permanent-magnet machines are operated
v = r i + ω λ + p λ (13)
at higher speeds than their surface-mounted counterparts be-
qs s qs r ds qs
cause the iron surrounding the mechanically weak permanent
v = r i − ω λ + p λ (14)
ds s ds r qs ds
magnet adds the structural strength required to withstand
high operating speeds.
and
v = r i + p λ (15)
0s s 0s os
MACHINE MODELS
and
To provide a more quantitative look at synchronous-machine
drives, it is first necessary to set forth the mathematical mod-
v# = r# i# + p λ# i ∈ 1 . . . m (16)
kqi kqi kqi kqi

els of the various types of synchronous machines. These mod-
v# = r# i# + p λ# (17)
els have three parts: the voltage equations which relate the
fd fd fd fd
applied voltage to the ohmic voltage drop to winding resis-
tance and the time rate of change of flux linking the various and
windings; the flux linkage equations which relate the flux
# # # #
linking the individual windings to the current in the windings
vkdi = rkdi ikdi + p λkdi i ∈ 1 . . . n (18)
and electrical rotor position; and finally a torque equation

which predicts the electromagnetic torque in terms of the ma-
respectively, where m and n are the number of damper cir-
chine current and rotor position. The classical synchronous- cuits
used in the q and d axes and p is heaviside notation for
machine is based on Park’s transformation (2) which is ex-
differentiation with respect to time. The stator and rotor flux
pressed as
linkage equations are given by
 # # # #
λqs = Lls iqs + λmq (19)

2π 2π
cos(θ ) cos θ − cos θ +
r r r
    
3 3
λds = Lls ids + λmd (20)
f   f
qs 2  # 2π # # 2π #  as
    
f ds = sin(θ ) sin θ − sin θ + f bs
 
3 r r 3 r 3 and
f 0s   f cs
 1 1 1 
λ0s = Lls i0s (21)
2 2 2
(10)
and
In Eq. (10), f is a stator voltage v, stator current i, or stator
# # #
flux linkage #. Expressing the machine model in terms of
λkqi = Llkqi ikqi + λmq i ∈ 1 . . . m (22)
transformed variables results in considerable analytical sim-

λ = L# i# + λ (23)
plification. In particular, it eliminates rotor-position-depen-
fd lfd fd md
dent inductances from the machine description and also re- and
sults in a model wherein the various voltages, currents, and
flux linkages are constant in the steady state. In addition to
λ# = L# i# + λ i ∈ 1 . . . n (24)
kd lkd kd md
expressing the synchronous machine model in terms of trans-
i i i
formed stator variables rather than physical variables, it is

In
Eqs. (19–24), #mq and #md are the q- and d-axis magnetizing
also convenient to refer the field and damper winding vari-
flux
linkages defined as
ables to the stator by the appropriate turns ratio. In the case
of the field winding, this referral is made by defining the re-
# m #
ferred field voltage and current as
λmq = Lmq iqs + # ikqi (25)
i= 1
# Ns
vfd = N vfd (11) and
fd
# n #
and
λ = L i + # i + i# (26)
md md ds kdi fd
i= 1
# 2 Nfd
ifd = 3 Ns ifd (12)
Finally, electromagnetic torque, in terms of the transformed
variables, is expressed as
where Ns and Nfd are the effective stator and field turns, re-
3 P
T = (λ i − λ i ) (27)
spectively. The damper circuits are similarly referred. Be-
e 2 2 ds qs qs ds
cause the turns ratio is not needed to use the model and can-
not be measured in the laboratory, however, these referred
Figure 6 illustrates an equivalent circuit consistent with
variables are not defined here. Eqs.
(13–26). It is worth noting that the model is valid for
----------------------- Page 535-----------------------
iqs As
in the case of the wound-rotor machine, this model is valid
ω λ
r ds
for both steady-state and transient conditions. The model is
+ –
also valid for both surface-mounted and interior-mounted,
+ r i
s qs
permanent-magnet machines, although, in the case of sur-
L'lkq 1 L'lkq
face-mounted machines, Lq and Ld equal. Therefore, both of
v L
these symbols are normally replaced by the symbol Lss when
qs mq
working with surface-mounted, permanent-magnetic ma-
chines. It is also worth noting that, unlike a wound-rotor syn-

r' r'
– kq 1 kqm
chronous machine in which L # L , in the case of the interior
d q
permanent-magnet machine, Lq # Ld because the permanent

q axis
magnetic increases the reluctance of the d-axis magnetic
path. For most applications, the model in Eqs. (28–31) is
quite accurate. The reader should be aware, however, that it

ids ω λ L' r' is
only valid in machines with sinusoidal back-emfs. For anal-
r qs lfd fd
ysis of machines with nonsinusoidal back-emfs, the reader is

– +
+ rs Lls +
referred to (6,7).
i'
L' L' fd
lkd1 lkqn
POWER CONVERTERS
v L v'
ds md fd
The most important component of a synchronous machine

– r'kd1 r'kdn –
drive system is the power converter, which takes numerous
forms depending on whether the

input is ac or dc. In most
d axis
applications, the input power is directly obtained from a dc
Figure 6. Synchronous-machine equivalent circuit based on Park’s
source or else a dc source is constructed by rectifying the ac
transformation.
input to the drive. Therefore, it is appropriate to focus on con-
verting electrical energy from dc to ac, which is accomplished

by
a device called an inverter. The most prevalent topology
transient and steady-state conditions. At the same time, how-
for an inverter designed for a three-phase machine is the
ever, the reader is cautioned that the model has been set forth
three-phase, fully controlled, bridge converter depicted in Fig.
in a simple form. Important secondary effects, such as mag- 7,
showing the converter connected to a machine in a wye
netic saturation of the magnetizing path, are not included,
configuration. In this device, each of the transistors is a vari-
but are discussed in (3,4,5). For a detailed discussion of the
ety of fully controlled semiconductor devices (that is, they are
derivation of the model, the reader is referred to (1).
turned on and turned off at times specified by the gating sig-
Before concluding this section, it is appropriate to consider
nals T1–T6) including bipolar junction transistors (BJT),
briefly the models of the synchronous-reluctance and the per-
metal oxide semiconductor field effect transistors (MOSFET),
manent-magnet machines. In the case of the reluctance ma-
insulated gate bipolar transistors (IGBT), MOS controlled
chine, the model is identical to that of the wound-rotor syn-
thyristors (MCT). Regardless of the technology used to manu-
chronous machine except that the field current is set to zero
and the field voltage equation is eliminated from the machine
description. In the case of the permanent-magnet machines,
idc
the machine model is obtained by (1) eliminating all damper
currents and voltage equations from the machine description +
(because the rotor currents in this type of machine are nor-
T1 T2 T3
mally negligible) and (2) treating the field current as a con-
stant. After making these changes and combining the flux
linkage equations with the voltage equations,
vdc
+ + +
v = r i + ω L i + L pi + ω λ (28)
qs s qs r d ds q qs r m
v = r i − ω L i + L pi (29)
T4 vag T5 vbg T6 vcg
ds s ds r q qs d ds
– – –
and –
v = r i + L pi (30)
0s s 0s ls 0s
ias ibs ics
In Eqs. (28–30), Lq and Ld are the stator q- and d-axis self-

+ + +
vbs
inductances and #m is a constant related to the strength of the
vas vcs
permanent magnet (the fictitious, constant, field current), the
– –
number of stator winding turns per phase, and the machine
–
geometry. The torque equation becomes
3 P
T = [λ i + (L − L )i i ] (31)
Figure 7. Fully controlled, three-phase bridge inverter and wye-
e 2 2 m qs d q qs ds
connected load.
----------------------- Page 536-----------------------
idc
voltage drops, these voltages are expressed by
+
vag = #vdc T1 on, T4 off (38)
T1 T2 T3
0 T1 off, T4 on
vdc T2 on, T5 off

vdc
vbg = 0 T2 off, T5 on (39)
+ + +
v v v and
T4 ag T5 bg T6 cg
– – – –
vag = #vdc T3 on, T6 off (40)
0 T3 off, T6 on
– +
ibs
vas vbs In
Eqs. (38–40), it is assumed that one, and only one, active
i
semiconductor of each phase leg is conducting. Operation with
as
two
devices simultaneously conducting is normally not al-
+ –
ics lowed
because this would short circuit the dc source. Opera-
tion
with neither device on is occasional. However, analyzing
– v +
cs this
condition is quite involved and so the reader is referred
Figure 8. Fully controlled, three-phase bridge inverter with delta- to

(8,9,10).
connected load. At
this point, we have set forth the relationships necessary
to
predict the voltages applied to the machine, given the dc
voltage and on/off status of the semiconductor switches. It is

now
possible to use these relationships to explain the inverter
facture the transistors, they are always operated in the satu-
operation. There are two principal modes of operating the

rated or the cutoff region of their i–v characteristic, that is,
fully
controlled, three-phase bridge whereby it can be used to
completely on or completely off. Under these conditions, the
convert power from dc to ac. These modes are voltage-source

voltage across or current through the semiconductor device is
inverter (VSI) operation and current-regulated inverter (CRI)

always zero which prevents excessive power dissipation
operation. The most basic method of converting power from
within the semiconductors.
dc to
ac is to switch the transistors, as illustrated in Fig. 9,
To analyze the three-phase, fully controlled, bridge con-
where
a wye-connected load is considered. Using the stated
verter feeding a wye-connected load, it is convenient to note
gating signals and the relationships in Eqs. (32–34), one ob-

that, for any of the machines considered in the previous sec-
tains
the line-to-neutral voltage depicted in the figure. Be-
tion, the sum of the line-to-neutral voltages must be zero. Us-
cause
this mode of operation consists of a repetition of six
ing this property,
switching states, it is called a six-step operation. It is also
commonly called a 180# voltage-source inverter (180# VSI) op-

2 1 1
vas = 3 vag − 3 vbg − 3 vcg (32)
eration. The resulting line-to-neutral voltages, although ac,
possess considerable
harmonic content which results in in-
1 2 1
creased machine losses. Nevertheless, this strategy is often
vbg = − 3 vag + 3 vbg − 3 vcg (33) used
for permanent-magnet synchronous motor drives be-
cause
it is one of the least expensive strategies. For the pur-
and poses
of analysis, machines operated from the three-phase
bridge by this control strategy respond primarily to the fun-

1 1 2
damental component of the applied voltages except at very
vcg = − 3 vag − 3 vbg + 3 vcg (34) low
speeds (frequencies). The fundamental component of the
applied voltages is expressed as
In the case of a delta-connected machine, illustrated in Fig.
2
8,
v = v cos(θ ) (41)
as ,f und dc c
π
vas = vag − vbg (35)
2 v # 2π # (42)
vbs,f und = dc cos θc −

π 3
vbs = vbg − vcg (36)
and
and
2 v # 2π # (43)
vcs = vcg − vag (37)
vcs ,f und = π dc cos θc + 3
Using Eqs. (32–34) or Eqs. (35–37), the voltages applied to where

#c is an independent variable which is related to the
the machine are readily established in terms of the line-to- rotor
position for closed-loop switching controls, as described
ground voltages. Neglecting the forward diode and transistor in
later sections. In the event that the machine is delta-con-
----------------------- Page 537-----------------------
S1
T1
T1,T4
T4
S2
T2,T5
T2
T5
T3,T6
S3
T3
T6
d +
c
vag vdc
t –
Figure 10. Duty-cycle modulating, switch-control algorithm. The net
effect of the duty-cycle control is to directly control the amplitude of

vbg vdc the
machine phase voltages from the control signal d.
and
v
cg vdc
2 v # 2π # (46)
vcs ,f und = dcd cos θc +
π 3
v
1/3 dc
The
advantage of this control is that the amplitudes of the
v
vng 2/3 dc
applied voltages are readily controlled. The low-frequency
voltage harmonics are present just as in the case of a six-step
operation. In addition, the high-frequency harmonic content
increases but is less important because of the filtering action

v
1/3 dc of
the machine inductance. When the machine is delta-con-
vas
nected rather than wye-connected, the changes to the funda-
v
mental components of the applied voltage are the same as
2/3 dc
those made in the case of a six-step operation.
The primary disadvantage of a six-step operation and
duty-cycle modulation is the low-frequency harmonic content.

0 π 2π π 4π 5π 2π
There are several strategies for controlling the fully con-
3 3 3 3
trolled, bridge converter in which low-frequency harmonics
θ
c are
completely eliminated. A diagram illustrating one of these
Figure 9. Gating signals and inverter voltages for 180# voltage
strategies, sine-triangular modulation, is illustrated in Figure
source operation. The b- and c-phase voltages are identical to the a- 11.
There, instantaneous duty cycles for each of the three
phase voltage vas with phase lags of 2#/3 and

2#/3, respectively.
da +
nected Eqs. (41–43) are adjusted by multiplying the ampli-
– T1
tude by #3 and adding #/6 to each of the cosine terms.
The next level of sophistication in the control of the fully
T4
controlled bridge converter is applying duty-cycle modulation
to the switching signals, as illustrated in Figure 10. There,
db +
T2
the signals S1, S2, and S3 are identical to T1, T2, and T3 in
–
the case of a six-step operation. The commanded duty cycle d
is compared to a high-frequency triangular wave tr, which
T5
varies between zero and one. If the duty cycle is greater than
dc +
the triangular wave, the resulting line-to-neutral voltages are
T3
–
the same as in a six-step operation. Otherwise, they are zero.
The net effect is that the fundamental components of the ap-
T6
plied voltages become
t
2
Figure 11.
Sine-triangle modulating switch-control algorithm. The
v = v d cos(θ ) (44)
as ,f und π dc c
fundamental component of the resulting a- b- and c-phase line-to-
# #
2 2π
neutral voltages using this control strategy are in phase with and
vbs,f und = vdcd cos θc − (45)
proportional to the magnitudes of the instantaneous phase duty cy-
π 3 cles
d , d , and d provided that the signals constitute a balanced set.
a b c
----------------------- Page 538-----------------------
phases denoted d , d , and d are compared to a high fre-

i >i * + h
a b c
as as
quency triangular wave which varies between

1 and 1. The
result of this comparison yields the gating signal to each of
the inverter semiconductors. Assuming that the instanta-
T1 on T1 off
neous duty cycles sum to zero and that the absolute value of
T4 off T4 on
each of the duty cycles is less than 1, neglecting high-fre-

quency harmonics,
i <i * – h
as as
v = 1d v (47)
Figure 12. Hysteresis
current-regulated, switch-control algorithm.
as ,f und a dc
2
Using this algorithm, the converter switching is based on comparing
the
error between the commanded currents and the actual currents
1
vbs,f und = db vdc (48) to
a threshold level h.
2
and
by
the hysteresis level h, the lower transistor of the leg (T4)
1 is
turned on, and the upper transistor (T1) is turned off, so
v = d v (49)
cs ,f und 2 c dc
that the current decreases. Conversely, if the actual current
falls below the commanded current by an amount h, the lower

By commanding the duty cycles as
transistor (T4) is turned off, and the upper transistor (T1) is
turned on, which increases the current. For delta-connected

d = d cos(θ ) (50)
a c
machines, i* and i are replaced by (i*
i*) and (i
i ),
as as as cs as cs
d = d cos #θc − 2π # (51)
respectively, in the hysteresis control. The b- and c-phases are
b 3
controlled similarly. The net result of this switching is that,
ideally, the actual current is always within h of the com-

dc = d cos #θc + 2π # (52)
manded current. There are restrictions to this, however.
3
First, a step change in current command necessarily causes
the
actual current to deviate from the commanded current by
where d is the duty cycle (which is a constant as long as the an
amount exceeding the hysteresis level because the actual
amplitudes of the applied voltages are constant) the line-to-
current does not change instantaneously under an inductive
neutral voltages become or
motor load. Furthermore, even for steady-state conditions,
if
the actual currents are within h of the commanded cur-
1
v = dv cos(θ ) (53)
rents, the rms value of the fundamental component of the mo-
as ,f und 2 dc c tor
phase voltages must satisfy
1dv # 2π # (54)
vbs,f und = dc cos θc −
1
2 3
v < v (56)
s √6 dc
and
if
the machine is wye-connected or
1dv # 2π # (55)
vcs ,f und = 2 dc cos θc + 3
1
v < v (57)
s √2 dc
As with duty-cycle modulation, the amplitude of the applied

voltages is readily controlled. However, at the same time, if
the machine is delta-connected. If this constraint is not sat-
there are important differences. First, low-frequency harmon-
isfied, the actual current deviates significantly from the com-
ics are avoided. There is a price paid for eliminating these
manded current, a condition called a loss of current tracking.
harmonics. The maximum amplitude of the fundamental com-
Although fully controlled, three-phase bridge converters
ponent of the line-to-neutral phase voltages which is achieved are
the dominant technology for constructing inverters, other
is now limited to (1/2)vdc . Although it is possible to increase
technologies are used in industrial applications. Figure 13 de-
the amplitude further, such action introduces low-frequency
picts a three-phase, semicontrolled bridge converter con-
harmonics. For this reason, other techniques, such as space
nected to a voltage-behind-reactance ac source. The most im-
vector modulation (11), have been developed which also avoid
portant difference between the semicontrolled, three-phase
low-frequency harmonics but offer a greater maximum ampli-
bridge and its fully controlled counterpart is that semicon-
tude compared with sine-triangle modulation.
trolled semiconductors, and, in particular, thyristors, are
An example of a current regulated inverter control algo-
used rather than fully controllable semiconductors. Thyristor
rithm is hysteresis current control. This strategy assumes the
devices are considered semicontrollable because, although
existence of a current command signal for each phase. Then
they are gated on at any time, they actually turn on only if
the inverter semiconductors are switched so that the actual
forward biased and cannot be actively turned off. Turn off
currents are always within a prespecified bound (the hystere-
occurs whenever the current through the device attempts to
sis level) of the commanded current. Figure 12 illustrates the
become negative. Although this makes the control of the con-
switching-state transitional diagram for the a-phase of the in-
verter less straightforward and eliminates most of the modu-
verter where the commanded a-phase current is denoted i*.
lation strategies available in the fully controlled case, this
as
Whenever the actual current exceeds the commanded current

type of converter has the advantage that the maximum volt-
----------------------- Page 539-----------------------
Ldc
eas ebs ecs
– v + id
as
e
as Lac i + +
as
– + T1 T3 T5
– v +
as
ebs L
ac ibs
– +
vr vdc
T2 T3 T4 T5 T6 T1
– v +
as
e
cs L
ac i
cs
– + T4 T6 T2
– –
i
as
Figure 13. Semicontrolled, three-phase bridge converter. This con-

verter contains semicontrolled semiconductors with high voltage and
current ratings appropriate for high-power applications.
age and current ratings available in thyristor devices exceed

those available in fully controlled semiconductors. For this
reason, semicontrolled, three-phase, bridge converters are
most often used in high-power synchronous motor drives.
ibs
Semicontrolled bridge converters are used either as con-
trolled rectifiers, or as inverters. Figure 14 illustrates the op-
eration of the three-phase bridge converter in rectifier mode
with the dc current id constant. The upper trace illustrates
the three-phase sinusoidal voltage source which represents
the back-emf of the machine. The point at which the various
thyristors are turned on is indicated directly underneath this
trace. The control of this converter is tied to the amount of
delay from the time the line-to-neutral voltages cross each
i
cs
other to the point at which the individual thyristors are fired.
For example, valve three is fired at a firing angle # after ebs
becomes greater than eas , or, more formally, when
π
θc = α + (58)
3
Assuming the firing pattern shown and that the dc current is

constant, the resulting ac currents are illustrated in the next
three traces of Fig. 14. It is interesting to observe that, each
time a new thyristor is turned on, it causes the current in the
v
r
thyristor, which was turned on 2#/3 radians previously, to
pass through zero whereupon it turns off. This process is
called commutation, and the angle between the point where
one device begins to conduct and the device gated on 2#/3
π π π 2π 5π π 7π 4π 3π 5π 11π 2π
radians previously ceases to conduct is called the commuta-
6 3 2 3 6 6 3 2 3 6
tion angle u. For overly large dc currents or firing delays, it
θc (rad)
is possible that commutation does not occur (a commutation
failure) which prevents the converter from operating satisfac-
Figure 14. Operation of a three-phase semicontrolled converter in
torily. The final trace illustrates the rectified output voltage.
rectifier mode. The three-phase voltages represent the machine’s
back-emf. The arrows beneath these voltages indicate the rotor posi-
The most pronounced feature in the figure is the notch which
tion at which the various thyristors are turned on. The phase cur-
occurs during the commutation process.
rents are the result of the firing sequence assuming that the dc cur-
Analyzing the semicontrolled, three-phase, bridge con-
rent is constant. The notch in the rectifier voltage results from the
verter is normally quite involved, because a synchronous ma-
phase current commutation.
chine cannot be modeled accurately in a voltage-behind-in-
ductance form (unless the subtransient inductances are
equal, but this is actually never the case). However, neglect-
ing commutation, the average rectifier voltage is given by
√
√
3 3
vd = 2 E cos α (59)
π
----------------------- Page 540-----------------------
and the fundamental components of the ac current are given

φhm
by
2√3
bs axis
ias = id cos(θc − α + π ) (60)
ha
π
√
as'
ibs = 2 3 id cos #θc − α + π − 2π # (61)
ωrm
π 3
S bs qs axis
cs
and
θ
rm
ics = 2√3 id cos #θc − α + π + 2π # (62)

as axis
π 3
hb
bs' N cs'
From Eq. (59), for firing angles between 0 and # radians, the
average rectifier voltage is positive. Because the dc current
must be positive, power must be flowing out of the machine,
as hc
and so the converter acts as a rectifier. However, as the firing
cs axis
delay is increased past # radians, the average rectifier voltage
ds axis
is negative, and so power flows from the dc side of the con-
Figure 15. Hall-effect sensors mounted on a permanent-magnet syn-

verter to the ac side. Therefore, for firing angles greater than
chronous machine. In a practical machine, the sensors are mounded

# radians, the converter operates as an inverter. It may ap-
close to the permanent magnet on the end cap or buried in the stator
pear that the optimal phase delay is 2# radians, which maxi-
iron. The main purpose of this figure is to show the position of the
mizes power transfer as an inverter, and in practice the firing
sensors and the phase shift #hm .
delay angle is maximized for inverter operation. The possibil-
ity of commutation failure, however, limits the maximum
phase delay achievable in this type of converter (12).
control loop. This latter approach typically mitigates stability
issues at the cost of additional sensors and control complexity.
GENERAL CONTROL PHILOSOPHY
SENSOR REQUIREMENTS
Regardless of whether the machine is excited from a fully con-
trolled or semicontrolled inverter, the synchronous machine The
closed-loop control of synchronous motor drives requires
always operates at a speed corresponding to the frequency of a
rotor position sensor to # . For the 180# VSI switching strat-
r
the applied voltage and current providing a very easy means egy,
rotor position is determined by inexpensive Hall-effect
of speed control. In particular, by simply operating the in-
sensors mounted on the stator as shown in Fig. 15 where
verter at a frequency corresponding to the desired speed, the #hm
denotes the angle of mechanical shift of the sensors from
torque angle automatically adjusts itself so that the electro- the
reference position shown. These sensors put out a logic
magnetic torque is equal to the load torque, subject to the high
when under a South magnetic pole and a logic how when
conditions that (1) the load torque is greater than the mini-
under a North magnetic pole to produce the logic signals as a
mum and less than the maximum electromagnetic torque pro-
function of rotor position as shown in Fig. 16 where the elec-
duced in accordance with the torque versus torque angle char-
trical phase shift angle #h is related to the mechanical phase
acteristic, (2) the system is dynamically stable for the given
shift by the number of pole pairs or #h # (P/2)#hm . Note that
load torque, and (3) the rate at which the frequency is varied
and the way the load torque varies with speed are such that
transient stability is maintained during startup. Because the
torque angle automatically adjusts itself to the correct value
ha
to satisfy the load torque and because the speed must match
the applied frequency, this open-loop type of speed control is
attractive in that no rotor position or speed sensors are re-
quired. In practice, however, conditions (2) and (3) limit the
hb
use of this type of control. The alternative is closed-loop con-
trol in which rotor position #r is measured and #c is calculated
by
θc = θr + φv (63)
The added phase shift #v is calculated on the basis of a de-

0 π π 5π 7π 3π 11π 2π
6 2 6 6 2 6
sired torque angle (in fact, #v #
#; the difference in nomen-
θ +φ (rad)
clature is that #v is traditionally used by drive engineers,
r h
whereas # is used by power system engineers). In this ar-

Figure 16. Hall-effect logic signals. These signals are identical to the
rangement, the speed varies until the load torque is satisfied,
control signals required for 180# VSI operation and thus provide a
and the regulation of the speed requires an additional speed
convenient and inexpensive implementation of that control.
----------------------- Page 541-----------------------
the signals produced by the Hall-effect sensors are identical and

average output power is given by
to the signals of the 180# VSI control shown in Fig. 9 with
# # # . In most 180# VSI drives, the control signals gener-
P = T ω (67)
v h
out e rm
ated directly from the Hall-effect sensor signals provide a con-

venient and inexpensive implementation. In
Eq. (66) the power calculation is an approximation because
For high-performance switching strategies, such as sine- the
average of the product of two terms (the voltage and cur-
triangular modulation or hysteresis current control, in which
rent) is replaced by the product of the averages. This approxi-
the commanded voltage or current is a sinusoidal function of
mation works well in practice (18). Machine efficiency is cal-
rotor position, the rotor position must be measured continu-
culated by
ously. This high-resolution measurement usually requires an
optical encoder (13) or a resolver (14) which adds expense to
Pout
eff = (68)
the drive. Although the encoder or resolver can be eliminated
P
in
by using a full observer to determine rotor position, these
methods usually require an elaborate scheme for start-up If
the power electronic converter losses are neglected, another
(15,16), and cannot guarantee a bound on the initial error.
useful relationship applicable to any drive system is that the
Furthermore, these methods generally require knowledge of
average dc current is found from the input power by
the motor parameters. Recently a hybrid observer was devel-
oped which senses rotor position continuously with inexpen-
P
sive Hall-effect sensors (17). This observer does not require
idc = in (69)
v
dc
knowledge of motor parameters or an elaborate start-up
scheme. Furthermore, the error in estimating the rotor posi-
For
this discussion, it is assumed that the dc voltage is con-
tion is bounded by a limit which is independent of operating
stant.
conditions. Besides rotor position, some synchronous-machine
With these basic relationships now in place, the perfor-

drive controls require motor current or voltage measurements
mance of a variety of synchronous motor drives is investi-

to operate. For example, hysteresis current control requires
gated in detail. First, voltage and current source operation of

measuring the stator currents. Closed-loop, Hall-effect sen-
permanent-magnet, synchronous machine drives are ad-

sors, which produce electrically isolated current measure-
dressed. These drives are used widely in industry and are re-
ment, are often used for this purpose.
placing standard brush-type dc motors in many applications.
Because their performance is similar to dc motors, they are

GENERAL RELATIONSHIPS FOR ANALYZING DRIVE SYSTEMS
often times called brushless dc motors. The next configura-
tions considered are voltage and current source operation of

Before setting forth a quantitative analysis of synchronous-
reluctance motor drives. Because synchronous-reluctance mo-
machine drives, it is appropriate to set forth some concepts tors
are usually driven from the 60 Hz line in open-loop mode
and relationships which hold regardless of the type of drive to
obtain constant speed, this configuration is not as widely
considered. The first concept is that of average-value model- used
and so is only discussed briefly. Finally, wound-rotor
ing. In a synchronous machine, in which a balanced three-
synchronous motor drive using semicontrolled bridge convert-
phase set of voltages is applied, the resulting q- and d-axis ers
are considered.
voltages and currents are constant. However, in a drive sys-
tem, switching of the semiconductors results in q- and d-axis
VOLTAGE-SOURCE, INVERTER-FED, PERMANENT-MAGNETIC,

voltages and currents which are nonconstant. For many pur-
SYNCHRONOUS MACHINE DRIVES

poses, it is sufficient to treat the machine as if the q- and d-
axis quantities were constants equal to their average values.
One
of the most common synchronous motor drive configura-
In terms of ac variables, this is equivalent to representing
tions is based on a three-phase, fully controlled, inverter feed-

only the fundamental component. In the remainder of this
ing
a permanent-magnet synchronous machine. This arrange-
article, average-value quantities are designated with an over-
ment
is always operated in a closed-loop with respect to rotor
bar. Another useful concept is that q- and d-axis quantities
position. The operation of such a drive is illustrated in Figure

are readily related to the phasor representation of the ac vari-
17,
which depicts the performance of a 560 W brushless dc
ables. In particular, the rms amplitude f s and phase #s (rela-
motor drive whose parameters are listed in Table 1. In this

tive to rotor position) of a phasor representation of the funda-
system, Hall-effect sensors generate gate signals to the in-

mental component of stator quantity is related to the q- and
verter which is operated in the six-step mode. The sensors are

d-axis components by
arranged so that the phase advance is zero. In the study, the

1 # 2 2 dc
voltage is 267 V and the mechanical speed is 314.2 rad/s.
f s = √2 f qs + f ds (64)
Variables depicted include the motor a-phase voltage vas , a-
phase current i , and torque T predicted by a computer simu-
as e
φs = angle(f qs −j f ds ) (65)
lation. As can be seen, the a-phase voltage waveform is a
stepped approximation to a cosine waveform. The resulting

where f is a stator voltage, current, or flux linkage. The power
machine currents contain low-frequency harmonics which
into a synchronous machine is expressed in terms of q- and
produce low-frequency torque ripple. This is eliminated by
d-variables as
sine-triangle modulation (note that continuous rotor position
sensing is required to do this, however) as depicted in Fig. 18.

3 r r
P = (vr i + vr i ) (66) The
duty cycle is 0.9 and the dc voltage is adjusted to 391 V,
in 2 qs qs ds ds so
that the fundamental component of the applied voltage is
----------------------- Page 542-----------------------
300
300
4 ms
4 ms
)
V
) (
0
V s
a
( 0 v
s
a
v
–
300
–300
15
15
)
) A
A 0 (
0
( s
s a
a i
i
–15 –
15
3
3
)
) m
m ⋅
N⋅ 1.5 N
1.5
(
( e
Te T
0
0
Figure 17. Operation of a 180# voltage-source, inverter-fed, perma- Figure

18. Operation of a sine-triangle modulated, voltage-source,
nent-magnet, synchronous machine. The voltage is a stepped approxi-
inverted-fed permanent-magnet, synchronous machine. The voltage
mation to a cosine waveform which results in undesirable low-fre-
modulation improves the current and torque waveforms compared to
quency harmonics in the current and torque. the six-
step operation shown in Fig. 17.
the same as in Fig. 17. In this case, the voltage, current, and The
phase advance #v is physically determined in software
torque waveforms contain high-frequency instead of low-fre-
(and
considered an input) for continuous type of rotor position
quency harmonics. This mode of operation is advantageous in
sensing.
If Hall-effect devices are utilized, the phase advance
terms of both current and torque ripple. Furthermore, the
is
dictated by the physical position of the sensors. If the ma-
duty cycle is readily controlled to control the speed. Disadvan-
chine
windings are connected in a delta configuration, Eqs.
tages of this control are the expense of the rotor position sen- (70–71)
must be modified by multiplying the voltages by #3
sor and increased semiconductor switching losses.
and
adding #/6 to the cosine and sine terms. The fundamental
Analyzing this type of drive system begins with determin-
component of the rms voltage is computed from Eq. (64) and

ing the average-value of the applied q- and d-axis voltages.
Eqs.
(70–71) by
Transforming the fundamental component of the applied volt-
age to the rotor reference frame yields
1
v = mv (73)
v − mv cos(φ ) (70)
s √2 dc
qs dc v
and The
machine q- and d-axis currents are determined by aver-
aging
the machine voltage Eqs. (28–29) and solving for cur-
v − mv sin(φ ) (71)
ds dc v rents.
This yields
where m is a function of the modulation strategy used. In

1
particular
i = [r v − ω L v − r ω λ ] (74)
qs 2 2 s qs r d ds s r m
r + ω L L
s r q d
2 ◦
 for 180 operation



π and

m = 2 d for duty-cycle modulation (72)
1

π
2

i = [ω L v + r v − ω L λ ] (75)
r q qs s q m

ds 2 2 ds r

r + ω L L

1
s r q d
 d for sine-triangular modulation
2
The
average torque is calculated by averaging Eq. (31) which
yields
Table 1. Permanent-Magnet Synchronous Machine
Parameters
3 P
T = [λ i + (L − L )i i ] (76)
e 2 2 m qs d q qs ds
rs # 2.985 # Lls # 1.84 mH
Lmq # 9.51 mH Lmd # 9.51 mH
In
obtaining Eq. (76), the average of the product of the q- and
P # 4 # # 0.156 V · s
m d-axis
currents is approximated. Given the average machine
----------------------- Page 543-----------------------
25
25
φv = 0
20
20
m = 0.47
φ = π
m = 0.35
v
) 15
) 15 6
m
m
⋅
⋅
N
N
(
(
e m = 0.24
e
T 10
T 10 π
φv =
3
m = 0.12
5
5
φv = π
0
0
0 100 200 300 400 500 600
0 100 200200 300 400 500 600
ω (rad/s)
ω (rad/s)
rm
rm
Figure 19. Torque vs speed characteristics of a voltage-source, in-

verter-fed, permanent-magnet, synchronous machine drive as the
verter-fed, permanent-magnet synchronous machine drive as the
modulation index is varied. The applied voltage and, thus, the re-
phase advance is varied. The performance characteristics vary widely
sulting torque increase with the modulation index.
with the phase advance. A phase advance of zero provides the maxi-
mum stall torque whereas a phase advance of #/2 provides the maxi-
mum torque at high speeds.

voltages and currents, the remaining quantities of interest
are readily determined by Eqs. (64–69).
Figures 19 and 20 illustrate average torque and the rms
phase advance of #/2 provides the most torque at high speeds
value of the fundamental component of the stator current ver-
and zero torque at stall. Adjustment of the phase advance
sus rotor speed for several different values of the modulation
increases torque output at a given speed (1) or maximizes ef-
index for the drive whose parameters are listed in Table 1. In
ficiency. In a system in which rotor position in sensed contin-
this study, vdc is 300 V, and #v is zero. It can be seen that
uously, the phase advance may be varied with the operating
both torque and current decrease with speed, except that,
point. When Hall-effect sensors are used however, the phase
after the torque passes through zero, the current begins to in-
advance is fixed once it is selected. Optimizing for one op-
crease.
erating point generally degrades the performance at other op-
As it turns out, the phase advance has a pronounced effect
erating points.
on the torque-versus-speed characteristic of the synchronous
motor drive. This is illustrated in Fig. 21 which depicts the
torque-versus-speed of the drive with a dc voltage vdc of 300
CURRENT-REGULATED, INVERTER-FED, PERMANENT-
V and a modulation index m of 0.47 for several values of the
MAGNET, SYNCHRONOUS MACHINE DRIVES
phase advance. Setting the phase advance angle to zero pro-

vides the maximum stall torque (the torque at zero speed). A
Another common configuration of synchronous motor drives
is the current-regulated, permanent-magnet, synchronous
machine drive which consists of a fully controlled, bridge con-

25
verter connected to a permanent-magnet, synchronous ma-
chine with continuous rotor position feedback. In this case,
the inverter is hysteresis modulated on the basis of a- b- and

20 m = 0.47
c-phase current commands, which are calculated in accor-
dance with Fig. 22. (Note that current control is also possible
with sine-triangle modulation using an inner current-control

15 m = 0.35
loop.) In the figure, based on torque command T* and speed
)
A
#r , the current command synthesizer determines the q- and
(
s
i m = 0.24
d-axis current command (i* and i*) so that the desired torque
qs ds
10
is obtained. In the case of a nonsalient machine, a simple cur-
rent command synthesizer is given by

m = 0.12
5
T∗
i∗ = e (77)
qs 3 P
λm
2 2
0
0 100 200 300 400 500 600
∗
ids = 0 (78)
ω (rad/s)
rm
Figure 20. Rms current vs speed characteristics of a voltage-source,

In a nonsalient machine, this type of control maximizes effi-
inverter-fed, permanent-magnet, synchronous machine drive as the
ciency because the d-axis current does not contribute to aver-
modulation index is varied.
age torque. Occasionally, negative d-axis current is used to
----------------------- Page 544-----------------------
extend the speed range of the machine. In this case, the

3
amount of d-axis current injected is a function of speed,
torque, and dc voltage. Strategies for accomplishing this are

)
⋅
set forth in (19,20) for salient and nonsalient machines, re-
N 1.5
* e 8 ms
spectively. Once the q- and d-axis current command is estab-
T
lished, the abc variable current command i* is established
abcs
using the inverse of Park’s transformation given by Eq. (10).
0
One of the chief advantages of this drive system is that

6
torque is controlled very rapidly and precisely. Figure 23 il-
lustrates the performance of a current-regulated, permanent-
)
A
magnet, synchronous motor drive during a step change in
( 3
* q
torque command. The machine parameters are identical to
v
those used to produce Figs. 17 and 18, the dc bus voltage is
225 V, the hysteresis level is 0.6 A, and the current command
0
synthesizer is given by Eqs. (77–78). In this study, machine

5
is operating at a speed of 314.2 rad/s and a torque command
of 1 Nm. Then the torque command is stepped to 2 Nm. Vari-
)
A 0
ables depicted include the torque command T*, the q-axis cur-
(
e
s
a
rent command i*, the a-phase current i , and the electromag-
i
qs as
netic torque Te . The q-axis current command is directly

proportional to the torque command. Because of the high
–5
bandwidth of this type of control, the a-phase current and

3
torque rapidly change to track the new reference during the
step change in torque command.

)
⋅
To analyze current-regulated drives, for normal operating
N 1.5
e
conditions, it is assumed that the actual q- and d-axis cur-
T
rents are equal to the commanded currents:
0
iqs = i∗ (79)
qs
Figure 23. Operation of a hysteretic, current-regulated, inverter-fed,
permanent-magnet, synchronous machine. For this type of machine,

and
the q-axis current command is directly proportional to the torque
command. When the current command is stepped up, the phase cur-
i = i∗ (80)
rent increases. This increases the machine torque which closely
ds ds
tracks the commanded torque because of the high bandwidth of this

Then the average q- and d-axis stator voltages are expressed
type of control.
by the average of Eqs. (28–29) as
v = r i + ω L i + ω λ (81)
The machine torque is calculated from Eq. (76). Other quanti-
qs s qs r d ds r m
ties of interest are found from Eqs. (64–69).

and
Figure 24 illustrates the torque-versus-speed characteris-
tics for several values of commanded torque. As can be seen

v = r i − ω L i (82)
ds s ds r q qs
for this type of drive system, torque is independent of speed
in the low to mid speed range, as it must be if the actual
currents are equal to the commanded current. As the speed

Power +
increases, however, the actual currents no longer track the
vdc Inverter PMSM
commanded currents. The condition for which this loss
of
source
–
tracking is expected to occur is given by Eq. (56). When ana-

T1-T6
lyzing the region where current tracking is lost, a much more
iabcs
involved analysis is required. The interested reader is re-
Hysteretic
ferred to (18).
current control θ
r
i*
VOLTAGE-SOURCE INVERTER-FED, SYNCHRONOUS-
abcs
i*
T* qs
RELUCTANCE, MACHINE DRIVES
e Current
— 1
command Kr
i* s
ds
synthesizer
Although not as common as permanent-magnet, synchronous
ω
motor drives, synchronous-reluctance motor drives have ad-
r
vantages in terms of robustness (because there are no me-

p
chanically weak or temperature-sensitive permanent mag-

Figure 22. Current-regulated, inverter-fed, permanent-magnet mo-
nets). As with permanent-magnet motor drives, these devices
tor drive configuration. The commanded current is generated by the
are fed by voltage-source or current-regulated inverter modu-
current command synthesizer based on the commanded torque, ma-
lating strategies. For voltage-source, inverter-based modula-
chine speed, and machine parameters.
tion, the same relationships for analyzing the voltage-source,
----------------------- Page 545-----------------------
Table 2. Reluctance Machine Parameters
rs # 0.382 # Lls # 0.83 mH

T* = 3 N⋅m
Lmq # 13.5 mH Lmd # 39.27 mH
e
3
r# # 31.8 mH L # # 6.13 mH
kq 1 lkq1
r# # 0.923 mH L # # 3.4 mH
kq2 lkq2
r# # 40.47 mH r# # 4.73 mH
kd 1 lkd1
)
T* = 2 N⋅m
r# # 1.31 mH L # # 3.68 mH
m e
kd2 lkd2
⋅
N 2
Ns/Nfd # 0.02711 P # 4
(
e
T
T* = 1 N⋅m
e
1
alent to a delta angle of
#/4, produces maximum torque as
is expected considering Eq. (8) or the reluctance torque curve
of Fig. 3. At low speeds, however, the motor impedance is

0
mostly resistive (nonreactive) and Eq. (8) is no longer valid.
0 100 200 300 400 500 600
Setting the phase advance angle to

#/4 results in the maxi-
ω (rad/s)
rm
mum torque at stall. Values of phase advance between
#/4
Figure 24. Torque vs speed characteristics of a hysteretic, current-
and #/4 produce the intermediate performance curves shown
regulated, permanent-magnet, synchronous machine drive as com-
in Fig. 25.
manded torque is varied. Commanded torque is not achieved at high
speeds because of the increased machine back-emf and the limited dc
supply voltage.
CURRENT-REGULATED, INVERTER-FED, SYNCHRONOUS-
RELUCTANCE, MACHINE DRIVES
Direct control of the machine currents offers the same advan-

inverter-based, permanent-magnetic, synchronous motor
drives are used for the voltage-source, inverter-fed, reluctance
tage in the case of the synchronous-reluctance drive as it does
motor drives except that #m # 0. Specifically, Eqs. (70–72) are
in the case of permanent-magnet, synchronous machine
used to calculate the motor average voltages, the average of
drives, that is, rapid and precise control of the electromag-
Eqs. (74–75) are used to find the motor average currents, and
netic torque. As with the permanent-magnet motor drive, this
Eq. (76) is used to find the average torque. Remaining quanti-
is achieved by current regulation (such as hysteresis current
ties are readily determined from Eqs. (64–69).
control) or by having an inner control loop wherein voltage-
Figure 25 depicts the torque versus speed curves, as pre-
source, inverter-based modulation is used to obtain voltages
dicted by the average-value model, of a voltage-source, in-
required to drive the actual currents to the desired currents.
verter-driven synchronous-reluctance motor for several val-
The analysis of this configuration is similar to that of the CRI
ues of the phase advance angle. The motor parameters are
permanent magnet motor drive with #m # 0, namely, Eqs.
given in Table 2. The dc voltage and modulation index for this
(77–78) are used to calculate the average q- and d-axis cur-
study were 400 V and 0.48 respectively. As can be seen, the
rent, the machine voltages and torque are found from Eqs.
motor performance varies widely with the phase advance
(81–82) and Eq. (76), respectively. Other quantities of inter-
angle. At high speeds, a phase advance of #/4, which is equiv-
est are readily obtained from Eqs. (64–69). As in the case of
permanent-magnet synchronous machine drives with current-
regulated inverter control, torque is easily and rapidly con-

8000
trolled, up to the point wherein the voltage capabilities of the
inverter are insufficient to track the commanded currents.

φv = – π
4
6000
)
m
Voltage
⋅
N
regulator/
(
Te φv = – π
exciter
6
4000
vfd
φv = 0
π
Semi
2000 φv =
Current
4
controlled SM
source
π
bridge
φv =
6
T1-T6
0
0 10 20 30 40 50
θ
Gate r
ω (rad/s)
(Firing delay) β
rm
sequencer

verter-fed, synchronous, reluctance machine drive as the phase ad-
Figure 26. Synchronous motor drive using a semicontrolled three-
vance is varied. The performance varies widely with the phase ad-
phase bridge. The synchronous machine is supplied from a constant
vance.
current source.
----------------------- Page 546-----------------------
150
levels. One possible configuration is illustrated in Figure 26
where a current source supplies a semi-controlled bridge (Fig.

13)
which is, in turn, connected to a synchronous machine.
) This
current source is synthesized by a voltage source coupled
V
( 75
d with
an inductor and a closed-loop, current-control system. A
f
v
6.6 ms
voltage regulator/exciter monitors the terminal voltages of
the
machine and adjusts the field voltage of the synchronous
0
machine so that the flux is at an appropriate level. To this
2 end,
a constant-volts-per-hertz scheme is employed. The firing
of
the inverter thyristors is based on rotor position. In this
case, the gate sequencer operates as in Fig. 14 except that,

)
instead of being fired relative to the back-emf, thyristors are
A 1
( fired
from the rotor position. In particular, valve three is fired
d
f
i when
π
0
θr = + β (83)
3
0
In
Eq. (83), # denotes the firing delay instead of # as in the
previous discussion of semicontrolled converters because # is
)
defined relative to the back-emf of the machine, not the rotor
V –200
(
r
position. Firing relative to the rotor position is robust in that
V
the
firing sequence is unaffected by the harmonics in the
terminal voltage induced by the rectifier. An alternate ap-

–400
proach is to fire the converter based on the filtered terminal
250
voltages.
Figure 27 depicts the performance of a 3.7 kW LCC syn-
chronous machine drive whose parameters are listed in Table

) 3.
In this study, the dc link current is regulated at 20 A, the
V
( 0
terminal voltage is regulated at v # 133 V, the electrical ro-
s
s
a
v tor
speed is 377 rad/s, and the firing delay relative to the
rotor position # is 2.18 radians. Variables depicted include

–250 the
field voltage vfd , the field current ifd , the rectifier voltage
vr ,
the a-phase line-to-neutral voltage vas , the a-phase line
25
current i , and the electromagnetic torque T . The small vari-
as e
ation in the field voltage is the result of the harmonics in the
line-to-line voltage, and the harmonics in the field current

)
A 0
largely result from conservation of flux in the machine’s d-
(
s
a
i
axis. The notch in the rectifier and line-to-neutral voltages
caused by commutation is clearly evident. As can be seen, the
current waveforms are similar to the idealized waveform de-

–25
picted in Fig. 14. The large amount of distortion in the cur-

40 rent
waveforms leads to considerable torque ripple, as de-
picted in the final trace.

)
Figure 28 depicts the average electromagnetic torque as #
m⋅ 20 is
varied from 1.48 to 2.18 radians, as calculated from a com-
N
( puter
simulation of a variety of operating points (the analyti-
e
T
cal
solution is quite involved; the reader is referred to
(12,21,22,23). The lower limit of # is established by the ap-

0
proximate point where the average torque becomes positive,
and
the upper limit by commutation failure. Based on the pre-
Figure 27. Operation of a wound-rotor, synchronous machine from
a semicontrolled three-phase bridge. Positive torque and negative dc
voltage indicate motor operation. Commutation is evident in the
notches in the a-phase voltage.
Table 3. Wound-Rotor Synchronous Machine Parameters
rs # 0.382 # Lls # 0.83 mH
Lmq # 13.5 mH Lmd # 39.27 mH
r# # 31.8 mH L # # 6.13 mH
SEMICONTROLLED, INVERTER-FED, WOUND-ROTOR,
kq 1 lkq1
r# # 0.923 mH L # # 3.4 mH
SYNCHRONOUS MACHINE DRIVES
kq2 lkq2
r# # 40.47 mH L # # 4.73 mH
kd 1 lkd1
r# # 1.31 mH L # # 3.68 mH
Because of the high voltage and current capabilities of thyris-
kd2 lkd2
r# # 0.122 mH L # # 2.54 mH
tors, synchronous machine drives are often based on the semi-
fd lkd2
Ns/Nfd # 0.02711 P # 4
controlled rather than fully controlled bridge at high power
----------------------- Page 547-----------------------
30
8. R. R. Nucera, S. D. Sudhoff, and P. C. Krause, Computation of
steady state performance of an electronically commutated motor,
IEEE Trans. Industrial Appl., 25: 1110–1117, 1989.
9. S. D. Sudhoff and P. C. Krause, Operating modes of the brushless
dc motor with a 120# inverter, IEEE Trans. Energy Conversion,

20
5: 558–564, 1990.
)
10. S. D. Sudhoff and P. C. Krause, Average-value model of the
m
⋅
brushless dc 120# inverter system, IEEE Trans. Energy Conver-
N
(
e
sion, 5, pp. 553–557, 1990.
T
11. H. W. Van Der Broeck, H. C. Skudelny, and G. V. Stanke, Analy-

10
sis and realization of a pulsewidth modulator based on voltage
space vectors, IEEE Trans. Industry Appl., 24: 142–150, 1988.
12. S. D. Sudhoff and O. Wasynczuk, Analysis and average-value
modeling of
line-commutated converter-synchronous machine
0
systems, IEEE Trans. Energy Conversion, 8: 92–98, 1993.
1.48 1.66 1.83 2.01 2.18
13. DC Motors, Speed Controls, Servo Systems: An Engineering
β (rad)
Handbook, Eden Prairie, MN, Robbins & Myers/Electro-Craft,
1989.
Figure 28. Torque vs # of a semicontrolled, converter-fed, wound-
14. D. C. Hanselman, Resolver signal requirements for high accuracy

rotor, synchronous machine drive. For maximum efficiency, it is de-
resolver-to-digital conversion, IEEE Trans. Industrial Electron.,

sirable to operate with the largest value of # possible. The upper limit
37: 556–561, 1990.

of # is determined by commutation failure.
15. T. Furuhashi, S. Sangwongwanich, and S. Okuma, A position-
and-velocity sensorless control for brushless dc motors using an
adaptive sliding mode observer, IEEE Trans. Industrial Electron.,

vious discussion of firing delay, the reader may have sus-
39: 89–95, 1992.
pected that the range over which the firing angle should be
16. N. Matsui and T. Takeshita, A novel starting method of sen-
varied is approximately # to 2# radians. The difference is
sorless salient-pole brushless motor, Proceedings of the 1994 IAS
Conference, 1: 386–392, 1984.

caused by the fact that the firing angle controlled herein is
17. K. A. Corzine and S. D. Sudhoff, A hybrid observer for high per-

the firing angle relative to the rotor position rather than the
formance brushless dc drives, IEEE Trans. Energy Conversion,
firing angle relative to the back-emf of the machine. As can
11: 318–323, 1996.
be seen, it is desirable to operate with as large a firing delay
18. K. A. Corzine, S. D. Sudhoff, and H. J. Hegner, Analysis of a

as possible, while recognizing that, beyond a certain value,
current-regulated brushless dc drive, IEEE Trans. Energy Conver-
commutation failure occurs.
sion, 10: 438–445, 1995.
Besides commutation failure, another difficulty with this
19. T. M. Jahns, Flux-weakening regime operation of an interior per-
type of drive is start-up. In particular, at zero speed the ma-
manent-magnet synchronous motor drive, IEEE Trans. Industry
chine does not possess a substantial back-emf necessary for
Appl., 23: 997–1004, 1987.
the thyristor valves to commutate. Conditions and strategies
20. S. D. Sudhoff, K. A. Corzine, and H. J. Hegner, A flux weakening
for successful start-up are set forth in (24,25,26).
strategy for current-regulated surface-mounted permanent-mag-
net machine drives, IEEE Trans. Energy Conversion, 10: 446–
454, 1995.
BIBLIOGRAPHY
21. S. D. Sudhoff et al., Transient and dynamic average-value model-
ing of synchronous machine fed load commutated converters,
IEEE Trans. Energy Conversion, 11: 508–514, 1996.

