You are on page 1of 11

PROGRESS IN PHOTOVOLTAICS: RESEARCH AND APPLICATIONS

Prog. Photovolt: Res. Appl. 2002; 10:533–543 (DOI: 10.1002/pip.442)

Applications Design of Tracking


Photovoltaic Systems
with a Single Vertical Axis
E. Lorenzo1,*,y , M. Pérez2, A. Ezpeleta3 and J. Acedo4
1
Instituto de Energı́a Solar, ETSI Telecomunicación, Ciudad Universitaria, s/n, 28040, Madrid, Spain
2
Alternativas Energéticas Solares, Pol Industrial La Nava, 1, 31300, Tafalla, Naavarrsa, Spain
3
Energı́a Hidroeléctrica Navarra, C/Yanguas y Miranda, 1, 31002, Pamplona, Spain
4
Ingeteam SA, C/Pintor Maeztu, 2, 31008, Pamplona, Spain

Solar tracking is used in large grid-connected photovoltaic plants to maximise solar


radiation collection and, hence, to reduce the cost of delivered electricity. In particu-
lar, single vertical axis tracking, also called azimuth tracking, allows for energy gains
up to 40%, compared with optimally tilted fully static arrays. This paper examines the
theoretical aspects associated with the design of azimuth tracking, taking into
account shadowing between different trackers and back-tracking features. Then,
the practical design of the trackers installed at the 14 MW Tudela PV plant is pre-
sented and discussed. Finally, this tracking alternative is compared with the more
conventional fully stationary approach. Copyright # 2002 John Wiley & Sons, Ltd.

INTRODUCTION

S
olar tracking remains an interesting option for PV generation, especially when medium/large grid-
connected PV plants are concerned. High reliability and low maintenance requirements have been
demonstrated in several practical projects.1–2 The Toledo PV Plant, for example, where a 1000-m2
tracking system has been in routine operation since July 1994, with 100% availability. This can be interpreted
as a mean time between failures below 60 000 h, which speaks for itself.
Single, horizontal, North–South-oriented axis structures associated with flat-plate modules represent by far
the most extended tracking solution in current PV plants.2–4 Because of their inherent lack of shadowing in the
North–South direction, single tracking devices can drive large surfaces and, owing to the horizontal axis posi-
tion, associated wind loads tend to be relatively low. It involves a particularly simple and robust mechanical
construction, which is a major advantage of this type of tracking. Coming back to the Toledo PV plant, its
tracker is formed of four 250-m2 surfaces, each powered by a single 025-HP standard AC motor. North–South
oriented horizontal single axis tracking also plays a major role in solar thermal electric technology. In particular,
it has been selected for the famous LUZ-developed power plants,5 totalling 23  106 m2 aperture area, which
began operation in southern California in 1984.
However, the horizontal axis position limits energy collection by the tracking surface. This depends on the
solar climate and latitude site, , and can be quantified by comparison with the energy collected by an ideal two-
axis tracking, which represents the largest solar radiation potential for a particular location. Table I presents

* Correspondence to: E. Lorenzo, Instituto de Energı́a Solar, ETSI Telecomunicación, Ciudad Universitaria, s/n, 28040, Madrid, Spain.
y
E-mail: Lorenzo@ies-def.upm.es
Contract/grant sponsor: EC; contract/grant number: NNE5/1999/547.

Published online 17 September 2002 Received 31 December 2001


Copyright # 2002 John Wiley & Sons, Ltd. Revised 16 March 2002
534 E. LORENZO ET AL.