1. P. C. Krause, O. Wasynczuk, and S. D. Sudhoff, Analysis of Elec-
tric Machinery, Piscataway, NJ: IEEE Press, 1995.
22. S. D. Sudhoff, Analysis and average-value modeling of dual line-
commutated converter-6-phase synchronous machine systems,

2. R. H. Park, Two reaction theory of synchronous machines gener-

alized method of analysis-part 1, AIEE Trans., 48: 716–727, 1929.
23. S. D. Sudhoff, Waveform reconstruction from the average-value

3. C. H. Thomas, Discussion of analogue computer representation
model of line-commutated converter-synchronous machine sys-
of synchronous generators in voltage-regulation studies, AIEE
tems, IEEE Trans. Energy Conversion, 8: 404–410, 1993.
Trans., 75: 1182–1184, 1956.
24. T. Peterson and K. Frank, Starting of large synchronous motor

4. R. G. Harley, D. J. N. Limebeer, and E. Chirricozzi, Comparative
using static frequency converter, IEEE Trans. Power Apparatus
study of saturation methods in synchronous machine models, IEE
Syst., 91: 172–179, 1992.
Proc., 127, Part B (1): 1–7, 1980.
25. R. L. Steigerwald and T. A. Lipo, Analysis of a novel forced-com-
5. K. A. Corzine, B. T. Kuhn, and S. D. Sudhoff, An improved

mutation starting scheme for a load-commutated synchronous
method for incorporating magnetic saturation in the q-d synchro-
motor drive, IEEE Trans. Industry Appl., 15: 14–24, 1979.
nous machine model, Accepted for publication in IEEE Trans. En-
26. S. D. Sudhoff, E. L. Zivi, and T. D. Collins, Start up performance
ergy Conversion, 1997.
of load-commutated inverter fed synchronous machine drives,
6. T. W. Nehl and N. A. Demerdash, Impact of winding inductances
and other parameters on the design and performance of brushless

dc motors, IEEE Trans. Power Apparatus Syst., 104: 2206–
KEITH A. CORZINE
2213, 1985.
University of Wisconsin, Milwaukee
7. P. Pillay, Modeling of permanent magnetic motor drives, IEEE

SCOTT D. SUDHOFF
Trans. Industrial Electron., 35: 537–541, 1988.
Purdue University
----------------------- Page 548-----------------------
SYSTEM IDENTIFICATION 263
SYNCHRONOUS SYSTEMS CLOCK DISTRIBU-

TION. See CLOCK DISTRIBUTION IN SYNCHRONOUS
SYSTEMS.
SYNCHROTRONS. See SUPERCONDUCTING CYCLOTRONS
AND COMPACT SYNCHROTRON LIGHT SOURCES .
SYNTAX AND SEMANTICS. See LOGIC PROGRAMMING
AND LANGUAGES .
SYNTHESIS OF NONLINEAR CIRCUITS. See NONLIN-
EAR CIRCUIT SYNTHESIS USING INTEGRATED CIRCUITS.
SYNTHESIS (OR DESIGN) OF ANALOG PASSIVE
FILTERS. See FILTER SYNTHESIS.
SYNTHESIS (OR DESIGN) OF LOSSLESS TWO-
PORTS. See FILTER SYNTHESIS.
SYNTHESIS, HIGH LEVEL. See HIGH LEVEL SYNTHESIS.
SYNTHETIC APERTURE RADAR. See REMOTE SENSING
BY RADAR.
----------------------- Page 549-----------------------
} { { } }
HOME
●
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Thermal Energy Storage
Standard Article

Barrett A. Flake1
1Air Force Institute of Technology, Ohio

reserved.
Search
DOI: 10.1002/047134608X.W3042
❍ Acronym Finder
Abstract
Fundamentals
Applications
Design
Operation
Conclusion
%20Conversion/W3042.htm17.06.2008 16:47:40
----------------------- Page 550-----------------------
THERMAL ENERGY STORAGE 19
THERMAL
ENERGY STORAGE
Thermal
storage is a method of storing heat for future use.
Most often,
thermal storage is used when future heat sources
or sinks
are either not available at a sufficient rate or will be
more
expensive. A common application of thermal storage is
the
domestic hot water heater. Most household water heaters
use an
electric resistance element or a natural gas burner
as a heat
source. A resistance element may supply heat at
a rate of 6
kW, whereas a hot shower requires 20 kW to heat
6 L/min
of water from 16#C to 40#C. Some amount of hot
water must
be stored in the tank because the element cannot
heat the
water at a rate sufficient for a shower or most other
uses.
Heat is
stored in the hot water tank to meet the demand
for future
use. The number or duration of showers possible
depends
upon the quantity of water stored and the tempera-

----------------------- Page 551-----------------------
20 THERMAL ENERGY STORAGE
ture at which it is stored. To meet the desired water tempera-

Phase change materials (PCMs) used in thermal storage
ture of 40#C with 60#C stored hot water, 3.4 L/min of hot wa-
applications are almost exclusively from solid to liquid or vice
ter can be mixed 2.6 L/min with 16#C cold water. If the water
versa. Changing a liquid to the gas phase is accompanied by
is stored at 70#C, only 2.8 L/min of hot water is required.
order of magnitude changes in volume (or pressure if the vol-
As in the hot water tank example, the fundamental proper- ume
is held fixed) and storage facilities to handle such
ties of any thermal storage system are the amount of energy
changes would be highly impractical to construct. However,
stored and the temperature at which it is stored. If the tem-
compressed air energy storage has been used at a few electri-
perature of the storage medium is below ambient conditions, cal
plants where underground caverns are used to contain the
it is a lack of heat (a heat sink) that is stored. Storage of a
compressed air. The quantity of heat necessary to change
heat sink rather than a heat source can alternatively be
phase from solid to liquid (or from liquid to solid) is commonly
called cool or cold storage. In either case, the potential to
called the latent heat of fusion. For most substances, their
transfer heat to or from the sink or source depends upon the
latent heat of fusion is equivalent to the quantity of heat re-
temperature and mass of the storage medium.
quired for a large change in temperature of the substance in
Heat is stored by raising the storage medium temperature a
single phase. Water’s latent heat of fusion at atmospheric
or changing its phase (e.g., from solid to liquid) or both. Cer-
pressure (335 kJ/kg) is approximately 80 times greater than
tain storage media may be better suited for a particular appli- the
heat associated with changing liquid water by 1#C (4.18
cation depending upon their thermodynamic properties. The
kJ/kg).
specific heat, c, of a substance is a property relating the quan-
Chilled water storage systems (a heat sink or alternatively
tity of energy stored or removed due to an increase or de-
called a cooling source) have seen extensive use in building
crease in temperature (without a change in phase.) Specific
cooling systems. Usually, the cooling storage is used because
heat is a temperature and pressure dependent property. How- the
electricity used to operate the water chilling equipment
ever, most practical thermal storage media are used in a solid
(i.e., chillers) can be purchased at a cheaper rate during cer-
or liquid state where the variance of specific heat with pres-
tain hours of the day. Chilled water storage is also employed
sure is negligible. Also, the change in specific heat over mod-
when the building’s cooling requirements sometime exceed
erate temperature ranges can be neglected in many applica- the
capacity of the installed chillers (1).
tions.
Materials with high specific heats are advantageous for
use as thermal storage media. In many cases the volume oc-
FUNDAMENTALS
cupied by the storage medium is of more concern than the
weight. In these cases the specific heat-density product pro- To
illustrate the application of thermal storage, consider the
vides a better indicator in comparing alternative media
transport of energy, whether thermal, mechanical or electri-
choices. For example, aluminum has a specific heat nearly
cal, from a source to a building cooling system as shown in
twice that of steel (0.9 kJ/kg-K vs. 0.5 kJ/kg-K), but because
Fig. 1. There may be a number of energy conversions (e.g.,
of their different densities, steel has a 60% larger volumetric
from thermal to mechanical to electrical) as the energy moves
heat capacity (2,454 kJ/m3 3
from source to the end use. The original source could be from
-K vs. 3,935 kJ/m -K). Water, at
standard conditions, has a volumetric heat capacity of 4,175 the
combustion of coal or natural gas, or from heat liberated
kJ/m3-K. in
the fission of radioactive materials. It should be noted that
Thermal
storage in
building mass Building
Internal
heat gains
Thermal
storage
system
Rotating
Rotating
Vapor
Coal Steam shaft
Electricity Electrical Electricity Electric shaft compressior
Boiler Turbine Generator
distribution
motor refrigeration
network
system
Heat rejected
to the
environment
Figure 1. Without any intervening energy storage, the

required energy must be simultaneously
supplied through a chain of energy conversion processes.
----------------------- Page 552-----------------------
the energy flows in the diagram are of ‘‘utilizable’’ energy

which decreases in quantity at each conversion and transport
process. The energy transferred from source to end use is con-
served and, although not shown, significant quantities are
transported to the ambient environment in each conversion
and at the end process.
r
e d
l a
Without any intervening energy storage along the path
i o
h
shown in Fig. 1, the rate of energy transferred from the
C Storage L
source and the rate of all energy conversions is directly linked

to the rate of energy consumed by the refrigeration cycle. As
the cooling load met by the refrigeration cycle increases or
decreases, the source and all the conversions must respond
immediately by increasing or decreasing their rate of energy
transfer. (In any real energy conversion and transport sys-
tem, e.g., an electrical generation and distribution network, Figure
2. Schematic of a simple chilled water storage system. De-
pending
on the flow arrangement, the chiller can charge the storage
there is some energy storage; however, the amount is very
tank
or (possible simultaneously) serve the load. Alternatively, the
small in comparison with the quantity of energy transferred.)
load
can be served by the storage tank by itself or in combination
With energy storage systems, the link of simultaneous en- with
the chiller.
ergy generation, transfer and consumption can be decoupled.
There are great economic advantages to decoupling the link.
Without storage, all generation, conversion, and distribution
equipment and facilities must be of sufficient capacity to meet
the largest expected load (e.g., peak load). With storage some-
The
schematic in Fig. 2 demonstrates the use of a thermal
where between the source and the end use, not all equipment
energy
storage device between an energy source (or sink, as
and facilities must be sized to meet the peak. Smaller equip-
in the
case of the chiller) and a heat transfer requirement or
ment can be used with energy coming from storage during
load.
The load may be periodic, as in a building’s cooling load,
peak periods and energy stored during periods of low demand.
peaking
during the day and reaching a minimum during the
Also, the cost of electrical energy generally increases with the
evening. An example cooling load profile over 2 days given by

rate (demand) at which it is provided. With a storage system
q(t) is
shown in Fig. 3.
energy can be stored over periods of lower cost and then be
Without the storage tank, to meet the required load the

pulled from storage during periods of higher cost.
chiller
must operate exactly in synch with the load trajectory.
Different types of storage systems may be employed as the
The
opportunities for energy cost savings are limited to either
energy is transported from the source to the end use. Thermal
reducing the load or increasing the operating efficiency of the

energy storage is employed in transfers to or from a heat
chilling equipment. However, with the storage tank, the

source or sink. Other energy storage systems include electri-
chiller
operation can be decoupled from the load. With flow
cal and chemical energy storage (e.g., batteries and fuel cells)
controls among the load, chiller, and storage tank, various

and mechanical energy storage (e.g., flywheels.)
modes
of operation are possible. The full capacity of the
The electrical energy transfers shown in Fig. 1 are used to
chiller
could be used solely to charge the storage tank or could
operate a building’s cooling system (i.e., a chiller.) The cooling
simultaneously serve all or part of the load. Likewise, the

system must remove heat from the building at a rate suffi-
storage
tank could provide all of the load or some portion of
cient to keep the building’s internal environment within a de-
the
load with the chiller serving the remainder.
fined range of comfortable temperature and humidity. Some
of the heat sources that contribute to the cooling load include
heat conducted through exterior walls, heat from lighting in-
side the building, heat from building occupants as well as
heat from personal computers and monitors. Also, fresh air
used to ventilate the building must be cooled and dehumid-
Chiller capacity Cooling load
ified.
###
A building’s exterior and interior mass constitute a ther-
mal storage system. The rate of heat added or removed from
From
the building mass depends primarily upon the weather (i.e.,
storage
To To From
outdoor air temperature, incident solar radiation) and the q (t)
storage storage storage
building’s internal air temperature. The interior temperature
can vary over a relatively small range of comfortable interior
#
##yy##yy##
temperatures during occupied hours and over a larger range

during unoccupied hours. Thus the building mass can be em-
ployed as an active thermal storage system by controlling the
interior temperature. Braun (2) discusses the control of build-
6a.m 10 2p.m. 6 10 2a.m. 6 10 2p.m. 6 10 2a.m. 6

ing mass thermal storage as a means to reduce energy costs
Time
and peak electrical demand. Morris et al. (3) present results
from experimental and simulated optimal control of building Figure
3. An example building cooling load profile over a 48-h
##yy##yy##
thermal energy storage. period.
----------------------- Page 553-----------------------
Usually, because almost all buildings will experience peak colder

chilled water is always drawn or added to the system
late afternoon loads in the summer, electrical utilities also at one
end of the series and the warmer water added to or
experience peak demands that heavily task their generating drawn
from the other end.
and distribution resources. Accordingly, utilities have a time
of day rate structure with relatively high costs during high-
Ice
Storage
load hours and lower costs during low-load hours. With ther-
mal storage, the chiller can run either solely or more often As with
chilled water storage, ice storage uses the heat trans-
during the lower rate periods charging the storage tank to fer fluid
as the storage medium. However, with ice storage,
meet future loads that occur during the higher rate period. the
relatively high latent heat of fusion of water (335 kJ/k#C
In addition to time of day rates, utilities often assess a at 0#C)
is utilized. Ice storage systems differ in the way the
charge for the customer’s largest electrical demand during ice is
formed and stored (5). In external ice-on-coil systems,
some billing period (e.g., during one month.) Using thermal ice
forms on coils submerged in a water storage tank. Al-
storage can also reduce electrical demand peaks. Usually, the though
the source of cooling in most ice storage systems is a
storage is charged by running the chiller overnight when
refrigeration cycle, a secondary fluid rather than the refriger-
other major electrical systems (e.g., lighting, personal com- ant is
circulated inside the coils. Typically an ethylene glycol
puters, etc.) are off, resulting in a reduction in the total build- and water
mixture is used as the secondary fluid. In external
ing electrical demand. Also, the longer the time available to ice-on-
coil systems, ice is built up on the coils, growing as
charge the storage tank, the lower the required chiller capac- thick as
7 cm. The water in the tank is aerated or otherwise
ity (rate of charging) and therefore the lower the chiller agitated
so that the ice forms on the coils uniformly. Exces-
power demand. sively
thick ice formation creates ice bridges, restricting wa-
To charge the storage tank adequately for future load re- ter
circulation around the coils and impeding heat transfer.
quirements, a predicted load profile must be assumed. If it is When
building ice externally on the coils, a thickness limit
desired that the storage tank serve all of the load in Fig. 3 is
imposed, as the increasing thickness of the ice tends to in-
from 6 a.m. to 10 p.m. on the second day, the storage tank sulate
the coil. A colder coolant temperature inside the coils
would need to be charged with a total capacity equal to or is
required to continue forming ice as the thickness increases.
greater than the integrated load over the 6 a.m. to 10 p.m. The lower
coolant temperature requires the refrigeration sys-
time period. tem
evaporator to operate at a lower temperature and conse-
quently
reduces cycle efficiency.
In
discharging the ice storage, warmer water is circulated
APPLICATIONS
about the
coils, melting the ice from the outside of the forma-
tion
(external melt).
Chilled Water Storage In
internal melt ice-on-coil systems, the secondary heat
In many large cooling systems, heat is transported from cool- transfer

fluid is used in charging and discharging modes. In
ing loads by chilled water. The chilled water is usually in a the
charging mode, the secondary fluid circulates between the
closed loop system where a refrigeration cycle provided the coils in
the tanks and a heat exchanger coupled with the re-
cooling source. A simple storage method is to use the chilled
frigeration system evaporator. In discharging the store, the
water as the storage medium. Typically, large water tanks secondary
fluid is rerouted through another heat exchanger
are used to store the water (4). In some facilities the on-site coupled
with the cooling water loop that serves the cooling
water storage for the fire suppression system is used for load.
The ice melts from the inside (at the coil external sur-
chilled water storage. face) as
the secondary fluid exchanges heat from the cooling
As chilled water is drawn from a storage system, such as water to
the ice storage tank. The internal melt system is
the tank in Fig. 2, an equivalent amount of warmer water
advantageous over the external melt system because the re-
from the load is added to the system. To have an effective
frigeration cycle does not have to operate at the lower temper-
chilled water storage system the colder chilled water must be atures
needed to freeze water on the outside of the formation.
separated from the warmer water it displaces. When a single Ice
harvester systems use a separate tank to store the ice
tank is used for chilled water storage, the natural stratifica- (6). Ice
is built up in cycles on the external surface of a verti-
tion of colder water at the bottom of the tank and the warmer cally
positioned refrigerant evaporator. After building up a
water at the top can be employed as a separation method. In certain
thickness of ice on the coils (approximately 0.5 to 1.0
order to maintain a sharp vertical gradient in temperature cm), the
refrigerant is valved such that the hot gas from the
(thermocline) between the warmer and colder water masses,
compressor enters the evaporator for a short period of time.
steps must be taken to minimize mixing in the tank. Baffles The hot
gas melts a thin film of ice at the evaporator external
may be placed in the tank to minimize horizontal mixing. surface
causing the ice build-up to fall into an ice storage tank
Also, using many distributed inlet and outlet nozzles helps to below.
Alternatively, in some systems the ice is scraped away
minimize mixing as water is added or removed from the tank.
mechanically. Chilled water is pumped from the storage tank
Since water’s density at standard atmospheric pressure is to the
heat exchangers serving the cooling load, then back
greatest at about 4#C, care must be taken to not charge the over the
evaporator coils.
tank with water at or below this temperature. Other
systems use small containers of water (e.g., plastic
Thermal storage tanks may also use a membrane (dia- balls
filled with water) to store ice. The method of freezing
phragm) to keep the cold and warm water masses separate. and
melting the ice is similar to the internal melt on coil stor-
The membrane can rise and fall as the tank is charged and age
systems. A secondary heat transfer fluid is circulated
discharged. The cold and warm water masses can also be sep- around
the water-filled containers and the refrigerant evapo-
arated using a series of smaller tanks or compartments. The rator
when freezing the ice. During the melting cycle, the sec-
----------------------- Page 554-----------------------
ondary fluid is circulated around the containers and another gions

with long periods of subzero weather where air source
heat exchanger coupled with the chilled water loop serving heat
pumps are not advantageous over resistance heat. Also,
the cooling load. in the
extreme cold climates, electrical utility peaks occur
Phase-change materials (PCMs) other than water may also during
the early morning hours of extremely cold days. The
be used in cool storage systems (7). Typical building air condi- rockbed
is charged during lower rate off-peak periods in the
tioning systems use approximately 7#C water in coils to cool daylight
hours and discharged at night.
air flows from approximately 24#C to 13#C. Using a PCM with
a phase-change temperature of approximately 6#C rather Building
Mass Thermal Storage
than water at 0#C allows the refrigeration cycle to operate at
a warmer evaporator temperature and therefore use less en- Building
mass can also be utilized for thermal storage (10).
ergy input per kilojoule of cooling. Typically mixtures of inor- The
building structure and internal elements such as floor
ganic salts and water having phase change temperatures slabs and
walls become the storage media used in conjunction
near 8#C are used in encapsulated containers. with the
building’s air conditioning system. During the night,
Some building air conditioning systems that use ice stor- the
building may be precooled to a limited extent using the
age (at 0#C) take advantage of the cooler thermal sink and air
conditioning system and cool outdoor air if available.
circulate smaller amounts of relatively low temperature air Lower
electrical rates can be taken advantage of when op-
through the building (1). The lower temperature air condi- erating
the refrigeration system during the night. Also, be-
tioning system uses relatively smaller duct work and less fan cause the
night air temperatures are usually cooler than the
energy. Because the low temperature air can cause fogging or day, the
refrigeration systems run more efficiently rejecting
heat to
the cooler outdoor air.
unwanted condensation and be poorly distributed if intro-
duced directly into a room, the cold air is usually mixed or There
are practical limits on using a building’s mass for
entrained with warmer air before being blown into the space. cooling
storage. The building’s internal surfaces should not be
so cool
that occupant comfort is decreased or so cool that there
Solar Heating Systems is a
possibility of condensation on surfaces when ventilation
air is
brought into the building. However, because of the large
Some means of thermal storage is required in solar heating building
mass, even a few degrees of decreased building mass
systems as the availability of solar energy is seldom in synch
temperature can contribute to decreasing cooling require-
with heating requirements. Solar heating systems are gener- ments
during the day. The use of PCMs in the building mate-
ally classified as passive or active systems. Passive solar heat- rials,
such as wallboard, is a method of increasing thermal
ing systems rely on natural circulation in transporting heat storage
capacity while limiting temperature changes.
to storage whereas active systems use forced circulation (i.e.,
utilizing pumps or fans). Turbine
Inlet Cooling
Solar water heating systems, active or passive, typically
use water storage tanks for thermal storage. Simple domestic Many
electric generating utilities use gas turbines to help
solar water heating systems may use existing electric or natu- meet peak
loads. Installing a gas turbine peaking plant is of-
ral gas fired water heaters for storage. When solar energy is ten more
economical for a utility than investing in increasing
insufficient to meet the water heating load, the electrical ele- larger
base load plant capacities even though the gas turbine
ment or natural gas burner will maintain the hot water sup- may
have a higher cost-per-kilowatt-hour rate. Warmer
ply. With an air/water heat exchanger, solar water heating weather,
which contributes to higher electrical demand, also
systems can also serve air heating loads. reduces
the power available from a gas turbine.
Various passive solar heating systems may be employed to A gas
turbine can lose 25% of its generating capacity as
store some amount of incident solar energy during the day for the air
inlet temperature increases from 15#C to 38#C. During
use overnight (8,9). A large area of glazing (windows) on the the peak
demand periods, it is the plant’s capacity and not
south face of a building (in the northern hemisphere) can be the cost
of energy that is of the greatest value. Ice storage
combined with an interior wall facing the glazing. During the systems
have been employed to precool turbine inlet air dur-
daylight hours, incident solar energy warms the wall to the ing peak
load periods (11). Electrical energy from the same
extent that the wall cools and warms the air in the house turbine
or any other source on the grid is used during off-
during the hours of darkness. Heat transfer from the wall to peak
hours to build the ice store. Although the overall cost
the air in the house is primarily by convection and in some per
kilowatt-hour may be increased with the investment and
cases the air flow can be controlled to minimize losses during operation
of an ice storage/inlet air cooling system, the in-
the storage period and to enhance heat transfer during the crease in
available capacity is usually of greater value to the
discharge period.
generating entity.
Electric Heat—Rockbed Storage

DESIGN
Rockbed thermal storage systems are usually employed in
conjunction with electrical resistance household heating sys- Thermal
storage systems may be designed to meet all or part
tems. With time-of-day electrical rates (on-peak and off-peak of a
given load over some period of time. For example, in
pricing), the thermal storage system can heat the rock mass chilled
water storage systems, the water storage tank may be
during the lower rate period for discharge during the high- sized to
meet all the building’s cooling load during occupied
rate period. The primary advantage of the electrical resis- hours
with the chiller system charging the entire store over-
tance heater/rockbed combination is the ability to use off- night.
Alternatively, in a situation where the building’s chill-
peak electricity. The rockbed storage is typically used in re- ing
system is undersized, a thermal storage system may be
----------------------- Page 555-----------------------
sized only to meet the difference between the building’s peak cient
rate or at relatively low cost) and discharge it over an-
cooling load and the undersized chiller’s capacity. other.
However, in most thermal storage systems, the rate
Storage size can be determined by assuming a load profile and
duration of charging and discharging the store can be
that must be met and the capacity of the thermal energy
controlled. To maximize the benefits of thermal storage, an
source. An example cooling load profile over a 48-h period is optimal
control strategy that minimizes cost or some other
shown in Fig. 3. One possible profile for operating the chiller desired
criterion should be employed (15).
is also shown. The quantity of energy storage required is indi- In
some cases, when the heat source or sink either will not
cated by the integrated area below the required load curve be
available in the future or will be available only at great
and above the available chiller load curve. In order to charge cost,
the optimal strategy is obvious. In a solar heating sys-
the store, the integrated area where the chiller capacity ex- tem,
the incident solar energy is free, and, if the heating re-
ceeds the required load must exceed the area indicating
quirement exceeds the maximum amount of heat storage, the
storage. store
should be charged as much as possible. If the on- and
With thermal storage, the chiller is decoupled from the off-peak
rates of an electrical rate structure differ by orders
load and can have almost any profile of operation so long as of
magnitude, a chilled water system should be charged fully
the integrated area associated with charging the store ex- (or at
least to some amount greater than the expected next
ceeds the area associated with storage discharge. However, day’s
cooling load) when the lower rate is available.
there is at least one profile of chiller operation that minimizes In
most applications, the best or optimal control strategy
energy costs over the subject time period. The storage indi- is not
entirely obvious, but may be determined through simu-
cated by this optimal chiller operation profile is the optimal lation
or other analysis. A general outline of an approach us-
energy storage requirement over the subject period. But for a ing
simulation follows:
valid economic analysis, operation over a longer period, such
as the anticipated life of the system, along with equipment
Develop a model of the system including the storage unit
costs must be considered.
and the load served by the store. Often weather-depen-
Models of the thermal storage system, chiller system, and
dent or otherwise time-varying loads can be predicted
predicted load profiles can be used to estimate operating costs
using statistical models.
over a given period of time. Weather data used in the simula-
Identify controlled variables such as the mass flow rate of
tion can be real measured data or meteorological data which
the fluid entering the store. Identify the external time
simulates a typical year. Various storage charging and dis-
varying uncontrolled variables such as the predicted
charging schedules along with various storage sizes and
heat-transfer load to be served by the storage system.
chiller capacities can be simulated to determine an optimal
In some cases the thermal loads may be calculated from
storage size. a
building model where the external variables will in-
Electrical rates that differ depending on the time of day
clude the outdoor weather conditions.
are usually the primary influence in determining the opti- The
model should reasonably agree with the actual pro-
mum tank size as well as in determining the best operating
cess. Model parameters, if present (such as rate param-
schedule (or profile) for charging and discharging thermal
eters), can be determined using measured data for pro-
storage systems. Before deregulation of electrical utility sys-
cesses already operational. If measured data are not
tems in the United States, a common electrical rate structure
available, manufacturers frequently provide perfor-
applied to commercial or industrial customers included time
mance data in tabular or graphical form.
of day rates (usually one ‘‘on-peak’’ rate and one ‘‘off-peak’’
Define the objective function to be minimized, such as elec-
rate) for energy use (kW-hr) and a demand charge for the
trical costs over a given period. The cost function should

highest rate of energy use (kW) over some billing period. More
include time varying rates if applicable.

recently, industrial and commercial users have been able to
choose real time pricing (RTP) for their electrical rate sched-
Identify constraints on the control variables and any
ule. In RTP, electrical energy rates vary each hour according
model-dependent variables. For example, a particular
to electrical market supply and demand.
piece of refrigeration equipment may only provide
Optimum tank size and storage operation under RTP can
chilled water between a defined lower and upper limit
differ significantly from optimal values under a time of day
on temperature.
rate structure. For example, if RTP rates are significantly
Create a computational model of the combined system,
lower over a weekend, it may be economically advantageous
storage unit, and load. The computational model should
to have a storage system capable of meeting loads over more
take as input, predictions of the load profile as well as
than one day. Daryanian and Bohn (12) investigate and com-
time varying values of the controlled variables. Model
pare optimal sizing of storage systems under both RTP and
output should include calculated values of the objective
time of day electrical rates.
function over time.
High temperature thermal storage systems may also be de- Use
an optimal control algorithm such as dynamic pro-
signed for use by electrical generation plants. Somasundaram
gramming that can determine a trajectory of control
et al. (13,14) investigate the application of thermal storage in
variable values that minimizes the objective function
power plants as well as applications in cogeneration systems.
(e.g., energy costs) over a particular time period (e.g., 24
h). Although a study period of 30 days may produce a
particular minimum energy cost, shortening the time

OPERATION
period (i.e., horizon) to 24 h, for example in a building
cooling storage system, could greatly decrease the diffi-

A simple operating strategy is to charge the store over one
culty of the problem solution while only slightly degrad-
period (when the heat source or sink is available at a suffi-
ing the results (15).
----------------------- Page 556-----------------------
THERMAL MAGNETORESISTANCE 25
CONCLUSION
sign, Analysis and Improvement of Energy Systems, New York:
ASME, 1991.
In the United States as well as other industrialized nations,

12. B. Daryanian and R. E. Bohn, Sizing of electric thermal storage
electric utilities have been converted from government- or in-
under real time pricing, IEEE Trans. Power Syst., 8 (1): 35–43,
vestor-owned regional monopolies to open market competi-
1993.
tors. In an open market, electrical energy is a commodity
13. S. Somasundaram et al., Integrating thermal energy storage in
priced according to its supply and demand. However, unlike
power plants, Mech. Eng., 115 (9): 84–90, 1993.
most other commodities, marketed electrical energy is not
14. S. Somasundaram et al., Cost evaluation of diurnal thermal en-
stored in any significant quantities.
ergy storage for cogeneration applications, Energy Eng., 89 (4):
Although electrical energy is not normally storable by elec-
8–22, 1992.
tricity consumers, with thermal storage, relatively lower cost
15. G. P. Henze, R. H. Dodier, and Moncef Krarti, Development of a
electricity available during low-demand periods can be used
predictive optimal controller for thermal energy storage, HVAC
to generate or transport heat into sources or sinks for use
R. Res., 3 (3): 233–264, 1997.
during periods of higher demand and cost. As the number and
16. J. S. Caldwell and W. P. Bahnfleth, Chilled water thermal energy
capacity of thermal storage systems increase, the large varia-
storage without electric rate incentives or rebates, J. Arch. Eng.,
3 (3): 133–140, 1997.

tions in electrical demand will decrease and therefore the pre-
dictable difference in rates for on-peak or off-peak usage will
17. M. K. Drost et al., Central station thermal energy storage for
peak and intermediate load power generation, Energy, 17 (2):

decrease. The influence of an open market electricity supply
127–139, 1992.
on the installation of thermal storage systems is considered
by Caldwell and Bahnfleth (16).
BARRETT A. FLAKE
This article has presented the basic uses and operation of
Air Force Institute of Technology

thermal energy storage systems. In addition to the applica-
tions listed, there are other unique thermal storage systems
including those used in storing high-temperature heat
sources used in power generating cycles (17). Although the
media (typically molten salts) and temperatures in power cy-
cle thermal storage are significantly different than chilled wa-
ter or ice storage, the fundamental use and operation of the
thermal storage is the same.
BIBLIOGRAPHY
1. C. E. Dorgan and J. S. Elleson, ASHRAE design guide for cool

thermal storage, ASHRAE Trans., 100 (1): 33–38, 1994.
2. J. E. Braun, Reducing energy costs and peak electrical demand
through optimal control of building thermal storage, ASHRAE
Trans., 96 (2): 876–888, 1990.
3. F. B. Morris et al., Experimental and simulated performance of

optimal control of building thermal storage, ASHRAE Trans., 100
(1): 402–414, 1994.
4. K. O. Homa, C. W. Sohn, and S. L. Soo, Thermal performance

of stratified chilled water storage tanks, HVAC R. Res., 2 (2):
158–170, 1996.
5. ASHRAE Handbook: Heating, Ventilating, and Air-Conditioning

Applications. Atlanta: ASHRAE, 1995.
6. D. E. Knebel, Optimal design and control of ice harvesting ther-

mal energy storage systems, Proc. ASME Natl Heat Transfer
Conf., Minneapolis, MN, 1991, pp. 1–9.
7. I. O. Salyer and A. K. Sircar, Review of phase change materials

research for thermal energy storage in heating and cooling appli-
cations at the University of Dayton from 1982 to 1996, Int. J.
Global Energy Issues, 9 (3): 183–198, 1997.
8. J. A. Duffie and W. A. Beckman, Solar Engineering of Thermal

Processes, 2nd ed., New York: Wiley, 1991.
9. I. Dincer and S. Dost, Perspective on thermal energy storage sys-

tems for solar energy applications, Int. J. Energy Res., 20 (6):
547–557, 1996.
10. I. Andersen and M. J. Brandemuehl, Heat storage in building

thermal mass: A parametric study, ASHRAE Trans., 98 (1): 910–
918, 1992.
11. K. J. Cross et al., Evaluation of thermal storage options for com-

bustion turbine inlet air cooling, in Thermodynamics and the De-
----------------------- Page 557-----------------------
} { { } }
HOME
ABOUT US
●
CONTACT US
HELP

Your Librarian
Profile
Thermionic Conversion

Standard Article

Mohamed S. El-Genk1 and Dmitry V. Paramonov2
1University of New Mexico, Albuquerque, NM
2Westinghouse Science and Technology Center, Pittsburgh, PA

Search
reserved.
DOI: 10.1002/047134608X.W3043
❍ Acronym Finder
Abstract
Background
Work Function
Modes of Operation
Conversion Efficiency
Types of Thermionic Converters
Engineering Aspects of Thermionic Conversion

Application to Space Electric Power Generation
Application to Terrestrial Space Electric Power Generation

%20Conversion/W3043.htm17.06.2008 16:48:12
----------------------- Page 558-----------------------
THERMIONIC CONVERSION 49
THERMIONIC
CONVERSION
Thermionic
(TI) energy conversion relies on the emission of
electrons by
solids heated to a sufficiently high temperature.
Thermoelectron emission phenomenon was discovered by

Thomas Edison
in 1883 and is still known as the Edison effect.
Herring and
Nichols (1) published a review of thermionic
emission as
early as 1949. The interest in thermionic conver-
sion was
resurrected in the mid-1950s, when Hastopoulos (2)
published a
description of two thermionic converters. The op-
eration of
thermionic converters was first demonstrated in the
1950s by P.
Marchuk in the USSR (3) and by Houston (4) and
Wilson (5) in
the United States. Major advances in thermionic
technology
had been made through the 1950s and 1960s in
the USSR,
United States, France, Germany, and many other
European
countries.
In the
1970s and 1980s, funding for the development of
thermionic
technology in the United States was significantly
decreased.
Research continued, however, but at a reduced
level, to
demonstrate the operation of multicell thermionic
fuel
elements at General Atomic and Los Alamos National
Laboratory.
Basic research on thermionic conversion also con-
----------------------- Page 559-----------------------
50 THERMIONIC CONVERSION
Emitter Collector from

the collector at constant temperatures. If joule losses in
the
electrodes and the leads connecting the converter to the
external load are ignored, then the work done by the electrons
circulating through the load could be considered adiabatic.

The
potential difference between the electrodes drives the
flow
of the electrons through the external load. A converter
terminal voltage equals the difference between the electric po-

Heat source Load Heat sink
tentials of the emitter and the collector, minus any internal
voltage drop due to the flow of the electrons through the inter-
Electron
electrode gap. Therefore, thermionic converters have a poten-

Figure 1. A schematic of a thermionic converter. tial
for achieving a large fraction of Carnot efficiency. How-
ever,
this requires the following:
1. Minimizing parasitic thermal losses from the emitter

tinued by Raser’s and Associates, Space Power Inc., Therma-
surface (by lateral conduction in the electrodes and

core, Thermoelectron, Inc., and a few universities in the
leads; by thermal radiation to the collector surface and

United States. The USSR had continued the development of
other structure materials in the gap, such as insulation

thermionic technology at a high level, focusing on both
and metal-ceramic brazes; and by conduction in the

multicell and single-cell fuel element concepts for integra-
plasma-filled gap)
tion into space nuclear reactor power systems. In 1987 and
1988, the USSR announced the successful operation of two
2. Minimizing the voltage drop in the interelectrode gap
thermionic power systems in space for 6 and 12 months,
3. Minimizing joule losses in the electrodes and the leads
respectively. These power systems, which used in-core
to the load
multicell thermionic fuel elements (TFEs), became known
4. Minimizing the collector work function
later in the West as TOPAZ-I. A number of Soviet TOPAZ-
II systems, which employed in-core single-cell thermionic
WORK
FUNCTION
fuel elements, had been built and ground tested successfully
in late 1970s and in the 1980s, using tungsten electric
Although electrons having an energy within the Fermi con-

heaters, instead of nuclear fission, but they had never been
duction band are free to move inside a metal, for them to es-
deployed in orbit.
cape
the metal surface they must have sufficient energy to
Following the end of the Cold War, a joint technology pro-
overcome a potential barrier, which is characteristic of the

gram (1991–1996) was initiated between the United States
type
of the metal. This potential barrier, associated with the
and Russia and included participation from the United King-
electron’s positive image charge inside the metal and the un-
dom and France. The objectives of the program were to (1)
compensated surface charge, is equal to the difference be-

perform full-scale tests of a number of TOPAZ-II systems, us-
tween
the metal’s Fermi energy level and the electron’s poten-
ing tungsten heaters instead of nuclear fuel; (2) explore the
tial
in vacuum and is known as the metal’s bare work
use of a TOPAZ-II type power system for future space power
function, # (typically expressed in electronvolts).

and propulsion missions; (3) advance the technology of insula-
The metal bare work function depends on the crystallo-

tion materials and ceramic-metal brazes; and (4) develop and
graphic orientation of the surface (e.g., monocrystalline or

test advanced high-power density TFEs. This program contin-
polycrystalline) and the type of the metal (6–13). The sorption

ued successfully until it was terminated in 1996. The joint
of
rarefied vapor, such as cesium and barium, onto the metal
tests of the TOPAZ-II system were performed in a special fa-
surface lowers its thermionic emission work function, de-

cility that was built for that purpose at the University of New
pending on the surface temperature and the vapor pressure

Mexico in Albuquerque.
of
the rarefied vapor (Fig. 2) (14). The sorption of very low-
pressure oxygen (#10#2 Pa) onto a refractory metal surface

BACKGROUND has
also been shown to increase its bare work function (15).
When
cesium is sorbed onto such a surface, the oxygenated-
A thermionic converter consists basically of two metal elec-
cesiated work function of the metal is usually lower than its
trodes separated by a small gap of either vacuum or rarefied-
cesiated work function, resulting in a higher electron emis-
vapor plasma. One of the electrodes, maintained at a high sion.
The potential difference between the effective work func-
temperature is referred to as the cathode or emitter, and the tions
of the emitter and the collector also increases, increas-
other, maintained at a lower temperature, is referred to as ing
the terminal voltage and the electric power output as well
the anode or collector. The thermal energy supplied to the as
the conversion efficiency of the converter.
emitter from a heat source enables some electrons to escape
The emission current density, J, is given by the well-
its surface. These electrons traverse the vacuum or plasma known
Richardson–Dushman (16–18) equation:
gap, separating the emitter from the collector, and are cap-
2 2 −(eϕ/kT )
tured by the collector. They then return to the emitter
J (A/cm ) = AT e (1)
through an external load, hence producing electrical power
(Fig. 1). In
Eq. (1), e and me are the electron charge and mass, respec-
The thermodynamic cycle that describes the operation of a
tively, k is the Boltzmann constant, and T is the absolute tem-
thermionic converter resembles that of Carnot, in which ther-
perature of the metal surface. For an ideal, isothermal sur-
face,
the constant, A (A/cm2 2
mal energy is ideally supplied to the emitter and removed
#K ), derived theoretically by
----------------------- Page 560-----------------------
5
Table 2. Bare Work Functions of Typical Electrode Materials
Bare Work Bare Work
Electrode Function, Electrode Function,
Material # (eV) Material # (eV)

4
Emitter
) W
polycrystal 4.54 Nb polycrystal 4.0
V (typical) W
(100) 4.63 Pt polycrystal 5.32
e
(
n W
(110) 5.25 Pt (111) 5.7
o
i
t W
(111) 4.47 Re polycrystal 5.0
c 3
n
Ir polycrystal 5.27 Re (1011) 5.75
u
f Collector
k
Ir (110) 5.42 Ta polycrystal 4.12
r (typical)
o
Ir (111) 5.76 Ta (100) 4.15
W Polycrystal W
Ir (100) 5.67 Ta (110) 4.8

2 Monocrystal W
Mo polycrystal 4.3 W2 C 2.6–4.58
Polycrystal Mo
Mo (110) 4.95 ZrC 2.1–4.39
Mo (111) 4.55 Cs 1.8
Ni polycrystal 4.1 Ba 2.2

1
1 2 3 4 5 6
Electrode temperature / Cesium reservoir (T/TCs)
Figure 2. Work functions of refractory metals in cesium (14).

The collector always operates near the minimum cesiated
work function, while the cesiated emitter work function
should be significantly higher (Fig. 2), in order to maximize

Richardson-Dushman, was given as
the terminal voltage and the conversion efficiency (7,10).
2 3
The emitter temperature and the operating cesium reser-
A = (4πem k /h ) = 120.4 (2)
e
voir temperature (or vapor pressure) are, therefore, selected
to
reduce the effect of the space charge near the emitter and
In Eq. (2), h is Planck’s constant. The actual value of A ,
at
the same time increase electron emission. For a given emit-
however, is determined experimentally and depends on the
ter temperature, low cesium pressure increases the effective

surface characteristics and the experimental method used
emitter work function and, hence, the terminal voltage but
and thus could vary slightly from one source to another. Ta-
fails to eliminate the space-charge effect, lowering the elec-

bles 1 (16) and 2 list typical values of the bare work function
tron emission and increasing the voltage drop in the gap. The
and the constant, A , for a number of typical electrode materi-
space-charge effect can be eliminated through the combina-

als. As Table 1 indicates, the actual values of A could be much
tion of using a small interelectrode gap and/or introducing

lower than the theoretical value of Richardson–Dushman
cesium at the proper vapor pressure into the gap. The Cs ions
[Eq. (2)].
for the plasma in the gap are produced mainly by surface ion-
All refractory metals, capable of prolonged operation at
ization at low vapor pressure and by impact, or volume ioniza-

typical emitter and collector temperatures between 1800 K
tion, at high cesium vapor pressure. The plasma density, n,

and 2000 K, and 800 K and 1000 K, respectively, have high
due to Cs ionization is given by the Saha–Langmuir equation:

bare work functions (in excess of 4 eV) (Tables 1 and 2). The
electrode work functions are adjusted during operation by
−e (E −ϕ )
n g i i
controlling their respective temperatures and the cesium va-
= e kT (3)
por pressure in the interelectrode gap (Fig. 2). When Cs
ntot g o
atoms are sorbed onto a refractory metal surface, having a

higher bare work function higher than that of Cs (1.89 eV),
where n and ntot are the plasma density and total density of
they are ionized (cesium’s first ionization potential, Ei # 3.89
charged and neutral particles, respectively, and gi and g o are
eV), producing a local electric field that reduces the metal’s
statistical weights for ions and atoms, respectively. The ion-
effective work. Since the high sorption energy of cesium and
ization of Cs atoms can occur through a single step (Cs
#
# #
the high bare work of an electrode material result in a low e
# Cs
2e ) or a multistep (Cs
e # Cs*
# #
#
cesiated work function, monocrystal tungsten is the preferred e
; Cs*
e # Cs
2e ) ionization (where Cs* denotes a Cs
electrode material for thermionic converters (Tables 1 and 2).
atom in the excited state). In the case of the impact ioniza-
tion, the plasma concentration distribution in the interelec-
trode gap is influenced by the various transport effects (dis-

Table 1. Bare Work Functions and Emission Constant for
cussed later) as well as the local net rate of ion generation
Selected Electrode Materials (16)
(ionization-recombination rate), (x), which can be given as
(3)
Electrode Bare Work A Constant
Material Function, 2 2
# (eV) (A/cm # K )
ni (x ) # n2 (x ) #
Molybdenum (Mo) 4.2 55
#(x ) = 1 − i (4)
τ n2 (T )
Nickel (Ni) 4.61 30
i e
Tantalum (Ta) 4.19 55
Tungsten (W) 4.52 60
where # is the effective ionization time, n (x) is the local
i i
4.54 46
W
Ba 1.6 1.5 plasma
concentration, and n(Te) is the equilibrium plasma
concentration [Eq. (3)], at Te, the local electron temperature.
----------------------- Page 561-----------------------
MODES OF OPERATION
sen discharge, can be written as
This section describes the different types of thermionic con-

j i = r me (6)
verters and the mode of electron discharge in each. The earli-

j e mCs
est types of converters are those that employ vacuum or
quasi-vacuum discharge. Because of the inherently very low
where j e and j i are the electron and ion currents, respectively,

emission current, their use is limited to special-purpose appli-
and mCs is the mass of a Cs atom. The minimum bare work

cations such as radiation-hard integrated circuits. For electric
o
function of the emitter, # , required for complete space-charge

power conversion applications, however, in which an electric
E
2 2
ionization, is given as
power density of 1 W/cm to 15 W/cm or higher is required,
# #
other discharge mode converters are more suitable, such as
o 1 1 2AT 2 r me
those that operate in the Knudsen and ignited modes of dis-
ϕE = Ei + kTE ln E (7)
charge. The major features of the different discharge mode
2 2 eno mCs
converters as well as the advantages and limitations of each

are discussed in the following subsections.
where no is the flux of neutral Cs atoms to the emitter and
T
is the emitter temperature. When j #j #m / m , there are
E i e e Cs
Vacuum and Quasi-Vacuum Modes of Discharge
excess ions and the converter is said to operate in the over-
The most familiar thermionic converters are those that oper-

compensated mode. In the undercompensated mode, how-
ever, j #j #m / m , and the rate of ion generation is not suf-

ate in the vacuum and quasi-vacuum modes of discharge.
i e e Cs
However, because of their low power density, they are of very
ficient to neutralize the space charge fully (hence the emission
limited practical use. Vacuum discharge converters have no
current becomes space-charge limited).
Cs in the interelectrode gap, which is kept very small, on the
Assuming complete space neutralization (ideal Knudsen
order of 1 #m or less, to avoid the formation of a space charge.
mode converter), the relation between the interelectrode volt-
In these converters, the discharge current density is typically
age and net emission current density, j (the voltampere char-
very small (mA/cm2). This would require a very large elec-
acteristics) can be readily determined from the following
trode area or operating at very high emitter temperature to
equations:
achieve a meaningful electric power output (#mW/cm2).
− e ϕ E e (ϕE −V )
High-temperature operation in vacuum converters, however,
2 kT 2 − kT
j
E = ATE e E and j C = ATC e c for V ≤ ϕE − ϕC
is not possible because it risks closing the gap, due to the
thermal expansion of the electrodes, and shorting the con-
(8a)
verter.
e (ϕ +V ) e ϕ
C C
In the quasi-vacuum mode converters, Cs is introduced j
= AT 2e− kTE and j = AT 2e− kTC for V > ϕ − ϕ
E E C C E C
into the interelectrode gap, but its pressure is kept quite low
(8b)
(a few millitorrs), just enough to adjust the electrodes’ work
function, as discussed earlier, but not sufficient to neutralize
and
the space charge near the emitter surface. Very little or no
surface or impact ionization of cesium takes place in the gap,
j = j E − j C (8c)
since the electron mean free path, # , is much larger than the
e
interelectrode gap size, d(#e # d). Consequently, the space

In
these equations, #C and TC are the collector work function
charge forming at the emitter surface repels the emitting elec-
and temperature, respectively; and j E and j C are the electron

trons, limiting the converter’s discharge current. In other
emission current densities from the emitter and the collec-

words, the space charge increases the effective emission work
tor, respectively.
function by a certain amount, #, that depends on the extent
The voltampere characteristics of an ideal, Knudsen mode
of the space-charge region near the emitter surface. In this
converter is shown in Fig. 3. In the saturation region of the

type of converter, the emission current can still be calculated
voltampere characteristics, all the electrons from the emitter

using the Richardson-Dushman relation [Eq. (1)], but after
are accelerated in the gap and reach the collector; however,

accounting for the space-charge effect, as follows:
those emitted by the collector are partially repelled back.