Table I. Yearly solar energy collection in several locations


Location ( ) Clearness indexa Ga0 b, (kW h m2) Two-axis/Ga0 c Horizontal/Ga0 d Azimuth/Ga0 e Static/Ga0 f

Medellin 62 0469 1708 122 118 108 1


Morelia 197 0410 1470 120 114 108 1
Cayro 306 0637 2040 154 140 144 104
El Paso 315 0689 2190 163 146 150 106
Tudela 421 0596 1680 163 143 156 110
Freiburg 48 0428 1100 142 126 136 107
St Petersburg 595 0453 942 172 143 164 118
Ice-Island 80 0580 900 271 194 25 167
a
Ratio (global horizontal/extraterrestrial) irradiation.
b
Global horizontal irradiation.
c
Ratio (two axis tracking/horizontal) irradiation.
d
Ratio (one North–South horizontal axis tracking/horizontal) irradiation.
e
Ratio (one azimuth axis tracking/horizontal) irradiation.
f
Ratio (optimally tilted static/horizontal) irradiation.

some examples calculated on a yearly basis. This table has been compiled from solar radiation data contained in
the H-World database,6 except for the case of Tudela, where radiation data has been provided by the Regional
Meteorological Services of Navarra. Although it does not affect the central message of this paper; it is worth
mentioning that significant data differences are found for the same location, when consulting different sources
of solar radiation data.
The major motivation for the development of other one-axis tracking alternatives is to overcome this limita-
tion, while keeping the mechanics fairly simple. This is the case for the 14 MW PV plant installed in Tudela
(Spain), which is formed of 400 azimuth trackers. This type of tracker rotates around its vertical axis, in such a
way that the azimuth of the receiver surface is always the same as the Sun’s azimuth, while its tilt angle remains
constant. Calculation of the solar radiation collected by ideal tracking is rather straightforward, the basic rules
are well defined in classical books on solar radiation.7 Some results in Table I show that azimuth tracking repre-
sents an energy collection increase, in comparison with horizontal axis tracking, of about 10%. It can also be
seen that the advantages of tracking energy increase for both latitude and clearness index. In the case of Tudela,
the energy collected by an ideal azimuth tracker is about 40% higher than that corresponding to an optimally
tilted static surface.
When several trackers are arranged together, mutual shadows give rise to a design optimisation problem. The
lower the spacing between adjacent trackers, the lower is the gross land occupation and, therefore, the lower the
land-area-related costs (land, civil works, wiring, etc.). On the other hand, the larger the impact of shadowing,
the greater is the detrimental effect on the electricity generation of the PV plant. This paper first examines the
theoretical aspects of this problem, by relating the collection of energy to the relevant design parameters,
namely, the tilt angle  and the aspect relation (length/width) b of the single tracked surfaces, and the spacing
between adjacent trackers in North–South and East–West directions, lNS and lEW, respectively. The so-called
Back-tracking8 features are also considered as a mean of reducing shadowing impact. Then, a cost-optimisation
exercise is performed for the particular case of the recently installed 14 MW Tudela PV plant, and some general
conclusions are outlined.

GEOMETRY OF SHADOWING

Let us consider a set of azimuth trackers arranged as shown in Figure 1. Note that the ground cover ratio (GCR),
defined as the ratio of total PV module area to total gross land area, is given by:

b
GCR ¼ ð1Þ
lNS lEW

Copyright # 2002 John Wiley & Sons, Ltd. Prog. Photovolt: Res. Appl. 2002; 10:533–543
SINGLE-AXIS TRACKING PV SYSTEMS 535

Figure 1. Design parameters of a tracking field: (a) tilt angle ; (b) aspect relation b; (c) spacing between
adjacent trackers in North–South and East–West directions, lNS and lEW

Figure 2. (a) Solar coordinates: azimuth S, and elevation  S; (b) incidence angle of the beam radiation S

Figure 2 shows the solar angular coordinates: solar azimuth S, and solar altitude  S; and the incidence angle of
the beam radiation on the tracking surface S. Straightforward geometrical considerations lead to:


S ¼  S   ð2Þ
2

East–West shadowing
Depending on the Sun’s position, partial shadowing between two adjacent trackers may occur, mainly in the
early morning and late afternoon. Figure 3 shows a case of shadowing in the East–West direction. The horizon-
tal projections of two adjacent tracked surfaces and their corresponding mutual shadowing are presented in
Figure 3(a), and additional geometrical details are given in Figure 3(b). Note that shadowing occurrence
requires two simultaneous conditions. First, the shadow should point to the adjacent tracker and, second, its
length should be large enough to reach it. This can be described by introducing a parameter FSEW, being 1 when
ideal tracking leads to shadowing occurrence and 0 otherwise. Thus:

ðLEW cos SÞ <1 and ðLEW sin SÞ < s ¼ s1 þ s2 ) FSEW ¼ 1 ð3Þ

Copyright # 2002 John Wiley & Sons, Ltd. Prog. Photovolt: Res. Appl. 2002; 10:533–543
536 E. LORENZO ET AL.