2 2
Only those electrons with energy higher than (#E # #C # V)
J (A/cm ) = AT exp(ϕ + #ϕ)/kT ) (5)
would reach the emitter. In the obstructed region, the situa-
tion reverses. According to Eq. [8(c)], the resulting net current

Knudsen Mode of Discharge
equals the difference between the emitter and collector elec-

In Knudsen mode converters, the interelectrode gap size is
tron currents.
significantly larger than in vacuum and the quasi-vacuum
In the Knudsen discharge, a Cs pressure of only a few
mode converters. Also, the Cs pressure in the gap is suffi-
millitorrs (or tens of pascals) is needed to neutralize the space
ciently high to partially, or completely, neutralize the space
charge fully. However, such low Cs pressure does not suffi-
charge by the Cs ions generated by surface ionization at the
ciently lower the emitter work function to produce the high
emitter surface. In Knudsen discharge, the Cs pressure is not
current density required for energy conversion applications.
high enough, however, to produce volume ionization by the
When the Cs pressure is increased, while #e is still larger
collisions of electrons with neutral Cs atoms in the gap (in
than the gap size, d, losses of electron energy due to collisions
Knudsen mode #e #d). The condition for a complete space-
with neutral Cs atoms become significant and the operation
charge neutralization (i.e., n # n ), when operating in Knud- of
the converter transitions to the diffusion mode of discharge.
e i
----------------------- Page 562-----------------------
j
characterized by the ion generation in the gap by volume ion-
Saturation region Obstructed region
ization of Cs atom due to the collision with emitted electrons.
In
this mode of discharge, the interelectrode voltage is low
)
2 enough
and the Cs pressure is high enough that the emitted
m
c Transition point
electrons gain sufficient kinetic energy, as they are acceler-
/
A
( Saturation ated
in the sheath near the emitter surface, to cause volume
y
t
i current
ionization of Cs atoms in the gap.
s
n
When the load voltage is slightly lower than the ignition
e
d j E
t
voltage, Vign , the overcompensated plasma in the interelec-
n
e trode
gap does not occupy the entire gap and a small region,
r
r
u j C with
a negative space charge, V*, exists near the emitter.
C
This
space charge is called virtual cathode because the effec-
tive
emitter work function now becomes #eff # #E
V*. The
0
corresponding portion of the
voltampere characteristics is
ϕ E – ϕ C V
called the obstructed region [Figs. 4(a) and (b)]. As the con-
Open circuit
voltage verter
voltage is reduced further, V* is also reduced, until
Load voltage (V) the
transition point or the ‘‘knee’’ is reached, at which the ion
generation is just sufficient to neutralize the space charge

near
the emitter (i.e., V* # 0). In the saturation region the
ϕ C plasma
throughout the gap is overneutralized and all emitted
ϕ C
electrons reach the collector. In addition, the high concentra-
tion
of positive Cs ions near the emitter creates a strong posi-
ϕ E ϕ E ϕ C ϕ E
tive
electric field on the emitter surface, increasing the satu-
V V ration
current, due to the Schottky effect, above the emission
V
current given by Eq. (1). The voltampere characteristics of an
ignited mode converter and the corresponding electron motive

Transition point Saturation region Obstruction region
diagrams are shown in Fig. 4.

Figure 3. The voltampere characteristics of an ideal, Knudsen mode
The performance and the plasma behavior in the ignited
converter and the corresponding electron motive diagrams. mode
converters are described by the following transport
equations for electrons and ions, respectively (3):
Diffusion Mode Discharge

dV n T
j = −eu n − eD e − u n (1 + β )k e (9a)
The diffusion mode of discharge occurs when the emission
e e e dx e dx e e e dx
electrons undergo frequent collisions with neutral Cs atoms

in the gap (diffusion) but the contribution of volume ioniza- and
tion to current discharge is negligible. Hence, the diffusion
dV n T
mode voltampere characteristics and the electron motive dia-
j = −eu n + eD e + u n (1 + β )k e − R (9b)
i i i dx i dx i i i dx i
grams are qualitatively similar to those of the ignited mode.
Because of the significant losses of electron energy, diffusion
where
#e and #i are the electron and ion collision coefficients,
mode discharge is not desirable in thermionic conversion for
and Ri
is the friction force due to ion-electron collisions. The
electric power generation, due to the low conversion effi-
corresponding electron and ions continuity equations are

ciency. However, thermionic converters may operate in this
mode when Cs pressure and/or load voltage are off optimum
d je d ji
and when it is desirable to operate at very high current
= − = e#(x ) (10)
dx dx
density.
In diffusion mode discharge, as the load voltage is lowered, The
equations for the electron energy flux, Se, and its energy
the emitted electrons gain more kinetic energy as they are
balance are
accelerated by the interelectrode voltage drop, V. When
their energy becomes sufficiently high to cause volume ioniza-
## # #
5 kT T
tion, the operation mode of the converter transitions to the
Se = −j e + βe e + V − λe e (11a)
2 e dx
ignition mode discharge. The ignited mode discharge is char-
acterized by overcompensated plasma in the interelectrode and
gap, due to the high ionization rate of Cs atoms and the low
ion diffusion rate as compared with that of the electrons. The
dSe
emitter, V , and collector, V , plasma voltage drops change
= −Ei #(x ) − #Srad − #Sei − #Sea (11b)
E C
dx
their sign as compared with the diffusion mode (i.e., the emis-
sion electrons are accelerated near the emitter and deceler- In
these equations, #e is the thermal conductivity of electrons,
ated near the collector) (Fig. 4). and
S , S , and S are the losses of electron energy by
rad ei ea
radiation and by electron–ion and electron–atom collisions,

Ignited Mode of Discharge
respectively.
This is the operation mode of choice in the state-of-the art
When supplemented by the appropriate boundary condi-
thermionic converters for electric power applications. It is tions,
electron energy fluxes to and from plasma at both elec-
----------------------- Page 563-----------------------
)
2
m
Transition point
c
/
A
(
,
Saturation
y
t
i
current
s
n
e
Ignition
d
t
n
e
Obstructed Unignited
r
r
region (diffusion)
u
C
mode
V V
ign L
Load voltage (V)
∆V ∆VE ∆V
Vd ∆VE
Vd ∆VE Vd ∆VC
∆VE
∆VC ϕ C
ϕ E ∆VC ϕ E
∆VC ϕ E ϕ C ϕ E
ϕ C
ϕ C
Figure 4. Voltampere characteristics of V

V V V
ignited mode discharge and the corre-

sponding electron motive diagrams. Saturation region
Transition point Obstructed region Unignited mode
trode surfaces, Eqs. (9) to (11) could be solved for the distribu-
trodes. The latter depends on the emitter temperature and
tions of T , n, V, j , j , and S across the interelectrode gap, the
electrodes geometric mean temperature, T # (T
e e i e
m E
and the net current discharge as functions of the interelec-
TC)0.5, and is given as
trode voltage.
# = (1/# (T ) + 1/# (T ) − 1)−1 (15)
eff E E C m
CONVERSION EFFICIENCY
The conduction heat flux through the Cs-vapor-filled inter-
electrode gap is given also in terms of the emitter TE , and the

The total heat flux from the heat source to the emitter sur-
collector, TC, temperatures, using the following empirical rela-

face, qt , is the sum of several energy quantities: tion
(17) as
qin = qec + qrad + qCs + qcon + qlead − 0.5 qj (12) q

= k (T − T )/{d + 1.15 × 10−5 (T + T )/P )} (16)
cond Cs E C E C Cs
In this equation, qec accounts for the electron cooling of the In

Eq. (16), d is the interelectrode gap in cm, PCs is the cesium
emitter, qrad accounts for thermal radiation to the cooler col-
pressure in the gap in torr [P 8 #0.5
Cs # 2.45 10 (TCs) exp( #

lector surface, qCs accounts for the heat conduction through
8910/TCs)], TCs is the cesium reservoir temperature in kelvin,
the Cs-vapor-filled interelectrode gap of the converter, qcod and
kCs is the thermal conductivity of cesium at the effective
and qlead account for the conduction losses through the elec-
temperature, T*, given as
trodes and the leads, respectively, and qj is the heat generated
∗ 3 3 2
in the leads by joule effects. The electron cooling heat flux,
T = 2/3{(T − T )/(T − T ) } (17)
E C E C
qec, is given as
For
optimized leads, q and q are given in terms of the con-
Iead j
J 2 −e ϕ
version efficiency of the converter, as follows (18):
qec = (2kT + eϕ)T e kT (13)
e
q = J [L (T2 − T2 )(2 − η/4η)]0.5 (18a)
lead o E C
The thermal radiation heat flux from the emitter to the collec-
tor through the interelectrode gap, qrad , is given as and
q = σ # (T4 − T4 ) (14)
q = J [L (T2 − T2 )(η/2 − η)]0.5 (18b)
rad eff E C
j o E C
In this equation, # is the Stefan–Boltzmann constant (5.67 In

these equations, L is the Lorenz universal number for met-
10#12 2 4
#8 2 2
W/cm # K ), and #eff is the effective emissivity of the elec- als.
It varies from 2.2 to 2.9 10 V /K for pure metals at
----------------------- Page 564-----------------------
273 K and changes very little with temperature above 273 K,

Emitter (cathode)
increasing typically about 10 to 20% per 1000 K. The voltage
1800 K to 2300 K
drop in a pair of optimized leads can also be expressed in
terms of the converter efficiency as (18)
Collector (anode)
800 K to 1200 K Heat

V = [L (T2 − T2 )(η/2 − η)]0.5 (19)
source
lead E C
Therefore, the conversion efficiency of a thermionic converter,

with optimized leads and negligible lateral heat conduction in
Interelectrode
the electrodes (isothermal electrodes), can be expressed as
gap
Emitter
η = (V − V )J /q (20)
lead o in
where V is the measured terminal voltage of the converter.
This equation is solved for the efficiency after substituting for
the input heat flux to the emitter, qin [Eqs. (13) to (19)]. The
Electric load
total heat rejection to the heat sink from the collector and the
leads, qrej , is given as
qrej = (1 − η)qin (21)

(a) (b)
Figure 5. Schematics of a planar and a cylindrical TI converter. (a)

TYPES OF THERMIONIC CONVERTERS
Planar TIC. (b) Cylindrical TIC.
Thermionics technology for converting heat to electricity has

been under development in the United States, USSR/Russia,
clear fission reactor or a solar concentrator) [Figs. 5(a)
France, Germany, and many other countries for more than
and (b)]. A cylindrical TIC [Fig. 5(b)] could be assembled
four decades. These static converters had been developed
in single-cell (Fig. 6) and multicell configurations. These
mostly for space power applications. In such applications, typ-
configurations are preferred for compact arrangement
ical emitter (cathode) and collector temperatures are 1850 K
in a nuclear reactor core, where heat is generated by
to 1900 K and 800 K to 900 K, respectively, and the interelec-
nuclear fission in the fuel pellets stacked inside the cy-
trode gap is typically 0.1 mm to 0.5 mm wide. The gap is filled
lindrical emitter tube [Fig. 5(b)]. Cylindrical thermionic
with cesium vapor at a pressure as high as 3 torr (400 Pa).
elements can also be used outside the nuclear reactor
The electrodes in thermionic converters (TICs) are usually
core, where heat is transported to the inside of the emit-
made of high-temperature refractory metals, such as tung-
ter surface using high-temperature heat pipes. The heat
sten, molybdenum, or rhenium. These materials have rela-
pipe working fluid could be sodium or lithium, de-
tively high bare work functions (4 eV to 5 eV), but their cesi-
pending on the reactor operation temperature.
ated work functions could be as low as 1.2 eV. The advantages
Planar TICs [Fig. 5(a)] can be heated in space using
of using TICs in space power and propulsion systems are as
energy from a parabolic solar concentrator. Unlike a nu-
follows:
clear reactor TI power system, a solar TI space system
typically has a thermal energy storage unit. Excess

1. There is an absence of moving parts, which potentially
thermal energy, while in the sunny portion of the orbit,
increases the converter’s operation lifetime, up to 10
is stored in a thermal energy storage unit. The stored
years, as well as the system’s reliability.
thermal energy is then recovered and transported to the
2. The hottest temperature in a TI power system is limited
TICs by the working fluid or by conduction while in the
to the emitter (or cathode) assembly (Fig. 5), thus
eclipse portion of the flight orbit. Such an energy stor-
allowing the use of nonrefractory metal, such as stain-
age unit adds to the total mass of a TI solar space power
less steel, as structure material for the coolant loop and
system, which is not an issue with terrestrial solar TI
the heat rejection radiator.
power systems.
3. There is low specific mass, since TICs can be operated
5. Thermionic converters are inherently radiation hard de-
at very high power density, in excess of 20 W/cm2 . The
vices and thus could be used in satellites operating
peak electric power density, however, depends on the

within the Van-Allen ionizing radiation belt, or in
electrode temperatures, the type of the converter, the
power systems for deep space exploration missions.
electrode material, the size of the interelectrode gap,
and the cesium vapor pressure in the gap. The tempera-
Depending on the desired operation conditions, however,
ture on the outside of the collector (or anode) assembly
there are several types of TICs that have been developed and
(Fig. 5) of about 800 K is low enough to use stainless
operated successfully. Some have been tested and operated in
steel as the structure material but high enough to reject a
laboratory setup, under well-controlled conditions. A few
heat by radiation to space at high heat flux, resulting
have been integrated into a nuclear reactor power system
in a lower radiator size and mass.
(TOPAZ type systems) and were operated in space for up to
4. Thermionic converters could be planar or cylindrical, for a
year by Russia in 1997. The latter were in-core multicell
optimum integration with the heat source (e.g., a nu-
thermionic fuel elements, cooled by a sodium (23 wt%)-
----------------------- Page 565-----------------------
potassium (78 wt%) molten alloy (NaK), which has a freezing
temperature of 261 K. Except for the emitter assembly, the
rest of the TOPAZ system operates at or below 800 K. Hence,
stainless steel has been used as the structure material for the
coolant loop. The NaK coolant is circulated through the sys-

tem
using a passive electromagnetic pump, transporting the
Bellows
excess heat from the nuclear reactor core to the system’s ra-
diator.
Nb collar
In the following sections, a number of advanced TICs, such

Monocrystal Al O
2 3 as
high-temperature cesium TICs, with and without oxygen-
insulator seal
Nb collar
ated electrodes, and cesium–barium high-temperature TICs,
are
described. These converters have the potential of deliv-
ering high conversion efficiency (#15%) and high electric
power density (#10 W/cm2).

Emitter connection
In the advanced, high-temperature

TICs, the emitter and
Nb collar
collector temperature could be as high as 2000 K to 2500 K
Monocrystal Al2O3 and
1200 K to 1400 K, respectively. In addition, a brief review
insulator seal and
an update on the development of low-temperature planar
Nb collar
TICs with grooved electrodes is presented. In these convert-
Collector connection
ers, the emitter and collector temperatures range from 1600
K
to 1880 K and 800 K to 1200 K. The effect on the TIC
Sc2O3 spacer
performance of introducing inert

gases, such as helium and
Polycrystal Mo
collector
xenon, into the cesium interelectrode gap is also discussed.
Plasma sprayed
Finally, the operation principle of microgap TI converters is
alumina insulator
described.
UO2 fuel region

High-Temperature Cesium Converters
Sc O spacer
2 3
These converters operate at an

emitter temperature of 2000
Monocrystal Mo
collector K
to 2300 K and a collector temperature of 1000 K to 1200 K
and
have a relatively small
interelectrode gap (#0.5 mm).
Interelectrode gap
They have the potential for operating at conversion efficiency
of
about 12 to 15% and electric power density up to 15
W/cm2 2
to 20 W/cm . These very high

emitter temperatures,
Sc O spacer
2 3
however, necessitate using low vapor pressure refractory ma-
terials, such as rhenium, as the emitter material to minimize
material loss from the emitter surface. The issues of material
loss from rhenium (Rh)

electrodes at such high emitter tem-
Collector connection
peratures, and in the presence of gaseous traces, such as oxy-
gen, have not yet been investigated. However, the sublima-

Nb collar
tion of Rh would be much lower than tungsten, due to its
lower vapor pressure.
Nb collar
High-Temperature Cesium Converters
with Oxygenated Electrodes
Emitter connection
Many investigations have attempted to increase the conver-
sion efficiency of thermionic converters, beyond the typical 10

Nb collar to
12% of conventional or common TICs, either by increasing
Monocrystal Al O
2 3 the
emitter temperature up to 2100 K to 2300 K and/or intro-
insulator seal
ducing tracer amounts of oxygen in the cesium vapor filled

Nb collar
interelectrode gap (typically #0.5 mm wide). When operating

at
such high emitter temperatures, however, lifetime issues,
40 mm
such as gradual loss of emitter material and the effects on
the
TIC performance of the emitter material deposits on the
Figure 6. Longitudinal cross section in a fully assembled TOPAZ-I,
collector surface, should be addressed.
single-cell thermionic fuel element.
The results of laboratory investigations of thermionic con-
verters with isothermal tungsten emitters have demonstrated
that introducing tracer amounts of oxygen or cesium oxides
into the interelectrode gap can significantly improve the con-
verter performance (12,19,20). The measured improvements

in
the conversion efficiency and the electric power density of
the
TICs were attributed to the following two effects:
----------------------- Page 566-----------------------
1. An increase in the emitter oxygenated bare work func- 270
tion (11,12,21,22) which increases electron emission due

PCs = 3 torr
to the low oxygenated-cesiated work function of the )
W
emitter. Such low emitter work function not only de- ( 240
4 torr
d x
a a
o m
creases the optimum cesium pressure, corresponding to l
P
the maximum TIC electric power output, but also per- r,

1.5 torr
e 210
mits the use of a larger interelectrode gap, therefore in- w

o
creasing the converter reliability. p
d
2. A decrease of the minimum work function of the collec- a 180
o
l
tor surface when covered with tungsten oxide deposits m
1 torr Qth = 3500 W
from the emitter surface, which increases the load u
m
i
voltage. x 150
a
M
In recent laboratory tests (20), conversion efficiencies of
0.75 torr
2 120
18% to 22% at power densities up to 15 W/cm (Fig. 7) have

–10 –9 –8 –7 –6
10 10 10 10 10
been reported. These tests were performed using a planar
O/Cs
TIC. Oxygen vapor pressure up to 8 10#6 Pa was intro-
duced into the cesium-filled interelectrode gap. The emitter Figure

8. Calculated effect of oxygen on maximum electric power
temperature was #1800 K, the collector temperature was density
of a planar converter (23,24).
680 K to 800 K, and the cesium pressure in the interelec-
trode gap varied from 0.5 torr to 1.5 torr (67 Pa to 200 Pa).
The TIC had an emitter made of a CVD (chemical vapor depo- 4.
Emitter tungsten loss depletes the emission layer on
sition) tungsten coating onto molybdenum and an electropol-
the emitter surface (typically tungsten), which can
ished molybdenum collector. It has been shown that the intro-
eventually result in the base material diffusing to the
duction of a larger amount of oxygen (#10#6 Pa) into the
surface, hence reducing the emitter’s bare work func-
interelectrode gap, however, accelerates the emitter material

tion and increasing its cesiated work function.
loss (23,24). It also degrades the converter performance due 5.
Tungsten oxide deposits on the collector surface in-
to the deposit of the oxides of the emitter tungsten on the
crease the effective emissivity of the electrodes, low-
collector surface (Fig. 8). The degradation in the TIC perfor-
ering the emitter temperature and, hence, the emission
mance, in this case, can be attributed to several effects:
current and conversion efficiency.
1. Oxide deposits on the collector surface lower the collec-

The
net contribution of all or some of these effects, how-
tor cesiated work function, which increase the load volt-
ever,
strongly depends on the TIC design and operation condi-
age, thus lowering the electric power output and the
tions.
These conditions include the input thermal power, the
conversion efficiency.
interelectrode gap size, the type of electrode material, the

2. Tungsten oxides increase the thermal losses and electri- load
resistance, the cesium pressure, and the partial pres-
cal current leakage, when deposited onto spacers and sures
of the gaseous contaminants, such as air, oxygen, water
insulator surfaces, lowering the electric power output vapor,
or carbon dioxide.
and conversion efficiency.
3. Tungsten oxide deposits could eventually short circuit Cs–
Barium (Ba) High-Temperature Converters
the electrodes, if chipped off, during thermal cycling or
vibrations, and bridge the electrodes (this process is The Cs–
Ba converters operate at the same or higher emitter
more likely in a small interelectrode gap due to swelling
temperatures than the high-temperature Cs converters, but
of the thermionic fuel element [TFE] emitter). at much
lower Cs pressure (#10#2 torr versus several torrs in
high-
temperature Cs converters). In addition to neutralizing
the
space charge, Cs vapor serves as the transport media for
15 the
electrons from the emitter to the collector. The Ba vapor
pressure, on order of a millitorr, is sorbed readily onto the

Cs + O TIC
)
2 Cs TIC emitter
surface, lowering its effective emission work function.
m
c In
other words, Ba in these converters replaces Cs in the
/
W 10 high-
temperature Cs converters in regulating the work func-
(
y
t tions
of the electrodes. Consequently, it is possible to operate
i
s
n Cs–Ba
converters in the Knudsen mode, where scattering
e
d 5 Westinghouse losses
are almost nil and the maximum electric power output
r
e Bettis could
be derived at a terminal voltage that is equal to the
w
o contact
potential difference, (#Eo # #co).
P
The
optimum current density from the emitter surface,
0 which
corresponds to the transition from the unignited to the
0 0.2 0.4 0.6 0.8 1.0
ignited
mode of discharge, Jo, is independent of the type of
Interelectrode gap (mm)
2
the
emitter material ( 1.5 A/cm ) but increases with the ce-
Figure 7. Reported electric power densities for TICs with oxygen- sium
pressure, raised to the one-half power [i.e., Jo # (PCs)1/2,
ated electrodes (20). up to
2 10#2 torr] (25). These investigators have demon-
----------------------- Page 567-----------------------
strated experimentally that Jo is almost independent of PCs in

the range from 2 10#2 torr to 6 10#2 torr. Beyond 6
) Data
10#2, however, Jo decreases as the cesium pressure increases
2 20 Predictions
c
because of the scattering of electrons by cesium atoms in the
/
W
interelectrode gap.
( 10 1.7 eV
t
The emitter temperature strongly affects the value of Jo .
i
s
#2
n
At PCs # 10 torr, Jo almost doubles in value with increasing
e 5
d Gap size — 1 mm
the emitter temperature in 100 K increments, starting at

r
e
1900 K. For example, at TE # 1900, 2000, and 2100 K, the
w Collector work
2
o 2
measured J was 2, 4, and 8 A/cm , respectively (25). The
P function — 2.2 eV
o
optimum emitter work function, #Eo , also increases as the
1
emitter temperature increases (e.g., #Eo , # 3.27, 3.65, and 4.04
1900 2000 2100 2200 2300 2400
eV at emitter temperatures, TE , of 2000, 2500, and 3000 K, re-
Emitter temperature (K)
spectively).
The only effect the emitter material has on Jo is to change
Figure 9. Electric power densities versus emitter temperature for
the value of the barium pressure at which Jo occurs. This Cs–Ba
thermionic converters (28).
change is due to the difference in the sorption energies of the
barium atoms onto the different electrode materials; the
higher the sorption energy of barium, the lower is the opti-
temperatures. Also, since Cs–Ba converters operate at a very

mum barium pressure. For example, because the sorption en-
high
electric power density (Fig. 8), large ohmic losses occur
ergy of barium onto tantalum (Ta) is higher than onto tung-
in
the converter leads and electrodes. To reduce these losses,
sten (W), the optimum barium pressure for the former ( 4
10#2 torr) is higher than that for the latter ( 1.6 10#2 torr).
only planar electrodes, with a relatively small surface area
2
(#10
cm ), could be used. While planar electrodes are suitable
In Cs–Ba converters, the optimum PCs depends on the size
for
use in conjunction with solar concentrators, either in space
of the interelectrode gap. Experimental data have shown that
#2 or
terrestrial applications (28) and may be used in topping
for a gap size of 1.0 and 0.1 mm, the optimum PCs was 10
#1
cycles in fossil power plants, they are unsuitable for use with
torr and 10 torr, receptively (26). For these gap sizes and
a
nuclear reactor heat source.
collector temperatures of 1250 K to 1350 K, the optimum col-
lector work function, #co, was found to be 2.3 eV. This collector
work function corresponds to the maximum electric power
Grooved Electrodes, Low-Temperature Converters
density of the Cs–Ba converter. The maximum electric power
When
a TI converter is operated in the ignited mode of dis-
and the corresponding conversion efficiency increased almost
charge, increasing the cesium pressure increases the electric

exponentially with the emitter temperature, particularly in
power
output up to a point beyond which scattering losses
the range from 2000 K to 2300 K. The maximum power den-
cause
the electric power output to decrease. The optimum ce-
sity and conversion efficiency also increased with increasing
sium
pressure, corresponding to the peak electric power, de-
the optimum cesium pressure. Experimental data have shown
that at PCs of 10#2 torr, when the optimum gap size was 1.0 pends
on the type of the electrode material and temperatures.
For a
given cesium pressure, raising the emitter temperature
mm, the maximum power density increased from 1.0
W/cm2 2 or
lowering collector temperature increases the emission cur-
to 12 W/cm , as the emitter temperature increased
from 1900 K to 2300 K, respectively. When the gap size was rent
and the potential differential across the external load.
reduced by an order of magnitude to 0.1 mm (corresponding The
temperature of the collector, however, should not be
optimum PCs # 0.1 torr), the maximum power density was lower
than that corresponding to the minimum cesiated work
2
function of the selected collector material.
significantly higher, increasing from 2.3 W/cm to 30
W/cm2, as the emitter temperature increased from 1900 K to
On the other hand, increasing the emitter temperature in-
2300 K, respectively.
creases the material loss from its surface, by sublimation and
The optimum emitter and collector temperatures for the
formation of volatile oxides, when oxygen pressure in the gap
Cs–Ba converters are about 300 K to 400 K higher than those #10#8
torr (23,24,29–31). High emitter temperatures also im-
in high temperature Cs converters. These high electrode tem- pose

a challenge to the integrity of the insulation material
peratures make it possible to achieve a conversion efficiency and
the metal-ceramic brazes in the converter, hence shorten-
that could be as high as 60% of Carnot, at the same emitter ing
its performance lifetime. Furthermore, increasing the
(hot) and collector (cold) temperatures. After accounting for
emitter temperature increases the optimum cesium pressure,
radiation and conduction losses in the gap, the calculated con- which
results in a high back emission from the collector, low-
version efficiency ranges form 18 to 25% for TE of 2300 K ering
the converter’s dielectric strength.
(25,27). Ender et al. (28) have reported electric power densi-
Recently, there has been growing interest in developing
ties up to 15 W/cm2 (Fig. 9). low-
temperature Cs converters for terrestrial applications,
Due to the very high emitter temperatures, Cs–Ba con- such
as in topping cycles of fossil electric power plants or die-
verters suffer from the same limitations on their operation sel-
fired engines. In these applications, the source tempera-
lifetime (less than three years), as high-temperature cesium ture
could be several hundred degrees lower than in high-
converters discussed earlier, due to the rapid loss of emitter
temperature Cs converters (1600 K to 1800 K). A practical
material (28). The operation lifetime of Cs–Ba converters is
approach to enhance the performance of these lower-tempera-
also limited by finding an appropriate insulation material ture
thermionic converters is to use grooved electrodes. When
that is compatible with both Ba and Cs vapor at such high
grooved electrodes converters operate in the so-called hybrid
----------------------- Page 568-----------------------
e– setup
and procedures, reported results were inconclusive and
sometimes contradictory (39).

Ignited
Shimada (34,35) had tested a thermionic converter with
discharge
molybdenum (Mo) emitter that had 0.5 mm square longitudi-
nal
grooves, with a 1.0 mm separation. The nongrooved por-
Grooved – Unignited
emitter e discharge tion
of the emitter was coated with rhenium, which has a high
– bare
work function and hence lower cesiated work function
e
than
Mo. Shimada reported a 300% increase in electric power
Smooth
output, compared to a smooth electrode converter. Tskhakaya
collector
et
al. (36) tested a converter with a smooth zirconium oxide
collector and an Mo emitter that had grooves similar to that

e– of
Shimada’s, but the tops were coated with platinum. They
compared the electric power output with that for a converter

(a) (b) that
had a smooth platinum emitter and zirconium oxide col-
Figure 10. Low-temperature cesium converters with macrogrooved

lector.
emitter. (a) Hybrid operation in a TIC having a macrogrooved emit-
Although the performance of the former was better than
ter. (b) Macrogrooved emitter design (32). the
latter, Tskhakaya et al. (36) were unable to duplicate the
large
increase in electric power output reported by Shimada
(34,35). Their optical observations of plasma discharge con-

mode [Fig. 10(a)], the discharge breakdown, or ignition, in the firmed
that, while the grooves supplied ions to the rest of the
grooves occurs at low cesium pressure and provides gap
[Fig. 10(a)], the converter did operate in the hybrid mode
described by Shimada.
1. electrons for cesium ionization in the nongrooved por-
Tskhakaya et al. (36) results also showed that the opti-
tion of the emitter surface, which operates in the diffu- mum
gap size for their grooved emitter converter (1.6 mm)
sion mode was
higher than that for a smooth electrode converter (1.0
mm).
They have also tested a converter with a grooved emit-
2. Cesium ions, which diffuse from the grooved to the non-
ter
and a zirconium oxide collector that had the same grooves
grooved portion of the emitter surface, contributing to
as
the emitter. In these experiments, the collector’s grooves
the neutralization of the space charge, hence lowering
were
laid in the perpendicular direction to the emitter’s
the optimum cesium pressure (32,33) [Fig. 10(a)].
grooves.
Atamasov et al. (37) had tested a converter consisting of a

Several experiments have been performed using electrodes
single-crystal, smooth tungsten emitter and five microgrooved

with macro- and microgrooved electrodes for a wide range of
collectors in the same test device. The interelectrode gap was

electrode temperatures, cesium pressures, and interelectrode
0.35
mm wide. They applied an axial magnetic field to elimi-
gap sizes, with mixed results (34–38). In general, TI convert-
nate
edge effects and electron scattering between various col-
ers with grooved electrodes have performed better than those
lectors. The longitudinal grooves in all collectors were 0.1 mm

with smooth electrodes (23,24). The best performance has
wide
and 0.15 mm apart but had different depths of 0.1 mm
been reported for emitter temperatures of 1700 K to 1750 K,
to 0.5
mm. The measured peak electric power was 20% to 60%
showing an increase of 25% to 50% in the conversion effi-
higher
than with a smooth collector and occurred at 0.1 V to
ciency and the electric power density (Fig. 11). However, due
0.2 V
higher terminal voltage. Recently, Lee et al. (38) per-
to the lack of systematic studies using the same experimental
formed
a series of experiments with four electrode combina-
tions,
namely (1) smooth electrodes, (2) microgrooved emitter
and
smooth collector, (3) smooth emitter and microgrooved
1.6
collector, and (4) microgrooved electrodes. The emitter
grooves were 0.075 mm wide, 0.25 mm deep, and 0.075 mm

/ 1.5 Lee et al. (38) apart,
while the collector grooves were 0.1 mm wide, 0.25 mm
) ) Tskhakaya and Yarygin (1982)
e deep,
and 0.125 mm apart. The emitter was made of tungsten
c e Atamasov and Ivanova (1990)
a c
f a 1.4
chloride, the collector material was niobium, and the size of
r
f
u r Kulandin (1987)
s u the
interelectrode gap varied from 0.125 mm to 0.5 mm. Their
s
d
e h 1.3
results showed a 0.1 V increase in the output voltage of con-
p t
o
o verter
(1), a small increase in the electric power output and a
l o
e m
v s 1.2 50%
increase in the emission current density of converter (3),
e (
d r
( e and no
discernible difference in the performance of converter
r w
e o 1.1 (4)
compared to the converter (1).
w
o P
Recently, El-Genk and Luke (32) performed experiments
P
1.0 to
characterize the performance of developed electrodes TI
converters [Fig. 10(b)], as a function of the cesium pressure,

0.9 (PCs #
10 Pa to 240 Pa). They tested three configurations: (1)
1300 1500 1700 1900 2100 2300 2500 2700 2900
smooth
electrodes converter, (2) grooved emitter converter,
Emitter temperature, TE (K)
and
(3) grooved collector converter, at the same cesium pres-
Figure 11. Comparison of electric power of grooved and smooth elec- sures
and interelectrode gap size (1.5 mm). Results showed
trode converters (39). that
the grooved emitter converter gave higher electric power
----------------------- Page 569-----------------------
output than the smooth electrode converter at low Cs pres- tion of

Carnot, a Cs–Ba or a high-temperature cesium con-
sure, similar to previous results of Atamasov et al. (37) and verter,
in which the collector and emitter temperatures are
Lee et al. (38). Furthermore, the peak electric power output 1200 K
and 2500 K, respectively, would have to deliver a con-
for the former (0.4 W/cm2) occurred at a lower cesium pres- version
efficiency of 12.7%. Similarly, a conventional thermi-
sure (120 Pa) than the peak electric power for the latter (0.31 onic
converter operating at an emitter temperature of 1900 K
W/cm2 at 150 Pa). and a
collector temperature of 800 K would have to deliver a
This reduction in the cesium pressure at the peak electric

conversion efficiency of 14.2%, to realize the same fraction
power of the grooved emitter converter was apparently due of
Carnot (24.5%).
to the space-charge neutralization at the emitter surface by The
results of El-Genk and Momozaki (33) have also
diffusing ions from the grooves to the tops portion of emitter shown
that lowering the collector temperature to 1073 K sig-
surface [Fig. 10(b)]. The electric power output of the grooved
nificantly increased the electric power density and conversion
emitter converter was more than a factor of 2 higher than
efficiency. At emitter and collector temperatures of 1573 K
that of the smooth electrode converter at PCs # 57 Pa, decreas- and
1073 K, respectively, the measured peak electric power
ing to only 5% at 240 Pa (32). The peak electric power for the density
and peak conversion efficiency were more than 20%
grooved emitter converter (0.4 W/cm2) occurred at PCs # 120 and
12% higher, respectively, than those for identical con-
Pa, while that of the smooth electrodes converter (0.325 verter

with a larger gap of 1.5 mm. Results suggested that
W/cm2) occurred at a higher PCs # 150 Pa. lowering
the collector temperature to about 800 K could fur-
They also reported that at cesium pressure less than 110 ther
enhance the performance of grooved electrodes convert-
Pa, the electric power output of the grooved collector con- ers, in
terms of both higher efficiency and electric power den-
verter was also higher than that of the smooth electrode con- sity.
Further research on grooved electrode converters could
verter, but lower than that for the grooved emitter converter be
rewarded handsomely by opening a broad range of poten-
(32). At higher cesium pressure, the electric power of the tial
application for commercial electric power generation.
grooved collector converter was the highest, and peaked at
the same cesium pressure as the smooth electrodes converter
(150 Pa). The electric power output of the grooved collector Effect
of Inert Gases on the Performance of Cesium TICs
converter was 1.4 times that of the smooth electrodes con-
There
was a lot of interest in the 1960s and 1970s in investi-
verter, which is the same as the increase in the grooved col- gating
the effect of inert gases, such as helium and xenon,
lector surface area. Since the use of a grooved collector does on the
performance of Cs converters. Inert gases affect the
not affect space-charge neutralization at the emitter surface,
transport and distribution of Cs ions in the interelectrode

the electric power output for the grooved collector converter
gap.
While chemically inert, they have a low scattering cross
peaked at the same cesium pressure as the smooth electrode
section
for slow electrons, and their high ionization potential
converter.
impedes
surface ionization as well as ion diffusion, but not
For emitter and collector temperatures of 1573 K and 1073
electron
diffusion. Also, due to the very large temperature
K, respectively, the conversion efficiency of the grooved collec-
gradient
in the gap ( 2000 K/mm), the relative diffusion of
tor converter was more than 70% and 20% higher than that
the
inert gases and Cs ions could alter the Cs ion density at
for the smooth electrodes converter at PCs of 60 Pa and 240
the
emitter by forcing them to diffuse into the colder region
Pa, respectively (32). The conversion efficiency of the grooved
of the
gap, hence altering the saturation emission current.
emitter converter was almost twice that of the smooth elec-
The
results of earlier work have been contradictory, with
trode converter at PCs # 60 Pa, but the same at 240 Pa. The
some
showing an increase in the electron current (or a de-
efficiency of the smooth electrode converter was significantly
lower than those of the other two converters at low cesium crease
in internal voltage drop) and others showing the oppo-
site
effect.
pressure, but approached that of the converter with a grooved
emitter and a grooved collector as PCs increased. The peak
Kaplan and Merzenich (40) and Bekmukhambetov et al.
conversion efficiency of 5.8% for the converter with a grooved (41)
reported that the addition of xenon or krypton (up to 60
collector occurred at PCs # 140 Pa, at which the conversion
torr) to cesium in the interelectrode gap has increased the
efficiencies of the grooved emitter and smooth electrodes con- electron
current by as much as 15% to 20%. They attributed
verters were 5.16% and 4.9%, respectively. The peak conver- this
increase in the electron current to a reduction in the in-
sion efficiency of the grooved emitter converter (5.22%, which ternal
voltage drop required for volume ionization. Kondrat’ev
represents 16.4% of Carnot) occurred at a lower cesium preset al.
(42) have shown that increasing the xenon pressure in
sure (110 Pa) than the smooth electrode and the grooved col- the
interelectrode gap from 3 torr to 10 torr increased electron
lector converters (32). current;
however, a further increase in xenon pressure caused
El-Genk and Momozaki (1998) have continued the work of the
electron current to decrease. They attributed the results
El-Genk and Luke (32), but at higher electrode temperatures to two
competing mechanisms: (1) higher ion density, since
and PCs and smaller interelectrode gap size (0.5 mm). They xenon
tends to impede the diffusion of cesium ions out of the
varied the emitter temperature from 1573 K to 1773 K and
interelectrode gap; and (2) increased electron scattering due
PCs from 50 Pa to 500 Pa. The collector temperatures were to the
presence of high-pressure xenon.
1073 K and 1173 K. A peak electric power density of 1.2
Kaplan and Merzenich (40) reported a decrease in the elec-
W/cm2 and a peak conversion efficiency of 8.8% were tron
current when neon was added to the cesium in the inter-
achieved at a relatively low emitter temperature of 1773 K
electrode gap. Rufeh and Kitrilakis (43) reported a systematic
and a high collector temperature of 1173 K. This peak conver-
reduction in the electron current (or increase in internal volt-
sion efficiency represents 24.5% of Carnot, at the same age
drop) as the pressure of argon gas in the interelectrode
emitter and collector temperatures. To obtain the same frac- gap was
increased from 0 torr to 100 torr. Gverdtsiteli et al.
----------------------- Page 570-----------------------
(44) showed a significant increase in electron current, how- in a

decrease in the internal voltage drop (or increase in the
ever, in the presence of argon in the interelectrode gap. electron
current). These results suggest that low electrode
Rufeh and Lieb (45) performed a series of experiments us-
temperature Cs converters, such as those with grooved elec-
ing a thermionic converter that had a rhenium emitter and a trodes,
could benefit greatly from the addition of inert gases,
molybdenum collector, and they varied the interelectrode gap such as
xenon. Since the optimum cesium pressure in these
from 0 mm to 2.5 mm. The emitter temperature was 1800 K,
converters would be lower than in conventional and high-tem-
the cesium pressure was about 1 torr, and the inert gas pres- perature
Cs converters, the benefit of adding inert gases in
sure was varied in small steps from 0 torr to 200 torr. Unlike the
interelectrode gap could be significant, particularly at
some of the earlier work, they reported a consistent decrease low-
collector temperatures (600 K to 1000 K).
in the electron current (or increase in the internal voltage
drop) when inert gases (xenon, krypton, or argon) were intro- Microgap
Converters
duced into the interelectrode gap. The decrease in the elec-
Quasi-
vacuum and Knudsen mode converters have long been
tron current increased as the pressure of the inert gas in-
perceived as of very little promise because of the difficulty

creased. Rufeh and Lieb (45) attributed the reported increase
with
maintaining an interelectrode gap size on the order of
in electron current by earlier investigators to the presence of
few
microns. However, recent advancements in microelectron-
impurities, such as oxygen, in the inert gases. At very low
#2 ics,
precise machining, monocrystalline material fabrication,
cesium pressure, on the order of 10 torr, an increase in the
and
other manufacturing technologies have made it possible
output voltage or electron current was reported when inert
to
develop microgap converters. The operation of laboratory-
gases were introduced into the interelectrode gap (44,46).
scale
microgap converters, with 4 #m to 10 #m interelectrode
Most recently, Tschersich and Niekizch (47) have investi-
gap, has
been demonstrated successfully at Space Power Inc.
gated the effect of xenon on the voltage drop in the interelec-
by
Fitzpatrick et al. (47a) and at the Russian Institute Luch
trode gap of a cesium converter. They performed experiments
by
Nikolaev et al. (1996). The test converters had monocrys-
using a variable-spacing converter that had a polycrystalline
talline
electrodes to achieve high thermionic performance and
tantalum emitter and a polycrystalline niobium collector.
provide
dimensional stability. During operation, the elec-
They used a method similar to that of Rufeh and Lieb (45) to
trodes
are separated by a microsized gap due to the thermal
ensure that oxygen impurities were not introduced with the
expansion of ceramic spacers recessed in the collector (Fig.

xenon gas. The values of PCsd, where d is the size of the inter-
#3 #2 12). An
applied helium gas pressure on the collector assembly
electrode gap, were 4 10 torr-cm to 5 10 torr-cm (PCs
provided
precise control of the interelectrode gap size and
from 0.13 torr to 1.84 torr) and the pressure of xenon was
minimized parasitic heat losses.

increased from 0 torr to 80 torr. Their results showed a re-
Stable operation of these microgap converters has been

markable reduction in the internal voltage drop when the
#2
demonstrated for hundreds of hours. Ways enhance the per-
value of PCsd, was below the optimum value ( 5 10 torr-
formance
of these converters in the future include the fol-
cm). For example, at PCsd of 2.6 10#3 torr-mm (d # 0.2 mm
lowing:
and PCs # 0.13 torr), increasing the xenon pressure from 0
torr to 50 torr decreased the internal voltage drop from 1.38
1.
Using monocrystalline emitters, with built-in heat
V to 0.8 V (a 0.58 V decrease). At the same cesium pressure,
pipes, to maintain uniform electrode temperature, thus

when the gap size was increased five times to 1.0 mm, the
preventing structural distortion

internal voltage drop decreased from 0.715 V to a minimum
of 0.55 V (a decrease of 0.17 V) as xenon pressure increased 2.
Implementing active and precise control of the gap size
from 0 torr to 20 torr).
using piezoelectric transducers (47b)
Near the optimum PCsd, the change in internal voltage 3.
Using highly reflective coatings for the collector surface,
drop as well as its minimum value, due to the presence of
maintaining high emitter temperature and hence high
xenon, was relatively small and negligible, respectively. Be-
electron emission
yond the optimum PCsd, there was no effect on the internal
voltage drop due to the addition of xenon. The lowest voltage
Because the space charge in a Knudsen mode thermionic
drop measured in the absence of xenon was 0.55 V at a PCsd
converter is at least partially compensated for by the Cs ions
of 2 10#2 torr. The results of Tschersich and Niekizch (47) generated
at the emitter surface, the gap size, emitter tem-
have confirmed earlier results that introducing inert gasses
perature, and emission current can be significantly increased
increases the electron current (or reduce the internal voltage with
very little loss of the electron energy. Typical Knudsen
drop) in Cs converters. These results, however, are in dis- mode
microgap converters have an interelectrode gap that is
agreement with those of Rufeh and Lieb (45). about
20 #m and are operated at emitter temperatures of
In summary, it may be argued that the introduction of in-
ert gases in cesium converters would improve their perfor-
mance (i.e., increase the electron current or reduce the inter-
Heat rejection structure
nal voltage drop) at low cesium pressure, due to the retention
He gap
of Cs ions by inert gases and negligible scattering losses. At
high cesium pressure, there should be no effect if scattering
Spacer
losses are compensated for by the retention of Cs ions. Hence,
Collector
the introduction of inert gases would improve the perfor-
mance of Cs converters due to two effects: (1) an increase in
Emitter
the plasma density and enhanced plasma distribution in the
gap, and (2) a change in the potential distribution, resulting
Figure 12. A schematic of a microgap converter.
----------------------- Page 571-----------------------
35
Quasi-vacuum
mode (d = 10 m)(Fitzpatrick et al. 1993 & 1998)µ
µ
Quasi-vacuum
mode (d = 6 m)(Nikolaev et al. 1996)
µ
Knudsen mode (d
= 20 m, eff = 0.05)(Nikolaev et al. 1996)
30
µ
Quasi-vacuum
mode (d = 4 m)(Nikolaev et al. 1996)
µ
Knudsen mode (d
= 20 m, eff = 0.17)(Nikolaev et al. 1996)
)
% 25
(
y
c
n
e
i
c
i
f
f 20
e
n
o
i
s
r
e
v
n
o 15
C
10
Figure 13. Calculated efficiencies of 5

1200 1400
1600 1800 2000
quasi-vacuum and Knudsen mode micro-
gap converters.
Emitter temperature, TE (K)
1700 K to 1900 K. At the higher emitter temperatures, how- The

collector material should have a cesiated work func-
ever, radiation heat transfer becomes significant, reducing tion
that is as low as possible at the operating collector tem-
the fraction used for electron emission and resulting in a
perature and at the cesium vapor pressure in the interelec-
lower efficiency and electric power density. As delineated in trode
gap to maximize the load voltage. Low emissivity (or
Fig. 13, using a thin, highly reflective (# # 0.05) copper or high
reflectivity) collector is also important, since it directly
silver coating of the collector surface is expected to increase affects
the emitter temperature through the thermal radia-
the converter efficiency into the 25% to 30% range. tion
transport between electrodes. Because the collector oper-
ates at
a relatively low temperature (800 K to 900 K), mate-
rial
loss is of no consequence, but it is important that the
ENGINEERING ASPECTS OF THERMIONIC CONVERSION
collector material remain chemically stable in cesium vapor.