Figure 3. East–West shadowing in ideal tracking. Horizontal projection of: (a) trackers and shadows; (b) a meridian plane

where

s1 ¼ b cos  and s2 ¼ b sin  cot S ð4Þ

Because of the finite dimensions of the PV field, East–West shadowing unavoidably occurs at in some times
during the year ( S ! =2 and S ! 0, close to sunrise at the equinoxes). When shadowing occurs, it can be
avoided by moving the surface’s azimuth angle away from its ideal value, just enough to get the shadow border-
line to pass through the corner of the adjacent surface (Figure 4). The new surface’s azimuth, 0 , is given by:

0 ^C
¼ S  FSEW A ð5Þ

^ C is the azimuth correction angle, which can be found by analysing the triangles SS1S2 and T1S2T2.
where A
From the first:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s ¼ s2 þ s2 þ 2s1 s2 cos A ^C ð6Þ
1 2

0
Figure 4. Surface azimuth necessary to avoid East–West shadowing

Copyright # 2002 John Wiley & Sons, Ltd. Prog. Photovolt: Res. Appl. 2002; 10:533–543
SINGLE-AXIS TRACKING PV SYSTEMS 537

s2
sin ^S2 ¼ ^C
sin A ð7Þ
s

^  ^ C  ^S2
1¼  S þA ð8Þ
2
and from the second:
b 2 ¼  þ ^S2
T ð9Þ
2

b 2 ¼ lEW sin ^1
sin T ð10Þ

^ C . It is worth noting that:


The solution of the system formed by Equations (6–10) leads to the value of A

 ^ C ¼ cos1 ðlEW cos


¼ )A SÞ ð11Þ
2

which represents a maximum limit for the A ^ C value for the general case. Because such a maximum always
avoids shadowing, it can be used as a rough approximation to the general case.
This non-ideal tracking mode is called back-tracking, and has sometimes been implemented in horizontal
single-axis tracking.8–9 Figure 5 shows the evolution of the surface’s azimuth at an equinox. One can note that
the surfaces begin each day facing South, and gradually rotate towards the East, to avoid shadowing, until some
time in the morning, when they reverse direction and rotate West to minimise the beam incidence angle. Then,
there is no difference between ideal and back-tracking strategies, until the low afternoon Sun’s elevation and
ideal tracking strategy would produce shadowing again. Once more the back-tracking reverses direction and
gradually returns to face the South.

North–South shadowing
Figure 6 shows the horizontal projection and a tracked surface with its corresponding shadow for a general case.
North–South shadowing requires, first, that the shadow falls behind the front line of the rear trackers row, and,
second that shadow points toward some tracker, and not towards the free space among them. A conservative
approach (reasonable for low lEW values) consists of assuming that shadowing occurs just when the first

Figure 5. Surface azimuth angle plotted against solar time

Copyright # 2002 John Wiley & Sons, Ltd. Prog. Photovolt: Res. Appl. 2002; 10:533–543
538 E. LORENZO ET AL.

Figure 6. North–South shadowing; horizontal projection of a tracker and its shadow

condition is fulfilled. This can be described by defining a parameter FSNF, being 1 when North–South shadow-
ing occurs and 0 otherwise. Thus:

0
lNS < s sin ^1 þ sin ) FSNS ¼ 1 ð12Þ

ENERGY PRODUCED BY THE PV PLANT

For simplicity, let us consider that all the solar cells of a single-tracked surface are associated in series, that the
cell size is negligible compared with the surface size, and that the PV field is large enough to neglect border
effects. Note that in the hypothetical situation of only beam radiation, the simple shadowing event would annul
the output of the whole PV array. Thus, neglecting the albedo, the electrical power from the PV field is given by:
 0 0

P ¼ A FSNS Bð; Þ þ Dð; Þ ð13Þ

where  and A are the global efficiency of the PV plant, and Bð; 0 Þ and Dð; 0 Þ are the beam and diffuse
irradiances incident on the tracked surfaces. The annual electrical energy produced by the PV plant, EPV, is
given by the integral of P over the whole year, and is easily calculated from widely available horizontal solar
radiation data, following classical procedures to translate from horizontal to inclined surfaces.10
In particular, we have developed an hourly-based software application that uses as input the twelve monthly
mean values of the horizontal global daily irradiation. It considers the diffuse global correlation proposed by
Collares Pereira and Rabl,11 the Liu and Jordan method12 to estimate hourly irradiation values from the daily
irradiation, and the model proposed by Hay and McKay13 to estimate the circumsolar and isotropic components
of the diffuse solar radiation. Then, the circumsolar radiation has been treated as beam radiation for considera-
tion of shadowing. On the other hand, we have estimated the global efficiency of the PV plant by considering
separately each phenomenon that affects its behaviour, as indicated by the following expression:

 ¼  TC AL INV ð14Þ

where * is the efficiency of the PV array under standard test conditions (STC), TC takes into account the
dependence of the efficiency on the operating temperature of the solar cells,14 AL takes into account the angular
reflection losses at the PV module surface,15 and INV considers the DC-to-AC conversion losses.16

Copyright # 2002 John Wiley & Sons, Ltd. Prog. Photovolt: Res. Appl. 2002; 10:533–543
SINGLE-AXIS TRACKING PV SYSTEMS 539

Figure 7. Relative yearly energy production of ideal one-axis trackers plotted against tilt angle for several locations

In this way, a value for annual energy production can be calculated for each PV field configuration, i.e., for
each set of , b, lEW and lNS values, given the characteristics of the PV modules and the inverter composing the
PV plant. Obviously, the largest possible energy yield corresponds to the case of only one tracked surface,
which excludes mutual shadowing and allows for fully ideal tracking. Just for this case, Figure 7 shows the
relative variation of the energy yield, as a function of the angle of inclination of the tracked surface, for several
locations. It can be seen that, irrespective of the site, a maximum occurs for an inclination close to the latitude
plus 10 . It is very interesting that the sensitivity of the annual capture of energy for the inclination angle is very
low. A value of approximately 04% loss for each degree of deviation from the optimum value is roughly indi-
cative of the situation. A descriptive empirical expression is:
 
 EPV ðÞ 
  2
E ð Þ ¼ a0 þ a1 ð  opt Þ þ a2 ð  opt Þ ð15Þ
PV opt l¼1

where  is the site latitude, opt ¼  þ 10 ; a0 ¼ 1; a1 ¼ 032  102 , and a2 ¼ 104 . It is interesting to note
that, because the product A is scarcely related to the inclination angle, such functions are essentially
independent of the PV module and inverter type. Hence, they describe not only the energy produced by the
PV field, but also the energy collected by the PV array.

OPTIMAL DESIGN

A rough approximation to the produced energy unit cost CE can be made by assuming that it is a linear function
of the ratio between the total investment cost and the yearly energy yield. Furthermore, the total investment cost
can be divided into a term related to the gross land area (wiring, civil works, fencing, land, etc.) and another
term independent of it (PV modules, tracking structures, inverters, etc.). Thus:
a
CE ¼ ð p1 p þ lEW lNS NT p2 Þ ð16Þ
EPV
where p1 is the cost per unit of peak power of the PV system excluding land, p* is the STC power of the PV
array, p2 is the land area related cost per unit area, NT is the number of trackers composing the PV field, and a is
a parameter accounting for the evolution of the economy (discount rate, inflation, etc.). The cheapest concei-
vable energy cost corresponds to the hypothetical case of only an optimally tilted ideal tracker and no land
occupation. It is worth using this case as a reference. Thus:

REF
 
CE EPV 1 p2
¼ 1þ ð17Þ
CEREF EPV GCR  GSTC p1

Copyright # 2002 John Wiley & Sons, Ltd. Prog. Photovolt: Res. Appl. 2002; 10:533–543
540 E. LORENZO ET AL.