The
collector material should have a high electric conductiv-
The electrode and structural and insulator materials used in
ity,
to minimize joule losses, and high thermal conductivity,
thermionic converters need to satisfy strict and sometimes to
ensure uniform electrode temperature. The electrodes,
contradictory requirements. For example, the emitter mate-
however, should be thermally insulated at their edges to min-
rial should possess the following properties: imize
thermal losses. In addition to the emitter materials
mentioned earlier, other low work function materials, such as

1. High bare work function, to provide low cesiated work Nb,
LaB6, Ba, or Zr oxides, Ni, and even stainless steel, could
function and/or efficient surface ionization be used
for the collector.
2. Stable mechanical and surface properties, to minimize
Other important structural components of a thermionic
the loss rate of emitter material when operating at high
converter are the interelectrode gap spacers and electric insu-
temperatures and in cesium vapor lators,
which are usually manufactured of ceramic material
3. Low emissivity, to minimize thermal energy transport that
should satisfy the following requirements:
to the collector by thermal radiation, increasing electron
emission and achieving high conversion efficiency 1.
Chemical and mechanical stability at operation temper-
atures and in a cesium environment

4. High electric conductivity but low thermal conductivity,
to minimize joule losses and axial heat conduction, re- 2.
Linear thermal expansion coefficients similar to other
materials used in the converter

spectively, in the electrodes
3.
Low electric and thermal conductivities
5. Structural strength at high temperatures, which favors
using monocrystalline instead of polycrystalline emitter 4.
Low release rate of volatile and gases during operation,
material. The former also has a higher bare work func-
to avoid contaminating the cesium vapor in the inter-
tion than the latter.
electrode gap
5.
High fracture strength in order to withstand thermal
The high-temperature emitter materials that satisfy these re-
cycling and vibrations
quirements are therefore limited to tungsten (W), molybde- 6.
Good chemical compatibility with emitter and collector
num (Mo), tantalum (Ta), and rhenium (Rh).
materials
----------------------- Page 572-----------------------
Table 3. Causes of Failure and Performance Degradation of TFEs
Open Short
Cause
Circuit Circuit Degradation
Emitter:
Change in position, emission coating
Separation
Damage or deformation
Loss of vacuum tightness, release of fission products
Contamination and mass transfer
Interelectrode gap:
Lack of cesium
Penetration of contaminants (from outside or from
nuclear fuel)
Increase in heat losses
Spaces:
Destruction or damage
Decrease of contact thermal resistance
Deposition of conducting layers
Collector assembly:
Deposition of contaminants and reaction products
Change of cooling conditions
Increase in thermal resistance deformation
Insulation failure
Emitter-collector commutation:
Destruction
Cross-section reduction
The most commonly used ceramic materials in TI converters
Characterization
are oxides of Al, Be, Zr, Mg, Th, and Sc.
Some internal, unmeasurable characteristics of a thermionic
converter or a TFE, which are extremely important for as-

Lifetime and Performance Degradation
sessing any performance degradation during operation and
Cylindrical (TFE) and planar thermionic converters may ex- for

on-line performance optimization, can be inferred from
perience gradual degradation in performance due to changes
measurable parameters. For example, considering any two
in the operation and material conditions of the converter or
points (1 and 2) on two static (constant thermal input power
at the system level. Internal causes of performance degrada- to
the emitter) voltampere characteristics (Fig. 14), the differ-
tion are (1) changes in the work function and the effective
ence in the total heat removal from the emitter at the same
emissivity of the electrodes; (2) presence of gaseous contami-
nants in the interelectrode gap; (3) mass transport of emitter
material to the collector surface; and (4) changes in the elec-
trode temperatures. Examples of system level causes of per-
formance change are (1) improper operation or a failure of the
RA
cesium supply system and the introduction of gaseous con-
calc
j A (RA , ϕ E, ϕ C)
taminants or ambient air into the interelectrode gap; (2)
j exp
changes in heat rejection, increasing the coolant and the col-
A A
lector temperatures; and (3) a loss of vacuum tightness. )

A
(
R
Generally, there are two types of a TFE or thermionic con- y
B
t
j 1 2
verter failures: open and short circuit. In an open circuit fail- i
1 exp
s
j
n
T T B calc
ure, the internal electrical resistance of the converter rapidly e
1 2 j (R B, ϕ E, ϕ C)
d
B
increases as the voltage and current on the external load ap- t
B
n
e
q
proach zero. For TFEs connected in series to a constant load, r
2tot
r
u
a short circuit in one of the TFEs would cause a reduction in C
q 1tot
the load voltage and current. Also, the slope of the static (con-
stant power) voltampere characteristics will increase. While
both open circuit and short circuit failures in a TFE lead to
abrupt changes in the voltampere characteristics, gradual
vL1 vL2
changes in the TFE performance cause a slow change in the
0
Load voltage, VL (V)

voltampere characteristics. Table 3 provides a list of various
causes of failure and performance change in electrically and
Figure 14. A schematic for determining the effective emissivity and
nuclear fission heated TFEs (partially adapted from Ref. 48).
work functions of the electrodes.
----------------------- Page 573-----------------------
current density, j , can be written as

ment. In the former, the thermionic fuel elements (for exam-
1
ple, see Fig. 6) are arranged typically in a compact, triangular

q − q = (q − q ) + j (V − V ) + 2k (T − T )j
lattice. The nuclear fuel pellets are staked inside the emitter
2tot 1tot rad2 rad1 l L2 L1 2 1 l
e
tubes of the TFEs. The heat generated from nuclear fission is
+ (qCs2 − qCs1) + (qstr2 − qstr1) − 0.5(qJ2 − qJ1 )
then transferred by radial conduction through the fuel pellets
(22)
(typically UO2, 90 wt% enriched), to the emitter tubes of the
TFEs (typically at 1800 K to 1900 K). The heat transferred to

In this equation, the subscripts rad, Cs, str, and J refer to the
collector assembly (Fig. 6) is then removed by circulating
radiation heat transfer, conduction heat transfer in the Cs- a
liquid-metal coolant (typically NaK or Na).
filled interelectrode gap, end heat losses in the electrodes, and
In the ex-core arrangements, the TICs are placed outside
joule heating, respectively. When the total input power, q2tot, the
nuclear reactor core, and heat is transported, typically
is slightly higher than q1tot, the collector temperatures at
using Na heat pipes, to the inside of the emitter tubes. The
points 1 and 2 are approximately equal and the effective
heat from the collector assembly of the TFEs is removed ei-
emissivity of electrodes, #, can readily be determined. The two
ther by a circulating liquid-metal coolant or by heat pipes and
power levels, q1tot and q2tot, should be close enough, however,
rejected to space by thermal radiation through the radiator
so that the corresponding collector temperatures can be as-
assembly. The heat rejection radiator of a thermionic space
sumed almost the same. Therefore, the calculated effective
nuclear reactor system is typically the most voluminous and
emissivity corresponds to the average emitter temperature at
massive component of the system. The ex-core arrangement
points 1 and 2 in Fig. 14. Equation (22) can thus be used to has
the advantage of a compact nuclear reactor core, which
calculate the effective emissivity of the interelectrode gap at
means smaller reactor mass and smaller size and mass of the
different combinations of T1 and T2 and determine its depen-
shadow radiation shield. The disadvantage, compared to the
dence on the average emitter temperature, for system perfor- in-
core arrangements, however, is that the reactor coolant has
mance analyses during testing and operation lifetime. To to
operate at very high temperatures, requiring using high-
apply Eq. (22), however, the contributions of heat conduction
temperature refractory alloys as the structure material. On
through the interelectrode gap and end losses through the the
other hand, in the in-core arrangement the highest tem-
electrodes should be calculated using, for example, Eqs. (16)
perature in the system is limited to the fuel stack and the
and (18a), respectively.
emitter tube. The rest of the system operates at or below the
The effective work functions of the electrodes can then be
collector temperature, typically 800 K to 1000 K, hence
estimated on the basis of the experimental voltampere char-
allowing using stainless steel as the structure material for
acteristics, assuming linear characteristics, which is typically the
rest of the system. More details of the design of thermi-
the case in ignited mode discharge, within a certain range of
onic space nuclear power systems could be found elsewhere
load voltages, using the reverse solution of Eqs. (13) and (22).
(49–61).
In this approach, the electrode temperatures, the cesiated
work functions, and the gap size are related to the measured
load current and voltage. Assuming linear voltampere charac-
APPLICATION TO TERRESTRIAL SPACE

teristics, load resistance (R and R ), and corresponding cur-
A B
ELECTRIC POWER GENERATION
rent densities (j A and j B), the cesiated work functions are
found from the following equalities:
The
interest in using
thermionic conversion in a number of
j exp = j calc (RA , ϕE , ϕC ) (23a)
commercial and military terrestrial applications began as
A
early as the 1960s (62,63). The high emitter temperature of
j exp = j calc (RB , ϕE , ϕC ) (23b)
thermionic converters makes them attractive for use in top-
B
ping cycles of terrestrial fossil fuel power plants, where high

In these equations, j exp and j calc are the experimental and cal-
temperatures exist in combustion chambers (64). Thermionic
culated current densities (at points A and B in Fig. 14), re-

conversion had been considered for a topping cycle for various
spectively. Solving Eqs. (23a) and (23b) for the different elec-
types of gas and steam turbine power plants (65–67) and co-
trode temperatures and different cesium pressures,
generation plants (68–70). The idea of implementing a therm-
dependencies of the electrodes’ effective work functions on the
ionic topping cycle is simple: Hot gases in the combustion
emitter and collector temperature as well as the vapor pres-
chamber heat the emitter of thermionic converters up to
sure can be obtained based on actual, on-line measurements
1600 K to 2000 K to achieve satisfactory thermionic perfor-
of the voltampere characteristics. The determined values can
mance. The collector is cooled ( 800 K to 900 K) with forced
then be compared with initial values of the electrodes’ work air
flow before entering the combustion chamber (Fig. 15).
functions to determine any changes as a function of opera-
The work performed in the early 1980s under the Depart-
tion lifetime.
ment of Energy’s (DOE) thermionic technology program dem-
onstrated that using a thermionic topping cycle in a gas tur-

APPLICATION TO SPACE ELECTRIC POWER GENERATION
bine power plant could increase the plant’s overall efficiency
and
the electric power by several percentage points (67,72).
Historically, the interest in thermionic conversion technology In
1984, when this program reached the demonstration stage,
has been driven by its potential use in nuclear reactor space the
DOE suggested that the US industry take over the fi-
power systems. In these systems, the nuclear reactor serves
nancing of the terrestrial thermionics, but industry was not
as the heat source and could be coupled with the thermionic yet
ready to invest in this type of advanced electric power
conversion elements, either in an in-core or ex-core arrange-
generation.
----------------------- Page 574-----------------------
Fuel
BIBLIOGRAPHY
Combustor
1. C. Herringand M. H. Nichols, Thermionic emission, Rev. Mod.

Topping
Phys., 21: 185–270, 1949.
cycle Emitter
2. G. N. Hastopoulos, The thermoelectron engine, Sc.D. disserta-
tion, MIT, Cambridge, MA, 1956.

Load
Load
3. F. G. Baksht et al., Thermionic Converters and Low-Temperature
Collector
Turbine
Plasma, Rep. No. DOE-tr-1, Washington, DC: U.S. Dept. Energy,
Tech. Inf. Center, 1978.
4. J. M. Houston, Theoretical performance of the thermionic energy

Generator
conversion, J. Appl. Phys., 30 (4): 1959.
Air heater
5. V. C. Wilson, Conversion of heat to electricity by thermionic emis-
sion, J. Appl. Phys., 30: 475, 1959.

Compressor To bottoming
Air inlet cycle
6. K. P. Luke, Thermionic work function of a single crystal, Sc.D.
thesis, MIT, Cambridge, MA, 1964.
7. V. S. Fomenko, in G. V. Samsonov (ed. and trans.), Handbook of

Figure 15. A schematic of thermionic converter being used in the
Thermionic Properties, New York: Plenum, 1966.

topping cycle of a gas turbine, electric power generation plant (71).
8. D. Steiner and E. P. Gystopoulos, An equation for the prediction
of the bare work functions, Proc. 27th Annu. Conf. Phys. Electron.,
Recent advances in natural-gas-fired gas turbines, such as
Cambridge, MA, 1967, pp. 160–168.
integrated gasification combined cycle (IGCC) and pressur-
9. G. N. Hastopoulos and F. Rufeh, Thermodynamic correlation of
ized fluidized bed combustion (PFBC) technologies, have
work function, Energy Convers., 10: 135, 1970.
made using a thermionic topping cycle that employs state-of-
10. G. N. Hastopoulos and E. P. Gystopoulos, On the relation be-
the-art or near-term advanced TI technology even less attrac-
tween work function and other thermophysical properties, Proc.
tive. Recent studies of using thermionic topping cycles have
Thermionic Convers. Spec. Conf., Miami Beach, FL, 1970, pp.
shown that, even using advanced thermionic technology (for
203–209.
example, microgap converters), the efficiency of combined cy-
11. T. Alleau and M. Bacal, Influence of oxygen adsorption and oxy-
cle power plants that employ advanced GTs, such as Westing-
gen-cesium coadsorption on (100) tungsten work function, Proc.
house 501ATS, can be increased only by less than 1.5 percent-
Thermionic Convers. Spec. Conf., Miami Beach, FL, 1970, pp.
434–440.
age points (71). Such a small performance gain does not
justify using a thermionic topping cycle, considering the tech-
12. J.-L. Denisot, Electron Emission Experimental Study of a (100)
Tungsten Single Crystal Covered wtih Cesium and Oxygen, CEA-
nology development required and the risks involved. As far as
`
R-4508, France; Commissariat a l’Energie Atomique, 1973.

using a TI topping cycle in coal-fired steam power plants, the
13. N. S. Rasor, Thermionic energy conversion, Appl. At. Collission

relatively low efficiency of thermionic converters makes these
Phys., 5: 170–220, 1982.

far less competitive with IGCC and PFBC gas turbine electric
power generation plants.
14. M. S. El-Genk, H. Xue, and C. Murray, Transient and load-follow-
ing characteristics of a fully integrated single-cell thermionic fuel

One of the most recent advancements in the field of terres-
element, J. Nucl. Technol., 102: 145–166, 1993.

trial thermionics has been made by the Dutch-Russian joint
15. W. R. Wilkins and E. P. Gyftopoulos, Transport phenomena in

venture ECS (73,74). A number of natural-gas-fired cogenera-
low energy plasmas, J. Appl. Phys., 37: 3533–3540, 1966.

tion devices, ranging from a small domestic cogeneration unit
16. D. W. Culp, Principle of Energy Conversion, New York: McGraw-

to a large industrial burner, have been designed. The program
Hill, 1997.
culminated in manufacturing and testing a thermionic mod-
17. S. Kitrilakis and M. Meeker, Experimental determination of the

ule heated by a natural-gas-fired recuperative burner. The
heat conduction of cesium gas, Adv. Energy Convers., 3: 59–68,

ECS thermionic converter has a CVD silicon carbide protec-
1963.
tive coating, but the collector is water cooled. The emitter
2
18. G. N. Hastopoulos and E. P. Gyftopoulos, Thermionic Energy Con-
area is 38 cm and the interelectrode gap is 0.4 mm to 0.5 mm
version, Vol. 1, Cambridge, MA: MIT Press, 1973.

wide. This converter produced a maximum electric power out-
2
19. V. A. Korykin, V. P. Obrezumov, and V. I. Vybyvanets, Variation
put of 69 W ( 1.8 W/cm ) while operating at an emitter tem-
of emission-adsorption properties in an operating TEC collector,

perature (T ) of only 1623 K. For an assembly consisting of
E
Proc. 24th IECEC, Vol. 2, 1989, pp. 1161–1164.
12 cylindrical thermionic converters having emitter surface
2
20. C. B. Geller et al., Final Report of the High Efficiency Thermionics
area of 38 W/cm each and operating at TE # 1600 K, an elec-
(HET-IV) and Converter Advancement (CAP) Programs, Rep. No.
2
trical power output of 625 W ( 1.37 W/cm ) and overall con-
WAPD-T-3106, West Mifflin, PA: Bettis Atomic Power Labora-
version efficiency of 8.4% were projected (75).
tory, Westinghouse Electric Corp., 1996.
For more than a decade, Space Power Inc. (SPI), in cooper-
21. J. L. Desplat, W (100) work function change during adsorption of
ation with Scientific-Industrial Association ‘‘Luch’’ of Russia,
oxygen, cesium, and oxygen-cesium co-adsorption, Proc. 3rd Int.
has been developing combustion-heated microgap thermionic
Conf. Thermionic Electr. Power Generation, Juelich, Germany,
converters with active gap control for cogeneration applica-
1972, pp. 1397–1407.
tions. A conversion efficiency of about 13% at an emitter tem-
22. B. Gunther et al., Summary Report on Oxygen Additives, Rep.
perature of 1300 K is projected. Other potential terrestrial
TEF-3056-2. Thermo Electron Corp., 1974.
applications of thermionic conversion include automotive
23. D. V. Paramonov and M. S. El-Genk, An analysis of emitter mate-
waste heat recovery (76) and high-temperature solar power
rial transport in thermionic converters, AIP Conf. Proc., 361 (3):
systems in remote areas.
1137–1148, 1996.
----------------------- Page 575-----------------------
24. D. V. Paramonov and M. S. El-Genk, Effect of oxygen on the oper-

47. K. G. Tschersich and E. A. Niekizch, Xenon induced changes of
ation of a single-cell thermionic fuel element, J. Nucl. Mater., 256:
the I-V curves of a thermionic converter, Proc. IECEC, 1975, pp.
218–228, 1998.
376–381.
25. V. I. Babanin et al., Optimization of a Knudsen Cs-Ba thermionic

47a. G. Fitzpatrick et al., Demonstration of close-spaced thermionic
converter, Sov. Phys.—Tech. Phys., 23: 444–451, 1978.
converters, Proc. 28th IECEC, American Chemical Society, 1.573–
26. Yu. K. Szhenov, A. F. Piskunov, and V. A. Osadchii, Teplofiz. Vys.
1.580, 1993.
Temp., 10: 738, 1972.

47b. G. Fitzpatrick et al., Close-spaced thermionic converter with
small gap, AIP Conf. Proc. 361, M. S. El-Genk (ed.), 1299–1304,

27. V. I. Babanin et al., Potential distribution in a two-gas thermi-
1996.
onic converter at current saturation, Sov. Phys.—Tech. Phys., 17:
1695–1700, 1973.
48. V. V. Sinyavskii et al., Design and Testing of Thermionic Fuel
Elements, Moscow: Atomizolet, 1981 (in Russian).

28. A. Y. Ender et al., Ultra-high temperature thermionic system for
space solar power application, AIP Conf. Proc., 301: 861–867,
49. N. Rasor and R. C. Howard, Nuclear thermionic space power sys-
1994.
tems, Proc. 18th Annu. IAS Propulsion Meet., Inst. Aerosp. Sci.,
Cleveland, OH, 1963.

29. V. M. Gun’ko and P. V. Smirnova, High temperature oxidation of
tungsten in presence of cesium vapor, J. Phys. Chem., 57: 2581–
50. W. J. Leovic and C. Z. Kamien, Solar thermionic power systems
2583, 1983.
development, Prog. Astronaut. Aeronaut., 6: 1966.
30. V. A. Korykin and V. P. Oberzumov, Dynamics of changes in

51. C. D. Sawyer, P. R. Hill, and D. R. Wilkins, Multimegawatt
thermionic energy converter collector properties under conditions
thermionic reactor systems for space applications, Proc. 2nd Int.
of mass transfer in the interelectrode gap, Proc. 25th IECEC,
Conf. Thermionic Electr. Power Generation, Stresa, Italy, 1968,
1990, pp. 332–339.
pp. 171–184.
31. D. V. Paramonov and M. S. El-Genk, An analysis of Ya-21 therm-

52. J. J. Ward, R. Breitwieser, and R. M. Williams, Conceptual de-
ionic fuel elements test results, AIP Conf. Proc., 361: 1395–
sign of a 150 Kwe out-of-core nuclear thermionic converter sys-
1400, 1996.
tem, Proc. 9th Annu. Thermionic Conf. Spec. Meet., Miami Beach,
FL, 1970, pp. 179–184.

32. M. S. El-Genk and J. Luke, Performance comparison of thermi-
53. Gulf General Atomic Co., Development of a Thermionic Space

onic converters with smooth and micro-grooved electrodes, J. En-
Power System, Final Summ. Rep. (Gulf GA-A126o8), San Diego,

ergy Convers. Manage., 39: 1–13, 1998.
CA: Gulf Atomics, 1973.

33. M. S. El-Genk and Y. Momozaki, Performance evaluation of a
54. G. M. Gryaznov et al., Thermionic reactors-converters for space

thermionic converter with a macro-grooved emitter and a smooth
NPS, At. Energ., 66 (6): 374, 1989.

collector, AIP Conf. Proc., 420: 308–316, 1998.
55. N. G. Gunther, Characteristics of the Soviet TOPAZ II Space Power

34. K. Shimada, Low arc drop hybrid mode thermionic converter,
Systems, Rep. SPI-52-1, San Jose, CA: Space Power Inc., 1990.
Proc. 12th Intersoc. Energy Convers. Eng. Conf. (IECEC), Soc. Au-
tomot. Eng., 1977, pp. 1568–1574.
56. J. G. Jacox et al., Small ex-core heat pipe thermionic reactor con-
cept (SEHPTR). Proc. 25th IECEC, Monterey, CA, 1990.

35. K. Shimada, Recent progress in hybrid mode thermionic con-
verter development, Proc. 14th Intersoc. Energy Convers. Eng.
57. J. C. Mills et al., A fast spectrum heat pipe cooled thermionic
Conf. (IECEC), 1979, pp. 1890–1893.
power system, AIP Conf. Proc., 246: 504–509, 1992.
36. V. K. Tskhakaya, L. P. Chechelashvili, and V. I. Yarygin, Hybrid

58. T. H. Van Hagan and B. R. Lewis, TFE fast driver reactor system
operating mode of a thermionic converter with a grooved collec-
for low-power applications, AIP Conf. Proc. 246: 498–503, 1992.
tor, Sov. Phys.—Tech. Phys., 28: 869–870, 1983.

59. H. S. Rhee et al., Space-R thermionic space nuclear power system
37. V. D. Atamasov et al., Experimental study of cesium arc thermi-
with single cell incore thermionic fuel element, AIP Conf. Proc.,
onic converters with multicavity anodes, Sov. Phys.—Tech. Phys.,
246: 120–129, 1992.
29: 42–45, 1984.

60. W. R. Determan and T. H. Van Hagan, Space power reactor in-
core thermionic multicell evolutionary (S-PRIME) design, AIP

38. C. Lee, D. Lieb, and G. Miskolczy, Ceramic composites for thermi-
Conf. Proc., 271: 1235–1239, 1993.

onic fuel elements, Proc. 25th IECEC, 1990, pp. 294–299.
61. M. S. El-Genk, A Critical Review of Space Nuclear Power and

39. D. V. Paramonov and M. S. El-Genk, A review of cesium thermi-
Propulsion 1984–1993, New York: American Institute of Phys-

onic converters with developed emitter surfaces, J. Energy Con-
ics, 1994.
vers. Manage., 38: 533–549, 1996.
62. L. J. Lazaridis, P. G. Pantazelos, and I. Shai, Design of a Gas-

40. C. Kaplan and J. B. Merzenich, Proc. Thermionic Convers. Spec.
Fired Thermionic Power Supply for Domestic Furnaces, Paper No.

Conf., Cleveland, OH, 1966, p. 333.
66-WA/ENER-3, New York: ASME, 1966.

41. E. S. Bekmukhambetov et al., Cesium thermionic converter with
63. L. J. Lazaridis and P. G. Pantazelos, Test of flame-heated thermi-

additives of the inert gases, Sov. J. Tech. Phys., 96: 1481–1482,
onic diode, Proc. 10th Annu. Power Sources Conf., Atlantic City,
1966 (in Russian).
NJ, 1966.
42. F. V. Kondrat’ev, G. V. Sinyutin, and V. F. Tikhonov, Effect of
64. F. Huffman, Overview of terrestrial thermionics, Proc. 18th

xenon addition on the operation of a cesium thermionic converter,
IECEC, Paper No. 839027, 1983, pp. 179–185.

Atomnaya Energiya, 23:208, 1967.
65. G. O. Fitzpatrick, E. J. Britt, and G. Carnasciali, Increased cen-

43. F. Rufeh and S. S. Kitrilakis, Proc. Thermionic Convers. Spec.
tral station power plant efficiency with a thermionic topping sys-

Conf., Houston, TX, 1996.
tem, Proc. 12th IECEC, 1977, pp. 1602–1609.
44. I. G. Gverdtsiteli et al., Proc. 2nd Int. Conf. Thermionic Elect.
66. G. Miscolczy, G. Gunter, and A. E. Margulies, Conceptual design
Power Generation, Stresa, Italy, 1968, p. 1097.
of a thermionic topped steam electric generator plant, Proc. 13th
45. F. Rufeh and D. Lieb, The influence of inert gases on the I-V
IECEC, 1978, pp. 1893–1897.
characteristics of a cesium diode, J. Energy Convers., 10: 149–
67. G. Miscolczy et al., Thermionic combustor application to com-
153, 1970.
bined gas and steam turbine power plants, Proc. 16th IECEC,
46. R. Ya. Kucherov et al., J. Sov. Phys.—Tech. Phys., 15: 1937, 1971.
1981, pp. 1956–1961.
----------------------- Page 576-----------------------
THERMISTORS 67
68. G. Miscolczy and D. Lieb, Cogeneration using a thermionic com-

bustor, Proc. 17th IECEC, 1982, pp. 1918–1923.
69. G. Miscolczy et al., Cogeneration burner module design, Proc.
18th IECEC, 1983, pp. 186–191.
70. R. S. Dick, Thermionic cogeneration burner assessment, Proc.
19th IECEC, 1984, pp. 2307–2312.
71. D. V. Paramonov and M. D. Carelli, A thermionic topping cycle
for advanced gas turbines, Proc. 33rd IECEC, American Nuclear
Society, Chicago, 1998.
72. G. O. Fitzpatrick et al., Thermionic converters for increased per-
formance in gas turbine power plants, Proc. 16th IECEC, 1981,
pp. 1962–1967.
73. L. R. Wolff, Small scale thermionic cogeneration systems: Com-
mercialization perspectives, Proc. Thermionic Energy Convers.
Spec. Conf., Goteborg, Sweden, 1993, pp. 57–63.
74. V. I. Yarygin et al., Combustion heated cold seal TEC, Proc. 32nd
Intersoc. Energy Convers. Eng. Conf. (IECEC), 1997, pp. 2264–
2270.
75. V. I. Yarygin et al., Test of a TEC-module, Proc. 29th Intersoc.
Energy Convers. Eng. Conf. (IECEC), 1994, pp. 1061–1066.
76. R. Svenson and L. Holmid, TEC as electric generator in an auto-
mobile catalytic converter, Proc. 31st IECEC, IEEE, 1996, pp.
941–944.
MOHAMED S. EL-GENK
University of New Mexico
DMITRY V. PARAMONOV
Westinghouse Science and
Technology Center
----------------------- Page 577-----------------------
} { { } }
●
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Thermoelectric Conversion

Standard Article

Hal Edwards1 1 2
, Qian Niu , Alex de Lozanne
1Texas Instruments, Inc., Dallas, TX
2The University of Texas at Austin, Austin, TX

Search
reserved.
DOI: 10.1002/047134608X.W3044
❍ Acronym Finder
Abstract
Thermoelectric Effects
Thermoelectric Refrigeration
Thermoelectric Power Generation
Efforts to Improve Thermoelectric Efficiency
Cryogenic Thermoelectric Refrigeration
Thermoelectric Refrigeration Using Quantum-Scale Devices
Thermoelectric Refrigeration Using a Superconducting Tunnel Junction
Prospects and Applications for Cryogenic Thermoelectric Refrigeration
Concluding Remarks
Acknowledgments
----------------------- Page 578-----------------------

----------------------- Page 579-----------------------
80 THERMOELECTRIC CONVERSION
THERMOELECTRIC CONVERSION
Thermoelectric effects offer a means of converting heat to

electric
power using only solid-state components. By supply-
ing
electrical power to a thermoelectric material, it may also
be used
to produce refrigeration. Because a thermoelectric
system
has no moving parts, it may be reliable and compact.
However,
currently available power generators and refrigera-
tors
based upon thermoelectric effects are too expensive and
inefficient for large-scale applications. Today, thermoelectric
generators are used to produce electric power in satellites and
thermoelectric refrigerators are used to cool small heat loads

or in
situations requiring portability.
In
this article, we review the physics of thermoelectric ef-
fects
and survey some present and potential applications of
thermoelectric technology. Then an explanation of the ther-
moelectric effects and the operation of thermoelectric genera-

tors and
refrigerators is given. There is no attempt to be com-
prehensive here as the detailed theory of thermoelectric

effects
has been well described elsewhere (1,2).
Following this, we briefly discuss the present status of
thermoelectric refrigeration and power generation. These
technologies are commercially active, with thermoelectric op-

tions
dominating several niche markets. We also describe sev-
eral
lines of research that attempt to improve the efficiency
of
thermoelectric materials.
Finally, a new research field, cryogenic thermoelectric re-
frigeration, is described. Here, the efficiency can be quite

high, so
that the temperature may be reduced by a large fac-
tor.
Also, we describe proposed designs for thermoelectric re-
frigeration using quantum-scale devices as well as recent ex-
perimental verification of these concepts in superconducting

tunnel
junctions.
THERMOELECTRIC EFFECTS
Thermoelectric effects occur because electrons and holes

(which
are vacancies in the electron sea) in conductive mate-
rials
carry energy. Thermal energy at a temperature T (in K
or
Kelvin, measured from absolute zero) is distributed among
all
parts of a conductive material—the atoms making up the
crystal
lattice, the electrons, and the holes. As a result, elec-
trons
and holes have a range of energies; the electronic ‘‘en-
ergy
spread’’ in a metal is a few times k T, where k is Boltz-
B B
mann’s
constant. A thermoelectric effect occurs when the
most
energetic (or ‘‘hot’’) electrons and holes are transported
preferentially during current flow in a conductive material.

In a
sense, the hot electrons and holes ‘‘evaporate’’ from one

----------------------- Page 580-----------------------
THERMOELECTRIC CONVERSION 81
part of a material to another, cooling the region that they
Heat load/source
leave in the same way as the evaporation of water has a cool-
ing effect.
R
There are several distinct but related physical effects that
arise from this heat capacity of electrons and holes. When an
n Q p
electric current is driven through a thermoelectric material

(for instance, by applying a voltage across the thermoelectric
En I I Ep
material), a heat flow occurs as well. This is called the Peltier
Heat sink
effect, which is the basis of thermoelectric refrigeration.
When one end of a piece of thermoelectric material is hotter
(a)
than the other, a voltage develops across the material. This
is called the Seebeck effect, which is used in thermoelectric
Heat load/source
power generation. Finally, when a temperature gradient ex-
ists inside a thermoelectric material, a gradient in the electric
potential develops due to the Thomson effect. We will not dis-
Q Q Q
cuss the Thomson effect here, since it is beyond the scope of
n p n p n p
this article, but more details may be found in Refs. 1–3.
These effects are known in the science of thermodynamics
I I I I
as reversible processes (1), in the sense that they may operate
Heat sink
in the forward or in reverse directions. For instance, by re-
versing the direction of current flow in a thermoelectric refrig-
(b)
erator, the direction of heat flow is reversed, thereby causing
heating instead of cooling. The efficiency of thermoelectric re-
frigerators and generators is limited by certain irreversible
phenomena (1). The two main irreversible effects are thermal
kBT
conduction and electrical resistance.
When a temperature gradient exists in any material, heat
flows through the material to minimize the temperature gra-
(ii) E F
dient. This is clearly an irreversible process, otherwise objects
(i) (iii)
would spontaneously develop cold and hot spots until they
(iii) Eg E F (i)
cracked or evaporated. Thermal conduction in a thermoelec-

E F (ii)
tric refrigerator represents a heat leak. In a thermoelectric
generator, it is manifest as a power loss.
The passage of electricity through a conducting material is
accompanied by the generation of heat, as the electrons and
En n R p Ep
holes lose energy to various mechanisms such as the crystal
lattice. Similar to thermal conduction, this resistive process
(c)
is irreversible. If it were reversible, then a warm object would
Figure 1. Diagram of a thermoelectric device. (a) Heat flows verti-

spontaneously emit electric current and cool itself to absolute
cally and electric current flows from left to right. (b) Placing multiple
zero. Also similar to thermal conduction, resistive losses rep-
thermoelements electrically in series and thermally in parallel in-
resent an irreversible heat leak in a thermoelectric refrigera-
creases the cooling power. (c) Representation of the microscopic mech-
tor and a power loss in a thermoelectric generator.
anism of thermoelectric effects in terms of electron energy states in
Under some conditions, thermoelectric effects take on an
each material. Vertical scale is electron energy.
irreversible nature when they are driven past the ‘‘small-sig-
nal’’ limit (4–6). In the following, we will discuss situations
under which these extreme conditions are of benefit to ther-
voir’’ that is attached to a heat source (in the case of a thermo-
moelectric refrigeration.
electric generator) or to a device to be cooled (in a thermoelec-
A generic thermoelectric circuit is shown in Fig. 1(a). The
tric refrigerator). En and Ep are ‘‘electrodes’’ that are used to
thermoelements n and p embody the thermoelectric effects in
electrically connect the thermoelectric circuit to an external
this generic circuit. In the positive thermoelement p, electric
electric circuit (a current source to drive thermoelectric re-
current I and heat Q flow in the same direction. In the nega-
frigeration or an electric load in the case of thermoelectric
tive thermoelement n, heat is conducted in the direction oppo-
generation). En and Ep also are thermally sunk to a heat sink
site to current flow. In the circuit of Fig. 1(a), the thermoele- to
maintain their temperature at T.
ments n and p are said to be thermally in parallel, as heat
When the thermoelectric circuit of Fig. 1(a) is in operation,
flows through each in parallel, and electrically in series, be-
electric current flows from En, through n, through R, through
cause electric current must traverse them sequentially. To en- p,
to Ep . Heat Q flows from R, through n and p, to En and
hance the thermoelectric performance of this circuit, it can be
Ep . In the case of a thermoelectric refrigerator, a voltage is
seen from Fig. 1(b) that many more such elements may be
applied between Ep and En to drive an electric current which
added so that they are electrically in series, and thermally
causes the flow of heat Q from the heat load to the heat sink
in parallel. by
the thermopower of n and p. For thermoelectric power gen-
The circuit is completed by metallic elements R, En and
eration, a source of heat is applied to R, which causes heat Q
Ep , which serve as conductors and heat sinks. R is a ‘‘reser- to
flow down. Due to the thermopower of n and p, electric
----------------------- Page 581-----------------------
current I flows up through n and down through p, the net

conductivity of n and p is not shown here, since it is mostly
result of which is to drive electric current from E to E ,
due to the crystal lattice rather than the electrons.
n p
which may be used to drive an electric load.

Any thermoelectric effect is caused by differences in the
THERMOELECTRIC REFRIGERATION
energy levels electrons occupy in a material, and hence may
be visualized by tracing the path of electrons through these By
applying electric current to the thermoelectric circuit of
different energy levels. Such a diagram is shown in Fig. 1(c),
Fig. 1(a), heat is extracted from R. This refrigerator is driven
which represents the energy levels (vertical scale) that elec- by
the Peltier effect. Specifically, if a current I is driven
trons occupy when they are traveling through the circuit com-
through a thermoelectric material, a flow of heat
ponents of Fig. 1(a).
Several aspects of solid-state physics must be understood
Q = #I (1)
to interpret Fig. 1(c). First, electrons are thermally excited so
that they are continually absorbing and emitting energy from
accompanies the flow of current; # is known as the Peltier
other electrons and the crystal lattice. The typical amount of
coefficient of the thermoelectric material.
energy exchanged is about kBT, as mentioned earlier. The
For a negative thermoelement such as n in Fig. 1(a), # is
magnitude of kBT is shown in Fig. 1(c) for reference. If we
negative. Thus, the electric current and heat current flow in
could employ ‘‘Maxwell’s demon’’ (3) to capture the electrons
opposite directions in n, so that while the current flows up,
when they were excited to a high energy and use their excess
heat flows down. For a positive thermoelement such as p in
energy to perform work, then heat would be extracted from
Fig. 1(a), # is positive. In this case, the electric and heat cur-
the metal, resulting in a cooling effect. This is how a thermo-
rents flow in the same direction; in Fig. 1 (a), both heat and
electric refrigerator works.
electric current flow downward through p. Thus, heat flows
Consider an electron in R, at an energy called the ‘‘Fermi
downward through both n and p so that both contribute to
level’’ EF, which is essentially the average energy for the elec-
cooling R.
trons that participate in electrical transport (3). Because elec-
In a real thermoelectric refrigerator, there are other
trons belong to the particle class called Fermions, they fill
sources and flows of heat that must be considered when try-
energy states one at a time until EF is reached. Thermal exci-
ing to understand the refrigerating behavior. If the thermo-
tations occur for energies around E , with a range of k T.
electric refrigerator is cooling a heat load (e.g., a semiconduc-
F B
Some states below EF are vacated by an electron, which has

tor diode laser or a solid-state infrared detector), which is
absorbed energy to be excited to an empty state. These vacant
producing a heat flow QL , then this heat flows into R and adds
states are called holes, and can move about in the same way to
the heat load that the thermoelements must carry away to
as electrons. One can imagine this as a percolation process,
accomplish a given amount of refrigeration.
although in terms of quantum mechanics electrons and holes
In an ideal thermoelectric material, there would be no elec-
are described similarly (3). The excitation energy of a hole
trical resistance, and hence no Ohmic heating. However, this
is opposite to that of an electron; thermally excited holes lie is
not the case for real materials. If the electrical resistance
below E . of
n is R and that of p is R , then Ohm’s law (3) dictates that
F
n p
2 2
To understand the microscopic operation of a thermoelec-
the amount of heat I R is generated in n and I R is generated
n p
tric circuit, we can trace the path of an electron through Fig. in

p. Since this heat is generated uniformly throughout the
1(c). (i) First, the electron in R is thermally excited to above
thermoelectric materials, it is generally assumed that half the
EF . Then the electron may move to the conduction band of the
heat is transmitted up to R, and half goes down to En and
semiconductor n. This band is shown with a slope because the Ep
.
electrical resistance of n causes a voltage drop with distance
As the thermoelectric refrigerator operates, the flow of
(ii). Finally, the electron is deposited in the electrode En,
heat Q causes a temperature difference #T to develop be-
where it gives up its ‘‘extra’’ energy relative to E . An analo-
tween R and E and E . In an ideal thermoelectric material,
F
n p
gous path for a hole is shown in Fig. 1(c) as well.

the thermal conductivity would be zero, so that no heat
One thing to note is that the direction of current flow for
leaked back across to the cold side. However, thermoelectric
both the electron and hole paths in Fig. 1(c) is to the right
materials always have a significant thermal conductivity. As-
because the charge of the electron is negative and the charge
sume that the values of thermal conductance for n and p are
of a hole is positive. An analogy between electrostatic poten- K
and K , respectively. Then a heat leak of K #T flows up-
n p n
tial energy (as is shown in this diagram) and gravity might

ward through n and Kp #T flows through p.
be drawn here; it is as though electrons slide down slopes in
Without a heat load, the net heat that flows from R to En
the energy bands and holes percolate up like bubbles. Thus, is
no ‘‘extra’’ current source must be connected to R for a current
Q = −# I − K #T − 1I 2R (2)
to traverse the circuit. It is also important to note where the
n n n 2 n
different energy-exchange mechanisms are taking place. In
Step (i), the electron and hole absorb thermal energy from the
and that flowing from R to Ep is
other electrons or the crystal lattice of R, which in turn ab-

1 2
Q = # I − K #T − I R (3)
sorbs heat from the heat source in the case of thermoelectric
p p p 2 p
generation or the heat load in the case of thermoelectric re-
The cooling power of the thermoelectric refrigerator is the

frigeration. In Step (ii), the electron and hole give up heat by
sum of Q , Q , and the heat load #Q

Ohmic heating due to the electrical resistance of n and p.
n p L
Finally, in Step (iii), the electron and hole release their excita-
Q = Q + Q − Q = (# − # )I − (K + K )#T
cooling n p L p n n p
tion energy to the electrodes E and E , which then release
(4)
n p
1 2
the heat to the heat sink. The heat leak due to the thermal
− 2 I (Rn + Rp ) − QL
----------------------- Page 582-----------------------
Qcooling represents the heat flow that the thermoelectric refrig- we

consider the case when the thermoelements have similar
erator draws from R and the heat load. If Qcooling is positive at
properties, namely
any temperature, then R will drop in temperature due to the
net cooling action. If Qcooling is positive in the absence of a tem-
Rn = Rp ≡ R ; Kn = Kp ≡ K ; −αn = αp ≡ α ;
(9)
perature difference #T, then R will cool until it sits at a tem-
−# = # ≡ #
n p
perature below the ‘‘ambient’’ temperature T.
The coefficient of performance of a thermoelectric refrig-
so
that the coefficient of performance simplifies to
erator is defined to be the ratio of the cooling power to the
electrical power P required to drive the current I through the
#I − K #T − 1I 2R
refrigerator
ϕ = 2 2 (5## )
Iα#T + I R
Qcooling
ϕ = P (5) We can
further simplify Eq. (5
) using the Kelvin relationship
P is computed by the relationship P # IV , where V is the

# = Tα (10)
voltage drop across the device. To compute V, we need to de-
scribe the other major thermoelectric effect—the Seebeck which
may be derived from the Onsager relation in the theory
effect. of
irreversible thermodynamics. Equation (10) is only valid
A contact potential Vcontact generally exists at a junction of for
small V and #T. This topic is beyond the scope of this
dissimilar materials. In thermoelectric devices, this contact
article; more details may be found in Refs. 2–4.
potential and its temperature dependence are utilized. In par- We
can find the value of I at which Eq. (5
) reaches its
ticular, the contact potential between a thermoelectric mate-
maximum value by taking the derivative of # with respect to
rial (e.g., n or p) and an ordinary metal (e.g., R, E , or E ) is
n p I and
setting it equal to zero. After performing this operation
given by and
substituting this current value into Eq. (5
), we arrive at
the
optimal value of the coefficient of performance
Vcontact{thermoelectric,metal} = αT + V0 (6)
where V0 depends upon the particular materials chosen and
#1 + ZT − T 
 
T − #T T − #T
the text in curly brackets indicates the direction of positive
ϕmax =  # 
contact potential, in this case from thermoelectric material
#T  1 + ZT + 1  ###
to metal. The linear temperature coefficient # in the contact
 
(5 )
potential is called the Seebeck coefficient. Because the junc-
#1 + ZT − ϕ −1
Carnot
tions at R and those at En and Ep sit at different tempera-
= ϕCarnot  #1 + ZT + 1 
tures, a net potential drop occurs across each thermoelement
V = V {E , n} + V {n, R}
n contact n contact where
T # 1/2(T # (T # #T)) is the mean temperature in the
= −V {n, E } + V {n, R}
thermoelements and Carnot # (T # #T)/#T is the Carnot ther-
contact n contact
= −(αn T + V ) + (αn (T − #T ) + V )
modynamic maximum efficiency of a refrigerator (1–3). The
0 0
quantity
= −αn #T ;
(7)
Vp = Vcontact{R, p} + Vcontact{p, E }
α2
p
Z = (11)
= −Vcontact{p, R} + Vcontact{p, E }
RK
p
= −(α (T − #T ) + V ) + (α T + V )
p 0 p 0
is
known as thethermoelectric figure of merit, and has the
= α #T
p
dimensions of inverse temperature. Thus, the quantity
Due to the electrical resistance of the thermoelements, there

2
α
is also an Ohmic voltage drop I(R # R ) in the circuit; thus
ZT = T (12)
n p
RK
the total power supplied to the thermoelectric refrigerator is
P = IV is
known as the dimensionless figure of merit, which summa-
2 (8) rizes
the thermoelectric performance of a material at temper-
= I (αp − αn )#T + I (Rn + Rp ) ature
T. More analytical details of the efficiency of a thermo-
electric refrigerator may be obtained from Refs. 1 and 2. Here,

The coefficient of performance becomes we
will restrict ourselves to general comments regarding the
status
of thermoelectric technology.
Qcooling
ϕ = The
material parameters of the thermoelements enter Eq.
P
1 2 (5# ) (5
) only through Z . In this sense, the figure of merit summa-
(# − # )I − (K + K )#T − I (R + R ) − Q
p n n p 2 n p L rizes
the thermoelectric performance of a thermoelectric ma-
=
I (αp − αn )#T + I 2 (Rn + Rp )
terial. Although Eq. (11) includes the electrical parameters R
and
K , we can see that it may also be defined in terms of
Generally, one would like to know the best coefficient of
intensive material parameters. For instance, suppose that the
performance that a given material system can supply. For
thermoelements are of uniform cross-sectional area A and
this case, we consider the limit with no load (QL # 0). Also, length
L . Then the thermal conductance and electrical resis-
----------------------- Page 583-----------------------
tance may be written as
A L
K = κ ; R = ρ (13)
L A
where is the thermal conductivity and # is the electrical

resistivity of the thermoelements. The figure of merit then
becomes
2
α
Z = (11# )
ρκ
which contains only intensive quantities, and thus the ther-

moelectric figure of merit is itself an intensive material pa-
rameter. The dimensionless figure of merit becomes
Figure 3. Photograph of a commercially available thermoelectric re-
frigerator using Bi Te semiconductors as thermoelements. Product
2 3
2
ZT = α T (12# )
literature from Marlow Industries, Inc.
ρκ
Maximizing ZT is thus seen as the central goal of thermo- mary

of the thermal and electrical properties of conductive
electric materials research. In practice, it has been difficult to
materials as a function of electron concentration. It can be
exceed ZT # 1 (at room temperature). There has even been seen
that electrical and thermal conductivity tend to increase
speculation that ZT # 1 represents a fundamental physical with
the electron concentration, whereas the thermopower
limit (1,2). In any case, ZT must be in the range of 3 to 5 for tends
to decrease. These factors lead to a peak in thermoelec-
thermoelectric refrigeration to compete with conventional va- tric
figure of merit in the semiconductor range.
por-compression refrigeration (1). Efficiency is important in a
The best materials for thermoelectric refrigeration are
refrigerator, both in terms of the economic and environmental
compound semiconductors. The most common material is bis-
costs of power consumption and in terms of the amount of muth
telluride (Bi Te ). This material is used in commercially
2 3
heat rejected for a given amount of cooling. Thus, thermoelec-
available thermoelectric refrigerators (such as the one shown
tric materials have not yet offered a strong competitor in the in
Fig. 3), and may attain ZT values approaching 1. In fact,
area of large-scale refrigeration or air conditioning.
thermoelectric refrigerators (such as the one depicted) have a

The practical difficulties in maximizing ZT are in a com-
growing market in cooling solid-state electronic devices such

promise between the material parameters that comprise ZT .
as
diode lasers and infrared radiation detectors. They are also
For instance, metals have a low electrical resistivity but also
dominant in areas where compactness or portability are re-

a low thermopower and a high thermal conductivity. The
quired, such as in portable refrigerated food containers or bio-

least thermally conductive materials are often electrically in-
logical-specimen transport vessels. Also, thermoelectric re-

sulating.
frigerators can be operated in reverse to heat a sample; thus

As it turns out, the best thermoelectric materials have
been heavily doped semiconductors. Figure 2 shows a sum-
active temperature control is possible, as utilized in biologi-
cal
applications.
Another practical aspect of thermoelectric technology is

Insulators Semi- Semimetals Metals the
use of multiple stages in commercial devices. These can
conductors offer
improved efficiency because the design and materials
properties can be optimized for the operating temperature of

σ each
stage. Lower temperatures can be achieved this way in
α
thermoelectric refrigerators.
K
THERMOELECTRIC POWER GENERATION
The
circuit of Fig. 1(a) may be used as a thermoelectric gener-
ator
by heating R. Heat QS is supplied from a heat source to
the
hot electrode, which causes a flow of thermoelectric cur-
rent
I . The thermal conductance K of the thermoelements
allows heat to leak from the hot electrode to the ambient-
temperature heat sink, and Ohmic heating of the thermoele-

α 2 σ ments
dissipates some of the generated power. The efficiency
ZT =
K # of
a thermoelectric generator is defined to be
P IV
η = = (14)
Qs Qs
Carrier concentration
where
P is the electrical power supplied to a load through the
Figure 2. Schematic graph of the properties of thermoelectric mate-
ambient-temperature electrodes and V is the total voltage
rials as a function of electron concentration (1). drop
across the generator. By following a line of reasoning
----------------------- Page 584-----------------------
analogous to the forementioned one for thermoelectric re- (4–10).