Table II. Input data for the design of the Tudela PV plant

Latitude( ) 421
Reference year, m ¼ 1    12
Global horizontal daily irradiation (W h m2 1826, 3184, 4637, 5777, 6611, 7183, 7494, 6489, 5000, 3247, 2118, 1582
Maximum daily temperature ( C) 69, 74, 126, 154, 194, 246, 261, 280, 210, 137, 95, 91
Minimum daily temperature ( C) 07, 10, 36, 54, 91, 127, 130, 146, 112, 70, 33, 32
PV module dimensions (mm) 1222  538
PV array STC efficiency (%) 133
Inverter efficiency parameters16 k0 ¼ 001, k1 ¼ 0025, k2 ¼ 005
Cost parameters p1 ¼ 66 s/Wp; p2 ¼ 6 s/m2

This ratio represents the relative unit energy cost associated with a particular PV field configuration, and is
particularly well suited to judge the merit of its design.

The Tudela PV plant


Table II summarises the required input data for the optimisation design of the Tudela PV plant. The most
obvious criterion is to seek to lower the unit energy cost, i.e., to select the configuration leading to the minimum
value of the ratio described by Equation (17). A simple approach consists of the successive optimisation of each
configuration parameter. In this way, an inclination equal to opt is first selected. Figure 8 shows the cost ratio as
a function of b and lEW, for a given lNS. It is noticeable that the lowest b gives the lowest the energy cost. How-
ever, there are practical limits to b. On the one hand, values lower than b ’ 05 imply lengthy pedestal structures,
which are inherently unstable in strong winds. On the other hand, all the PV modules of a single series string
should be installed in the same tracker to avoid mismatching losses.
In our case, the pre-established DC nominal voltage of the PV field (450 V) requires each string to be formed
by 36 PV modules, for which 8 different arrangements can be imagined: 1  36, 2  18, 3  12, 4  9, 6  6,
9  4, 12  3 and 18  2. Moreover, each arrangement can adopt two different physical dispositions, depending
on the PV module side chosen as the length. Hence, 16 values for b, ranging from 0012 to 817, are possible.
We have finally selected a 3  12 arrangement leading to b ¼ 05675, which is shown in Figure 9.
Coming back to Figure 8, it can be seen that the optimum value of lEW is about 23. Now, Figure 10 shows the
REF
variations of the cost ratio as a function of lNS, for  ¼ opt ; b ¼ 05676 and lEW ¼ 23. The ratio (EPV/EPV ) is
also shown. Again, an optimum lNS value around 23 is observed. Hence, the strict optimum design corresponds
to a GCR ’ 01, which is much larger than usual values in common PV fields, and has adverse aesthetic
implications (trackers too sparsely placed), which can not be neglected in highly visible demonstration

Figure 8. Relative energy production (dashed lines) and cost ratio (dotted lines) plotted against East–West spacing; tracking
aspect relation b is used as parameter

Copyright # 2002 John Wiley & Sons, Ltd. Prog. Photovolt: Res. Appl. 2002; 10:533–543
SINGLE-AXIS TRACKING PV SYSTEMS 541

Figure 9. Selected tracker for the Tudela PV plant: (a) front; (b) back. The driven mechanism can be observed

Figure 10. Relative energy production (dashed line) and cost ratio (dotted line) plotted against North–South spacing

projects, such as the present case. Because of this, we also have analysed other less land-consuming alternatives.
Fortunately, they are largely favoured by the low sensitivity of the unit energy cost to the GCR. The same is
true for the energy yield. For example, the figures of merit for the optimal case are (CE/CEREF ) ¼ 1166 and
REF
(EPV/EPV ) ¼ 0923, while for lEW ¼ 1547 and lNS ¼ 2012, the values are (CE/CEREF ) ¼ 12436 and
REF
(EPV/EPV ) ¼ 085 and GCR ¼ 0182. In other words, selecting this case implies a 67% energy cost increase,
but a 41% reduction of the total occupied land. We have finally selected this case because it adapts particularly
well to the available terrain. Figure 11 shows a general view of the plant. Table III presents the yearly perfor-
mance parameters for this case, as estimated by the model. The corresponding performance ratio,17 defined as
the ratio between the final yield and the on-plane irradiation, is 082, which is rather optimistic when comparing
it with current experimental values. The reason lies in the different losses not considered by the model: wiring,
dirt, mismatching, supplied PV power below nominal values, etc. Together, they will probably amount to an
11% energy production decrease, leading to a performance ratio of about 073.