Later, we will discuss the new field of cryogenic ther-
frigeration, the efficiency of a thermoelectric generator may moelectric
refrigeration (4,5,7,8). Here, we briefly summarize
be expressed in terms of the dimensionless figure of merit of recent
work in thermoelectric conversion at higher tempera-
the thermoelement materials tures,
which could impact the broadest range of applications.
A
notable improvement in
materials synthesis over the
#T  #1 + ZT − 1  last few
decades has been thin-film deposition with high crys-
η = #
 talline quality and purity by using methods such as molecu-
 
T
T + #T 1 + ZT + T + #T lar-beam
epitaxy (11). With this capability, traditional semi-
# conducting
materials may be formed into structures not
# #
(14 )
1 + ZT − 1 available
to the original designers of thermoelectric devices in
#
= ηCarnot the 1950s
and 1960s.
1 + ZT + (1 − ηCarnot)
Accordingly, several groups have begun exploring the ther-

where T and Z are defined in the foregoing and #Carnot # moelectric
properties of multilayer structures composed of
layers of
the Bi2Te3 and related compounds traditionally used
#T/(T # #T) is the Carnot thermodynamic limit on the effi-
as
thermoelectric elements (9). This activity was initially of a
ciency of a generator (1–3). As discussed in the foregoing,
theoretical nature, and it was observed that the thermoelec-

ZT is around 1 at best. For this value, # ranges from
0.2#Carnot to 0.3#Carnot. For a temperature difference of 100#C tric figure
of merit of these materials in a properly designed
multilayer
may be increased from its bulk value. This was
with T # 300 K, #Carnot # 0.25 and the thermoelectric efficiency
confirmed
experimentally, but the figure of merit of the entire
with ZT # 1 would be under 5%. Thus, thermoelectric power
structure,
including nonthermoelectric layers, has not been
generation is not competitive with more conventional means.
demonstrated yet to exceed the bulk value of the thermoelec-

For comparison, a solar cell may have a generating efficiency
tric
materials.
of 8%, an automobile engine may have 15% efficiency, and a
It has
recently been proposed (6) that thermionic effects
steam-generating plant may achieve 20 to 40% efficiency (1).
might be
used to overcome limitations imposed by the ‘‘linear-
What would it take for thermoelectrics to compete in power
response’’
regime implicitly assumed in the traditional theo-
generation? Suppose for a moment that ZT for a given mate-
retical
understanding of thermoelectric effects as described in
rial was optimized over the temperature range 300 to 600 K,
earlier.
In a thermionic effect, electrons or holes are ther-
and that T # 300 K and the heat-source temperature is 600
K. Then #Carnot # 0.5 and, for ZT # 1, the generator becomes mally
excited across a large potential barrier and then trans-
ported
away. In terms of the microscopic picture of Fig. 1(c),
11% efficient. For ZT # 3, the generator becomes 20% effi-
the
conduction would become thermionic if the barrier be-
cient, the point at which it begins to compete. Thus, if thermo-
tween R
and n or p were much higher than kBT. To overcome
electric materials are to be competitive with more conven-
the
potential barrier, each electron or hole must absorb a
tional systems for mainstream power-generation
large
amount of energy from the crystal lattice; thus a large
applications, ZT must be improved by at least a factor of
amount of
heat is extracted for a relatively smaller amount of
three. Furthermore, the materials must be inexpensive,
electric
current than is the case for a traditional thermoele-
whereas semiconductors tend to be fairly expensive. Thus, the
ment. With
thermionic effects, there is also the possibility of
prospects for large-scale power generation using thermoelec-
introducing a vacuum barrier or insulating material between

trics are not great at present.
electrodes
to reduce heat leaks.
As in refrigeration, the present applications of thermoelec-
Another
exciting area of materials research is along the
tric power generation take advantage of the compactness and
traditional lines of bulk thermoelectric devices, but in new

reliability of solid-state components. The most prominent ap-
material
systems. The skutterudites are a new thermoelectric
plication is in power generation for satellites, in which a radi-
system
proposed recently (10). These materials have a com-
ation source such as a radioactive isotope is used to supply
plex
crystal structure which frustrates heat flow, and can be
heat to the generator. The material system of choice is the
doped to
provide electrical conductivity. Initial measurements
alloy of silicon and germanium, which has a lower thermal
of the
dimensionless figure of merit in this system have ex-
conductivity than the crystalline forms due to enhanced pho-
ceeded 1,
with values of 1.4 to 1.5 being reported by several
non scattering, and is thermally stable enough to operate at
groups.
Because the skutterudites family is unexplored and
a high hot-electrode temperature. These generators have been
fairly
extensive, there may exist a skutterudites thermoelec-
found to be extremely reliable, and nuclear power provides
tric
material with a sufficiently high figure of merit and ade-
sufficient heat to overcome the relatively low generating effi-
quate
thermal and mechanical stability to compete in more
ciency (1).
large-
scale applications.
EFFORTS TO IMPROVE THERMOELECTRIC EFFICIENCY

CRYOGENIC
The traditional approach to optimizing the efficiency of ther-
moelectric conversion has been incremental. For instance, At
temperatures near absolute zero, the reduction in ambient
commercial devices today are still based upon Bi2Te3 and re- thermal
energy allows quantum-mechanical effects to domi-
lated compounds, developed in the 1950s and 1960s, with ad- nate the
properties of some types of matter, giving rise to ex-
vances mostly coming in the form of material purity or means
traordinary phenomena. Some metals lose all electrical resis-
of attaching thermoelements. tance and
fall into the superconducting state (3). The viscosity
Recently, a number of workers have become interested in of liquid
helium vanishes, causing it to flow over the sides of
revisiting the basic assumptions of thermoelectric conversion a
container (3). This superfluidity of liquid helium gives rise
----------------------- Page 585-----------------------
metallic reservoir at the center is cooled by the removal of

hot
electrons and holes and their deposition in electrodes that
Hole –
reside at the ‘‘ambient’’ temperature, which here may refer to
k T tunneling the
phonon temperature or the true ambient temperature of
B 0 channel
E F + the
medium surrounding the refrigerator structure. Physi-
Electron
kBT tunneling
cally, such a refrigerator would probably be comprised of thin
channel metal
films or semiconductor layers on a substrate; the dia-
gram
of Fig. 4(a) shows electron energies, rather than physi-
cal
structure.
Electrode Metallic Electrode
In this light, the vertical position of an electron in Fig. 4(a)
at ambient reservoir at ambient
corresponds to its energy. The ‘‘sea’’ of electrons in the various
temperature to be cooled temperature
structures corresponds to the Fermi sea of electrons in a

(a)
metal. It is at the surface EF of this Fermi sea that thermal
excitations—in the form of thermally excited electrons and
holes
—are contained. Thus, it can be seen that a hole below
(i) the
‘‘sea level’’ of EF carries the same amount of thermal en-
ergy
as an electron excited an equal amount above EF . If this
T > T (ii) Fermi
sea of electrons is at temperature T, the average exci-
0
0
tation energy of its electrons and holes from EF is roughly
T
c kBT
(3). In addition, we use the term ‘‘hole’’ hereafter to refer
i
n
o to a
vacant electronic state in a metal, which is more general
r
t
c than
the definition used for p -type semiconductor material
e (iii)
l
E (3).
This nomenclature simplifies the understanding of a ther-
Phonon E
moelectric device.
heating
The voltage differences between different pieces of metal,
O K g (E ) such
as the electrodes and the reservoir of Fig. 4(a), are re-
flected in differences between their Fermi levels. By applying

O K Ambient T a
voltage to this structure, an electric current flows. The
(b)
structures marked ‘‘electron tunneling channel’’ and ‘‘hole
tunneling channel’’ will be explained in more detail in what

Figure 4. (a) Energy-level diagram and (b) temperature regime of
follows, but the basic idea is the following. Because the tem-
operation of a cryogenic thermoelectric refrigerator. Reproduced from
perature of an electron sea is defined by the energies to which
Ref. 5.
electrons and holes are excited from the Fermi levels, remov-
ing
the excited electrons and holes should cool the remaining
to thermodynamic properties, which are exploited in the dilu-
electrons. The alternative to this selective process is to allow
tion refrigerator, the most common means of performing ex- both
electrons and holes to move arbitrarily through both in-
perimental work at temperatures between 0.01 and 1 K.
terfaces, resulting in a loss of control over the direction of
The low-temperature properties of matter of interest here heat
flow. The preferential control over the direction of heat
are related to phonons, the quantum-mechanical manifesta- flow
during electrical transport is the basis of thermoelectric
tion of vibrations in the crystalline lattice. Phonons are the
refrigeration.
predominant repository of heat in solids at temperatures
Thus, for thermoelectric refrigeration to be possible, it is
more than a few kelvins above absolute zero. Near absolute
required to preferentially remove excited charge carriers from
zero, however, electrons in a metallic solid hold most of the near
the Fermi level. The conventional way of doing this is to
heat, and hence thermal energy. Thus, cooling the electrons use
a negative thermoelement, which carries thermally ex-
in a metal at these low temperatures should provide an effec- cited
electrons away from the reservoir, and a positive ther-
tive means of refrigerating the entire metal.
moelement that removes thermally excited holes. The terms

Because it is a vibration or deformation wave in the crys-
‘‘negative’’ and ‘‘positive’’ arise from the sign of the charge

talline lattice, a phonon has a wavelength. At higher tempera-
carriers. Another interpretation of these terms is that, for a

tures, the predominant wavelength—that of phonons that
positive thermoelement, heat and electric current flow in the

carry the thermal energy kBT—is shorter. Near absolute zero,
same
direction, whereas the opposite is true for a negative
this phonon wavelength becomes so long that it barely inter-
thermoelement.
acts with electrons in the solid, analogous to the way that
broad swells on the sea do not overturn a small boat as do
As was mentioned in the foregoing, an important differ-
abrupt breakers. ence
between the thermal properties of matter at room tem-
This ‘‘decoupling’’ of the electrons and the phonons in a
perature and cryogenic temperatures is the fact that the elec-
solid enables the operation of certain devices based upon ‘‘bal- trons
hold most of the heat within 1 K of absolute zero.
listic’’ electron transport (12). It also allows the temperature
Combined with the fact that electrons and phonons and other
of the electrons in small devices to be controlled indepen-
degrees of freedom of a metal tend to decouple at cryogenic
dently of the phonon temperature. This is the basis of the
temperatures, this gives rise to some exotic possibilities for
first experimental demonstrations of cryogenic thermoelectric
thermoelectric refrigeration. Figure 4(b) shows a diagram of
refrigeration, to be discussed in the following. the
ambient temperature T versus the electronic tempera-
The structure and operation regime of a cryogenic thermo- ture
To for a ‘‘generic’’ thermoelectric refrigerator as shown in
electric refrigerator are depicted in Fig. 4 (5). In Fig. 4(a), the Fig.
4(a), in which we assume that the electrons are cooled
----------------------- Page 586-----------------------

independently of the phonons. There are several temperature

(a) QDR
regimes of interest.
In the upper left-half of the #T, To# diagram of Fig. 4(b),
DL R DR
the electronic temperature exceeds the ambient; thus this re-

VL Area A VR
gion is not of interest in a discussion of refrigeration. In the
lower right-half, the electrons are cooler than the ambient,
and there are situations one can image. First, if the ambient
(b)
temperature T is too high, the electrons will absorb more en-
∆ EDL δ
ergy from phonons than can be removed by the electron and
hole tunneling channels, and hence the electrons will be
µR
heated. This situation is depicted for the hashed region in the
µ 0
lower right corner of the diagram.
µL ∆ EDR
The unhashed area to the center of the #T, To# diagram of

Fig. 4(b) is thus the region in which the refrigerator is opera-
tional. In this region, there are several additional types of
behavior, which come into play according to the rate at which
VL DL R DR VR
the electrons can ‘‘thermalize,’’ in other words, refill states

EDL– µ 0= µ 0 – EDR ≡ ε ; µ R – µ 0 = µ 0 – µ L ≡ eV b
emptied by the electron and hole tunneling channels. In re-
gime (i), the electron distribution g (E) [shown as a function of Figure
5. (a) Schematic of the physical structure and (b) energy-level
diagram
of a quantum-dot thermoelectric refrigerator. The quantum
electron energy E on the right-hand side of Fig. 4(b)], is con-
dots
act as thermoelements removing hot electrons and holes from
tinuous, and has a form expected for the thermally excited
the
reservoir R by resonant tunneling. This cools the remaining elec-
electron distribution near the Fermi level (3). However, as the trons
in R. Reproduced from Ref. 5.
temperature decreases [regime (ii)], the hot electrons and
holes are removed at a rate approaching that at which the
electrons scatter among themselves to smooth their energy
distribution. Finally, at the lowest temperatures [regime (iii)], A
quantum dot is a piece of metal so small that the individ-
the bands of energy that are aligned with the electron and ual
quantum states, which electrons occupy, are separated by
hole tunneling channels will be fully saturated by the cooling an
energy comparable to or in excess of the thermal energy
process. In essence, the electrons are not only decoupled from kBT. In
Fig. 5(b), the individual quantum states in the quan-
the phonons, but from each other, as far as the refrigeration tum
dots are denoted by horizontal lines. One property of
process is concerned; and individual sets of energy states may such
states is that electrons and holes may travel through
be selectively cooled even when ‘‘hot’’ electrons and holes exist them
from states aligned in energy with them, in a process
at other energies. called
resonant tunneling. Thus, if a quantum state of a quan-
Figure 4(b) describes the case when the electrons and pho- tum dot
is aligned in energy with the thermally excited elec-
nons are decoupled. To apply such a refrigerator, the ‘‘metallic trons
above the Fermi sea in R , the electron may undergo
reservoir to be cooled’’ would have to be able to exchange elec-
resonant tunneling to be removed from R and deposited in
trons with any heat load, which requires integral design and the
electrode, where its excess energy is dissipated as heat. A
fabrication. Cryogenic thermoelectric cooling would be most similar
process occurs for holes resonantly tunneling through
useful when the electrons and phonons are coupled and the the
other electrode.
cooling power is strong enough to achieve bulk cooling. In the The
net effect of these processes is the cooling of R as ther-
following, we will show two examples of cryogenic thermoelec- mally
excited electrons and holes are removed. In addition, it
tric refrigerators and illustrate the temperature regime over can be
seen that electric current passes from left to right.
which these limits apply. Thus,
this structure is a Peltier refrigerator with the left-
hand
quantum dot serving as the negative thermoelement
and the
right-hand quantum dot serving as the positive ther-
moelement. Now, of course, one might wonder under what cir-
USING QUANTUM-SCALE DEVICES
cumstances such a device produces a significant cooling effect.
Most
quantum-dot studies have taken place at or below 1
The promising nature of cryogenic thermoelectric refrigera- K,
although quantum-dot devices have been demonstrated to
tion was suggested in 1993 (4), when it was proposed that work at
room temperature as well (13). At 1 K, kBT is 8.6 #
quantum dots be used as thermoelements. Although this 10#5
eV or 1.6 # 10#23 J. This is the amount of heat one could
quantum-dot refrigerator has not yet been realized in prac- expect

to remove from a piece of metal at 1 K by removing a
tice, it is a good example because its efficiency can be shown typical
electron that is thermally excited above the Fermi
to approach the Carnot limit (5). Figure 5(a) shows the physi- level.
Resonant-tunneling through quantum dots typically
cal structure and Fig. 5(b) shows the energy-level diagram of
produces electric currents in the range of 10#6 A, which is
one such quantum-dot thermoelectric refrigerator. The indi-
equivalent to the passage of 5 # 1012 electrons/s. Thus, a cool-
vidual electronic states of the quantum dots are separated by ing
power of 8 # 10#11 W could be extracted by such a device.
#, and each is lifetime-broadened by #. These are the only

Compared to a 100 W light bulb, this seems minuscule. How-
important properties of quantum dots for this device—the ever,
the heats exchanged at 1 K are in the picowatt range,
quantum dots may actually be larger or smaller than the res- and
electrons hold most of the heat at 1 K and below—there
ervoir R to be cooled, depending upon the material system simply
is not much heat to be spread around so near abso-
and the temperature range of interest (6). lute
zero.
----------------------- Page 587-----------------------
The models of Ref. 5 indicate that a quantum-dot thermo-

Cu reservoir
electric refrigerator could cool a micrometer-sized metallic
(electrons at
Silicon nitride T , lattice at T )

reservoir of electrons in the 1 K regime; here, such a small
0 ph
membrane
collection of electrons is not strongly coupled to the crystal
A1 electrode Pb electrode
lattice and hence is not heated by the absorption of phonons.
Thus, quantum-effect thermoelectric refrigeration might be a
viable means of cooling microelectronic components that rely
upon sub-1 K material properties (e.g., superconductivity or
ballistic transport) in which a temperature difference between
electrons and phonons is acceptable.
As it turns out, a thermoelectric refrigerator with quantum
dots as thermoelements may be capable of bulk cooling at
temperatures a thousand times lower (1 mK) or below (5).
Due to the fine scale of the energy levels of quantum dots,
r2 r1
Substrate at
this might be an effective means of microkelvin or nanokelvin
temperature T
cooling of metals. Such refrigerators have not been tested,
however, thus their true performance is an open question.
(a)
However, the thermoelectric transport properties of various
quantum structures have been measured and found to be con-
sistent with the assumptions of the theoretical simulations of
Ref. 5. The greatest experimental challenge in the realization
of a quantum-dot thermoelectric refrigerator remains the ac-
2∆ Pb
curate measurement of the temperature of the electrons in R .
2∆Al
THERMOELECTRIC REFRIGERATION USING A

SUPERCONDUCTING TUNNEL JUNCTION
Al Cu Pb
electrode reservoir electrode

Although resonant tunneling through a single quantum state
(b)
is simple to model and may be capable of refrigeration near
the Carnot limit, there are other low-temperature phenomena
Figure 6. (a) Schematic and (b) energy-level diagram of the normal-
which may be exploited in the lab to perform refrigeration.
insulator-superconductor (NIS) thermoelectric refrigerator. Hot elec-
One such system is a tunneling junction between a supercon-
trons and holes in the Cu reservoir R tunnel into quasi-particle states
ducting metal and one in the ‘‘normal’’ state. A normal-insu- in
the Al superconducting thermoelement, cooling R . The Pb contact
lating-superconductor (NIS) refrigerator was demonstrated
electrode is used to inject charge into R , but has no net heating or
cooling effect. The structure is shown fabricated on a silicon nitride

(7) shortly after cryogenic thermoelectric refrigeration was
membrane to minimize the heat leak due to substrate thermal con-
originally proposed (4).
duction. Reproduced from Ref. 5.
The NIS refrigerator is depicted in Fig. 6. In a supercon-
ducting metal, electrons form into pairs, which in turn co-
alesce into the superconducting condensate, a macroscopic
moelectric refrigerators discussed in this article. In this case,
quantum state of matter from which an energy of 2# is re- the
second aluminum electrode would be the positive thermo-
quired to excite them (3). When the required energy has been
element.
applied to an electron or hole, a subtle excitation called a
Because tunneling junctions between metal thin films may
quasi-particle is formed. This is represented in an energy- have
a macroscopic surface area, the current and hence the
band diagram in Fig. 6(b) as a ‘‘forbidden gap’’ of 2# about
cooling power of such a device may be scaled to larger values
the Fermi energy. Because paired electrons in the supercon- than
the quantum-dot thermoelectric refrigerator discussed
ducting condensate have a lower energy than quasi-particles, in
the foregoing. A measurable temperature reduction of 10
one can say that quasi-particles carry heat. mK
below 100 mK was demonstrated in an initial experiment
In the NIS structure, refrigeration is accomplished by (7).
An extra NIS tunneling junction (not shown) was used
transport of thermally excited electrons above the Fermi level as
a thermocouple to measure electron temperature. In this
in R into quasi-particle states in the superconductor [in Fig.
experiment, the metal thin films comprising the refrigerator
6(b), the aluminum electrode serves this purpose]. Quasi-par- were
deposited on a bulk substrate, and were patterned on
ticles are transported through the Al electrode to be deposited the
micron scale; thus it was assumed that the electrons were
(along with the heat they carry) in a heatsink (not shown in
cooled but the phonons remained at 100 mK, due to the de-
the figure). Thus, heat is removed from R to the ‘‘external’’
coupling of electron and phonon degrees of freedom in small,
world through the quasi-particle states and R is cooled by the cold
metals.
‘‘negative thermoelement’’ Al electrode. In this experiment
More recently, cooling from 300 to 100 mK was achieved
(7), a lead counterelectrode with a thin, high-conductance by
an improved NIS refrigerator (8). This device was fabri-
tunneling barrier (to maintain Fermi-level alignment) was
cated on a membrane [as depicted in Fig. 6(a)]. Thus, there
used to inject electrons at the Fermi level. However, a second was
not a strong thermal coupling to the ambient-tempera-
aluminum electrode identical to the one shown could be used ture
substrate, and the crystal lattice in the metal films com-
in its place to extract thermally excited holes into hole-quasi-
prising the refrigerator could be cooled along with the elec-
particle states, in a configuration analogous to the other ther-
trons, thereby reducing the heat leak due to phonon
----------------------- Page 588-----------------------
absorption. The feasibility of this experiment was theoreti- using

thermionic effects instead of reversible thermoelectric
cally predicted in Ref. (5). These devices are being developed effects
(6), enhanced efficiency might be achieved by high en-
to cool advanced radiation detectors for use in astronomy, but ergy
barriers and a departure from the traditional linear-re-
any electronic device that must operate at temperatures be- sponse
regime of thermoelectric behavior. Therefore, the cool-
low 1 K could benefit from this type of refrigerator. ing or
power generation process would be far more efficient
The significance of this result is that thermoelectric cooling than would
be expected from examination of the material
has been achieved in which the kelvin temperature was re- figure of
merit.
duced by a factor of 3. To accomplish this with a bulk mate-
Cryogenic thermoelectric refrigeration represents a re-
rial, the figure of merit would have to be in the range of search
frontier and, accordingly, applications are still in the
ZT # 2 to 4 (1,2). Thus, previous speculations (1,2) that research
phase. However, although commercial applications
ZT # 1 represents a fundamental limit of physics may now remain
distant, cryogenic thermoelectric refrigeration may
be laid to rest. provide an
enabling technology for scientific work such as the
mK-range
radiation detector used in astronomy as discussed
in this
article. And, of course, if sufficient advantages are re-
PROSPECTS AND APPLICATIONS FOR CRYOGENIC
alized by
cryogenic thermoelectric refrigeration, it may give
rise to new
commercial applications. For instance, liquid ni-
trogen is
used to cool components in cellular-telephone base
The success of the NIS refrigerator indicates the promise that
stations.
If a thermoelectric refrigerator could be used to cool
thermoelectric refrigeration holds for the sub-1 K regime. Any
from 77 K
to lower temperatures, new classes of electrical
material interface or device structure, which conducts electric
components
could be utilized in widespread applications such
current but has a significant energy dependence, should be
as this.
capable of thermoelectric refrigeration at some temperature
With
even greater extrapolation, one could imagine a cas-
scale (5). One particularly interesting material system is the
caded
series of thermoelectric refrigerators, for which the ‘‘hot
III/V semiconductors, which may be formed into heterostruc-
side’’ is
at room temperature and the ‘‘cold side’’ is at a cryo-
tures with custom-tailored energy band structures and high
genic
temperature. An example of an application where such
carrier mobilities. These structures have been used for power
a device
might receive widespread commercial application in
devices, and thus can carry large electric currents, and vari-
the future
is in quantum computers, which should be capable
ous structures could be envisioned that would produce a sig-
of
massively parallel computations but may have to be kept
nificant cooling effect.
cold to
preserve quantum phase coherence.
Cryogenic thermoelectric refrigeration may also lead to un-

CONCLUDING REMARKS expected
scientific discoveries; for instance, it is not known
whether
there exists a class of metals incapable of supporting
Widespread application of thermoelectric energy conversion a
superconducting state or if all metals will go superconduct-
remains a goal of materials science. While certain niche appli- ing at a
low enough temperature. At present, experiments
cations support a healthy, small Peltier-cooler industry, ther- have only
been conducted to the range of 10#5 K (15). Direct
moelectric energy conversion will not compete economically

thermoelectric cooling may be able to cool metals further, per-
with conventional methods until inexpensive, robust materi- haps
answering this fundamental question in our under-
als with a higher figure of merit are available. However, a standing
of the electronic structure of matter. Perhaps the
good thermoelectric material would be very useful. For in- most
fundamental contribution of this field so far is that
stance, as high-performance microelectronics continues to ZT # 1 is
not a universal physical limit, and that there may
pack more transistors on a chip, heat loads are expected to still be a
highly efficient thermoelectric material system wait-
exceed 100 W in several generations (14). This will require ing to be
discovered.
some form of direct cooling, and an efficient, low-cost thermo-
electric alternative is attractive compared to water cooling,
ACKNOWLEDGMENTS
which may be hard to maintain reliably in a consumer end
product.
One observation that leads to cautious optimism is that We would
like to acknowledge useful technical discussions
the present commercially available thermoelectric systems with John
Martinis (NIST, Boulder), Michael Nahum (Har-
are based upon decades-old materials science developments. vard
University), and Paul Warburton (Kings College).
Many advances in computational materials modeling, materi-

als synthesis (such as molecular-beam epitaxy), and materials
BIBLIOGRAPHY
characterization (such as high-resolution transmission elec-
tron microscopy and scanning-probe microscopy) have oc- 1. R. R.
Heikes and R. W. Ure, Jr., Thermoelectricity: Science and
curred in the meantime, enabling the exploration of complex
Engineering, New York: Interscience Publishers, 1961.
material systems such as the high-temperature superconduc- 2. H. J.
Goldsmid, Thermoelectric Refrigeration, New York: Ple-
tors and colossal magnetoresistive materials. The skutter- num,
1964.
udites discussed in this article are an example of the fruits of 3. N. W.
Ashcroft and N. D. Mermin, Solid State Physics, Philadel-
modern materials science. As further complex material sys- phia:
Saundes College, 1976.
tems are explored, the right combination of low thermal con- 4. H. L.
Edwards, Q. Niu, and A. L. de Lozanne, A quantum-dot
ductivity, high electrical conductivity, and large thermopower
refrigerator, Appl. Phys. Lett., 63 (13): 1815–1817, 27, 1993.
may yet be found in an inexpensive, robust material. 5. H. L.
Edwards et al., Cryogenic cooling using tunneling struc-
New device concepts in thermoelectric conversion may tures
with sharp energy features, Physical Rev. B, 52 (8): 5714–
overcome some traditional barriers as well. For instance, by 5736,
1995.
----------------------- Page 589-----------------------
90 THERMOPILES
6. A. Shakouri and J. E. Bowers, Heterostructure integrated therm-

THERMOELECTRIC POWER. See TEMPERATURE
ionic coolers, Appl. Phys. Lett., 71 (9): 1234–1236, 1997.
SENSORS.
7. M. Nahum, T. M. Eiles, and J. M. Martinis, Electronic microre-
THERMOMETERS. See TEMPERATURE SENSORS.
frigerator based on a normal-insulator-superconductor tunnel
junction, Appl. Phys. Lett., 65 (24): 3123–3125, 1994.
8. M. M. Leivo, J. P. Pekola, and D. V. Averin, Efficient Peltier re-

frigeration by a pair of normal metal/insulator/superconductor
junctions, Appl. Phys. Lett., 68 (14): 1996–1998, 1996.
9. P. J. Lin-Chung and T. L. Reinecke, Thermoelectric figure of

merit of composite superlattice systems, Physical Rev. B, 51 (19):
13244–13248, 1995; J. O. Sofo and G. D. Mahan, Thermoelectric
figure of merit of superlattices, Appl. Phys. Lett., 65 (21): 2690–
2692, 1994; L. D. Hicks, T. C. Harmon, X. Sun, and M. S. Dressel-
haus, Experimental study of the effect of quantum-well struc-
tures on the thermoelectric figure of merit, Physical Rev. B, 53
(16): R10493–R10496, 1996; L. D. Hicks and M. S. Dresselhaus,
Effect of quantum-well structures on the thermoelectric figure of
merit, Physical Rev. B, 47: 12727, 1993.
10. B. C. Sales, D. Mandrus, and R. K. Williams, Filled Skudderite

antimonides: A new class of thermoelectric materials, Science,
272: 1325–1328, 1996.
11. A. Y. Cho and J. R. Arthur, Molecular beam epitaxy, Progress in

Solid State Chemistry, 10: 157, 1973.
12. H. Ehrenreich and D. Turnbull (eds.), Solid State Physics, 44:

Semiconductor Heterostructures and Nanostructures, New York:
Academic Press, 1991; Supriyo Datta, Electronic Transport in
Mesoscopic Systems (Cambridge Studies in Semiconductor Phys-
ics and Microelectronic Engineering, 3), Cambridge: Cambridge
University Press, 1997.
13. M. A. Reed et al., Observation of discrete electronic states in a

zero-dimensional semiconductor nanostructure, Physical Rev.
Lett., 60 (6): 535–537, 1988.
14. The National Technology Roadmap for Semiconductors, to be pub-

lished by the Semiconductor Industry Association, San Jose, CA.
15. Frank Pobell, Solid-state physics at microkelvin temperatures: is

anything left to learn, Phys. Today, 46 (1): 34–40, 1993.
Reading List
R. R. Heikes and Roland W. Ure, Jr., Thermoelectricity: Science and

Engineering, New York: Interscience Publishers, 1961.
H. J. Goldsmid, Thermoelectric Refrigeration, New York: Plenum,

1964.
N. W. Ashcroft and N. D. Mermin, Solid State Physics, Philadelphia:

Saundes College, 1976.
H. L. Edwards et al., Cryogenic cooling using tunneling structures

with sharp energy features, Physical Rev. B, 52 (8): 5714–5736,
1995.
H. Ehrenreich and D. Turnbull (eds.), Solid State Physics, 44: Semi-

conductor Heterostructures and Nanostructures, New York: Aca-
demic Press, 1991.
C. Wu, A Silent Cool, Science News, 152 (10): 152–153, September

6, 1997.
C. Wood, Materials for thermoelectric energy conversion, Rep. Prog.

Phys., 15 (4): April 1988.
L. Kouwenhoven, Coupled quantum dots as artificial molecules, Sci-

ence, 268: 1440, 1995.
S. Washburn, A superconducting siphon, Nature, 373: 106, 1995.
HAL EDWARDS
Texas Instruments, Inc.
QIAN NIU
ALEX dE LOZANNE
The University of Texas at Austin
----------------------- Page 590-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Turbogenerators
Standard Article

William R. McCown1 1 1
, Aleksandar Prole , Robert J. Nelson ,
Joseph D. Hurley1 1 1
, Richard B. Chianese , Ross Guttromson

1Siemens-Westinghouse Power Corp.
Search
reserved.
DOI: 10.1002/047134608X.W3009
❍ Acronym Finder
Abstract
Fundamentals
Excitation Source
Construction Features
Ventilation and Power Density
Basic Mechanical Design Considerations
Thermal Design
Dynamics and Vibration
Applicable Standards and Codes—Summary of Basic Requirements
Turbogenerator Applications
Siting of Turbogenerators and Transmission Interconnections
----------------------- Page 591-----------------------
Operation of Turbogenerators

----------------------- Page 592-----------------------
TURBOGENERATORS 595
A
B C
C B
A A
B C
TURBOGENERATORS
C B
The modern generator, which evolved from the discovery of

magnetic induction by Michael Faraday in 1831, is a machine
used to convert the mechanical energy of rotation into electric
Magnetic stator Magnetic rotor
energy. As a high-speed rotating device, the generator is sub-
core-laminated forging
ject to various vibratory, fatigue, and tensile stresses. As an
electromagnetic device, it is subject to the dielectric stresses
Figure 2. Simplified two-pole generator cross section.
associated with insulated high-voltage equipment and to the

inherent heat transfer problems resulting from various elec-
FUNDAMENTALS
tric and mechanical losses. In addition to these basic machin-
ery problems, there are various other issues to consider, such
as designing the generator to be suitable for the turbine Modern
turbogenerators usually operate at one of four rota-
which drives it, meeting special terminal voltage require- tional
speeds:
ments specified by the purchaser, addressing stability prob-

lems of the machine connected to the power system, meeting •
3600 rpm (60 Hz) or 3000 rpm (50 Hz), 2 pole (usually
with fossil-fueled steam or combustion turbines)

environmental requirements, and transporting the generator
to the station site. •
1800 rpm (60 Hz) or 1500 rpm (50 Hz), 4 pole (usually
Since the installation of the first alternating current (ac)
with steam turbines at nuclear stations)
central power station turbine generator in 1903 at the Hart-
ford Electric Light Company, the most predominant single
These speeds are generally based on the most efficient tur-
factor has been the tremendous increase in generator power bine
operating speeds, given the steam conditions, but be-
ratings required by the demands of the industry. The highest cause
of the requirement to connect to a system of fixed fre-
power rating of single shaft generators has been increased quency
(50 or 60 Hz), the speed ns is constrained to
seven hundred fold from the 2000 kW, 1200 rpm generator in
120f
1903 to the 1500 MW generators available in 1998. Figure
ns =
1 shows a typical installation of a 400 MW hydrogen-cooled
P
generator with a brushless excitation system.

where f
is the electric frequency (Hz), P is the number of gen-
erator
field poles (even number), and ns is the synchronous
speed
(rpm).
This
means that the operating speed of a directly con-
nected
turbine must match the system frequency and the se-
lection
of the number of field poles. Some smaller applications
(less
than 70 MW) use a gear drive to allow separate optimi-
zation
of generator and drive speeds.
With
these rotational speeds, turbogenerators are designed
with
round-rotor field windings and support structures to
support
the windings with the centrifugal forces involved. A
basic
cross section of a two-pole generator is shown in Fig. 2.
The
rotor field winding is wound in machined slots in an alloy
steel
magnetic rotor forging that forms a rotating electromag-
net when
a dc current is applied to it. The resulting magnetic
flux
penetrates the rotor across the ‘‘air-gap’’ into a magnetic
stator
and closes back on itself as illustrated in Fig. 2. Fara-
day’s
law states that a rotating flux field passing the station-
Figure 1. Hydrogen-cooled generator with brushless excitation ary
stator conductors generates a voltage in each conductor
system.
proportional to the rate of change of flux through the conduc-

----------------------- Page 593-----------------------
596 TURBOGENERATORS
tor. As the rotor turns, the voltage in each conductor peaks,
goes to zero and reverses, creating approximately a sine wave.
For the two-pole generator shown, one complete rotor revolu-
tion generates one complete electric cycle.
All large alternating current generators have the stator
conductors connected as three separate windings (or phases)
spaced 120# apart around the periphery of the stator. Thus
the voltages generated in each of these windings is 120# out
of phase with the others. The cross section of Fig. 2 shows a
typical configuration with two conductors per stator slot, each
phase labeled A, B, or C. The top coil bars are connected to
bottom coil bars of the same phase in the ends of the machine
forming a coil-type winding. The ends of each winding are
brought out to terminals, usually at one end of the generator.
Although all six terminals are usually accessible from outside
Figure
4. Two-pole hydrogen inner cooled generator axial cross
the generator housing, three are usually connected together, section.
forming the neutral connection and creating a ‘‘Y’’-connected

generator as shown in Fig. 3(a). The neutral is often grounded
through a large resistance to limit the current due to a phase- where IP
is the current in each phase (A) and VP is the phase-
to-ground fault. There is little or no current through this to-
neutral voltage in each phase (V).
ground connection during normal operation. A ‘‘delta’’ connec- A
commonly used formula for a ‘‘Y’’-connected generator,
tion is also possible as shown in Fig. 3(b) but not common assuming
balanced currents, is
for generators.
When the stator windings are connected to a load, the cur-
Power Output (kilowatts) = √3I V × PF
p T
rents that flow are also 120# out phase and thus create an
additional rotating flux wave. The wave travels at the speed where Ip
is the current in one phase (A) and VT is the phase-
of the rotor (synchronously) and generally opposes that of the to-phase
voltage (kV).
rotor flux at some angle depending on the operating phase
angle of voltage and current at the terminals of the generator.
EXCITATION SOURCE
Thus, to maintain a regulated voltage at the machine termi-
nals, the dc field current in the rotor windings (also called
There
are several common approaches for providing dc cur-
excitation current) must be adjusted depending on the stator
rent to
the generator rotor. One is to transfer dc current from
current and voltage magnitude and the phase angle differ-
a
stationary source to the rotor via stationary carbon brushes
ence. A common parameter describing this voltage and cur-
riding
on rotating ‘‘collector’’ or ‘‘slip rings.’’ Another means is
rent phase angle is called power factor (PF), which is
with a
‘‘brushless excitation system,’’ which eliminates the
PF = cos θ need
for brushes or slip rings. Excitation systems are dis-
cussed
in a later section.
where is the phase difference between stator voltage and
current.
CONSTRUCTION FEATURES
The real power output of a three-phase electric generator
in watts is given by
A
typical hydrogen-cooled generator axial cross section is il-
Power Output = 3IP ×VP × PF = 3IPVP cos θ
lustrated in Fig. 4. The rotor is made of a solid high-strength
magnetic
forging with slots machined to contain the field
winding.
When thestator phase currents are balanced, a
(a)
measurement on the rotor would observe no time-varying
magnetic
flux because the rotor and flux rotate together. The
A B stator
core is made of laminated ‘‘punchings’’ typically about
0.5 mm
thick that are insulated from each other. The stator
core
carries a time-varying flux that creates the voltages and
currents
in the stator winding. The electric conduction path
C in the
stator core is broken by laminating the core steel to
minimize
eddy currents in the stator core, which only gener-
(b) ate heat
and losses. The rotor and stator windings are usually
made of
copper, but aluminum is occasionally used. Bearings
B and
bearing supports required at each end of the generator
need
bearing lubricating oil, usually from the same source as
A C
for the
turbine bearings. For generators that use hydrogen
cooling,
a hydrogen-tight frame is required and hydrogen
gland
seals at each end of the generator where the rotor shaft
Figure 3. (a) ‘‘Y’’ connected generator. (b) ‘‘Delta’’ connected gen- must
extend through the frame to minimize hydrogen leak-
erator. age.
The stator core must be mounted to the frame and in
----------------------- Page 594-----------------------
TURBOGENERATORS 597
turn to a foundation. For large generators, the core is often slots.
One of the most substantial loads imposed on the rotor
spring-mounted to the frame to isolate core vibration induced is the
force needed to retain the windings in the slots at op-
by the rotating flux field from the frame and foundation. Ro- erating
speed. This is accomplished by placing a metal wedge
tor shaft lateral and torsional dynamics must be carefully cal- at the
top of the rotor winding in each slot. Loads from the
culated to avoid resonant modes at operating speed.
windings and the wedge are transferred to the ‘‘tooth’’ of the
unmachined portion of the rotor between the winding slots.

The
copper conductors must exit the slots at the end of
VENTILATION AND POWER DENSITY
the
rotor body, traverse the periphery of the shaft, and enter
another
slot in the rotor body to complete the circuit for the
The removal of heat from the generator is one of the most
winding. These ‘‘end-windings’’ are not contained within slots

important design challenges as ratings are increased. Ventila-
as are
the straight parts of the winding, but they still must
tion circuits must be designed to avoid any local overheating
be
supported against the enormous rotational forces. End-
in any of the components that generate heat and in other
windings are mechanically supported by high-strength cylin-

nearby components. Smaller generators (i.e., under 200 MW)
drical
retaining rings. The retaining rings are mounted on the
are air-cooled and simply take ambient air and discharge it
back to the surroundings (open-air-cooled or ‘‘OAC’’) or use a rotor
body by shrink-fits and/or mechanical attachments.
closed, recirculating air circuit with air-to-water heat ex- They
extend over the end-windings and provide the required
changers (cooler) in the path, called totally enclosed water-to-
centripetal forces to keep the copper conductors and insula-
air cooling (TEWAC). tion in
place during operation.
To design a generator with a significant increase in rating
Power to drive the turbogenerator comes from combustion
without a corresponding increase in size, better coolants must
turbines, steam turbines, or a combination of combustion and
be used. Hydrogen is commonly used for larger generators be- steam
turbines. The generator rotor must transmit the full
cause of its high heat capacity and heat transfer capabilities. torque
of all of the turbines in the shaft system. In addition
A hydrogen-cooled generator must use a closed cooling path to the
steady torque from the turbines, the shaft must be de-
with a cooler and requires reasonable safety precautions. Hy- signed
for the dynamic torque generated by a certain degree
drogen does not support combustion at purities above 80%, so of
start-stop cycles, out-of-phase synchronization, system
hydrogen purity inside the generator must be kept well above faults,
and other abnormal operating conditions that may oc-
this level. As generator ratings become even larger, direct wa- cur
over the generator’s design life (often as long as 40 years).
ter cooling is used for some of the active components, particu- The
generator stator surrounds the rotor and provides the
larly the stator winding. The largest ratings usually utilize support
for the stator windings (or bars) that produce the
hydrogen-cooled generators with direct water cooling of the
electrical output of the machine. The stator windings are
stator winding. A special water treatment and cooling system placed
into slots on the inside the stator core. As in the rotor,
is needed for this water to keep the water deionized and the
windings are retained in the slots by wedges. The wedges
therefore ‘‘nonconducting.’’ Figure 5 illustrates the variation must
keep the windings tight within the slots and transfer
in megavoltamperes per active generator volume for different the
electromagnetic forces from the bars into the ‘‘teeth’’ of
ventilation categories. the
core.
The
stator core is the stationary part of the generator’s
magnetic circuit. It is composed of thousands of individual

BASIC MECHANICAL DESIGN CONSIDERATIONS
punchings (thin sheets of silicon steel) stacked to form a long

hollow
cylinder. The individual punchings are designed and
There are two major mechanical components in a turbogener-
stacked
in a precise pattern so that the final core has slots for
ator, the rotor that produces the magnetic field and the stator
the
windings, passages for the cooling gas, longitudinal holes
that produces the electric output of the machine.
for
through-bolts to keep the core tight, and provisions for
Rotors are long cylindrical shafts supported on hydrody-
supporting the outside of the core to the frame. The stator

namic bearings. The main body of the rotor has long longitu-
core
must support the windings against the magnetic and
dinal slots machined into its outer surface. The copper wind-
electrical forces and transfer these forces to the stator frame.

ings that produce the magnetic field are inserted into these
The
stator frame is the main structural component of the
stator.
It must transfer the forces and torque from the core
into
the foundation of the building or supporting structure.
) 2000
A The
powerful magnetic field of the rotor attracts the core steel
V Direct-cooled with hydrogen,
M to the
north and south poles of the rotor body. This causes
( water-cooled stator, 4-pole
r 1500 the
core to assume an oval shape. As the rotor turns, the oval
o
t Direct-cooled with hydrogen,
a shape
of the core moves to align itself with the magnetic poles
r water-cooled stator, 2-pole
e
n 1000 of the
rotor body. This rotating oval induces vibrations in the
e
g Direct-cooled with
frame
at the attachment to the core.
m hydrogen, 2-pole
u 500 If
the vibrations are significant, they could damage the
m
i Air-cooled, 2-pole
foundation supporting the generator and cause metal fatigue
x
a
M 0 of the
frame support. The noise from uncontrolled vibrations
0 1 2 3 4 5 6
2 can
also become excessive. Small generators directly connect
Per unit active core volume, D L
the
core to the frame because the magnitude of the core defor-
Figure 5. Generator maximum megavoltamperes versus active vol- mation
is small. Large generators with high power densities
ume for different types of cooling. must
provide flexible mounting for the core so that vibration
----------------------- Page 595-----------------------
598 TURBOGENERATORS
from the core is not transferred to the frame and the foun- ducted
through the core steel to the vents where the heat is
dation.
transferred to the gas. The frame is designed to direct the
Some frame designs include the bearings that support the
circulation of the cooling gas, and it is cooled by direct contact
rotor. Then the weight of the rotor must be carried by the with the
gas.
frame and transferred to the foundation. Other generator de- The
windings may be cooled by one of three methods: con-
signs have pedestal bearings mounted outside the generator ventional
cooling, direct cooling, or liquid cooling.
frame. The pedestals support the rotor and transfer the loads In
conventional cooling the heat generated in the copper is
to the foundation, allowing a smaller frame. conducted
through the insulation that surrounds the wind-
Hydrogen-cooled generators use pressurized hydrogen gas ings
into the core where it is conducted through the steel to
to remove the heat generated in the copper conductors and the
vents. This is the simplest cooling method, and it has been
the core. The frame must be designed as a pressure vessel to used
since the beginning of the electrical age. Heat conduc-
retain the hydrogen within the generator and prevent its es- tion
through insulating materials is not very efficient, so this
cape to the atmosphere. Because hydrogen and oxygen com- method
of cooling seriously limits the power density of the
bine to form explosive mixtures, the frame must be designed
generator.
to contain any explosion that might occur. The explosion pres- Direct
cooling is achieved by providing cooling passages
sure may be many times the hydrogen pressure, so large gen- within
the windings. This is normally accomplished by em-
erator frames are typically made from steel plate several bedding
nonmagnetic metal tubes within the winding. The
inches thick. The frame of hydrogen-cooled generators must
circulating gas is forced through the cooling tubes. Conduc-
also support coolers to remove the heat picked up by the hy- tors
within the windings are electrically isolated from each
drogen gas. The coolers are usually finned-tube heat exother
and the cooling tubes by a thin covering of insulation.
changers. Hydrogen gas flows across the fins mounted on The
cooling tubes are in direct contact with the thin layer of
banks of tubes through which a liquid coolant passes. The
insulation resulting in much improved heat transfer.
generator frame must support the weight of the cooler, and it Liquid
cooling is the most efficient method of removing
must also provide gas passages for the hydrogen to circulate. heat from
the copper conductors. The winding is designed so
The hydrogen gas is circulated through the generator by ei- that
approximately every third copper conductor is hollow. At
ther an external motor driven blower or a rotor-mounted fan each end
of the winding, all of the hollow conductors within a
or blower.
half-coil
(or bar) are joined together into a header. Liquid
coolant
is pumped into a manifold at one end of the generator.
Insulated
tubes connect the manifold to the header at one end
THERMAL DESIGN
of the
winding, and coolant passes through each of the hollow
conductors. Heat from the conductors is transferred to the liq-

Very large currents are carried by the copper conductors
uid
coolant. The header at the opposite end of the winding
within the generator. The currents heat the copper to high
collects
the liquid coolant from each hollow conductor and the
temperatures, and this heat must be removed if the generator
coolant
flows through another insulated tube into another
is to function reliably in continuous operation.
manifold.
The coolant from the manifold passes out of the
The rotor windings are normally cooled by forcing gas (air
generator, and the heat is rejected to the atmosphere through
or hydrogen) through passages around or within the copper
a cooler.
conductors. The gas picks up the heat from the conductors

and then is expelled from the rotor. Then the circulating gas
is exhausted from the generator (open air cooled designs), or
DYNAMICS
AND VIBRATION
it passes through a cooler (TEWAC or hydrogen-cooled de-
signs) which removes the heat from the gas.
Several manufacturers have developed liquid (oil or wa- The shaft
systems of large turbogenerators are subject to two
ter)-cooled rotor windings to improve the heat transfer from types of
vibration. Lateral vibration is characterized by the
the copper conductors and improve the efficiency of the gener- beam-
bending of the shafts between the bearing supports.
ators. Liquid-cooled rotors must have complex mechanisms to Torsional
vibrations arise from the axial twisting of the rotor
circulate the cooling liquid into and out of the rotor while it sections
with respect to each other.
is turning at high speed. The connections within the rotor The
lateral natural frequencies of shaft systems are called
windings must also be leakproof to prevent the cooling liquid the
critical speeds because the vibration of a rotating shaft
from contaminating the inside of the generator. The complica- system
can become very large at certain rotational speeds
tions of liquid-cooled rotors have limited the number of units when the
natural frequency of the shaft system matches the
that utilize this cooling method. rotating
speed of the shaft. Small imbalances in the shaft sys-
Heat is also generated in the stator components during op- tem
excite large vibratory motion at critical speeds. Clearly,
eration. Large alternating currents are generated within the it is
essential that a turbogenerator shaft system be designed
stator windings. These currents heat the copper conductors to so that
the operating speed does not match a critical speed.
high temperatures. The rotating magnetic field also heats the It is
also important to eliminate as much of the residual
core and the frame. All of this heat must be removed if the imbalance
in the shaft system as possible. To accomplish this,
unit is to operate continuously. turbine
and generator shafts are designed and machined to
Usually the core and frame are cooled by the cooling gas precise
tolerances to eliminate as much mass imbalance as is
(air or hydrogen) which circulates through the generator. The
practical. Numerous balance planes are designed into the ro-
core is constructed with axial and/or radial vents for the cool- tors and
shafts so that residual imbalance can be reduced by
ing gas. Heat generated in the windings and the core is con- adding
balance weights to the shafts.
----------------------- Page 596-----------------------
TURBOGENERATORS 599
Torsional vibrations in shaft systems may arise due to un-

applications of turbogenerators. They are routinely refer-
balanced loads in the three-phase electric system. Phase im- enced
worldwide.
balance results in an oscillating torque on the shaft system at
Following are some of the specific standards that define
twice the electric system frequency (120 Hz for 60 Hz systems the
requirements and procedures for designing and testing
and 100 Hz for 50 Hz systems). Short circuits, transmission turbine
generators:
line switching, and out-of-phase synchronization also briefly
excite shaft system frequencies that lie near the electric sys- •
ANSI C50.10 ANS for Rotating Electrical Machinery—
tem frequency (50 Hz or 60 Hz).
Synchronous Machines
If one of the torsional natural frequencies of the shaft sys- •
tem is at or close to twice the electric system frequency, the
Cylindrical-Rotor Synchronous Generators
magnitude of the torsional vibrations become very large, pos-
•
ANSI C50.14 ANS Requirements for Combustion Gas
sibly large enough to produce fatigue cracking of shafts, tur-
Turbine Driven Cylindrical Rotor Synchronous Genera-

bine blade, or other rotating components.
tors
Torsional vibrational problems are generally more serious
than lateral vibrational problems for two reasons. First, there •
is no way to ‘‘balance’’ torsional vibrations. The magnitude of
Hydrogen-Cooled, Combustion-Gas-Turbine-Driven, Cy-
the lateral vibration is controlled by how close the critical
lindrical-Rotor Synchronous Generators
speed is to the operating speed and by the magnitude of the •
IEEE Std 115 IEEE Guide: Test Procedures for Synchro-
residual imbalance. If the magnitude of a lateral vibration is
nous Machines
too large, the system can be balanced to reduce the amplitude •
IEC 34-1 Rotating Electrical Machines Part 1: Rating
of vibration. Torsional vibrations can arise in turbogenerator
and Performance
shaft systems that have natural frequencies very near single
•
IEC 34-2 Rotating Electrical Machines Part 2: Methods
or double line frequency. This is not something that can be
for determining losses and efficiency of rotating electrical

easily corrected and often requires major design modifications
machinery from tests

to correct.
Another reason why torsional concerns are difficult to re- •
IEC 34-3 Rotating Electrical Machines Part 3: Specific
solve is because there is very little damping present in tor-
requirements for turbine-type synchronous machines
sional vibrations. The oil films in the bearings provide sub- •
IEC 34-4 Rotating Electrical Machines Part 4: Methods
stantial damping for lateral vibrations, but not for torsional
for determining synchronous machine quantities from
vibrations.
tests.
The best approach in avoiding torsional vibrational prob-
lems is in the initial (or upgraded) design of the turbine-gen- The
standards specify basic operating requirements, define
erator, so that the rotor system is designed to avoid excitable rated
and off-rated output, describe normal and emergency
torsional natural frequencies near single or double line fre-
operating conditions. Definition of rating is principally related
quency. Once operating, torsional vibration problems are re- to
temperatures reached by various generator components,
solved only by making substantial design changes to the gen-
primarily insulation. Insulating materials used in turbogen-
erator rotor or the turbine shafts. The stiffness of the shaft erators
are categorized in classes according to the tempera-
system can be changed by reducing section diameters, or ture
they withstand without damage. For each class of insula-
mass can be added at locations along the shaft. tion,
temperature limits are specified for various components
of the
machine along with the method of measurement. The
intent
of these limits is to provide reliable operation of the
APPLICABLE STANDARDS AND CODES—SUMMARY machine
throughout its design life. In addition to the base
OF BASIC REQUIREMENTS rating,
peak and peak reserve capabilities are specified. Tem-
peratures in the generator, at these special ratings, are al-

Turbogenerators are usually designed and built to certain na- lowed
to be higher than at base rating at the cost of ac-
tional or international standards. Standards provide a base
celerated insulation life consumption, which is generally
for objectively comparing machines of different designs. They
considered acceptable if it is of relatively short duration and
also allow customers to specify and acquire products with con-
infrequent (exceptionally hot summer or exceptionally cold
sistent features without going deep into the details and intri- winter
day, unexpected outage of other generators, etc.).
cacies of design and manufacture. The
conditions in which a generator operates are deter-
Standards are usually established by national, interna- mined by
the temperature and pressure of the available cool-
tional or professional organizations. The most commonly used ing
medium and the electrical parameters (voltage and fre-
standards applied to turbogenerators are established by the quency)
at its terminals. Enclosed machines operating in
American National Standards Institute (ANSI), the Interna-
pressurized hydrogen gas are typically operated with con-
tional Electrotechnical Commission (IEC), and the Institute stant
hydrogen pressure and cold gas temperature. The speci-
of Electrical and Electronics Engineers (IEEE). ANSI and fied
value of this temperature is such that it can be readily
IEC are the two main governing sets of standards. ANSI is
maintained in most typical situations. If this is not possible,
dominant in North America (predominantly 60 Hz), and the
agreements must be made for special design features. Air-
IEC standards are widely used elsewhere (predominantly 50 cooled
machines depend on their environment for pressure
Hz, except in Brazil, South Korea, Japan, Taiwan, and a few and, in
some cases, for cold air temperature. The standards
other places that use 60 Hz power). IEEE standards deal specify
how the rated output of the generator must be ad-
more with specific technical aspects of testing, control, and justed
so that the highest temperatures reached by its compo-
----------------------- Page 597-----------------------
600 TURBOGENERATORS
P
jX
P E · E S
P = δ
E · sin( )
X
I
a
P
max
E F ET
PT B
A
Figure 7. Equivalent circuit.
δA δ
the angle # is, that is, the larger P is with respect to P .
A max T
This ratio can be improved by making the variable X smaller.

Figure 6. Steady-state stability.
This parameter is the total reactance between the internally
generated voltage in the generator and a point in the power
system that is considered sufficiently strong and remote from

nents stay within required limits for the insulation class, that
the generator not to be affected by the generator’s perfor-
is, at higher altitude and/or higher cold air temperature the
mance. The internal reactance of the generator is a significant
rating of the generator may be reduced.
portion of this reactance. Transmission lines and transform-
Even during perfectly normal operation, the voltage at the
ers represent the balance. Similar considerations are in place
generator’s terminals and the frequency at which it operates
during transient conditions.
can normally be maintained only within a range around the
Parameters that determine the generator’s interaction
nominal values rather than exactly at those values. The stan-
with the power system are its reactances, time constants, and
dards specify the voltage and frequency ranges in which the
some other parameters. Reactances of a synchronous genera-
turbogenerators are required to operate reliably. The speci-
tor refer to its representation by a simple equivalent circuit
fied temperature limitations can be exceeded during off-nomi-
shown in Fig. 7. A device as complex as a turbogenerator,
nal voltage and/or frequency operation within these ranges,
consisting of a strongly nonlinear magnetic circuit and a num-
as long as they do not pose a threat to the generator.
ber of mutually moving, coupled electric circuits, cannot be
Emergency situations, such as lightning strikes, short cir-
represented by constant parameters in a simple linear equiv-
cuits, erroneous switching, and short-time overloads, are in-
alent circuit. Therefore, several separate sets of parameters
evitable in a power system. Standards define the types and
are, defined for the circuit in Fig. 7, that properly represent
severities of the disturbances that turbogenerators must
the machine’s behavior in various situations.
withstand and the acceptable amount of damage they can suf-
The subtransient reactance and subtransient time con-
fer from extreme conditions.
stant X
and T
describe the machine in the first few cycles
d d
following a major disturbance (sudden short circuit, out-of-

Reactances and Machine Parameters
phase synchronization). These parameters control the genera-
tor’s current and its rate of decay immediately after the short
The primary purpose of a turbogenerator is to deliver electric
circuit and determine the demands on the circuit breakers

power to a power system. This delivery, however, must satisfy
and the electromagnetic forces experienced by all equipment

certain conditions to make it acceptable by the system, partic-
between the generator and the short-circuit location.

ularly a modern, complicated system. The generator must re-
After the initial period, transient reactance and time con-

liably run in step (synchronously) with the system, it must
stant X # and T# better describe the machine and its ability to

properly support the voltage in its vicinity, and must not un-
d d
recover synchronous operation with the system after the fault

necessarily increase the short-circuit severity in its vicinity.
has been cleared. The synchronous reactance X d or its related

The concept of stability can be understood by considering
representation, the short circuit ratio, determine how close to
Fig. 6. The power PE delivered by the generator to the system
the stability limit the generator operates during normal

is proportional to the voltage EF induced in the generator, the
steady-state operation. The rated power factor determines the

voltage ES of the power system and the sine of the angle of
rotor rotation # representing the time by which the peak of
the generator internal voltage precedes the peak of the re-
mote system voltage. This power has a maximum Pmax taking
Steam from boiler
place when the generator voltage precedes the system voltage
or nuclear plant
by one-quarter cycle (90#). The power PT supplied by the tur-
bine is constant for a particular setting of the turbine’s gover-
nor. In normal steady-state operation, delivered power PE
must be equal to the turbine power PT (minus generator loss
ST Generator
which is relatively small), and the angle # adjusts itself ac-
cordingly (# ). The generator is said to be stable if this bal-
A
ance is established at the rising slope of the power curve,
Electrical
point A in Fig. 6. The equilibrium at point B is unstable be-
output
cause an increase in angle # causes a reduction of delivered
power. The generator’s operation is more stable the smaller
Figure 8. Simple cycle steam turbine drive.
----------------------- Page 598-----------------------
TURBOGENERATORS 601
magnetic loss due to ac magnetic fields
electric loss due to current flow through windings

('Free')
steam Heat-
mechanical loss from friction in bearings and ambient gas
recovery
steam
ventilation loss, that is, power to move the coolant needed
generator Hot CT
for heat removal
exhaust
According to their dependence on machine output, the
losses are constant losses and load-dependent losses.

ST Generator CT Generator
Because neither of the two classifications is very conve-
nient for practical purposes, the standards define the sched-

CT
ule
of losses with precise definitions. Such a schedule is given
Electrical starting
Electrical motor here
with a very brief description. More detail can be found
Fuel output
output
elsewhere, such as ANSI C50.10 or IEC 34-2.
Figure 9. Dual-shaft (standard) combined-cycle arrangement.

2
•
Armature I R loss: load-dependent electric loss due to
load current flowing through the dc resistance of the ar-
mature winding
generator’s ability to supply reactive power to the system, and •
Stray load loss: load-dependent electric and magnetic
contribute to the local area voltage support.
loss caused by load in various members of the machine
(eddy loss in windings, additional loss in core, rotor sur-

Losses and Efficiency
face loss, loss in various structural parts)
•
Core loss: magnetic loss caused by the alternating mag-
Although the efficiency of synchronous turbogenerators is
netic flux in the stator core (constant for operation at

very high (98% and higher), considering and minimizing
constant terminal voltage)

losses is of great importance. The losses in a turbogenerator
•
Field I 2R loss: load-dependent electric loss due to dc cur-
are the main limitation on the output power, that is, the out-
put is limited by the ability to remove the heat generated (re-
rent flow through the rotor winding
sulting from losses) in the generator. From an investment •
Exciter loss: sum of all losses in the equipment that pro-
perspective, every kilowatt of loss saved is an extra kilowatt
vides exciting power for the generator
available for sale. For both these reasons, it is important to •
Frictional and windage loss: generally constant mechani-
predict a turbogenerator’s losses precisely and to minimize
cal loss caused by friction in the bearings and gland seals
them.
(friction) and friction of all rotating parts against the gas
Various types of losses in a generator are related to design
that surrounds them (windage)
features, material properties, and output. The relationships •
Ventilation loss: constant power needed to move the cool-
are complex, nonlinear, and often partially uncertain (proper-
ing media (air, hydrogen, water) through the machine’s
ties of actual materials that eventually become part of the
cooling passages
generator are known only within some tolerances at the time
of design).
Testing
According to their physical nature, the losses are catego-
rized as follows:
Numerous tests are performed during manufacturing and in-
stalling a turbogenerator. Components of a generator are
tested during the manufacturing process to assure that they

are
of high quality and that the completed generator meets
the
final requirements. Particular attention is directed to
nondestructive structural testing of the rotor forging, dielec-

Heat- tric
properties of armature coils, and proper performance of
('Free') recovery the
stator core.
steam steam
Once manufacturing is completed, running tests are often

generator Hot CT
performed, of which there are two types. Commercial testing

exhaust
is
performed to verify the performance of the generator and
compliance with contractual agreements. These often apply

only
to the first of a design but depending on the contract,
ST Generator CT
could be required for subsequent units. Most commercial tests
are
specified by standards. Development testing is much more
Fuel
detailed and is usually performed only on the prototype for a
Aux.
Clutch Static-start power new
design or if a substantially new design feature is intro-
power and
duced. The types of tests and measured variables depend on
controller the
type of development that is being investigated.
Large turbogenerators cannot be tested at their normal op-

Electrical
erating conditions in the manufacturing plant. No manufac-

output
turer in the world owns a facility that could test a turbogener-

Figure 10. Single-shaft combined-cycle arrangement. ator
at its full rating because it requires a tremendously large
----------------------- Page 599-----------------------

602 TURBOGENERATORS
drive mechanism and electric load. Such tests are done only duces power
in a steam turbine. This configuration is referred
in the power plant after the generator is installed. Because a to as a
combined-cycle plant. Normally, the CT and the ST
power plant environment poses substantial limitations on drive their
own independent generators. If the CT and the ST
such testing, most manufacturers test generators in their own of the
combined-cycle plant are attached to opposite ends of
test facilities, using testing methods that validate the perfor- the same
generator, the configuration is called a single-shaft,
mance of the generator. These test facilities require a drive combined-
cycle arrangement.
mechanism sufficient to bring the generator to the required A few
fundamental issues must be addressed when select-
speed and compensate for losses. These tests are the steady- ing an
electric generator to match a particular prime mover.
state, open-circuit test; steady-state, short-circuit test; and First, the
generator’s real power output rating must meet or
sudden short-circuit test. exceed that
of the prime mover(s). Additionally, the generator
In a steady-state, open-circuit test the generator is driven must
provide reactive power to the power system as needed,
at constant rated speed without load (terminals open-cir- which is
usually specified as a power factor. The power output
cuited). Several consecutive values of excitation are applied, of many
combustion turbines varies according to their inlet
and the resulting terminal voltage is measured. The shaft in- air
density, and thus is higher on cold days and lower on hot
put power and excitation power are also measured. This test days. One
way to assure that the generator capability meets
determines the open-circuit saturation curve (correlation be- or exceeds
that of the CT is to map both CT output and gener-
tween induced voltage and applied excitation current) that ator
capability against ambient air temperature and com-
provides information for calculating the excitation required at pare them.
load and, to a significant extent, validates the magnetic cir- A second
important consideration is to determine the type
cuit design. The measured power supplies information for de- of
generator needed. Most synchronous generators rated over
termining the losses that are independent of the load (core 50 MVA use
water cooling, hydrogen cooling, or air cooling for
loss, friction and windage, ventilation). the stator
winding. The rotor cooling method can be specified
The steady-state, short-circuit test is performed with the
independently, but because water-cooled rotors are relatively
generator’s terminals short-circuited (zero voltage) and the rare, most
machines that use water cooling for the stator
field excited until a desired armature current is established. winding use
hydrogen cooling for the rotor. Consequently, a
The rotor is driven at rated speed. Excitation current, arma- ‘‘water-
cooled’’ generator, as the term is often used, generally
ture current, drive power, and excitation power are measured. describes a
generator with a water-cooled stator winding and
This is repeated at several current values. This test supplies hydrogen
cooling for the rotor winding, core, and other active
information about load-dependent losses, mainly the stray parts.
Water- and hydrogen-cooled generators have the ad-
load loss. vantages of
being more efficient and less susceptible to envi-
Both open-circuit and short-circuit, steady-state tests are ronmental
contamination. They are also capable of higher
usually performed at a few selected voltage or current levels, power
densities. Their disadvantages are cost, complexity,
respectively, for as long as it takes to reach thermal steady and
maintenance. Because of advances in technology, today’s
state. Then the temperatures reached by various machine air cooled
generators obtain relatively high power densities,
parts are recorded. This supplies information about the ade- making them
a very inexpensive and attractive alternative to
quacy of the thermal design. water- and
hydrogen-cooled generators for sizes under 150
The adequacy of the ventilation design, including pres- MW to 200
MW.
sures at various key points in the generator and gas flows
Generator capability is defined by the output achieved
through the machine’s components, is determined in a sepa- while
keeping temperatures inside the generator within lim-
rate ventilation test when the machine is run unexcited at iting
values listed in industry standards. Therefore the ‘best’
constant speed while the measurements are taken. generator
is defined as the least expensive generator that
Sudden short-circuit tests are performed to measure the meets or
exceeds its design requirements. With that in mind,
machine’s transient parameters (transient and subtransient the
following concepts can be utilized to maximize a genera-
reactances and time constants), and to verify its ability to tor’s
capability:
withstand such events without structural or other damage, as
specified by standards. 1. Water
provided to the generator for cooling should be
Before shipping, testing is concluded with a thorough inas
cool as possible. This allows the lowest generator
spection and final dielectric tests of the stator and rotor wind-
temperatures and the highest generator capability.
ings, as required by standards.
2.
Electric power factor requirements for the generator
also
contribute to the generator size. As the rated power
TURBOGENERATOR APPLICATIONS
factor decreases for a given real power output, the re-
quired reactive power increases, as do the stator cur-

rent
and rotor current. These increased currents cause
Several different configurations of prime movers for genera-
tors are possible. The most common are the simple-cycle more
heat that restricts capability. Therefore, reducing
steam or combustion turbine and the combined-cycle configu- the
reactive power output requirement of a generator,
rations. A simple-cycle configuration consists of a steam tur- often
allows greater real power output capability.
bine (ST) or a combustion turbine (CT) directly driving a gen-

erator at synchronous speed. For added efficiency, the hot In a
simple steam arrangement, the boiler must produce
exhaust of one or more combustion turbines is recycled steam
before the steam turbine can produce power. After
through an air-to-water heat exchanger, called a heat-recov- steam is
available, the turbine generator unit can be brought
ery steam generator (HRSG) to create steam. The steam pro- up to full
speed, and power produced.
----------------------- Page 600-----------------------
TURBOGENERATORS 603
In a combustion turbine unit, the CT must be brought to variable

loss, which occurs naturally in an oversized genera-
roughly 40% of synchronous speed before it can support com- tor,
often results in higher efficiency at the generator maxi-
bustion. Spinning the CT to start it is often accomplished by mum
output. This is shown in Fig. 11. As generator size in-
using a starting motor. Once the CT has started and acceler- creases,
the generator efficiency curve shifts to the right.
ates itself, a clutch mechanism commonly disengages the Although
the curve maximum still peaks at approximately
starting motor from the turbine generator train. In some the
same efficiency value, peak efficiency is closer to maxi-
cases, the CT is started by a mechanism called ‘static start.’ A mum
output for the larger generator. In air-cooled generators,
static start package is a power electronic circuit that converts fixed
loss often constitutes a larger percentage of the total, so
auxiliary ac power into controlled, variable-frequency, three- maximum
efficiency is generally obtained by using the small-
phase ac. Then this is sent into the generator armature wind- est
generator that satisfies the load requirements. Because
ing and temporarily utilizes the generator as a synchronous of the
relatively large amount of power necessary to provide
motor, accelerating the CT up to starting speed. Once starting
ventilation to air-cooled machines, it is normally advanta-
speed is attained, the CT is used to bring the entire CT-gener- geous to
operate these machines at elevated temperatures, if
ator train up to rated speed. at all
possible, by using insulation with high thermal capabil-
Combined-cycle units normally have their CT unit(s) ity
(Class F or Class H insulation, for example).
started first. When the CT is operating at full speed, the ex-
haust is sent to the HRSG to make steam. After sufficient Abnormal
Operating Conditions
steam pressure is available, the steam turbine portion of the
The
following are among the anticipated abnormal operating
combined cycle can be started. If the combined-cycle unit has
conditions that affect the size and selection of a new gen-

a single-shaft arrangement, a clutch mechanism keeps the
erator:
steam turbine disconnected while it is being brought up to
speed. This clutch can be engaged and disengaged at rated
1.
Extended operation at other than nominal frequency
speed so that a single-shaft, combined-cycle plant can operate
as simple cycle or combined cycle as desired. 2.
Extended operation beyond normal voltage limits
3.
The existence of system-negative sequence sources
Generator Efficiency—Optimization 4.
Low power factor operation, including synchronous con-
denser operation
Optimizing a generator for efficiency is very important. For
example, an efficiency improvement of only 0.01% on a 250
Before discussing abnormal operating conditions, it is nec-
MW generator results in an additional 25 kW available for essary
to establish what normal operating conditions are.
sale. At 5 cents per kilowatt hour, this totals roughly $10,000 Normal
operating conditions are generally recognized as
per year, assuming a 90% availability factor. Amortized at those
specified in industry standards.
10%, the resulting energy savings are roughly $100,000 in The
generator terminal voltage is directly proportional to
present worth. Additionally, each additional kilowatt repre- the flux
density inside the generator core. One of the concerns
sents capacity that normally costs in excess of $500/kW. The of
operating with a continuous overvoltage is that the genera-
corresponding savings, considering both energy and capacity tor
internal flux density exceeds design levels, possibly dam-
savings, are roughly $4500 per kW of efficiency enhancement. aging
core steel laminations. If the generator voltage is too
Before an effective efficiency study can be done, a loss eval- low and
this results in underexcited operation, core end-
uation factor in $/kW saved must be established using appro-
priate parameters for the application. This evaluation factor
is a guide in determining if changes to an existing generator
Operating
design are warranted. The costs of the changes are compared
load of
to the dollar savings achieved by the efficiency enhancement.
Generator efficiency vs. load generator
Efficiency enhancements are occasionally obtained by uti-
lizing excess generator capability if it is available. For exam-
Curve shifts to
ple, consider a generator with excess capability (defined by
right by using a
its low operating temperatures). If the generator is hydrogen-
slightly larger
cooled using 4 atm of hydrogen pressure, the generator could y
generator
c
have its hydrogen pressure reduced to 3 or 3.5 atm, as the n
e
i
c
design allows. This would raise internal temperatures be- i
f
f
cause of reduction in heat transfer capability of the hydrogen, e
Efficiency
r
change due to
o
but it would also reduce hydrogen density and decrease wind- t
a
change in static
age loss, thereby reducing generator loss. Other modifications r
Efficiency
e
H gas presure
n
2 gain by
can sometimes be effected to enhance efficiency. For example, e
a stator or rotor can sometimes be rewound with an optimized G
Changing static using a
H2 gas pressure slightly

coil to enhance efficiency, but this most often results in re-
shifts the efficiency oversized
duced generator capability.
curve along this line generator
When selecting a new generator, efficiency may be en-
hanced by selecting a slightly oversized generator, particu-
Generator load
larly for hydrogen- and water-cooled generators, which have
relatively low fixed loss. Because the point of maximum effi- Figure
11. Variation in generator efficiency with size and gas
ciency is where fixed loss equals variable loss, a reduction in
pressure.
----------------------- Page 601-----------------------
604 TURBOGENERATORS
Generator
rated capability:
Rated MVA and

Overexcited
Adjusting generator lagging (overexcited)
field
current moves
operation
power factor; typical
(reactive the
generator along lagging power factor
power out of a
vertical line. Moving is 0.90 or 0.85
generator up and
down this line
(lagging pf) changes
generator
reactive loading, but

Generator also
causes a change in
reactive
generator terminal voltage
power
output
Generator real output (MW)
(MVAr)
Underexcited
Generator real
operation power
output
(reactive moves
along this line
power into
generator)
Generator maximum
(leading pf)
output at rated leading
(underexcited) power
Figure 12. Generator reactive capability
factor; typical leading
curve.
power factor is 0.95
region heating concerns can arise. Also, to achieve rated

current because it results in flux components that rotate in a
power output, reduced voltage corresponds to higher current
direction opposite to that of the rotor. The resulting flux vari-
that may be excessive. ations
due to negative sequence current cause rotor surface
Operation at reduced frequency causes concerns of core
heating because of induced eddy currents.
heating similar to those from operation at excessive voltage.
Unbalanced operation of a three-phase generator is typi-
Voltage in a generator is defined by the following relation- cally
specified by limits of negative sequence current. Stan-
ship: dards
require generator manufacturers to allow up to 10%
continuous negative sequence current, depending on the rat-

Volts ∝ d Flux ing
of the generator. Smaller continuous negative sequence
dt limits
are allowed for larger generators, which are assumed to
be
connected to high-voltage transmission systems with small
or, equivalently,
amounts of unbalance. Voltage unbalances during operation
are
not specified by industry standards but they induce nega-
Volts
∝ Flux tive
sequence currents and are taken into account in the limi-
Hertz
At normal operating frequency, the flux in the generator

core is dictated by the terminal voltage. When the system fre-
IEC34-1 Voltage (per unit)
quency is reduced, more flux is required to maintain the same
Generator voltage and 1.08
frequency band
terminal voltage, which could result in a high flux condition
in the generator core. Therefore, it is important to operate
within the voltage-frequency band as shown in Fig. 13, which
1.05
approximates the requirements of the IEC standards, or as
dictated by the manufacturer.
Continous
During start-up conditions, the frequency changes rapidly
operating Frequency
as the unit is brought to operating speed. At low frequencies,
zone (per unit)
the generator terminal voltage must be kept low (or zero) to
0.95 0.98 1.02 1.03

prevent overfluxing the generator core. During a typical
startup, field excitation is not applied until the generator
0.97
achieves synchronous speed. Precautions must be observed to
prevent overfluxing the generator core, particularly with the
unit off-line during which flux increases cannot be con-
0.95
Short-time
strained by armature reaction for corresponding increases in
operating
excitation level.
zone
Balanced three-phase current in the generator stator is
0.92
sometimes called ‘positive sequence’ current because it cre-
ates a flux wave that rotates in the same direction as the
rotor. When stator currents become unbalanced, the magni- Figure
13. Normal voltage and frequency conditions for turbogener-
tude of the current unbalance is called ‘negative sequence’ ators.
----------------------- Page 602-----------------------
TURBOGENERATORS 605
Rated over excited reactive power
Overexcited
operation
(reactive
power out of Operate generator as
"synchronous condenser"
generator
(lagging pf) along this line. Slight motor
action in real power axis
Generator and generator action in
reactive reactive power axis
power
output
(MVAr) Generator real output (MW)
Generator
real
Underexcited power
output is
zero
because no
operation
(reactive No prime mover
exists
power into operation
generator) region
(leading pf)
Rated under excited reactive power
Figure 14. Operation as a synchronous compensator.
tations. Current unbalances normally result from system

tor to have more field current than when operating underex-
load, transmission line, transformer unbalance, or from un-
cited. Higher field current causes higher internal generator
balanced system disturbances, not from the generator.
temperatures from rotor winding I2R losses. Consequently, re-
For generators, a lagging power factor indicates overex-

quirements for operation with lower lagging power factor typ-
cited operation and means that the line current out of the
ically require a larger generator.
generator is lagging (or behind) the voltage in time. Likewise,
Although operating with a leading power factor requires
a leading power factor indicates underexcited operation, or
less field current than with a lagging power factor, other is-
current leading the voltage in time. Typically, generators are
sues arise which limit the leading power factor to a value typ-
specified with lagging power factors in the range of 0.9 to
ically no less than 0.95. Magnetic flux in the end-region of a
0.85, whereas leading power factors are routinely specified as
generator can travel axially, entering the end of the core nor-
0.95. Lagging power factor operation allows the generator to
mal to its surface. This axial flux causes many unwanted eddy
compensate for inductive power system loads and support low
currents and results in localized high temperatures. The axial
system voltage by producing reactive power (or vars). A lead-
magnetic flux component from the stator circuit opposes that
ing power factor allows the generator to compensate for a ca-
of the rotor circuit, and together, the rotor and stator circuit
pacitive power system and hold down high system voltage by
produce very little axial flux. During underexcited operation
absorbing vars. In general, power systems are more inductive
(leading power factor), however, the stator circuit produces its
than capacitive, requiring generator operation with a lagging
normal amount of axial flux, which is counteracted by a
power factor. Leading (underexcited) power factor operation
weaker rotor component. This condition causes a net higher
is sometimes required in metropolitan areas with large un-
axial end-region flux and results in higher end-region local-
derground cable systems, particularly during light load, when
ized temperatures. If it is essential that the generator be ca-
there is a surplus of capacitive reactive power in the system
pable of leading power factors lower than 0.95, this may re-
from cable charging. Operating a generator overexcited (lag-
quire increasing the generator size or capability.
ging power factor), which is more typical, requires the genera-
Synchronous Compensator Operation
A generator without a prime mover is often called a ‘‘synchro-
nous condenser’’ in the

United States and a ‘‘synchronous
X ’’ X
compensator’’ elsewhere. Synchronous compensators are used
t
~
for steady-state and dynamic voltage control. They supply or
absorb reactive power to support and control local system

V’’ = V + jI (X ’’ + X )
voltage. Some synchronous compensators are synchronous
sys t
generators at ‘retired’ power stations. The losses

in the com-
Vsys
pensator are supplied by the power system itself through syn-
Turbogenerator Transformer
chronous motor action in the generator. As reactive power is
needed, excitation is adjusted to allow operation at zero real

Figure 15. Model used for short-circuit studies.
power output. Applying a generator for operation as a syn-
----------------------- Page 603-----------------------
606 TURBOGENERATORS
chronous compensator generally necessitates carefully evalu- put

power. The H constant for a typical turbogenerator (in-
ating the generator terminal voltage and step-up transformer cluding
the connected turbine and exciter) is typically on the
tap settings to optimize the performance of the compensator. order
of 2 to 4 MW-s/MVA of turbogenerator capacity. Al-
though
certain oscillatory stability incidents may last up to
several
minutes, most stability studies are concerned with the
System Studies and Modeling time
domain between 0.1 s and 10 s. An understanding of the
important considerations for stability analysis can be ob-

The precise modeling of turbogenerators requires very de- tained
from the ‘‘undamped-swing-equation,’’ which roughly
tailed representations to accurately reflect the internal opera-
describes the operation of the generator rotor after a distur-
tion of the machine, particularly if higher order characteris- bance:
tics, such as saturation, are considered. Because power
system simulations may involve representing many ma-
2 2
d δ/dt = (P − P )/(2H )
m e
chines, transmission lines, transformers, etc., to reflect the
complexities of modern interconnected power systems, simpli- where #
is a measure of the angular displacement of the gen-
fied models have been developed for power system equipment, erator
rotor with respect to a continuously rotating reference
including turbogenerators. These simplified generator models frame;
Pm is the accelerating mechanical power supplied by
provide accurate simulations for specific types of power sys- the
turbine; and Pe is the restraining electric output power of
tem studies. Consequently, the model used to represent a tur- the
generator.
bogenerator depends upon what is being studied in the
During steady-state operation, the electric output power
power system. equals
the mechanical input power, and the rotor angular ac-
For typical power system applications, studies are divided
celeration (d2 2
#/dt ) is zero, so that the machine remains sta-

into three types: (1) steady-state, or ‘‘load flow,’’ studies; (2) ble.
When the electric output power is reduced to levels and
stability studies; and (3) short-circuit studies. Power system
durations that permit the rotor angle to advance beyond the
analysis software normally has standard modules to analyze point at
which synchronism is maintained, ‘‘pole slipping’’ and
these three areas. system
instability occurs. Before the widespread use of com-
puters,
the solution of the swing equation was sometimes per-
1. Steady-state: Steady-state, or load-flow, studies ana- formed
graphically, using a diagram similar to Fig. 6, so the
lyze the system in the steady state and are concerned with solution
of the swing equation is sometimes described as ap-
common operating conditions, such as an outage of a local
plication of the ‘‘equal area criterion.’’ The criterion is that
transmission line or of another local generator. These studies the
accelerating area of the curve is less than the stabilizing
are generally performed to determine how the generator be- area.
The amount of time that the rotor can accelerate unre-
ing studied can be best used to enhance steady-state opera- strained
without losing synchronism is sometimes called the
tion of the system. They are sometimes used to determine lo-
‘‘critical fault clearing time.’’
cal power transfer capabilities and operating reserve For
typical studies that require simulations of less than 5
requirements for generation. For this type of study, the gener- s to
10 s, the generator is modeled electrically as a voltage
ator is modeled very simply using its reactive capability curve source
connected to the step-up transformer by the genera-
and its terminal voltage limits. Real power output of the gen- tor’s
transient reactance X #. The transient reactance domi-
erator is stipulated, and the reactive power is either stipu- nates
the generator’s behavior in this time domain, and the
lated directly or, more commonly, adjusted to maintain a de- model
that assumes this is said to use the ‘‘voltage behind the
sired voltage at a specific point in the power system. For
transient reactance.’’ Longer simulations may require use of
example, the reactive output of the machine may be selected the
voltage behind the synchronous reactance.
to maintain nominal voltage within
1% at the midpoint of Excitation systems are often of considerable
benefit during
the generator step-up transformer leakage reactance. The short
duration transient situations, so stability studies nor-
point at which the voltage is controlled is a matter of local mally
often use a fairly detailed depiction of the generator
practice, but it is usually not on the transmission system be-
excitation system. Stability studies are often used to deter-
cause of the concern that other voltage control equipment mine
power transfer levels between adjacent systems, for ex-
(tap-changing transformers, switched capacitors, etc.) may ample,
between New England and New York, and to deter-
have conflicting voltage control objectives. mine
spinning reserve requirements. Spinning reserve is the
2. Stability: Transient stability studies examine the inte- amount
of generation kept on line beyond that required for
grated performance of the system during and following distur-
satisfying load requirements to maintain safeguards against
bances. Turbogenerators are generally the focus of these stud- failure
of a major unit or generating plant.
ies because of the relatively large amount of stored 3.
Short circuit: Short-circuit studies are performed to
mechanical energy available from the spinning shaft of the examine
the capability of the power system to withstand ma-
turbine and generator. Unlike other power system compo- jor
faults, such as three-phase short circuits, single-phase-to-
nents, such as transmission lines, transformers, and capaci- ground
short circuits, etc. These studies concern themselves
tors, generators normally have several seconds of stored en- with the
capability of the system to isolate the faulted section
ergy that can be applied to stabilize the system during by
operation of relays and power circuit breakers. Modern
disturbances. Stability studies require knowledge of the gen-
protective relays and breakers sense and interrupt a short
erator’s mechanical inertia, often described using the inertia circuit
in just a few electric cycles, so the time domain of in-
constant (usually designated as the ‘‘H constant’’). The H con- terest
for this type of study is 0.01 s to 0.1 s. Use of the proper
stant is the amount of spinning energy available from the model
for turbogenerators is important for these studies be-
generator, turbine, and exciter as a multiple of the rated out- cause
generators are the source of the short-circuit current in
----------------------- Page 604-----------------------
TURBOGENERATORS 607
the power system. No other component has such a large

because of limitations on continuous loading (ampacity)
source of electromagnetic energy as the magnetic field of the
of generator circuit breakers.
generator. Indeed, often the only source of short-circuit cur-
3. Interconnections of unit-connected generators are gen-
rent is assumed to be generators (both turbogenerators and
erally via multiple transmission circuit breakers, typi-
hydrogenerators), although large motors and induction gener-
cally in a ‘‘ring-bus’’ or ‘‘breaker-and-a-half’’ scheme, as
ators are sometimes also added. The model used for the gen-
shown in Fig. 16, so that maintenance or problems with
erator in this type of simulation is that of a voltage source
a single circuit breaker or bus do not force the generator
connected to its step-up transformer by the generator’s sub-
out of service.
transient reactance. The subtransient reactance dominates
4. The interconnection of a generator to a transmission

the performance of the generator in this time domain and lim-
system generally adds three to five times the generator
its the generator fault current during the critical first few cy-
rating (MVA) to transmission system short-circuit inter-
cles after initiation of the fault. Although modern excitation
rupting requirements in the local area. For example,
systems are employed to reduce generator short-circuit cur-
addition of a 250 MVA plant to a 132 kV transmission
rents, excitation systems are normally not modeled for short-
system increases short-circuit interrupting duties of cir-
circuit studies, yielding conservative results.
cuit breakers at the point of interconnection by roughly
1000 MVA (4.4 kA) and increases interrupting duties

SITING OF TURBOGENERATORS AND TRANSMISSION
by a smaller amount for a considerable distance. Area
INTERCONNECTIONS
breakers must be examined to determine whether they
can accommodate this increase in interrupting duties.

Important considerations for siting of new generating sta-
tions include:
OPERATION OF TURBOGENERATORS
1. The availability, cost, and security of the operating sup-

An
excellent guide for operating turbogenerators is provided
plies (fuel, water, etc.).
by
The IEEE Guide for Operation and Maintenance of Turbine
2. The ability of the system to accommodate the genera-
Generators (IEEE publication 67). This document addresses
tion, that is, adequacy of local transmission to transmit day-
to-day operating and maintenance considerations, such
power and adequacy of the local distribution system to as
start-up and synchronization, changing load, adjusting
provide station service power. load
to recognize constraints of the reactive capability curve,
3. The ability of the local transmission circuit breakers to
shutdown, operation with unbalanced system voltages, and
accommodate increases in short-circuit current that oc- other
routine operating measures that apply to individual
curs after introducing new generation.
machines. The information in IEEE 67 should be considered
4. The need of the local system for enhanced voltage regu- a
supplement to manufacturer’s literature, which addresses
lation and stability that accompanies the introduction
operation of specific turbogenerators. The reader is referred
of new generation. to
IEEE 67 for specific operating information.
From a system perspective, operation of turbogenerators is

Sometimes these objectives conflict. For example, an area
normally dictated by economic considerations. Power stations
with poor voltage regulation often has low-capacity circuit use
different fuels, have different operating characteristics
breakers. The introduction of new turbogenerators improve and
efficiencies, different start-up and shut-down times, etc.
the voltage regulation capability, but the resulting short cir- All
of these factors must be considered to determine when to
cuit currents may exceed the interrupting and/or continuous shut
a unit down or start it up and how to load it during
capabilities of the existing breakers, necessitating upgrade.
operation. This is often called the ‘‘unit dispatch’’ problem,
The responsibility for establishing rules for interconnec- the
term ‘‘dispatch’’ describing the output (MW) loading level
tion of turbogenerators is the responsibility of the local util- for
each unit. It has been demonstrated that a power system,
ity, but a few generalizations can be made:
1. Generators with a capability in excess of 50 MW are

generally connected to the local transmission system by
using a step-up transformer. A simple ‘‘rule of thumb’’
is that the MVA rating of the plant should not exceed
roughly twice the kilovolt rating of the interconnected
transmission system. For example, a 132 kV transmis-
sion system normally accommodates a generating plant
of up to about 265 MVA without extensive reinforce-
ments because the output of the plant can be transmit-
ted by a single transmission line. Higher outputs typi-
cally require multiple lines or lines consisting of
bundled conductors.
2. Generators with rated outputs much above 150 MVA
(a) (b)
generally must be unit-connected (i.e., they must be
connected to the transmission system directly via the
Figure 16. Generator installed in (a) a ‘‘breaker-and-a-half’’ and
step-up transformer, with no generator circuit breaker) (b)
‘‘ring-bus’’ configuration.
----------------------- Page 605-----------------------
608 TURBOGENERATORS
consisting of multiple operating units, is optimally dispatched put

capability sufficient to satisfy all generator operating con-
when all of the units have the same incremental operating
ditions. In addition, a modern excitation system has the fol-
cost. The plant incremental operating cost is often referred to
lowing additional characteristics:
as the plant ‘‘lambda’’ (#), which has the units of dollars (or
other appropriate unit of currency) per incremental (MWH •
Transient forcing capability, in which the excitation sys-
megawatt-hour) of output. A related problem is the ‘‘unit com-
tem output is temporarily increased above rated condi-
mitment’’ problem, which determines when a given unit
tions to aid in recovery during and following power sys-
should be started up or shut down, considering economic and
tem disturbances, such as faults.
technical criteria. Solution of the unit commitment problem •
Feedback control to adjust the excitation system output
entails considering fuel costs, power supply contracts, projec-
automatically to maintain stable steady-state terminal
tions of system lambdas, estimates of costs associated with
voltage during changes in load and system conditions.
start-up and shut-down of individual units, and turbine load- •
Limiting and protective features to prevent operation be-
ing ramp rates and other operating restrictions. Normally, be-
yond the capability of the generator.
cause of its complexity, the unit commitment problem is
solved by a computer simulation that employs linear pro-
An excitation system can be generally divided into two ma-
gramming techniques to evaluate possible alternative courses jor
parts: the voltage regulator, which includes the signal-level
of action and to select the alternative with the highest likeli-
control, limiting, and protective functions, and the exciter,
hood of financial gain.
which includes the power conversion apparatus to supply the
To illustrate the considerations used in solving the unit
required dc output to the generator field. Over the years, the
commitment and unit dispatch problems, consider this ex-
following types of exciter designs have been utilized.
ample.
A plant consists of three units, each with a rated capability
Dc Generator-Commutator Exciter. Prior to the widespread
of 100 MW. Two of the units, units 1 and 2, are identical. The
availability of solid-state semiconductor power rectifier de-
fuel cost of each of these units is described by the equation
vices, the only practical means of supplying the dc current
L # 250 2P 0.1P2, where L is the unit fuel cost ($/h) and
required for the field winding of a large ac generator was with
P is the unit output (MW). The third unit has a fuel cost de- a dc
machine. The dc machine output is rectified by a multi-
scribed by L # 300 1.5P 0.15P2 . It is estimated that a
segmented commutator ring and brushes. These exciters are
unit start-up costs roughly $1500 for any of the three units,
driven by the generator shaft, either directly or through a
based on fuel costs and increased maintenance costs. The
speed reduction gear, or are driven separately by an induction
plant has a contractual obligation to supply a daily output of
motor. The exciter dc output supplies the main generator field
235 MW from 9:00 A.M. to 9:00 P.M. and 135 MW from 9:00 by a
collector assembly composed of slip rings and brushes.
P.M. to 9:00 A.M. How many units should be committed dur-
Although many dc generator-commutator exciters are still
ing each 12 h period and how should the units be loaded dur- in
service, maintenance problems and unavailability of re-
ing those periods?
placement parts make them increasingly popular candidates
The incremental operating costs of units 1 and 2 are given for
retrofitting with static exciters.
by # # dL /dP # 2 0.2P , and that for unit 3 is given by
n n n n
# # dL /dP # 1.5 0.3P . Between 9:00 A.M. and 9:00 P.M.,
3 3 3 3
Rotating Ac Exciter. Rotating ac exciters employ an ac gen-
it is required that all three units be committed to satisfy the
erator whose output is rectified and sent to the main genera-
output requirement of 235 MW. Solving the ‘‘unit dispatch tor
field. Some rotating ac exciters employ a rotor-mounted
problem’’ for equal incremental costs, #1 # #2 # #3 # $19.5/ field
winding and stationary armature. The three-phase ac
MWH when P 1 # P2 # 87.5 MW and P3 # 60 MW, so that
exciter output is rectified by one or more stationary rectifier
P 1 P2 P3 # 235 MW. Between 9:00 P.M. and 9:00 A.M.,
bridges, and the dc bridge output is sent to the generator field
only two units must be committed, but it may prove more
through brushes and slip rings. However, most rotating ac
economic to commit all three units to avoid the start-up cost.
exciters built today are brushless, so that they use rotating
Explicit evaluation of the possible alternatives (i.e., solving
rectifiers and do not require slip rings to provide generator
the ‘‘unit commitment problem’’), shows that the most eco- field
current.
nomic alternative is to commit all three units rather than
keep two units on line for 12 h, followed by a daily unit
Brushless Exciter. Figure 17 shows the arrangement of a
startup for the third unit. Then, between 9:00 P.M. and 9:00
typical brushless excitation system. The shaft-driven rotating
A.M., with all three units committed, the units should be as- ac
exciter utilizes a stationary field winding and rotating ar-
signed loadings of P 1 # P2 # 50 MW and P3 # 35 MW, on the
mature. The three-phase armature ac output is sent to a ro-
basis of equal incremental costs.
tating rectifier wheel whose dc output is sent directly to the
Obviously, a more realistic example, with unequally sized main
generator field, typically through a bore in the genera-
units, ramp rate and other operating restrictions, and provis- tor
shaft. The brushless exciter accomplishes this without
ions for complex fuel contracts, hourly power transactions,
brushes, slip rings, or commutators and thus eliminates the
and so on, would considerably complicate the problem. Com-
maintenance and carbon dust associated with them. The rec-
puter analysis is normally required for optimal economic op-
tifier wheel typically includes diodes and fuses connected in a
eration of all but the smallest power systems.
bridge rectifier configuration.
Most brushless exciters also employ a shaft-driven perma-

Excitation Systems
nent
magnet generator (PMG) as a pilot exciter for the volt-
An excitation system provides controlled dc current to the age
regulator. The voltage regulator includes a controlled rec-
generator field (rotor) winding. It must have a continuous out-
tifier bridge composed of silicon controlled rectifiers (SCRs)
----------------------- Page 606-----------------------
TURBOGENERATORS 609
Rotating elements
PMG Rectifier
N
CT
Ac
Generator
Exc.
S
wheel
Voltage regulator
VTs
Shunt
Power Automatic control,

amplifier limiting, and
protective elements
Ac voltage
adjuster
Ac Dc Auto.
Firing Manual
Dc voltage
circuits control
adjuster
Man.
Figure 17. Simplified block diagram of a
typical brushless excitation system.
which rectify the PMG output and control the level of dc volt- the
generator terminals, a means of momentarily flashing the
age to the brushless exciter field winding. The firing angle of
generator field is typically provided to initiate a buildup of
the SCRs relative to the PMG voltage, and thus the dc output
generator voltage each time the unit is started.
voltage, is determined by the voltage regulator control cir-
cuits. All excitation power in a brushless excitation system is
Compound Source Static Exciter. A compound source static
derived from shaft rotation.
exciter derives its power from both the voltage and current
output of the generator. The current source provides short-

Potential Source Static Exciter. A potential source static ex-
circuit support by supplementing the power to the exciter
citer, as shown in Fig. 18, is so named because it derives its
during periods of low terminal voltage, such as during system
power from a potential source (either the main generator ter-
fault conditions. This type of exciter is not often used because
minals or station service supply), and all power conversion of
the added cost, size, complexity, and losses of utilizing both
devices are stationary. The excitation power potential trans-
potential and current power transformers along with the ad-
former (PPT) isolates and steps down the ac source voltage to
ditional rectifying and control equipment.
a level that, when rectified, supplies the required excitation
voltage to the generator field. The PPT output is rectified and
Voltage Regulators
controlled by a solid-state power amplifier, which uses one
or more SCR rectifier bridges whose dc output is sent to the The
voltage regulator is the portion of the excitation system
generator field via brushes and slip rings. The voltage regula- that
provides the means of automatically controlling the out-
tor adjusts the SCR firing pulses to control the dc output. In put
of the exciter. In an excitation system with a rotating
a static excitation system that derives excitation power from
exciter, the voltage regulator controls the voltage to the ex-
Rotating elements
CT
Generator
Collector
Power
amplifier
PTT
VTs
Dc Ac
Auto.
Firing Automatic control,
circuits limiting, and
protective elements
Ac voltage
adjuster
Field Man.
flashing Manual
Dc voltage Figure 18. Simplified block diagram of a
circuit control
adjuster typical potential source static excitation
system.
----------------------- Page 607-----------------------
610 TURBOGENERATORS
citer field. In a static excitation system, the voltage regulator

(connected to the power grid, so that the grid frequency deter-
directly controls the generator field voltage by determining mines
the generator output frequency and speed), adjust-
the firing angle to the power amplifier SCRs. ments
in the turbine mechanical output result in correspond-
Modern analog voltage regulators utilize integrated cir- ing
changes in generator real power (or kilowatt) output.
cuits containing operational amplifiers and other electronic The
other means of controlling generator output is by ad-
devices. Digital microprocessor-based voltage regulators are justing
the excitation system output. Although turbine output
becoming more prevalent as such hardware becomes faster
controls the real power output of an on-line generator, the
and more inexpensive.
excitation level determines the generator reactive power (or
kilovar) output. Therefore adjustments in generator kilovar

Excitation Performance Characteristics output
are made through the voltage regulator.
In addition to supplying the generator with the appropriate

Generator Protection
excitation level for steady-state conditions, an exciter usually
has additional transient forcing capability to help the genera- Because
the consequences of generator damage resulting from
tor recover from severe faults and other disturbances. Excita-
abnormal external conditions, internal faults, and misopera-
tion systems that are faster responding and have higher max- tion
might be extensive, it is important to protect against
imum (ceiling) voltages have a greater potential to contribute such
damage with an adequate protection system. Similarly,
to generator transient stability. Excitation systems that have if
protection systems act to trip generators or other critical
higher ceiling voltages are typically larger and costlier than
equipment unnecessarily during system disturbances, the se-
their lower ceiling counterparts. Therefore, high ceiling volt- curity
of the entire power system can be jeopardized. Thus,
age excitation systems are typically used only when there is a
suitable generator protection system allows a generator to
an apparent need for transient stability improvement.
continue operation during certain types and degrees of abnor-
An excitation system has high initial response (HIR) char- mal
conditions but reliably trips the generator when nec-
acteristics if it can force its output voltage from rated condi- essary.
tions to ceiling in 0.1 s or less. A typical brushless exciter is

Most protective relays are available now in both electrome-
not HIR, although it is usually faster than other types of ro-
chanical and solid-state designs. The solid-state models are
tating exciters. A static excitation system is inherently HIR. finding
increasing acceptance, mainly because of their accu-
racy,
flexibility, and reduced maintenance. In addition, micro-
Generator Synchronizing
processor-based protective relaying systems, in which many
standard relaying functions are combined into one package,

Synchronization is the process of connecting a generator to are
becoming more prevalent.
the power grid. Successful synchronization for large genera-
Because both the voltage and current levels on modern
tors is accomplished by closing the generator output circuit large
generators are too high to be used directly by protective
breaker only while meeting certain conditions. The conditions relays
and other such devices, special instrument transform-
are that the magnitude, frequency, phase rotation, and phase ers
are required. Voltage transformers (VTs) are connected
angle of the voltage on the generator side of the synchronizing across
the generator ac terminals and have an appropriate
breaker match those on the system side of the breaker. Faulty turns
ratio to step down the sensed generator voltage to a
synchronization can cause severe equipment damage and sys- reduced
level (typically 120 V rms) at the secondary. Current
tem transients.
transformers (CTs) are usually of the window type, in which
Prior to synchronization, voltage magnitude matching re- the
single-turn primary winding is the generator conductor or
quires adjusting the generator excitation level so that the bushing
that passes through the opening of the CT and the
generator terminal voltage matches the system voltage.
secondary winding has the appropriate number of turns so
Matching the frequency requires that the turbine speed be that
secondary current is nominally 5 A rms at full load.
precisely adjusted such that the generator output frequency The
following are some of the protective functions utilized
is the same as or slightly higher than the system frequency. on a
large generator:
Phase rotation matching means that the order in which the
three-phase voltages reach their peaks are the same. This re- •
Generator differential current protection, in which CTs in
quires proper connections and verification of phase order
each phase on both ends of the stator winding are used
prior to initial start-up. Matching phase angle means that the
to quickly detect the presence of internal multi-phase
voltages in each phase on the generator and system side of
faults.
the breaker are in phase or reach their peaks at precisely •
Stator ground fault protection, in which a high-impedance
the same time. Synchronization is performed manually by an
grounded generator is protected against the potentially
operator or automatically by an automatic synchronizer.
damaging effects of a low-current, single-phase-to-
ground fault before it spreads to a much more severe

Generator Control
multi-phase fault.
Once a large turbine-generator is connected to a power grid, •

Negative sequence current protection, which protects the
there are two basic means utilized to control its output. One
generator from excessive heating resulting from induced
is through controlling the turbine mechanical output via
eddy currents on the rotor surface due to the effects of
valve control on a steam turbine, or fuel control on a combus-
unbalanced (or negative sequence) currents in the stator.
tion turbine, or gate control on a hydro turbine. When the •
Field ground protection, which protects the field winding
generator is off-line, the turbine control adjusts generator
on the rotor from the potentially damaging effects of
speed and electrical frequency. When the generator is on-line
grounds.
----------------------- Page 608-----------------------
TURBOGENERATORS 611
• Overexcitation protection, which protects the generator L. K.
Kirchmayer, Economic Operation of Power Systems, New York:
field winding from overheating due to excessive excita-
Wiley, 1958.
tion current. G.
Kron, Equivalent Circuits of Electric Machinery, New York: Wi-
ley,
1951.
• Volts-per-hertz protection, which protects the core of the
generator (and connected transformers) from the heating M.
Liwschitz-Garik and C. Whipple, Alternating Current Machines,
New
York: Van Nostrand, 1961.
effects due to excessive magnetic flux.
L.
Matsch, Electromagnetic and Electromechanical Machines, Intext,
• Loss of excitation protection, which senses the condition
Scranton, 1972.
in which generator synchronism is jeopardized because G. R.
Slemon, Magnetoelectric Devices: Transducers, Transformers,
of a loss or severe reduction in excitation level. and
Machines, New York: Wiley, 1966.
• Loss of synchronism protection, in which a generator that W. D.
Stevenson, Elements of Power System Analysis, New York:
loses synchronism with the system is tripped off-line be-
McGraw-Hill, 1982.
fore damaging pole-slipping takes place. F. S.
Tse, I. E. Morse, and R. T. Hinkle, Mechanical Vibrations, Bos-
• Motoring protection, in which negative real power is ton:
Allyn and Bacon, 1968.
sensed which would indicate the loss of turbine output D. C.
White and H. H. Woodson, Electromechanical Energy Conver-
sion, New York: Wiley, 1957.

power.
A. J.
Wood and B. Wollenberg, Power Generation, Operation, and Con-
trol, New York: Wiley, 1984.

For any particular application, the costs associated with
particular protection systems and the degree of protection
WILLIAM R. MCCOWN
they afford must be carefully weighed against the risks en-
ALEKSANDAR PROLE
countered for lesser degree of protection. The amount and
ROBERT J. NELSON
complexity of protection that is usually applied varies ac-
JOSEPH D. HURLEY
cording to the size and importance of the generator.
RICHARD B. CHIANESE
ROSS GUTTROMSON
Siemens-Westinghouse Power Corp.

BIBLIOGRAPHY
P. M. Anderson and A. A. Fouad, Power System Control and Stability,

TURBULENCE, PLASMA. See PLASMA TURBULENCE.
Iowa State University Press, 1977.
TURING
MACHINE. See COMPUTATIONAL COMPLEXITY
Applied Protective Relaying, Newark: Westinghouse, 1976.
THEORY.
T. Baumeister and L. S. Marks, Standard Handbook for Mechanical
Engineers, New York: McGraw-Hill, 1967. TV,
INTERACTIVE. See INTERACTIVE VIDEO.
TWIN
LEAD. See ANTENNA ACCESSORIES.
A. F. D’Souza and V. K. Garg, Advanced Dynamics, Englewood Cliffs,
TWO-
DIMENSIONAL CODES. See MARK SCANNING
NJ: Prentice-Hall, 1984.
A. E. Fitzgerald, C. Kingsley, and A. Kusko, Electric Machinery, New
EQUIPMENT.
York: McGraw-Hill, 1971. TWO-
DIMENSIONAL GRAPHICS TRANSFORMA-
C. M. Harris and C. E. Crede, Shock & Vibration Handbook, New
TIONS. See GRAPHICS TRANSFORMATIONS IN 2-D.
York: McGraw-Hill, 1976. TWO-PORT
PARAMETERS. See NETWORK PARAMETERS.
E. W. Kimbark, Power System Stability: Synchronous Machines, New TWO-
PORTS. See MULTIPOLE AND MULTIPORT ANALYSIS.
York: Wiley, 1956. TWT.
See TRAVELING WAVE TUBES.
----------------------- Page 609-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
●
HELP

Your Librarian
Profile
Waste-to-Energy Power Plants
Standard Article

David Gordon Wilson1
1Massachusetts Institute of Technology, Cambridge, MA

reserved.
Search
DOI: 10.1002/047134608X.W3046
❍ Acronym Finder
Abstract
The Problems of Solid-Waste Production and Treatment
The Opportunities
Alternative Methods of Recovering Energy From Wastes
Conclusion

%20Conversion/W3046.htm17.06.2008 16:49:40
----------------------- Page 610-----------------------
448 WASTE-TO-ENERGY POWER PLANTS
baseline level above which modern civilization produces
vastly increased flows of waste. A large proportion of these
‘‘modern-civilization’’ wastes is paper and cardboard, also de-

rived
from trees and grasses and also a principal source of
energy. It is called biomass along with industrial wastes such

as
wood chips and sawdust, agricultural wastes such as sug-
arcane residues, and animal-husbandry wastes such as feed-

lot
and chicken-house bedding. Industrial wastes with sub-
stantial energy content come from petroleum and chemicals

and
from their products, such as plastics, vehicle tires, and
used
oil.
The quantities are large and still increasing. In the hun-

dred
years from the early eighteenth century to the same pe-
riod
in the nineteenth century, the refuse yards to which all
the
solid wastes of the city of Edinburgh, Scotland were
brought remained the same size (1). The thrifty people in
charge sorted and sold or gave away everything that was
brought in. The industrial revolution changed that situation
radically. Municipal waste collections in Manchester, En-

gland
increased 50 times from the 1930s to the 1990s (2). Do-
mestic wastes collected from U.S. homes doubled in the 1960

to
1995 period to about 1.5 kg per person per day (kg/p/d)
(3).
These post-consumer wastes are those that are usually
considered when the solid-waste situation is discussed. How-

ever,
animal feedlot and manure wastes approach 20 kg/p/d
(that
is, 20 kg for every man, woman, and child in the United
States, not just for those connected with animal raising) (4).
Crop
wastes are about 8 kg/p/d. Mining wastes in the United
States have been estimated at 13 kg/p/d, but these have little
energy content, in general. A list giving broad estimates (de-
pending greatly on definitions of waste categories and subject

to
large uncertainties) is given in Table 1 (5). These estimates
are
sufficient, however, to indicate the problems and the op-
portunities.
Not included in this list are oil-refinery wastes, estimated

to be
capable of producing a steady output of 135 GW of elec-
tricity worldwide in 2010 (6), and the used oil (about 4.5 bil-
lion
liters in the United States in 1990) and old tires from
motor
vehicles (around 125 million per year in the United
States).
In this compendium, the predominant liquid wastes from

which
energy may be produced differ little in character from
one
country to another around the world, although the quan-
tities produced, both absolutely and per person, obviously dif-

fer
widely. The constituents of solid wastes vary considerably,
Table
1. Organic Wastes Produced in the United States
in
1971a
Mass (106 tonnes)b
Readily
Source Total Collectable
Agricultural crops and food wastes 390 23
Manure 200 26
Urban
refuse 130 71
WASTE-TO-ENERGY POWER PLANTS
Logging and wood manufacturing residues 55 5
Miscellaneous 50 5
Nature itself produces wastes such as dead branches, leaves,
Industrial organic wastes 44 5
and sun-dried grasses. Primitive societies, once they had
Manicipal sewage solids 12 1.5
learned to master fire, used the energy in these wastes for a The
total has a potential energy equivalent to 175 billion liters of oil per year.
cooking and heating. These wastes can be considered the b Dry
ash-free basis.

----------------------- Page 611-----------------------
WASTE-TO-ENERGY POWER PLANTS 449
however, both within large countries and from one country to city
of Cambridge, Massachusetts were officially given as 50
another. In the United States and elsewhere, the climate has cents
per tonne. In the 1990s the disposal costs (additional to
a large influence, with the municipal collections in the south
collection costs) for this urban region of the country were in
and west of the country being dominated by palm fronds at the
region of $100 per tonne.] We can afford to do much bet-
certain seasons of the year, for instance. In poorer countries ter.
The second concern has more validity: the as-yet unsolved
such as Lebanon, urban wastes have less paper and more food problem
of groundwater pollution. Modern solid wastes, even
wastes (generally disposed of in sink grinders in richer coun- those
collected from homes, are extremely heterogeneous, in-
tries), so that the moisture content is higher and the overall
corporating used batteries, containers with paints, varnishes,
quantity per person is lower (7).
cleaning solutions, pesticides, fungicides, unused medical
drugs,
and many other potentially toxic materials. Industrial
wastes
can include all types of noxious solids and liquids.
THE PROBLEMS OF SOLID-WASTE Rain
percolates through refuse dumps, dissolving some solids
PRODUCTION AND TREATMENT and
combining with some of the liquids to produce what is
known
as leachate. If this passes directly into an aquifer,
It is generally conceded that as countries move to what is
drinking-water wells in the vicinity can be rendered hazard-
generally regarded as the western model, the production of ous
within a short period of time. If there is no nearby aqui-
consumer goods, packaged food and drink, newspapers and fer, a
leachate ‘‘plume’’ spreads out underground and moves
magazines, and everything associated with automobile trans- slowly,
perhaps only a few meters per year. Soil scientists
portation increases very greatly; that the resulting solid-
estimate that in some cases these noxious plumes may not
waste flow increases enormously in quantity and variety, in- pollute
an aquifer for thousands of years. However, in most
cluding a large number of noxious and toxic materials; that a cases
the pollutants will begin to appear in regional water
large proportion of citizens are careless in disposing of these within
tens or perhaps hundreds of years. Many efforts are
wastes; and that the authorities responsible for collection and being
made to devise methods by which such plumes can be
disposal of wastes do not always use the highest standards
contained or diverted, but none has yet proven effective.
that could protect future generations from harm. The follow-
Rules have therefore been established for the safe land-
ing aspects are less-publicized aspects of the solid-waste filling
of solid wastes (so that a landfill differs from a dump
problem. in that
such rules are nominally followed). The pits are lined
with
waterproof clay or a membrane of urethane or other
Are We in Danger of Burying Ourselves in Trash? long-
lasting and supposedly impenetrable material. The lea-
We are not, in the overall sense, burying ourselves in trash. A chate
is collected at the lowest point of the liner and taken to
large proportion of the materials we use comes from quarries, a
treatment facility. However, public concern over alleged
mines, and the like, and we leave most of these sites unfilled. (and
often confirmed) serious health effects from polluted
A ‘‘satellite’s eye’s view’’ of the material-transportation net-
drinking water and perhaps from gaseous emissions from old
work of any country would show trains, barges, and road ve- dumps,
as well as public confidence that landfill safeguards
hicles taking materials from mines and quarries principally will
operate perfectly for hundreds of years, is so low that
into towns to be used in buildings and roads and industry. A it is
very difficult to get acceptance for the establishment of
new
landfills.
large proportion of these vehicles would be empty on the re-
turn trip. [Cities in the past did, however, frequently get bur-
ied in their wastes. Some medieval German cities avoided this
Concerns Over the Incineration of Solid Wastes
danger by requiring that wagons that had brought in produce
The
open burning of refuse was, in the distant past, accept-
take out wastes to be deposited in the countryside (1)]. A proj-
able in
rural areas but prohibited in towns. Controlled burn-
ect investigated in the early 1970s by the New York Central
ing in
an enclosed furnace with a high stack to disperse the
Railroad to transport compacted wastes from New York City
smoke
seemed to be an advanced solution for towns when the
to a large quarry near a small town in Ohio failed because
first
so-called incinerator was built in England in 1874. Incin-
the town residents, who had anticipated enduring about five
eration
was not practiced in the United States in the period
years of noise and dust in return for a golf course and other
up to
the 1970s to the extent that it was followed in Europe
recreation areas created as part of the agreement, discovered
and
elsewhere. However, in the late 1960s a ’’solid-waste cri-
that it would take 175 years at then-current rates of produc-
sis’’
was declared in the United States, and a period of intense
tion of New York solid wastes to fill the quarry. Even the
research and of the adoption of improved treatment measures

prospect of residing tax-free from the fees paid on the dis-
followed. (These beneficial activities were unfortunately cut
posed refuse was insufficient inducement for the townspeople
short
by the ‘‘oil crises’’ of 1973 to 1978). Partly as a response
to agree to the scheme.
to the
concerns over putting solid wastes in the ground, the
burning
of wastes, often with no attempt at energy recovery,
Concerns about Transporting Wastes to Landfills
became
popular. However, suspicions that the smokestack
There are two primary concerns over transporting wastes to
emissions from incinerators were also responsible for serious
landfill them in distant sites. One is of little importance: the health
effects grew. Consequently, it has been difficult to get
cost of doing so. We as a society have been parsimonious in public
acceptance of new incinerators, even those equipped
allocating resources to reduce the impact of our wastes, and with
sophisticated air-pollution-control equipment such as
the costs of transportation and of responsible treatment are
electrostatic precipitators. Gases such as dioxins and hydro-
considerable only in relation to the almost negligible costs of gen
chloride can pass through these units. If exhaust-gas wa-
the past. [In 1970 the landfill costs for the author’s then-home ter-
spray scrubbers (an expensive solution often requiring the
----------------------- Page 612-----------------------
use of additional fuel oil to reheat the gases before discharge) companies
can indeed charge a ‘‘tipping fee’’ for the accep-
are used, a potentially noxious sludge must be disposed of. tance of
wastes; the greater the degree of pretreatment, the
lower the
fee.)
The
final degree of pretreatment is to convert the wastes
THE OPPORTUNITIES into a
liquid or gaseous fuel of desirable characteristics. Doing
this has
two beneficial side effects. One is that the treatment
Problems lead to opportunities. The opportunities are particu- process
is in effect a chemical plant, out of which there is
larly attractive in the waste-to-energy area, although not in tight
control of all effluxes. The other is that the clean fuel so
the relatively unsophisticated incineration plants of the past. produced
can be burned in processes that have much higher
These involved, in general, so-called ‘‘mass burning’’ of un- potential
thermal efficiencies, and possibly much lower capi-
treated or minimally separated wastes and had low thermal tal costs,
than incinerators. These improved processes will be
efficiencies (expressed as the output of useful heat or useful discussed
below, after a review of the baseline mass-burning
power divided by the calorific value of the wastes). Low effi- waste-to-
energy plant.
ciencies imply high-temperature high-volume discharge of ex-
haust gases, difficult and expensive to treat. The attainment
ALTERNATIVE METHODS OF RECOVERING
of high efficiency in itself results, therefore, in a reduction of ENERGY
FROM WASTES
pollution, in that the exhaust flow and temperature are re-
duced. It is easier to incorporate exhaust-cleaning systems. Mass-
Burning Incinerators Raising Steam
Nitrogen oxide emissions will probably be greatly reduced. Di-
Incineration may be carried out without any attempt to re-
oxin emissions are also likely to be reduced. cover
energy from the combustion heat. Energy recovery in
To reduce emissions from incinerators, better control sys-
association with incineration should be welcomed, although
tems are needed. Even the most highly sophisticated control the power
generated from the wastes must perhaps be looked
system cannot greatly improve the combustion process in upon only
as a byproduct. Perhaps surprisingly, the costs of
mass-burning in which, for example, a piece of dry tissue pa-
incineration with and without energy recovery are similar in
per may be close to a stack of water-soaked telephone books. many cases
examined (see below) because cooling of the gases
One is consumed in a fraction of a second, while the other reduces
the cost of the gas-cleaning equipment required and
dries and smolders for hours. Therefore there is also a need thus
compensates for the added costs of heat-recovery
for better fuels. equipment.
This is an area in which we have retrogressed to some ex- In

cooler climates there is sometimes a market for the low-
tent. The incinerator near where the author grew up in Great quality
steam produced from incineration plants for district
Britain in the 1930s and 1940s had a so-called ‘‘picking belt’’ heating.
Electrical energy production from incineration nor-
on which all incoming refuse was loaded. The belt lifted the mally
uses systems based on steam turbines. Scores of such
refuse through two or three meters to a horizontal section plants
have been operating regularly all over the world. A
over an elevated floor and hoppers. Four to eight people steam-
cycle waste-to-energy plant would seem to be an im-
(called pickers ) would be stationed beside the belt, each with a provement
over mass-burn facilities with no energy recovery.
responsibility to remove useful or noxious or difficult-to-burn On the
other hand, steam-cycle plants have rather low effi-
items in a restricted category and to deposit them into a hop- ciencies
(15–25%) and thus produce a low power output for a
per. Thus an income was generated from the sale of ferrous given flow
of wastes. Therefore the additional costs associated
and nonferrous metals, newspapers, and so forth, and a more with power
production may not be justified. Figure 1 (7) gives
uniform feed went to the incinerator. The exhaust emissions

were also less liable to be contaminated with the products of
7000
combustion of paint and pesticide cans and the like. In the
5%
late 1980s and 1990s there has been a revival of the picking
15 ) 10%
belt, along with research and development work to automate )
m = Typical
g
b 6000
k
l solid wastes
it, to produce a more homogeneous combustible product.
This / / 15%
J
u
operation is called full-stream processing (3). M
t
( B
( 20%
A further improvement in the fuel is to use hammer mill e
e
u
5000
l
u
or other types of shredders to produce a more uniform fuel a
l
v
a 25%
v
from a stream out of which undesirable components have g 10
g Beirut
n
US
i
n
t
i
been separated. (Serious explosions have occurred in solid- a
t 4000
e
a
waste hammer mills into which partly full cans of gasoline h
e
h
e
and live ammunition, for instance, have been fed, so that sort- s
e
u
s
f
u
ing of the input is necessary.) The shredded waste, mostly e
f 3000 Noncombustibles
R
e
paper and plastic, can then be fed (normally after warming
R percentage = 30%
and moisturizing) to a briquetting or pelletizing unit. The 5
product is called refuse-derived fuel (RDF) or sometimes den-
2000
0 10 20 30 40 50 60 70
sified RDG (d-RDF) (8); it can be stored, transported, fed and
burned in the same way as coal. If the presorting is done ef-
Refuse moisture content (%)
fectively, it has much lower emissions than coal. It has been Figure 1.
Refuse heat values versus moisture content and percent-
frequently co-fired with coal in utility steam generators with age of
noncombustibles. (From Ref. 7.) Moisture content depends on
the purpose not only of reducing the emissions to below some presence
of food wastes, on climate, and on collection practices, and
regulatory limit but of reducing the cost of the fuel. (Power has a
large influence on heating value.
----------------------- Page 613-----------------------
)
2400 $
S 350
U
Moisture /
n
2000 content o
i 300
l
l
= 50% i
m
) (
D 40% t
P 1600 s 250
T o
( c
30% l
y a
t 20%
i t 200
c i
a 1200 p
p 10% a
c
a
c ) 150
d
t e
n 800 t
a c
l e
P r 100
r
o
c
400 3
9 50
9
1
(
0 t
n 0
0 12 14 16 18 20 22 24 26 28 a
l
0 500 1000 1500 2000 2500 3000
P
Plant overall efficiency (%)
Plant capacity (TPD)
Figure 2. Overall plant efficiency of steam-cycle plants versus refuse Figure 4.

Capital cost of waste-to-energy plants versus capacity.
capacity and moisture content (steam conditions 41 bar, 400#C). (From Ref.
7.) There is remarkably little variation in capital cost per
(From Ref. 7.) This chart emphasizes the effect of the data shown in tonne-per-
day capacity with size.
Fig. 1: moisture content beyond 20% has a strongly negative effect on
the efficiency of traditional steam plants, especially at lower capac-
ities. boiler-type
corrosion. [Useful information on RDF combustion
and
emissions is given by Lockwood (9).]
Modern
mass-burn facilities provide an efficient, environ-
mentally
tolerable, but expensive way to help dispose of the
refuse heating values versus moisture content and noncom-
ever-
increasing amount of solid waste. Figure 4 gives the cap-
bustibles content, with typical refuse compositions for Beirut
ital cost in
the United States versus capacity for a number of
and for the United States shown on the plot. Beirut is chosen
incineration
facilities; the cost per tonne-per-day (TPD) capac-
as an example of an urban area not totally overtaken by con- ity is found
to be fairly uniform. The capital cost of the latest
sumerism. The overall efficiency depends on the moisture con- U.S. steam-
cycle plants appears to be about $130,000 per
tent and the capacity of the plant as shown in Fig. 2. daily tonne
capacity [Engdahl (10)]. Furthermore, data indi-
Data on coal-, RDF-, and raw-refuse-burning facilities cate that
the prices of RDF and mass-burning water-wall in-
allow the generation of a graph such as shown in Fig. 3. Plant stallations
are indistinguishable. This similarity results from
efficiency is shown to be directly proportional to the kind of the extra
cost of RDF front-end preparation being offset by
fuel used: The better fuel strongly affects the boiler efficiency the
comparatively smaller furnace and emission-control sys-
through better combustion, as well as through permitting tem that is
required, according to Rigo and Conley (11).
higher steam pressures and temperatures without excessive To make
the steam-turbine waste-to-energy system eco-
nomical,
the capacity should be at least 500 TPD. Figure 5
34 120
32
Coal 100
30
) 80
% 28 RDF )
Tipping fee
$
(
60
y S
c 26 U
n ( 40
e
i 24 Raw refuse e
c s
i 20
f u
Electrical-R
f 22 f
e e
Material-R
r
l
l 0
a 20 n
Amortization-C
r o
e t
–20
v 18 r
o e
t p –40
O & M-C
n 16 t
a s
l o –60
P 14 Raw refuse C
RDF –80
12
Coal
–100
10
0 50 100 150 200 250 300 –120
Plant gross power output (MW)
0 500 1000 1500 2000 2500
Plant capacity (TPD)

Figure 3. Steam-plant overall efficiency versus output power for dif-
ferent fuel types. (From Ref. 7.) The different fuels result in different Figure 5.
Component costs versus steam-plant capacity. (From Ref.
plant overall efficiencies, principally as a result of their moisture con- 7.) The
tipping fee is the charge per tonne needed in US conditions to
tent (see Fig. 2). operate the
facility without a net loss.
----------------------- Page 614-----------------------
shows the component costs and revenues in the United States

experimental basis (14). The shale contributes energy and
versus plant capacity. The figure also shows the minimum acts as an
additive or absorbent to remove sulfur dioxide, hy-
disposal or ‘‘tipping’’ fee needed in the United States to bal- drogen
chloride, and other pollutants.
ance the operation, with electrical sales and material recovery
considered the principal income earners. A tipping fee is seen
Fueling Gas
Turbines with Wastes
to be necessary and decreases very slowly with capacity above
500 TPD: It is nearly equal to the electrical revenues at a Almost all
modern utility-plant construction entails gas tur-
very large capacity (over 2500 TPD). bines,
alone or, more usually, in combination with steam tur-
bines. The
firing temperatures (at entry to the turbine rotors)
reached
1700 K by the late 1990s, when steam cooling was
Co-firing of Wastes in Utility Boilers
adopted
for the high-temperature rotor blades. These high
Municipal wastes converted to RDF are being successfully
temperatures produced high turbine-exhaust temperatures,
fired along with conventional fuels in utility boilers. The con- sufficient
to produce high-pressure superheated steam with-
ventional fuel is generally coal, because RDF as pellets or bri- out
supplementary firing when the exhaust gases were fed to
quettes can be handled by similar equipment and has similar a steam
generator. This in turn supplies a steam turbine. The
residence times in the furnace. A typical proportion is to have combination
is called a combined-cycle gas turbine (CCGT). A
25% of the heat input from RDF. A Swiss enhancement of gas
turbine requires considerably cleaner fuel than does a
RDF, described by Haneda (12) as ‘‘epoch-making,’’ mixes cal- steam
plant, and all high-efficiency plants burn natural gas
cium compounds, presumably lime, with the RDF. This stabi- or a
refined fuel oil related to kerosine. It seems, therefore,
lizes the RDF pellets mechanically, prevents the degradation that if a
gas turbine, a CCGT, or a stand-alone unit is to be
that has been a problem with RDF, and produces ‘‘exhaust fueled by
wastes, these have (with two exceptions discussed
concentrations of hydrogen chloride and dioxins [that] were below) to
be converted to a gas or liquid fuel. A review of
virtually zero.’’ Because these two pollutants are the primary some
alternative processes and of the plants that have been
concern of people living near refuse-burning plants, this de- proposed
to use them is given later. First we make mention
velopment could indeed be significant. of some
approaches to direct burning of wastes.
The US Department of Energy, in connection with its pro- There
have been some unsuccessful attempts in the past
gram to encourage biomass conversion to electricity, gives the at
producing a gas turbine that could burn solid wastes di-
efficiency of the co-firing option at 1.81 MWh/t, versus current rectly, for
instance the CPU-400 process of Combustion Power
steam-raising incinerators averaging 1.03 MWh/t. Another Corporation
(15). A larger effort has been devoted to burning
way of expressing efficiency is that the highest thermal effi- coal in gas
turbines, either indirectly in the successful closed-
ciency of electrical generation from mass-burn incinerators is cycle
turbines developed by Escher-Wyss, discussed by Keller
under 25%, whereas when RDF is co-fired with coal in utility (16), or
directly in various open-cycle experiments funded by
plants the thermal efficiencies range from 35% to 43%, the the US
Department of Energy and reviewed by Webb (17).
levels for the steam plants themselves. None of
the experimental units had reached commercial via-
Another waste stream that has been co-fired with coal is bility by
the mid-1990s.
that of used automobile tires. A pulverized-coal boiler (which The
present author has been working on a modification of
required considerable modification in the feeding mechanism) a gas-
turbine cycle adopted for the US Navy: the intercooled-
in Toronto, Ohio burned up to 20% whole tires, one tire every
regenerative cycle (18). The compressor is split into two units
10 seconds (13), in a test of the process. There was a 36% separated
by a water-cooled intercooler. The compressor-de-
reduction in emissions of nitrogen oxides, a 28% reduction in livery air
then passes through a heat exchanger heated by
particulates, and a 14% reduction in sulfur dioxide. The heat the
turbine exhaust augmented by a second combustor, the
rate (efficiency) also improved. Another successful test used addition
of which is the principal modification to the cycle
fluidized-bed combustors to burn a mixture of coal and shred- (Fig. 6).
The fuel needed in the first, high-pressure (so-called
ded tires (with the wire bead removed). This is a combustor
‘‘topping’’) combustor is thereby greatly reduced. In the un-
in which the combustion air is fed through a grate at the bot- modified
intercooled-regenerative cycle there is only one,
tom with sufficient velocity to maintain the coal particles, high-
pressure combustor). The purpose of the modification to
limestone, and pieces of tire in an airborne (fluidized) state. this cycle
is to avoid contact of the solid-fuel constituents with
There are several types of fluidized-bed combustor; they are the highly
stressed turbine blading while allowing a high tur-
particularly effective in burning ‘‘dirty’’ fuels like coal, partly bine-inlet
temperature to be used to produce high efficiencies
because intimate contact is given with limestone and other (50% to
60%). About one-half the thermal input is through
absorbents to remove sulfur and other pollutants. Shredded the low-
pressure combustor, and about one-half is through
tires have also been burned in 560 MW cyclone boilers at Illi- the high-
pressure combustor. This design is named the sup-
nois Power’s Baldwin plant. The company estimated fuel sav- plementary-
fired exhaust-heated gas turbine (SFEHGT). The
ings of two-thirds of a million dollars annually while reducing refuse
combustor burns RDF in a fluidized bed together with
coal consumption by 80,000 tonnes and reducing sulfur diox- sorbents
such as lime. It is unlikely that hot-gas cleanup will
ide emissions by over 3000 tonnes (13). be needed.
However, it has been tentatively specified. Various
Three plants in the United States are totally fueled by hot-gas-
cleanup systems for coal combustion are being devel-
scrap tires, taking about 10 million tires annually (13). (About oped for
the US Department of Energy; they are reviewed by
120 million automotive tires are discarded annually in the Webb
(17). The process uses a moving-module regenerator
United States). patented by
M.I.T. To withstand the high temperatures trans-
Another fuel that has been co-fired with solid wastes is oil ferred from
the RDF combustor the heat-exchanging surfaces
shale, in the proportion of 75% RDF to 25% oil shale on an are of
ceramic honeycomb. The ceramic modules are assem-
----------------------- Page 615-----------------------
Cooling water
Intercooler
Inlet air
Power
Compressor Compressor
Turbine
Natural
gas
or
petroleum
Scrubber
Refuse
Burner
or RDF
Mechanical
link
Refuse
Hot-gas
burner
cleanup
Figure 6. Supplementary-fired exhaust-

Heat exchanger
heated gas turbine (SFEHGT) cycle dia-
gram. (From Ref. 7.) This is, loosely, an

IDF
Waste-heat boiler
intercooled-regenerative gas turbine with
the addition of a refuse burner in the tur-

To stack
bine exhaust, and an induced-draft fan
after a scrubber.
bled into two heat-exchanging ‘‘faces’’ as they are shuttled

increases the incineration-plant thermal efficiency by reduc-
around a closed loop (Fig. 7). Modules can be individually re-
ing the turbine back pressure. It thus allows hot gas to be
placed for servicing without shutting down the plant. The re-
expanded further, while cold gas is compressed at a lower cost
generator is described by Wilson (19). The gas leaving the in
power than the increment delivered by the turbine. It also
regenerator at just under atmospheric pressure passes to a
aids in simplifying the feed process for the RDF through re-
waste-heat boiler to produce low-quality steam or hot water;
ducing the pressure in the solid-waste combustor to slightly
doing so further cools the gases, thereby aiding in exhaust-
below atmospheric pressure. The use of two combustors thus
gas cleanup. No credit for the energy content of this byprod-
allows very high efficiencies to be obtained in a gas-turbine
uct has been assessed in calculating the overall thermal effi-
plant that has direct combustion of RDF and that does not
ciency. The gases leave the waste-heat boiler and pass to a
require an associated steam-turbine plant. This process is in
water-spray scrubber that (a) removes chlorides and other
the laboratory stage.
soluble and condensable pollutants from the exhaust flow and
(b) cools the gases to close to atmospheric temperature. A mo-
Alternative Technologies for the Conversion
tor-driven induced-draft fan takes the cool moist gas up to of
Wastes to Clean Fuel
atmospheric pressure. It could therefore be of fiberglass or
One gasification process should be mentioned first: natural

similar low-cost construction. Its use has two advantages. It
decomposition. Gas turbines have run on sewage gas at least
since the 1950s, and from at least the early 1980s large land-
fills have been capped and drilled to supply fuel gas, largely
methane, to gas turbines.

Modular regenerator
Three methods of improving on the slow natural processes
Compressor
have been strongly advocated, but have not been successfully
Gas combustor
put into practice to the end of the 1990s. Pyrolysis or starved-
air combustion involves heating solid wastes in

absence the
RDF of
sufficient air to achieve full combustion, an established pro-
combustor
cess (carbonization) for converting coal into coke and coal gas.
The products of the pyrolysis of solid wastes are a char and a

Turbine
gas. The carbonization process, as recommended by Beer (20),
is
particularly suitable to the SFEHGT cycle discussed above.
Modular regenerator
The two other processes are acid and enzymatic hydrolysis
of
organic wastes to produce ethanol. Lynd et al. (21) point
Figure 7. Conceptual sketch of a gas turbine plant. (From Ref. 7.)
out that the technology for ethanol production from cellulosic
An attempt has been made to show the size of the RDF combustor
materials is fundamentally different from that for production
(which could be an incinerator) in relation to those of other compo-
for food crops. (Fuel ethanol is successfully produced by fer-
nents.
mentation, in Brazil from sugar cane and in the United States
----------------------- Page 616-----------------------
from corn and other starch-rich grains.) Lynd et al. state that
bines the output and efficiency are increased through the use
1990 production cost of ethanol is similar for the two forms of of
a ‘‘reheat’’ combustor. This increases the gas temperature
hydrolysis. However, the enzymatic process was at an earlier
after the gas has expanded through the first turbine stage.
stage of development and was likely to be the more cost-effec-
There is more heat in the exhaust, and therefore more power
tive after further research. Neither had favorable conversion is
delivered by a turbine running on the steam generated
economics in 1990. The authors point out that there is a con-
from this heat.
siderable quantity of solid material left after hydrolysis that,
There is sufficient oxygen in the exhaust of a simple-cycle
after dewatering, could be used as a solid fuel. gas
turbine to support additional combustion. However, most
combined-cycle plants do not have supplementary firing. The

Gasification. The most promising methods of burning coal
temperature of the steam at the stop-valve of large steam tur-
(and, potentially, solid wastes) in gas turbines appear to in-
bines is around 840 K, a temperature limit set by the increas-
volve the use of gasifiers (e.g., the Lurgi), which have been ing
presence of free radicals that cause corrosive degradation
long established. They can be air-blown or oxygen-fed; the lat- of
the steels used in the steam-generator superheaters. It is
ter brings a large increase in the capital cost and a large de-
desirable that the steam reach, but not exceed, this tempera-
crease in the quantity of gas to be cleaned. Further reduction
ture. The increasing turbine-inlet temperatures of modern
in cleaning requirements can be brought about by incorporat-
gas-turbine plant match the required steam conditions with-
ing catalytic cracking to convert the tars and other heavy con- out
the need for further combustion. There is also benefit in
stituents to lighter fractions, according to Ghezzi et al. (22).
increasing the output of the gas turbine by incorporating in-
These authors state that the technology is as-yet experimen-
tercooling and reheat (which is the incorporation of secondary
tal for municipal solid wastes. Emsperger and Karg (6) claim
combustors along the path of the turbine expansion, between
that gasification (in the integrated gasification combined cy-
stages), thereby also increasing the temperature of the tur-
cle, described below) could be implemented in 1996 for the
bine exhaust gases.
large quantities of oil-refinery wastes.
Integrated-Gasification Combined Cycle. Another variation

Alternative Cycles for the Efficient is
the integrated-gasification combined cycle (IGCC) that in-
Conversion of Clean Fuels (23)
corporates a system producing gas from coal. Where the gas-
Combined Cycles. The combined cycle is the most-used

ifier is oxygen-fed the system must include an oxygen plant
variation of the basic gas-turbine cycle in the last few years in
addition to the gasification plant, leading to a capital cost
of the twentieth century. The simplest form is the combined-
reported as approximately three times that of a CCGT fired
heat-and-power plant, or CHP (Fig. 8). A gas-turbine engine by
natural gas. The ability to use a low-cost fuel, coal, in an
exhausts hot gas into a heat-recovery steam generator
environmentally benign manner will justify the additional
(HRSG). The steam from the HRSG is led to a process appli-
capital cost in certain circumstances in the 1990s, and pre-
cation (for instance, a paper-making plant) or to building or
sumably in more circumstances later when natural-gas prices
district heating. In a true combined-cycle plant the steam op- are
certain to rise. The 250 MW Demkolec plant in the Neth-
erates a steam-turbine plant (Fig. 9), and the plant is some-
erlands started trial operation in 1994, and the Wabash River
times called a CCGT plant, for ‘‘combined-cycle gas turbine,’’
plant in Indiana started trials in 1995. The capital cost of
although manufacturers like to devise their own names for
larger plants in the United States was estimated at about
their particular offerings. (For instance, GE uses ‘‘STAG,’’ for
$1600/kW; several other IGCC plants are in the advanced
steam and gas.) Sometimes the gas-turbine part is called the
planning stage (24).
topping cycle and the steam-turbine portion the bottoming cy-
cle. Most of the new generating plant being built around the
CCGT and PFB. Coal is also being used to power combined-
world is designed to this cycle. Efficiencies of the small plants
cycle gas turbines by using pressurized fluidized beds for com-
are in the range of 50%, while for the larger plants it can go
bustion, initially in Spain, Japan, and the United States. The
as high as 60%.
beds contain limestone and other sorbents that, together with
This high efficiency is likely to be first achieved by turbines
slag-melting on the walls and base of the bed, produce a hot
produced by ABB (a Swiss–Swedish company) working on a gas
that can pass through a gas-turbine expander without
combined cycle. In the company’s GT24 and GT26 gas tur-
causing more than minor erosion, corrosion, or deposition.
Stack
Turbine- Shaft-power
Compressor
expander load
Hot
Air
Combustor- Process
exhaust
burner or thermal
load
Figure 8. Combined heat and power (CHP) plant.
(From Ref. 23.) As shown, this is a plant for ‘‘clean’’
fuel, oil or gas. The hot exhaust can raise steam for
process plants or for district heating.
Fuel
----------------------- Page 617-----------------------
Gas Turbine "Topping" Cycle
Stack
Turbine-
Shaft-power
Compressor
expander
load (1)
Air Combustor-
Heat-
burner
recovery
steam
Shaft-power
Hot
generator Steam- load (2)
Fuel exhaust
turbine
Condenser
Steam-Turbine
Boiler
"Bottoming" Cycle
feed pump
Figure 9. Combined cycle plant. (From Ref. 23.)

This is another plant
using a ‘‘clean’’ fuel. In
this case, the steam raised in a heat-recovery steam
generator is fed to a steam turbine, increas-
ing the power output and thermal efficiency.
The prices forecast for the plants are 75% of those for the than
for the combined cycle because virtually all the water is
IGCC plants.
normally lost in the exhaust rather than being circulated
in
a closed system and because the specifications are more
Steam-Injection Gas Turbines. Steam injection in a location
stringent. Any dissolved solids that become deposited on the
where it will expand through the turbine blading with the
turbine blades or elsewhere could form corrosion sites or po-
combustion gases is a third use (besides expansion in a steam
tential blockages. However, Tuzson states that water-purifi-
turbine and, in advanced plants, cooling of the turbine blades)
cation cost is of the order of 5% of the fuel cost and is not,
of the steam generated in a HRSG. Steam may be injected
therefore, a decisive factor. The reliability of early steam-
upstream of or into the combustion chamber, or into the tur-
injected units has been high—for instance, 99.5%. Rather
bine nozzles anywhere along the expansion. The steam does
surprisingly, combustor-liner durability has been found to
less work the further along the expansion it is injected. In
increase.
comparison with the combined cycle, the steam-injected cycle
One of the advanced gas-turbine systems being developed
has the following advantages. A substantial increase in power in
Japan uses an intercooled-reheated gas turbine (the in-
can be obtained from the gas-turbine engine with no modifi-
tercooler is a water-spray direct-contact type) in which the
cation in the configuration of the expansion turbine itself. The
steam raised in the HRSG can power a conventional steam
part-load efficiency is improved. The production of NOx is re-
turbine, or the steam can be injected into the gas turbine (26).
duced. In a review of the status of steam-injected gas tur- The
output, 400 MW, and the predicted efficiency, 54.3%,
bines, Tuzson (25) stated that combined-cycle turbines have
place it outside Tuzson’s guidelines above.
demonstrated the highest power-generation efficiencies and A
gas turbine is a good candidate for steam injection if the
the lowest cost in sizes above 50 MW (although he also quotes
compressor has a wide range of operation because the in-
a study giving the power level below which steam-injection
creased flow creates a higher back pressure. A high pressure
systems become more attractive than combined cycles as 150
ratio and a high turbine-inlet temperature are also desirable.
MW). At lower power levels the steam-injected gas turbine
These conditions seem to favor the aircraft-derivative turbine.
becomes attractive because of the avoidance of the large cost
However, Tuzson points out that heavy-duty industrial tur-
of the steam turbine. A typical power gain from steam injec-
bines can accommodate concentrations of contaminants about
tion for a GE LM5000 gas-turbine engine was quoted as in- five
times higher than can the aircraft-derivative turbines.
creasing the engine output for 34 MW to 49 MW, together
There are many variations of these relatively simple forms
with an efficiency increase from 37% (simple cycle) to 41%. of
water/steam injection. El-Masri (27) proposed an in-
GE analyzed the gains that would be obtained from a combi-
tercooled-recuperative cycle in which the intercooler and an
nation of intercooling and steam injection for its LM5000 gas-
aftercooler are direct-contact water-injected evaporative units
turbine engine: a power increase from 34 MW to 110 MW and and
there is subsequent water injection into the recuperator
an efficiency improvement from 37% to 55%. The water-puri-
(Fig. 10). There is no steam generator. The results of his anal-
fication requirements are more demanding for steam injection ysis
show considerably higher efficiencies over the conven-
----------------------- Page 618-----------------------
Stack
5
Recuperative heat exchanger
Water
Air
2B
2
a
Water Combustor- Fuel
Combustor- 4A
1 burner
burner
4
CD
3
High-pressure
High-pressure
Low-pressure compressor
turbine Low-pressure Power
turbine
compressor
Cooling flows
2A 1A
Water Evaporative
intercooler
Figure 10. El-Masri intercooled recuperative cycle.

(From Ref. 23.) In
this cycle, water is in-
jected into the airstream instead of using an
intercooler, and water is also injected into the
combustor and into the recuperator. Thus power
output and efficiency are simultaneously in-
creased at very low cost.
Combustor-
burner
Fuel
Low-pressure High-pressure
Turbine-
compressor compressor
expander
Recuperator
Air
Humid air
Warm
water
Warm water
Mixer
Intercooler
Aftercooler
Economizer
Water
Pump
drain
Water
Figure 11. Humid-air cycle. (From Ref. 23.) Water

is injected downstream of the compressor,
increasing the mass flow through the turbine and
recuperator, and raising the efficiency and
power output.
----------------------- Page 619-----------------------
tional intercooled-recuperative cycle and over steam-injected

BIBLIOGRAPHY
cycles. D. D. Rao (cited in Ref. 28) proposed the humid-air
cycle (Fig. 11) in 1990: it incorporates a water-cooled surface
1. D. G. Wilson , History of solid-waste management, in D. G. Wil-
intercooler and a similar aftercooler, followed by a water-in-
son (ed.), Handbook of Solid-Waste Management, New York: Van
jection evaporator, a recuperator, and a combustor. Rao be-
Nostrand Reinhold, 1977.
lieves that the cycle has advantages over the steam-injected
2. R. Huxford, Recycling reappraised, In Municipal Engineer, Proc.
cycles.
Inst. Civil Eng., Vol. 109, 1995, pp. 35–39.
3. Anonymous, Refuse processing and resource recovery, Public

Chemical Recuperation. Energy can be recovered from the
Works, 127 (5): E23–E36, 1996.
turbine exhaust by means other than heat transfer to the
4. W. R. Niessen, Estimation of solid-waste production rates, In D.
compressed-air flow. A chemically recuperated gas turbine
G. Wilson (ed.), Handbook of Solid-Waste Management, New York:
(CRGT) has been proposed by Kesser et al. (29) in which a
Van Nostrand Reinhold, 1977.
simple-cycle gas turbine passes its exhaust to a heat-recovery
5. L. I. Anderson, Organic Solid Wastes Produced in the United
methane-steam reformer. This takes the place of an HRSG in
States, 1971, Bureau of Mines Inf. Circular 8549, 1972, p. 13.
a steam-injected plant. ‘‘The tubes in the reformer (unlike the
6. W. Emsperger and J. Karg, Power from Waste, Power Eng. J., 10
tubes in the HRSG superheater) are filled with a nickel-based
(1): 35–41, 1996.
catalyst that promotes a chemical reaction between steam
7. D. G. Wilson et al., A waste-to-energy recycling plant for Beirut,
and methane.’’ The gaseous fuel that results is a mixture of
Proc. Inst. Mech. Eng., J. Power Energy, 209: 63–70, 1995.
CO, H , and CO and is assumed to produce NO emissions as
8. H. Alter, The ‘‘recycling’’ of densified refuse-derived fuel, Waste
2 2 x
low as 1 ppm (versus 25 ppm for a steam-injected system).

Manage. Res., 14: 311–317, 1996.
The authors calculated the thermal efficiency of the cycle (at
9. F. C. Lockwood and J. J. Ou, Review: Burning refuse-derived fuel
around 47%) as less than the combined cycle but better than
in a rotary cement kiln, Proc. Inst. Mech. Eng., Part A, 207 (A1):
the steam-injected cycles for the power level of interest.
65–70, 1993.
10. R. B. Engdahl, Energy recovery from raw versus refined munici-

Inlet-Cooling Cycles. A somewhat similar approach to
pal wastes, In National Waste Process. Conf., ASME Solid Waste
chemical recuperation is to use the exhaust heat via an HRSG
Process. Division, 1986.
to operate an absorption chiller that can be used to cool the
11. H. G. Rigo and A. D. Conley, Waste-to-energy facility capital
compressor-inlet air [Ebeling et al. (30)]. However, a more
costs, In National Waste Process. Conf., ASME Solid Waste Pro-
widespread practice is to use power off-peak-load (in the case
cess. Division, 1986.
of the Butler–Warner plant in Fayetteville, NC, having eight
12. H. Haneda, Efficiency improvement options for municipal waste-
GE Frame-5 units, 4.3 MW is used off-peak, 148 hours per
fired power generation—recent development activities in Japan:
week) to operate a vapor-compression ice-making system. Ice
A review, Proc. Inst. Mech. Eng., Part A, J. Power Energy, 209:
81–100, 1995.
is melted during power peaks for four hours per day, five days
per week. An increase of hot-weather capacity of 29% was
13. L. Lamarre, Tapping the tire pile, EPRI J. 20 (5): 28–34, 1995.
given for the plant quoted. The modification paid for itself in
14. H. E. McCarthy, R. L. Clayson, and G. R. Short, Solid waste/
under a year.
oil shale combustion—an environmentally benign plant, IECEC
paper no. E1-163, ASME, 1995.
15. Anonymous, Solid-Wastes Management Technology Assessment,

CONCLUSION
New York: Van Nostrand Reinhold, 1975.
16. C. Keller, Closed-cycle gas turbines for all fuels, Escher-Wyss

If the energy in wastes is converted to heat to produce steam,
News, 39: 1, 1966.
the design of the steam turbine is almost solely a function of
17. H. A. Webb, Coal-fueled heat engines and gas-stream-cleanup

the size of the plant. The larger the power output, the more
systems, In Proc. 7th Annu. Contractors’ Rev. Meet., US Depart-
measures are worth taking to improve the efficiency (includ-
ment of Energy, 1990.
ing increasing the steam pressure and temperature). If the
18. D. G. Wilson, The supplementary-fired exhaust-heated cycle for
wastes are converted to a clean fuel, however, a higher effi-
coal, wood and refuse-derived fuel, Proc. Inst. Mech. Eng. Part A,
ciency and lower plant cost will occur if a plant based on a
J. Power Energy, 207 (A3): 203–208, 1993.
gas turbine, or a combination of gas and steam turbines, is
19. D. G. Wilson, Low-leakage and high-flow regenerators for gas-
chosen. Here there are many potential alternatives, including
turbine engines, Proc. Inst. Mech. Eng. Part A, J. Power Energy,
the possibility of burning the waste as RDF directly in one
207 (A3): 195–202, 1993.
cycle. Only some of the alternatives were fully developed and
20. J. Beer, Personal communication, Cambridge, MA: MIT, 1992.
viable in the mid-1990s. The brief descriptions given will, it
21. L. R. Lynd et al., Fuel ethanol from cellulosic biomass, Science,
is hoped, allow the reader to recognize and evaluate the prom-
251: 1318–1323, 1991.
ise of future developments.
22. U. Ghezzi, S. Pasini, and L. Degli Antoni Ferri, Incineration ver-
If the definition of wastes is expanded to include wastes
sus gasification: A comparison in waste-to-energy plants, IECEC
occurring in nature, the mixing of fresh water and salt water
paper no. RE-409, ASME, 1995.
at the mouth of rivers like the Amazon and at the edge of
23. D. G. Wilson and T. P. Korakianitis, Thermodynamics of gas-tur-
polar ice sheets could, theoretically, produce vast flows of en-
bine cycles, In The Design of High-Efficiency Turbomachinery and
ergy. The reader merely has to imagine the power required in
Gas Turbines, 2nd ed., Englewood Cliffs, NJ: Prentice-Hall, 1997.
desalination plants to accomplish the reverse process, and to
24. I. Stambler, Progress in IGCC and advanced cycles outlined at
divide this number by two or three to allow for thermody-
EPRI meeting, Gas Turbine World, Jan.–Feb., 16–33, 1966.
namic and practical irreversibilities, to arrive at an estimate
25. L. Tuzson, Status of steam-injected gas turbines, J. Eng. Gas Tur-
of the potential power production.
bines Power, 114: 682–686, 1992.
----------------------- Page 620-----------------------
458 WATER TREEING
26. K. Takeya and H. Yasui, Performance of the integrated gas and

steam cycle (IGSC) for reheat gas turbines, J. Eng. Gas Turbines
Power, 110, 220–232, 1988.
27. M. A. El-Masri, A modified high-efficiency recuperated gas-tur-

bine cycle, J. Eng. Gas Turbines Power, 110, 233–242, 1988.
28. S. S. Stecco, The humid-air cycle: Some thermodynamic consider-

ations, ASME paper no. 93-GT-77, ASME, 1993.
29. K. F. Kesser, M. A. Hoffman, and J. W. Baughn, Analysis of a

basic chemically recuperated gas-turbine power plant, J. Eng.
Gas Turbines Power, 116: 277–284, 1994.
30. J. E. Ebeling et al., Peaking-gas-turbine capacity enhancement

using ice storage for compressor-inlet-air cooling, ASME paper
92-GT-265, ASME, 1992.
DAVID GORDON WILSON

Massachusetts Institute of
Technology
----------------------- Page 621-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Wind Power

Standard Article

Frede Blaabjerg1 and Ned Mohan2
1Aalborg University, Aalborg East, Denmark
2University of Minnesota, Minneapolis, MN,

Search
reserved.
DOI: 10.1002/047134608X.W3049
❍ Acronym Finder
Abstract
Power in the Wind
Power in a Wind Turbine
Power Control Methods
Simple Wind Power Systems
Advanced Wind Power Systems

%20Conversion/W3049.htm17.06.2008 16:50:07
----------------------- Page 622-----------------------
WIND POWER 613
ing
offshore in order to get the maximum power from the
wind.
Offshore the landscape is flat with high wind speeds.
POWER
IN THE WIND
When
the wind blows, kinetic energy is stored in it. This en-
ergy
is transformed to mechanical energy by use of a wind
turbine. The transfer function from the wind to the mechani-

cal
shaft power Pv in a wind turbine is
Pv = 1 Av 3ρCp (1)
2
A = π(d/2)2 (2)
where
Pv is the mechanical shaft power (W), A is the total
2
area
swept by the wind turbine (m ), d is the turbine diameter
(m),
v is the wind speed (m/s), # is the mass density of air
(1.225 kg/m ), and Cp is the power coefficient for the turbine.
Equation (1) shows that the power in the wind is propor-
tional to the cube of the wind speed.
POWER
IN A WIND TURBINE
WIND POWER
The
only unknown parameter in Eq. (1) is the power coeffi-
Energy comes from various sources. The major sources of en- cient
Cp for the turbine. This coefficient is a characteristic of
ergy for human use are fossil fuels such as oil, coal, and natu- a
specific wind turbine construction, and it is normally shown
ral gas. Another important contributor is nuclear energy. All as a
function of the ratio # of the tip speed to the wind speed:
these sources produce environmental pollution. Alternatives
to them are renewable sources such as solar, ocean waves,
λ = (d/2)ωr (3)
and wind energy. In some areas of the world, wind can be an
v
important contributor to electricity production. It gives elec-
trical power without any emission or pollution. The energy where
#r is the rotor speed and # is the tip–wind-speed ratio.
from the wind is normally extracted by a wind turbine, which
The tip speed will typically be fixed if an induction genera-
converts the kinetic power in the wind to mechanical power, tor
is used and connected directly to the grid. Measured Cp
which in many applications again is converted into electrical
characteristics for two windmills are shown in Fig. 2. The
power by use of an electrical generator. Figure 1 shows photo-
measured curves show that Cp has an experimental maximum
graphs of a wind turbine and a whole wind turbine park. of
0.48–0.5. It can be shown theoretically that the absolute
There are two basic wind power systems. In a standalone
maximum Cp is 0.59. The Cp curve is used for power control
system where wind is a dominant source of generation, a of a
windmill and is an important consideration when a wind
backup source based on diesel engines is needed because the rotor
is designed aerodynamically. The Cp curve also shows
wind is not constant. In the other system, the wind-electric that
the maximum Cp can be obtained at only one wind speed
generator directly feeds power into a well-established electri- if
the blade tip speed is fixed.
cal power grid. Here also, if wind-electric is a significant por-
tion of the overall generation, fluctuations in wind speeds POWER
CONTROL METHODS
cause undesirable power system oscillations. To overcome
this, energy storage has to be available on a large scale, which Stall
Regulation
is usually not economical. Batteries, superconducting induc-
tors, and fuel cells are obvious means, but in general, storage The
most commonly used power control method at higher
systems reduce the overall efficiency and cause the price to wind
speeds is stall regulation, where the rotor profile of the
be too high. Another solution, to keep wind-electric generation wind
turbine controls the mechanical power and limits it. The
fairly constant, requires reducing the power production of the
position of the rotor profile is always fixed in relation to the
wind turbine system by control at high wind speeds. However,
windmill, and since the rotor speed is almost fixed (varies
this solution is not beneficial for the economy of the wind only
1% to 2% due to the slip of the induction generator), the
power system. wind
speed across the rotor profile will increase, and if the
In many areas wind power will be a fractional supplement speed
exceeds a certain limit, turbulence (stall) will result.
(
10%) to future energy production because of the necessity Then
power is lost and the Cp value decreases. If the wind
of a backbone power source. In some special areas where the
turbine is properly designed, stall regulation along the whole
energy transportation cost is high, the wind power can be a wind
turbine profile can be used to control the maximum
significant portion. Typical wind power systems are today in shaft
power Pmax according to the equation
the power range of 400–600 kW, but new commercial 1.5 MW
P = C (λ)P = constant (4)

systems are on the market now. Wind turbines are also mov-
max p v

----------------------- Page 623-----------------------
614 WIND POWER
(a)
(b)
Figure 1. Photographs of (a) a single wind

turbine, (b) a wind turbine farm.
where Pmax is the maximum allowable shaft power to the elec-

The control unit can work in different modes. One mode
trical generator. The main benefits of such a control system is
based on the measured active power to the grid; a search
are that it is simple, cheap, and reliable. Figure 3 shows a
algorithm is used in order to get the maximum output power.
measured power curve for a stall-regulated 500 kW windmill
Another method is to measure the wind speed and use a
obtained at different wind speeds. There is only a small over-
lookup table to give the pitch angle. It is then possible to ob-
shoot above the 500 kW maximum level. The power is obtain
more power from the wind over the whole range of
tained as an average over ten minutes.
speeds. Figure 5 shows a measured power curve for a pitch-
controlled 500 kW windmill.

Pitch Control
Figure 5 shows that a very constant power is measured at
An alternative method to control the power is pitch control,
higher wind speeds, but at lower wind speeds it is not signifi-
where the rotor turbine (pitch) changes position continuously.
cantly different to the stall-controlled windmill of Fig. 3. The
When the generator reaches its power limit at high wind power
is again measured as an average over ten minutes. In
speeds, the pitch position is changed so that a fixed mechani- both
stall and pitch control fluctuations occur due to wind
variations.
cal power is obtained and thereby also a fixed electrical
power. This also reduces the mechanical stress in the system
compared to stall regulation. A second effect is that at lower
SIMPLE WIND POWER SYSTEMS
wind velocities the highest possible Cp value can be obtained
all the time. For
low power (
1 MW) the cheapest generator on the market
This strategy gives a more flexible control system, but it is
widely used: the three-phase squirrel-cage induction gener-
also requires a more complex system to control the pitch. This ator
(IG), which can be connected directly to the grid without
is traditionally done by a hydraulic system in the windmill any
special power and control electronics. For higher power
but some systems are also electrically based. A basic control other
systems can be preferable because of the requirement to
system is shown in Fig. 4, where a control unit adjusts the cool
the rotor. The only power electronics needed are thyristor
pitch.
silicon-controlled rectifiers (SCRs) for limiting the inrush cur-
----------------------- Page 624-----------------------
WIND POWER 615
0.8 800
700
0.6 600
) 500
W
k
(
C r 400
p 0.4 e
w
o
P 300
200
0.2
100
0
0.0
0 5 10 15 20 25
0 5 10 15 20 25
Wind speed (m/s)
Wind speed (m/s)

Figure
3. Measured power curve for a stall-regulated windmill:
(a) Nordtank
NTK 500/37.
0.8
Induction
generator
0.6
(4-pole)
ω ω
Wind
r g
Gearbox Grid
C
p 0.4
Pitch Control
control unit
0.2
Wind
0.0
Wind
0 5 10 15 20 25
speed
Wind speed (m/s)
Figure 4. Pitch-controlled windmill.

(b)
Figure 2. Power coefficient curves for two windmills (measured): (a)

Nordtank NTK 500/37, (b) Vestas V39.
800
700
rent during startup. One alternative is the three-phase syn-
chronous generator (SG), but this generator is very stiff, and 600
without any special damping of wind power fluctuations,
which can give problems. The SG is also more expensive than ) 500
W
the IG. The most cost-optimal solution at present is therefore k
(
r 400
a standard four-pole squirrel-cage induction generator with a e
w
gearbox. Multipole IGs and SGs can also be used. Figure 6 o
P 300
shows the systems for fixed-speed operation.
Some systems use also an induction generator with pole 200
change, so the windmill can operate at two speeds. This gives
a better utilization of the wind, but the generator is more 100
expensive. Different techniques exist to get maximum energy
from the wind at a given place in the landscape; in particular, 0
0 5 10 15 20 25
it is not always best to have the highest Cp value of the tur-
bine and windmill at maximum wind speed. It is better to
Wind speed (m/s)
maximize Cp at the wind speed that has the highest proba- Figure
5. Measured power curve for a pitch-controlled windmill:
bility. VESTAS
V39.
----------------------- Page 625-----------------------
616 WIND POWER
Induction ADVANCED
WIND POWER SYSTEMS
generator
(4 pole) As
mentioned, the maximum Cp of the wind turbine can only
Wind ω r ωg be
reached at one point in fixed-speed operation. In order to
obtain
more energy from the wind, adjustable-speed systems
Gearbox Grid
can be
used, but they require power electronics.
A
number of different topologies are possible when ad-
justable
speed is used. The most promising alternative is a
(a) forced-
commutated, voltage-source ac–dc–ac converter with
an
induction generator, which is shown in Fig. 8. Another
ω ω
Wind r g
possibility is a current-source inverter, which has poorer dy-
Synchronous
namics,
as shown in Fig. 9.
Gearbox generator Grid
(4 pole) Up to
2 MW forced-commutated voltage-source systems are
available using modern power electronic devices; current-

source
converters are used in very
high-power applications.
Field In Fig.
8, the inverter to the grid should not produce any sig-
(b) nificant
harmonic currents. Another system for windmills
Induction that has
been discussed for many years is based on a doubly
generator fed
induction generator (DIG), where the rotor is wound and
(multipole)
connected by brushes to the outside world. Figure 10 shows a

ω
Wind r typical
topology where the energy in the rotor is fed back to
Grid the
grid.
Some
of the advantages in this system are that ideally it is
possible
with a standard 1 MW generator to get 2 MW output
(c) without
overloading the system, and the speed can be ad-
Figure 6. Fixed-speed windmill generator systems: (a) induction justed

from synchronous speed upwards. Of course, the gener-
generator with gearbox, (b) synchronous generator with gearbox, (c) ator
has a greater thermal load, which can reduce the output
induction generator without gearbox. power,
and it also operates at higher speed. The reactive
power
consumption in the stator can be compensated by the
dc–ac
converter. A transformer is necessary in order to opti-
mize
the size of the converter. The disadvantages are the
Disturbances to the Grid
brushes,
the wound rotor, and a more complex control
One of the main problems with windmills is that the induc-
strategy.
tion generator requires reactive power for magnetization, An
alternative to the system in Fig. 10 is a doubly wound
which loads the grid unnecessarily. The problem is tradition-
induction generator where no brushes are used, as shown in
ally solved by inserting capacitor banks between the windmill Fig.
11. A variable resistance is put into the rotor and con-
and the grid as illustrated in Fig. 7. trolled
by power electronics. The rotor resistance causes a
The number of capacitor banks depends on the require- speed
variation (e.g., 5%) and this can be used for power
ments of the grid utility company. In many cases there is only control.
The losses in the resistor can be dissipated outside on
one capacitor bank, but for new 1.5 MW windmills in Den- the
rotor shaft. The power electronics is built into the rotor.
mark the requirement is that the power factor should be at Two
obvious alternatives to the induction generator sys-
least 0.99 over the whole operating region. Therefore a num- tems in
Fig. 8 and Fig. 9 are a synchronous generator with
ber of capacitor banks are used in high-power applications. the
same converters. They are shown in Fig. 12 and Fig. 13.
The utility also requires the banks to be disconnected from The
synchronous generator offers higher efficiency, but is in
the grid if the grid breaks down, to be sure that the windmill general
more expensive. However, a degree of freedom exists
does not run as a standalone unit. due to
the possibility of controlling the field. For a long time
Induction
generator
ω ω
Wind r g
Gearbox
Grid
Reactive
power
control
Figure 7. Capacitor banks in a typical

windmill installation.
Bank 1 Bank 2
----------------------- Page 626-----------------------
WIND POWER 617
Induction
generator
ω ω ac/dc dc/ac
Wind r g
Gearbox
Grid
Figure 8. Forced-commutated voltage-
source ac–dc–ac converter for induction
generator.
Induction
generator
ω ω
Wind r g
Gearbox
Grid
Figure 9. Current-source converter for
induction generator.
DIG
ω ω
Wind r g
Gearbox
Grid
ac/dc dc/ac Step-up

Figure 10. Doubly fed induction genera-
tor with ac–dc–ac converter and step-up

Transformer
transformer.
DIG
ω ω
Wind r g
Gearbox
Grid
ac/dc
Resistor Figure 11. Doubly wound induction gen-
erator with variable rotor resistance.
ac/dc dc/ac
Wind
ω ω
r g
Gearbox Synchronous
Grid
generator
Figure 12. Forced-commutated voltage-
sound ac–dc–ac converter for synchro-
nous generator.
Wind
ω ω
r g
Gearbox Synchronous
Grid
generator
Figure 13. Current-source converter for
synchronous generator.
ac/dc dc/ac
Wind
ω
r Multipole
synchronous Grid
generator
Figure
14. Multipole permanent-magnet synchronous generator
with
voltage-source forced-commutated converter.
----------------------- Page 627-----------------------

618 WIND POWER PLANTS
M. Papadopoulos and P. Malatestas, Operational advantages from

Wind ω Multipole ac/dc dc/ac
the use of ac/dc/ac interface for the connection of wind turbines in
r
switched-
diesel power systems. Eur. Community Wind Energy Conf., 1990,
Grid
reluctance
pp. 586–590.
generator
M. P. Papadopoulus and L. Papathanassiou, DQ modelling of a vari-
able speed wind turbine equipped with a double output induction
generator. Stockholm Power Tech. Conf., 1995, pp. 194–199.

Figure 15. Multipole switched-reluctance generator with forced-
commutated converter.
S. J. Philips, W. L. James, and C. V. Nayer, Power electronic system
optimisation for remote area mini grid diesel systems. IEEE
IECON ’90, 1990, pp. 1029–1033.
R. D. Richardson and W. L. Erdman, Variable speed wind turbine,

attempts have been made to avoid the gearbox because there
U.S. Patent No. 5,083,039, 1992.
are losses (4% or more) in it, and one less component in the
A. J. Ruddell et al., A wind/diesel system with variable speed fly-
system gives higher reliability. In order to avoid the gearbox,
wheel storage, Wind Eng., 17 (3): 129–146, 1993.
a multipole generator is one possibility. Two such machines
A. C. Saramourtsis et al., Probabilistic evaluation of the performance
are a multiple permanent-magnet synchronous generator
of wind-diesel energy systems, IEEE Trans. Energy Convers., 9:
(MPSG) and a multiple switched-reluctance generator
743–752, 1994.
(MSRG). The systems are shown in Fig. 14 and Fig. 15 respec-
´
R. Spee, S. Bhowmik, and J. H. R. Enslin, Adaptive control strategies

tively.
for variable-speed doubly-fed wind power generation systems. In
Major problems with multipole generators include the com-
Proc. 1994 IEEE Ind. Appl. Soc. Annu. Meet., vol. 1, 1994, pp.
plexity of the windings and the size of the generator. A major
545–552.
drawback with the MSRG is the high reactive current, which
K. Uhlen, B. A. Foss and O. B. Gjøsøter, Robust control and analysis
leads to extra losses. Readers interested in these topics may
of a wind-diesel hybrid power plant, IEEE Trans. Energy Convers.,
refer to the Bibliography.
9: 701–708, 1994.
L. Xu and Y. Tang, A novel wind-power generating system using field

orientation controlled doubly-excited brushless reluctance ma-
BIBLIOGRAPHY
chine. In Proc. 1992 IEEE Ind. Appl. Soc. Annu. Meet., 1992, pp.
408–413.
B. S. Borowy and Z. M. Salameh, Optimum photovoltaic array size
M. Yamamoto and O. Motoyoshi, Active and reactive power control
for a hybrid wind/pv system, IEEE Trans. Energy Convers., 9: 482–
for doubly-fed wound rotor induction generator, IEEE Trans.
488, 1994.
Power Electron., 6: 1991, pp. 624–629.
B. K. Bose and M. G. Simoes, Fuzzy logic based intelligent control of a

˜
variable speed cage machine wind generation system. IEEE Power
FREDE BLAABJERG
Electronics Specialists Conf., 1995, pp. 389–395.
Aalborg University
C. Brothers et al., Harmonics on wind/diesel grids, Wind Eng., 17

NED MOHAN
(6): 323–331.
University of Minnesota
´
C. Brune, R. Spee, and A. K. Wallace, Experimental evaluation of a
variable-speed, doubly-fed wind-power generation system. IEEE
Conf. Record Ind. Applications Soc. Annu. Meet., 1993, p. 480–487.
J. Douglas, Renewables on the rise, EPRI J., 16–25, 1991.
J. H. R. Enslin and J. D. Van Wyk, A study of a wind power converter

with microcomputer based maximal power control utilising an
over-synchronous electronic scherbius cascade, Renewable Energy,
2 (6): 551–562, 1992.
P. L. Fraenkel, The development of wind-pumps, Int. J. Solar Energy,

14: 239–255, 1994.
J. I. Herrera, T. W. Reddock, and J. S. Lawler, Harmonics generated

by two variable speed wind generating systems, IEEE Trans. En-
ergy Convers., 3: 267–273, 1988.
R. M. Hilloowala and A. M. Sharaf, Modelling, simulation and analy-

sis of variable speed constant frequency wind energy conversion
scheme using self excited induction generator. IEEE Southeastern
Symp. Syst. Theor., 1991, pp. 33–38.
R. M. Hilloowala and A. M. Sharaf, A utility interactive wind energy

conversion scheme with an asynchronous dc link using a supple-
mentary control loop, IEEE Trans. Energy Convers., 9: 558–563,
1994.
T. A. Lipo, Variable speed generator technology options for wind tur-

bine generation, WEMPEC Summary Publ., 2: 214–220, 1991.
´
H. L. Nakra and B. Dube, Slip power recovery induction generators
for large vertical axis wind turbines, IEEE Trans. Energy Convers.,
3: 1988, pp. 733–737.
P. Novak et al., Modelling and control af variable-speed wind-turbine

drive-system dynamics, IEEE Control Syst., 15 (4): 28–38, 1995.
----------------------- Page 628-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Wind Power Plants

Standard Article

Nicholas Jenkins 1 and Andrew Vaudin2
1UMIST, 88, Manchester, England
2National Wind Power, Bourne End, Bucks, England

Search
reserved.
DOI: 10.1002/047134608X.W3050
❍ Acronym Finder
Abstract
Operating Experience of Wind Power Plants
Wind Power Plant Design
Wind Turbine Electrical Systems ()
Wind Farm Electrical Systems ()
Impact of Wind Power Plants on the Power System

%20Conversion/W3050.htm17.06.2008 16:50:37
----------------------- Page 629-----------------------
WIND
POWER PLANTS
Windmills have been used for many centuries for grinding

corn
and pumping water. Before the invention of the steam
engine
they were of primary economic importance. However,
with
the industrial revolution and then the widespread avail-
ability
of cheap electricity produced from large fossil-fired or
hydropower stations, the use of wind power declined. Pioneer-

ing
work was carried out to investigate the use of wind power
for
the generation of electricity some 50 years ago (1,2) but
this
was discontinued due to the then low cost of fossil fuel.
However, with the first oil shock of the early 1970s there was
a
reawakening of interest in wind power; it is from this time
that
the development of modern electricity generating wind
turbines may be traced.

In
response to the sudden increase in the price of oil, a
number
of countries, including the United States and the
United
Kingdom, initiated major government-funded re-
search
and development programs to develop very large wind
turbines. Prototype turbines of up to 4 MW output and 100 m

in
rotor diameter were constructed. Unfortunately, the diffi-
culties
of building reliable wind turbines of this size were un-
derestimated and, in almost all cases, the prototypes failed to

give
satisfactory long-term service. At the same time, many
private
companies developed small wind turbines, with out-

----------------------- Page 630-----------------------
WIND POWER PLANTS 619
Rated Shut-down rotor

is parked and turned out of the wind in order to reduce
mechanical loading on the rotor and tower.

600
Wind speeds vary continuously with both time and place
and
it is clearly important to ensure that any wind turbine is
)
located in an area of high wind speed if maximum energy
W
k
output is required.
(
r
For assessment of the annual energy that will be obtained
e
w from
a wind turbine it is usual to use hourly mean wind
o
P
speeds. A typical example of a discrete probability distribu-

tion
of hourly mean wind speeds over a year is shown in Fig.
Cut-in 2.
This shows the number of hours in the year when the
hourly mean wind speed was in a 1 m/s bin centered on the
integer wind speed. The continuous form of these probability

0 25
Wind speed (m/s)
distributions is conventionally described using Weibull pa-
rameters (3). The annual energy yield of a turbine can be cal-

Figure 1. Power curve of a 600 kW wind turbine showing: (1) cut-in
culated simply by summing the product of the power curve
wind speed of 5 m/s; (2) rated conditions of 12 m/s and 600 kW; and and
wind speed distribution for each wind speed bin, that is,
(3) shut-down wind speed of 25 m/s.
i=n
E = #H W
i i
i= 1
puts of, say, 50 kW and rotor diameters of 15 to 17 m. These
small wind turbines were installed in large numbers in the where
E is the annual energy yield in kilowatts per hour, H
United States in California and in Denmark where particular is
the hours in wind speed bin i and W is the output power at
financial and administrative arrangements had been made to that
wind speed in kilowatts; n is the cut-out wind speed or
encourage electricity generation by wind power. Although
maximum wind speed of the probability distribution.
many of these small turbines also failed to work properly,
over the next 20 years, the development of wind turbine tech-
nology was such that by the end of 1997 some 7500 MW of
OPERATING EXPERIENCE OF WIND POWER PLANTS
wind turbine capacity had been successfully installed world-
wide, and a number of wind farms with turbines of rotor di- Wind
power plants, or wind farms as they are often called,
ameters up go 45 m and with outputs of 600 kW had given
operate as unmanned, automatic systems. The wind turbines
satisfactory service for several years. The first units of the
generate electricity automatically, provided that the wind
next generation of wind turbines with rotor diameters of up to speed
is within the operating limits for the wind turbines and
60 m and rated outputs of 1.5 MW had also been constructed. that
there are no equipment faults. Manual intervention is
There are two basic architectures of a wind turbine: (1) only
required for routine maintenance and fault correction op-
the horizontal axis or propeller type, and (2) the vertical axis,
erations.
Darrieus or ‘‘egg beater’’ type. Although there are some wind
Modern wind farms all have a supervisory, control and
farms of small vertical axis turbines in California and there data
acquisition (SCADA) system that allows an operator to
is limited continuing interest in the technology, this architec-
monitor the performance of the wind farm remotely and to
ture is not of commercial significance. Modern horizontal axis
identify any faults that require manual intervention. Each
wind turbines usually employ three blades as this is the most wind
turbine operates autonomously under the control of its
aesthetically pleasing arrangement. The rotor is located up- own
computer-based controller and the SCADA system inter-
wind of the tower to reduce torque pulsations and noise due
rogates each controller in turn to monitor its operation and
to the blades passing through the wind shadow of the tower. any
alarm conditions.
The power curve of a wind turbine (Fig. 1) shows the rela- A
typical example of the performance of a wind farm is
tionship between the wind speed some distance upsteam of shown
in Fig. 3. The upper trace shows the wind speed mea-
the rotor and the electrical output from the turbine generator. sured
by an anemometer located on a mast, at the hub height
At wind speeds below cut-in (5 m/s in this case) there is not of
the turbines, and in the middle of the site. However, as the
enough power in the wind to overcome the frictional losses of wind
farm was several kilometers in size and located in hilly
the turbine drive train and the rotor is kept parked. Between
terrain, each turbine would be subject to a different wind
cut-in and rated wind speed the output power rises very rap-
speed. The lower trace shows the output power of the plant
idly. The power available in the wind is proportional to the
measured at the point of connection with the distribution sys-
cube of the wind speed. However, with a fixed-speed wind tur- tem
of the public utility. With the measured wind speed
bine the aerodynamic efficiency of the rotor varies with wind
above, say, 15 m/s, the wind farm produced its rated output
speed and thus the power curve between cut-in and rated of
7.2 MW while when the wind speed dropped below approxi-
wind speed does not follow the cubic relationship precisely. At
mately 5 m/s no power was produced.
rated power and wind speed (600 kW and 12 m/s in this ex-
Due to the intermittent nature of the wind resource it is
ample) the power developed by the rotor is limited either by not
viable for the power output from a wind farm to be sched-
changing the pitch angle of the blades (pitch control) or by uled
or dispatched by the electricity utility responsible for op-
allowing the blades to enter aerodynamic stall (stall control).
erating the generators of the power system. Hence the power
The choice of rated power and wind speed is made by the
output from a wind power plant is exported to the local elec-
designer of the turbine. At shut-down wind speed the turbine
tricity system on a ‘‘take all’’ basis. Appropriate energy pur-
----------------------- Page 631-----------------------
1300
1200 750
1100 700
650
1000
H 600
900
,
r W 550
a
e 800 , 500
y )
r 700 W 450
e k
p 600 ( 400
t
s u
r 500 p 350
u
t
o u 300
H 400 O
250
300
200
200
150
100 100
0 0
1 2 3 4 5 6 7 8 9 10 1112 13 1415 16 1718 19 20 21

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1718 19 20 21
Wind speed (m/s)

Wind speed (m/s)
Wind speed distribution

Power curve
Figure 2. Hourly mean wind speed distribution of a site

and power curve of a turbine each in
1 m/s bins.
chase and generation connection agreements are established visit a

week to the sites was required on average to meet both
between the owner/operator of the wind power plant and the planned
and unplanned maintenance requirements.
electricity utility to facilitate the operating regime at a partic- The
impetus behind the requirements for high reliability
ular location. and
availability is twofold: (1) that the wind turbines operate
Wind turbines are required to be highly reliable and this
automatically, often at remote locations; and (2) that the wind
is usually expressed in terms of availability. Availability is turbines
can only generate within certain wind speeds. It
defined as the fraction of time that a wind turbine is available would not
be practical to have frequent fault correction visits
to generate within a time period, irrespective of the wind con- by
maintenance teams. Also, any reduction in availability
ditions. Modern commercial wind turbines can achieve an- would
compromise the revenue generated by the power plant.
nual availabilities of well above 95%. For example, the annual The high
reliability of modern wind turbines reflects the de-
availabilities averaged over all wind turbines for two typical sign and
development process that has taken place since the
wind farms in the United Kingdom for the years 1994, 1995, 1970s.
High reliability is achieved in part by the design and
and 1996 were above 97%. During these periods one operator
specification of components such as gearboxes, generators and
Mean wind speed

30
25
20
s
/ 15
m
10
0
01/01 02/01 03/01
04/01 05/01 06/01 07/01 08/01
Date
Mean wind farm power output

8
7
6
5
W 4
M
3
2
Figure 3. Mean wind speed and output
1
power of a 7.2 MW wind power plant in
Wales during January. The power plant 0
consisted of 24 of the 300 kW turbines, each 01/01 02/01 03/01
04/01 05/01 06/01 07/01 08/01
of rotor diameter 33 m.
Date
----------------------- Page 632-----------------------
rotors. In addition, site-specific factors are incorporated into In

general, energy production will be maximized for sites
the specification of wind turbines for a wind farm such as the with
strongly directionally biased wind speed distributions by
predicted wind loadings, the likelihood of lightning damage to wind
turbine configurations where the wind turbines are ar-
rotors and towers, and the risk of icing of rotors and meteoro- ranged
closely spaced in strings perpendicular to the prevail-
logical instruments used for the control of the wind turbines. ing wind
direction. Sites with less directionally biased wind
The wind turbines are then designed appropriately for the speed
distributions will have energy production maximized
specific site. using wind
turbines arranged in grid layout with larger spac-
It is common to refer to an annual capacity factor of a ings
between wind turbines.
power plant. This is simply the annual mean output divided When
selecting a site for the construction of a wind farm
by the rated output. Unlike availability, the capacity factor of there are
many factors other than energy production that
a wind power plant depends on the wind conditions at the must be
taken into consideration before the site layout is de-
site. Typical capacity factors for wind farms on good sites in termined.
The final layout will be a compromise between
the United Kingdom are 30–35%. these
factors.
Designated Areas. Most countries will have areas desig-

WIND POWER PLANT DESIGN
nated by regulatory authorities for such reasons as
landscape value and interest of flora and fauna; often it

An assessment of the site wind resource is an essential part
will not be permissible to construct wind farms in these
of the information required by the designer of a wind farm. A
areas. Many wind farms will have areas of high ecologi-
comprehensive knowledge of the wind resource enables the cal
or archaeological interest within the sites where it
developer to predict the annual energy production from a
will not be possible to position wind turbines.
wind farm with various configurations of wind turbines and
Geotechnical Assessment. A detailed geotechnical assess-
site layout. The site wind-resource assessment will also assist
ment of a site will provide information on areas where
the developer in predicting the highest wind speeds (usually it
would be difficult to position wind turbines because of
quoted as 2 or 5 s gust values) at which the wind turbine
unsuitable ground conditions.
must be designed to survive. It is common practice in Europe Visual
Impact. The visual effect of a proposed wind farm
for wind turbines to have their design certified by certifying is
often a key factor in the decision of the relevant plan-
organizations independent of the wind turbine designers and
ning authorities to give consent for the construction of
manufacturers. the
wind farm. Various techniques exist to predict the
The wind resource at a site is essentially assessed by a
visual effects of proposed wind farms, including photo-
measure-correlate-predict technique whereby on-site wind
montages and computer-generated ‘‘zone of visual intru-
speed and direction measurements are collected for at least
sion’’ plots. Wind turbine positions will be moved within
one year, usually at the hub height of the proposed wind tur- a
site to produce an acceptable visual effect.
bines. The measurements are made by erecting meteorologi-
Audible
Noise. Many wind farms in Europe are positioned
cal mast(s) with appropriate instruments and data recording
in
relatively quiet areas with nearby residential dwell-
equipment. The wind speed data at the site are then com-
ings. The generation of noise by wind turbines is an is-

pared with data covering a similar period collected at a
sue
that will normally be addressed by the relevant
nearby national meteorological service station. The results of
planning authorities; planning consents will include

this comparison are applied to the long-term (i.e., longer than
limits on the sound power levels and tonal content of

ten years) wind speed distribution at the meteorological sta-
the
sound at points adjacent to the wind turbines.
tion to produce an effective long-term wind speed distribution
for the wind farm site. Other
Land Usage. Often wind turbines will be installed
on
land that will continue to be farmed or used for other
Predictions for the energy output from a wind farm are
based on the wind-resource assessment for the site estab-
purposes. The positions of wind turbines will be varied
lished using the measure-correlate-predict technique. How- to
accommodate these continuing uses.
ever, the predicted wind speed distribution will be for the
Recently, there has been increasing interest in the siting

points on the site where the meteorological masts are in-
of wind
turbines offshore with three wind farms having been
stalled. A developer then assesses the topographical effect on
installed
in shallow European waters by 1997. Installation
the wind speed distribution at other positions on the site
offshore
offers the possibility of constructing very large wind
away from the meteorological mast positions. Computer mod-
farms with
only limited environmental impact and so is being
eling can be used for this process and the more advanced
considered
seriously in a number of countries. Initially, off-
codes can also include the effects of the reduced wind speeds
shore
installation is likely to be more expensive than on land,
caused by the wakes of other wind turbines. The estimated
but the
additional costs are partly offset by the larger energy
wind speed distributions at each wind turbine site are then
yields
available from the higher wind speeds found over the
combined with the wind turbine power curve as shown in Fig.
2. The power curve for a wind turbine will be supplied by the sea.
wind turbine manufacturer based on type tests, and will vary
according to the configuration of the wind turbine. A devel- WIND
TURBINE ELECTRICAL SYSTEMS (4)
oper will chose a configuration of wind turbine (variable fac-
tors include rotor diameter, tower height and generator ca- The
drive train of a conventional fixed-speed wind turbine
pacity) that produces a power curve maximizing energy consists
essentially of the blades and hub, a main shaft and
production for a particular site. gearbox,
and an induction generator (Fig. 4). The drive train
----------------------- Page 633-----------------------
Main shaft
Hub
Gearbox
Generator
Yaw section
Figure 4. Cross section of the nacelle of a

modern three-bladedstall-regulated wind turbine.
Access is by a ladder through the yaw section. The nacelle
roof opens to allow maintenance staff
to stand.
is located on a bed plate, which forms the base of the nacelle

ever, wind turbines experience cyclic torque pulsations of up
at the top of the tower. The bed plate is mounted on the tower to
20% of the mean torque as the blades rotate in the nonuni-
by a yaw bearing which allows it to be orientated into the form
wind field. It is extremely difficult to build adequate
wind by a yaw drive mechanism. The electrical output of the
damper windings on to the pole faces of synchronous genera-
generator is taken to the base of the tower by flexible pendant tors
to damp these torque variations; thus if synchronous gen-
cables. These are arranged to allow the nacelle to rotate sev-
erators are to be used, then additional mechanical damping
eral turns before needing to be unwound. At the base of the
arrangements (e.g., a fluid coupling) are required in the drive
tower an electrical cabinet houses the power factor correction
train of the turbine. Hence, synchronous generators are now
capacitors, the electrical switchgear, including the soft-start
hardly used on fixed-speed wind turbines as it appears to be
unit, and the wind turbine computer based controller. A more
cost effective to use induction or asynchronous genera-
transformer is located either inside the tower base or adjacent tors
that have an intrinsic damping capability.
to it to transform the generator output to a higher voltage
Induction generators are not commonly used for large-
more appropriate for collection of the power throughout the
scale generation of electricity as they have a number of disad-
wind farm. Figure 5 is a schematic representation of a fixed-
vantages that follow from the fact that it is not possible to
speed wind turbine.
control their excitation. When an induction machine is con-
Some early fixed-speed wind turbines used synchronous
nected to the electrical supply it takes a large transient in-
generators as is common with other prime mover types. How- rush
of magnetizing current as the magnetic circuits are ener-
gized. In modern wind turbines this is controlled by ramping

the
applied voltage using a soft-start device. Usually, this
anti-parallel thyristor voltage regulator is not left in service
continuously but is shorted out by a bypass contactor once

N A the
induction generator is fully fluxed. Because its excitation
cannot be controlled, an induction generator operates on a

fixed
locus of exporting real power (P) while importing reac-
tive
power (Q). This is the well-known circle diagram (5) of
ss
any
induction machine; the generating quadrant for a 600 kW
Tx
unit
is shown in Fig. 6. Once the generator is fully fluxed and
Network
connected to the network by the soft-start device, local power
factor correction capacitors are connected in order to reduce

PFC
the
reactive power demand from the network and to shift the
operating locus as shown.
Fixed-speed wind turbines, as their name implies, operate

Figure 5. Schematic representation of a fixed-speed wind turbine.
(1) N represents the gearbox; (2) A represents the induction genera- at
a rotational speed fixed by the frequency of the electrical
tor; (3) PFC represents the power factor correction capacitors; (4) ss
network apart from the very small change in speed caused by
represents the antiparallel thyristor soft-start unit; and (5) Tx repre- the
slip of the induction generator. However, some manufac-
sents the turbine transformer.
turers prefer a variable speed concept where the rotational
----------------------- Page 634-----------------------
300 WIND
FARM ELECTRICAL SYSTEMS (6)
Without PFC
In
almost all cases the power collection systems within wind
farms
in Europe and North America use underground cable.
This
is to reduce visual impact of the wind farm and to avoid
the
hazards associated with the large cranes required for
r
erection of the turbines operating in the vicinity of overhead
A
lines. The voltage of the main power collection circuits de-
V
k pends
on the size of the wind farm and the voltage at the
With PFC point

of connection to the utility network but is typically in
the
range of 10 to 35 kV.
The generator voltage of even quite large wind turbines

(600
to 800 kW) is usually 690 V, 3-phase, 3 wire. This results
in
rather high currents in the pendant cables but appears to
be
the most cost-effective choice for the wind turbine manu-
600
facturer. However, it is clearly not sensible to transmit such

kW
currents any distance and thus each turbine has its own
Figure 6. Part of the circle diagram of an induction generator. The

transformer. These are conventional, oil-filled distribution
active power is exported and the reactive power is absorbed. The ef-
transformers usually with the 690 V star winding earthed.
fect of the power factor correction capacitors (PFC) in moving the The
electrical energy generated from a wind farm is generally
operating characteristic may be seen. able
to attract a premium price because of its environmental
benefits. Therefore it is often economic to oversize both trans-
formers and cables in order to reduce electrical losses.

speed of the aerodynamic rotor is allowed to vary. Although
The electrical protection of a wind farm is generally
this does allow higher aerodynamic efficiencies, these gains
straightforward. An induction generator cannot contribute
tend to be offset by the increased losses in the power elec-
sustained fault current to a three-phase short-circuit and will
tronic equipment required to convert the varying generator only
make a modest sustained contribution to unbalanced
output frequency to that of the network. The main advantage
faults. Therefore all the current-operated protection on a
of variable speed operation is that the transient torques wind
farm relies on fault current from the utility distribution
caused by the blades cutting the changing wind field can be
network. Each turbine has over/undervoltage and frequency
absorbed by rotor speed changes rather than having to be re-
protection built into its controller; this determines if the elec-
acted by the wind turbine. This becomes particularly signifi-
trical system is outside its normal operating limits and, if so,
cant at large ratings where the damping provided by induc- the
turbine will then be shut down. In the event of a short-
tion generators ceases to be adequate. Thus, with variable
circuit on the wind farm electrical system the current-oper-
speed operation, savings can be made in the mechanical ated
protection isolates the wind farm from the network, the
strength of the turbine to offset the cost of the power electron-
generators will then lose their load and overspeed, and be
ics. Some of the more innovative variable speed designs use
tripped out either by the over-frequency protection or, more
large diameter multipole synchronous generators that do not
probably, by the faster-acting mechanical overspeed protec-
require a gearbox. Figure 7 shows this arrangement with a tion
of the rotor.
voltage source converter providing the change in frequency. A
simple induction generator is not able to supply an iso-
Both the machine side and network side converters use
lated
load because it requires a source of reactive power to
insulated gate bipolar transistor (IGBT) bridges with a
operate. However, if reactive power is supplied by power fac-
pulsewidth modulation (PWM) switching pattern. This allows tor
correction capacitors, then it is possible to operate in the
the power to be fed into the network without low-order har- so-
called ‘‘self-excited’’ condition. Figure 8 shows the usual
monics and at any power factor within the rating of the con-
positive phase-sequence equivalent circuit of an induction
verter.
machine with power factor correction capacitors connected at

its
terminals (7). It may be seen that, if the rotor circuit and
all
resistance is ignored and X is much greater than X , then
m 1
R 1 jX 1 jX 2 R2
S
–jX c jX m R2(1–s)/s
dc Tx
Network
Figure 8. Equivalent circuit of an induction generator fitted with

Figure 7. Schematic representation of a variable speed wind turbine. power
factor correction capacitance. (1) X represents leakage re-
(1) S represents a multipole synchronous generator; (2) DC represents
actance; (2) R represents resistance; (3) Xm represents magnetizing
the dc busbar of the IGBT based voltage source converters, and (3)
reactance; (4) Xc represents capacitive reactance; (5) 1 indicates stator
Tx represents the turbine transformer.
circuit; (6) 2 indicates rotor circuit; and (7) s represents slip.
----------------------- Page 635-----------------------
a parallel resonant condition exists when Table 1.

Savings in Gaseous Emissions from Wind
Generation under Typical United Kingdom Conditions

Xc = Xm
Each Kilowatt-Hour Annual Saving for

1
= ωL
of Electricity Saves a 5-MW Wind Farm
ωC m
Emission (Grams of Oxide) (Tons of Oxide)
Carbon
dioxide 800 10,500–16,100
where C is the capacitance of the power factor correction ca-
Sulfur
dioxide 10 150–240
pacitors, Lm is the magnetizing inductance, and # is the angu- Nitrogen
oxides 3.4 50–80
lar frequency of the islanded system, which, in this simple
analysis, is assumed to be equal to the rotational speed of the
generator. In this self-excited condition the induction ma-
chine is able to sustain a terminal voltage without a connec- common
with the capacities of wind generation now in ser-
tion to the network and, in fact, as the rotational speed of vice.
Therefore, wind power has an impact on the power sys-
the turbine rises, very large transient overvoltages may be tem both
because of the nature of the wind resource but also
experienced. Many utilities are extremely concerned about because
of its location within the power system.
self-excited induction generators supplying parts of their net-

work that have become disconnected or ‘‘islanded’’ from the Impact on
the Generation System
main system under fault conditions. Thus, most wind farms The main
impact of wind power plants on the generation sys-
will have sophisticated protection to detect this loss-of-mains tem is
the saving in fossil fuel. This translates directly into a
or islanding condition. This protection is located at the point reduction
in gaseous emissions and hence in environmental
of connection of the wind farm to the public electricity supply damage
due to acid rain and climate change or global warm-
network and often uses relays either based on detection of the ing. The
extent of this reduction in environmental impact
rate of change of system frequency or on a rapid change in clearly
depends on the plant mix of the generation system
the position of the voltage vectors. Both conditions will occur into
which the wind farms are connected. However, typical
when a wind farm is islanded. savings
for the United Kingdom generating system, which is
Obtaining a low impedance connection to the mass of earth still
heavily reliant on coal-fired plants, are shown in Table 1.
at all points in a large wind farm is difficult. This is particu- The
proportion of generation due to wind power in all global
larly the case where the wind turbines are located on rocky, major
power systems remains low (generally much less than
high-resistivity ground. A low impedance earth or ground con- 5%) and
is unlikely to rise above 10% for many years. At these
nection is required for personnel and animal safety as well as levels of
penetration, the effect of the intermittent nature on
to protect the wind farm plant from the effects of both electri- the wind
is small as any changes in wind generation are
cal faults and of lightning (8). Usual practice is to install some masked
by the much larger changes in conventional central
local earth rods and conductors at each wind turbines site, generator
loading due to variations in customer demand and
but also to lay bare copper earthing conductor with the main plant
breakdowns.
power collection cables and to bond the earths of the entire In
addition to the saving in fuel it is also possible to allo-
wind farm together. This can result in extremely long hori- cate some
capacity credit to wind power plants (9). The capac-
zontal earth electrodes of up to, say, 5 km in length. These ity
credit indicates the capacity of other forms of generation,
conductors then have a significant self-inductance and series which is
no longer required because of the connection of the
resistance which need to be considered in the design and test- wind
power plants. It is usual to calculate this using a substi-
ing of the earthing system. It is also important to be aware of tution
technique based on loss-of-load-probability (LOLP). Al-
the possibility of external earth connections being transferred though
the results of such calculations depend on the type of
into the site via telephone circuits or other services because
generating system considered, a number of studies in differ-
these can lead to large potential differences in the earthing ent
countries indicate that at low penetrations of wind power
system under fault conditions. (less
than 1%) the capacity credit of wind power plants is at
least
equal to the mean annual output while it drops to half
this
value at penetrations of 10%.
IMPACT OF WIND POWER PLANTS ON THE POWER SYSTEM
Studies in a number of countries have also indicated that

at wind
power penetration levels of up to 10% there will be
Modern conventional power systems rely mainly on large cen-
no
significant problems with the stability of conventional gen-
tral power stations generating very large quantities of power
erating
plants.
(up to 2000 MW) often at considerable distances from the load
centers. The power is transported by a high-voltage intercon-
Impact on
the Transmission System
nected transmission grid network before being passed to dis-
tribution networks for supply to the customers. In general, Wind
power plants have only a modest impact on the trans-
wind power plants are much smaller. Typically, sizes range mission
system. There are no significant technical problems
from individual turbines located on the owner’s land (50 to caused
by wind power plants or of power flowing into the
600 kW) up to wind farms of 30 to 50 MW, although there are
transmission system from one part of the distribution system
some examples of larger wind farms in the United States. during
periods of light local customer load. The effect of such
These wind power plants are connected directly to the distri-
generation embedded in the distribution system is generally
bution system and most of the power generated is consumed to reduce
the mean value of load experienced by the transmis-
locally. At times of high wind speed and low customer load, sion
system but also to increase its variance. Therefore, under
power may flow into the transmission network but this is not most
circumstances the electrical losses in the transmission
----------------------- Page 636-----------------------
system will be reduced as less power has to be transported used

to predict its performance at the proposed site. In the
to customers. As the design of interconnected transmission United
Kingdom, this testing is not a mandatory requirement
systems is based not on normal operation but on outage condi- as, in
practice, flicker from wind turbines has not led to com-
tions (i.e., when one or more circuits have failed), there is an
plaints. Some United Kingdom research has indicated that
argument to be made that a reduction in the mean load even
connection of wind farms on very weak rural circuits has ac-
with an intermittent energy source such as wind will lower tually
led to a reduction in voltage flicker on the network as
the requirement for the transmission system plant. However, the
wind turbines increase the fault level.
this remains the subject of debate.
Harmonic distortion of the network voltage by wind power
plants
is only of significant concern if variable speed turbines
Impact on the Distribution System are
used. Early designs of frequency conversion equipment
used
thyristors and this gave rise to significant low-order har-
Wind power plants may have a significant impact on local
monic
currents. Modern converters use IGBTs with a PWM
distribution networks and some care is required to ensure
switching pattern and should not create low-order harmonics.

that the quality of power delivered to other customers of the
However, there will be significant harmonic energy at around

network is not degraded by the connection of wind farms (10).
the
switching frequency of the converter.
The issue of power quality can be considered under a number
In
addition to their effect on power quality, wind power
of headings, for example: (1) steady-state voltage variations;
plants
alter the way the distribution system functions. Rather
(2) dynamic voltage variations or flicker; and (3) harmonic
distortion. than
all the power being supplied from the bulk supply points
of the
transmission network, some power is now provided by
Most public utility distribution circuits were originally de-
generators embedded within the distribution network. This

signed to accept power from the transmission system and to
will
result in changes in the operating costs of the network
supply it to customers. The voltage control systems using
particularly with respect to losses. If the wind power plant

transformer tap-changers and network capacitors were ar-
supplies a local load, then it is likely that the distribution

ranged for this unidirectional power flow. Thus they may not
network
losses will be reduced; if it is necessary to transport
function as intended when the direction of the power flow is
the
power from a wind farm a long distance, then losses may
reversed and hence may need modification.
be
increased. Each case needs to be examined and the calcula-
The general expression for the voltage change in a lightly
tion is
not straightforward given the nonlinear nature of the
loaded radial distribution circuit due to an induction genera-
losses.
A simple approximate calculation technique is to ex-
tor is given by
amine a
number of scenarios of network load and to evaluate
V (PR −XQ ) the
network losses with varying wind farm outputs. An esti-
V = V mate
is then made as to how many hours in the year each
scenario would occur, and hence the impact of the wind farm
P is the real power exported by the generator and Q is the
throughout the year is determined.
reactive power absorbed by it; R and X are the real and imagi- In
a simple radial distribution circuit, wind power plants
nary components of the network source impedance. Thus it do not
provide capacity credit as the circuit is required to pro-
may be seen that the real power export from the wind power vide
reactive power to allow the wind farm to operate. In more
plant acts on the circuit resistance to raise the voltage while
complex, interconnected distribution systems, wind power
the reactive power drawn by the induction generators acts on plants
may make some contribution to the capacity of the
the circuit inductive reactance to lower the network voltage. system.
Therefore, depending on the circuit source impedance and the

circle diagram of the induction generator (Fig. 6), voltage
BIBLIOGRAPHY
changes may be quite modest. Variable speed wind turbines
can be arranged to operate at unity power factor (i.e., with
1. E.
W. Golding, The Generation of Electricity by Wind Power, Lon-
Q # 0); but although this results in minimum electrical losses
don: E. & F. N. Spon Ltd., 1955; reprinted with an additional
a significant voltage rise may occur at the wind farm.
chapter by R. I. Harris, New York: Halstead Press, Division of
Dynamic changes in the voltage supplied to customers can
John Wiley and Sons, Inc, 1976.
cause annoyance and complaints, particularly if incandescent 2. G.
C. Putnam, Power from the Wind, New York: Van Nostrand
light bulbs are used. The human eye and brain are extremely
Reinhold Co., 1948.
sensitive to the changes in the light intensity emitted by an 3. L.
L. Freris (ed.), Wind Energy Conversion System, Englewood
incandescent light bulb supplied with a varying voltage and
Cliffs: Prentice-Hall, 1990.
hence this effect is normally referred to as light ‘‘flicker.’’ The 4. D.
A. Spera (ed.), Wind Turbine Technology, New York: AMSE
effect of light flicker on humans is highly frequency-depen-
Press, 1994.
dent with a maximum sensitivity at around 9 Hz. All utilities 5. G.
McPherson, An Introduction to Electrical Machines and Trans-
will specify a maximum level of flicker which is permitted on
formers, New York: Wiley, 1981.
their networks. 6.
IEEE Standard 1094-1991, Electrical design and operation of
Voltage flicker on distribution systems may be caused by
windfarm generating stations.
wind turbines in several ways: (1) by current transients on 7. J.
Hindmarsh, Electrical Machines and Their Applications, Ox-
connection; (2) by the changes in the wind field at the rotor
ford: Pergamon Press, 1970.
caused by turbulence; and (3) by the cyclic effect of the aero- 8.
IEEE Standard 80-1986, IEEE Guide for safety in ac substation
dynamic rotor cutting the varying wind field. In Germany,
grounding.
every wind turbine design is type-tested to establish its per- 9. D.
J. Millborrow, Wind energy economics, Int. J. Solar Energy,
formance with respect to flicker. This type-test data are then
16: 233–243, 1995.
----------------------- Page 637-----------------------
626 WIND TURBINES
10. N. Jenkins and A. Vaudin, The electrical systems of five wind

farms, IMechE Proc. A, 209: 195–202, August, 1995.
NICHOLAS JENKINS
UMIST
ANDREW VAUDIN
National Wind Power
WIND POWER PLANTS. See WIND POWER.
----------------------- Page 638-----------------------
} { { } }
HOME
ABOUT US
CONTACT US
HELP

Your Librarian
Profile
Wind Turbines

Standard Article

René Spée1 and Shibashis Bhowmik2
1Maxwell Technologies, Corvallis, OR
2Southwest Research Institute, Corvallis, OR

Search
reserved.
DOI: 10.1002/047134608X.W3051
❍ Acronym Finder
Abstract
Basic Wind Turbine Characteristics
Conversion Systems for Wind Power Generation
Wind Turbine Control Systems
Wind Turbine Operation
Outlook and Further Information

%20Conversion/W3051.htm17.06.2008 16:51:03

Air Pollution Control - Arnulfo de Castro (Pada BAB I Saja) - FAJRI FAIZAL AINUL YAQIN

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Air Pollution Control - Arnulfo de Castro (Pada BAB I Saja) - FAJRI FAIZAL AINUL YAQIN

Uploaded by

Copyright:

Available Formats

----------------------- Page 1-----------------------

22• Energy Conversion

22• Energy Conversion

Air Pollution Control Abstract | Full Text: PDF (185K)

Cepstral Analysis of Speech Abstract | Full Text: PDF (137K)

Compressed Air Energy Storage Abstract | Full Text: PDF (616K)

DC Machines Abstract | Full Text: PDF (222K)

Diesel-Electric Generators Abstract | Full Text: PDF (226K)

Diesel-Electric Power Stations Abstract | Full Text: PDF (188K)

Electrochemical Electrodes Abstract | Full Text: PDF (169K)

Electrostatic Precipitators Abstract | Full Text: PDF (279K)

Electrothermal Launchers Abstract | Full Text: PDF (155K)

Engineering Workstations Abstract | Full Text: PDF (93K)

Fuel Cell Power Plants Abstract | Full Text: PDF (145K)

Fusion Reactors Abstract | Full Text: PDF (473K)

Gas Turbine Power Stations Abstract | Full Text: PDF (246K)

Geothermal Power Abstract | Full Text: PDF (250K)

Homopolar and Inductor Machines Abstract | Full Text: PDF (426K)

----------------------- Page 2-----------------------

22• Energy Conversion

HVDC Power Converters Abstract | Full Text: PDF (366K)

Hydraulic Turbines Abstract | Full Text: PDF (468K)

Hydroelectric Generators Abstract | Full Text: PDF (308K)

Hydrogen Energy Systems Abstract | Full Text: PDF (709K)

Hydrothermal Power Systems Abstract | Full Text: PDF (185K)

Magnetohydrodynamic Power Plants Abstract | Full Text: PDF (336K)

Ocean Thermal Energy Conversion Abstract | Full Text: PDF (191K)

Optical Waveguide Transitions Abstract | Full Text: PDF (199K)

Power System Control Abstract | Full Text: PDF (123K)

Primary Cells Abstract | Full Text: PDF (345K)

Quantum Storage Devices Abstract | Full Text: PDF (141K)

Short-Term Load Forecasting Abstract | Full Text: PDF (235K)

Solar Energy Conversion Abstract | Full Text: PDF (258K)

Solar Heating Abstract | Full Text: PDF (162K)

Steam Turbines Abstract | Full Text: PDF (173K)

Synchronous Generators Abstract | Full Text: PDF (204K)

Synchronous Motor Drives Abstract | Full Text: PDF (280K)

Thermal Energy Storage Abstract | Full Text: PDF (107K)

----------------------- Page 3-----------------------

22• Energy Conversion

Thermionic Conversion Abstract | Full Text: PDF (263K)

Thermoelectric Conversion Abstract | Full Text: PDF (190K)

Thermoelectric Power Abstract | Full Text: PDF (727K)

Turbogenerators Abstract | Full Text: PDF (246K)

Waste-to-Energy Power Plants Abstract | Full Text: PDF (158K)

Wind Power Abstract | Full Text: PDF (160K)

Wind Power Plants Abstract | Full Text: PDF (142K)

Wind Turbines Abstract | Full Text: PDF (157K)

----------------------- Page 4-----------------------

Wiley Encyclopedia of Electrical and

● Email this page Enter words or

● Print this page phrases

The sections in this article are

Global Electricity Generation and the Environment

Emissions From Fossil Fuels

Pollution Control Technologies

International Agreements and Country Initiatives

About Wiley InterScience | About Wiley | Privacy | Terms & Conditions

----------------------- Page 5-----------------------

AIR POLLUTION CONTROL 359

Electricity has a large share in the energy system. It con-

conversion, are site-speciﬁc and need the electrical transmis-

respectively, are subject to site constraints.