DISCUSSION
We fully understand the reader’s surprise when realising that, after the two rather tedious tasks of developing a
full behaviour model for the PV plant, and performing a cost optimisation exercise, our final design differs from

Copyright # 2002 John Wiley & Sons, Ltd. Prog. Photovolt: Res. Appl. 2002; 10:533–543
542 E. LORENZO ET AL.

Figure 11. General view of the Tudela PV field

Table III. Estimated annual energy performance of the Tudela PV plant


Model estimates

Horizontal irradiation (kW h m2) 1680


On-plane irradiation (kW h m2) 2366
Effective on-plane irradiation (kW h m2)* 2235
Array DC energy yield (kW h kWp1) 2100
PV plant AC final yield (kW h kWp1) 1941
Performance ratio 082
More realistic estimates
Other losses (%) 11
PV plant final yield (kW h kWp1) 1728
Performance ratio 073

*Considering shadowing and reflection losses.

Table IV. Optimal GCR versus different economic scenario


p2/p1* Optimal GCR

09 010
13 012
45 018

*In units of (currency/m2)/(currency/Wp).

the optimal one, and attaches importance to aesthetic considerations. We also understand the tendency to
believe that the usefulness of the model presented is restricted to clarifying the geometry of back-tracking
for azimuth tracking, and to estimating the energy performance of the PV field. However, the model can be
used to study other relevant problems, such as the sensitivity of the design to economic factors, and the
comparison of tracking with conventional static alternatives.
For the first point, and because PV modules and tracking structures are expected to become cheaper in
the future, we have analysed the variation of the optimal GCR with respect to the ratio p2/p1. Table IV shows
the corresponding results. Roughly, a five-fold increase in cost ratio is translated into a two-fold decrease of the
required area. It is worth mentioning that the particular circumstances of the Tudela PV project have led to very
low p2 and high p1, i.e., to a rather low p2/p1ratio. In our opinion, p2/p1 close to 13 is a more representative value
for possible commercial PV plants.
For the second point, we have performed a similar design optimisation exercise for a hypothetical PV field
composed of fully static conventional support structures, and also operating in Tudela. For the same values of p1
and p2, the result is an optimal GCR ¼ 038, but the corresponding energy cost ratio (CE =CEREF ) is 148.
Compared with the actual tracking plant, the static PV plant would require only about 60% of the land, but
the energy cost would be 28% higher.

Copyright # 2002 John Wiley & Sons, Ltd. Prog. Photovolt: Res. Appl. 2002; 10:533–543
SINGLE-AXIS TRACKING PV SYSTEMS 543

CONCLUSIONS
This paper has discussed the theoretical aspects of azimuth tracking, including shadowing and back-tracking.
From them, a model capable of analysing the relation between the configuration of tracking PV plants (tilt
angle, aspect relation and spacing) and their energy performance and energy cost have been also presented.
This model has been used on the design of the 14 MW PV plant, recently installed in Tudela (Navarra,
Spain). The finally selected value of the GCR is 0182, and the expected yearly energy yield is about
1700 kW h kWp1. The comparison with conventional static arrangements shows that azimuth tracking land
requirements are about 40% larger than conventional static arrangements, but the corresponding energy cost
can be significantly reduced, providing the tracking structure’s cost is kept close to the static structures.

Acknowledgements
This work has been supported by the European Commission, under contract NNE5/1999/547. The authors
would like to acknowledge the helpful collaboration of all the persons involved in this project. The Diputación
Foral de Navarra has been kind in providing the radiation data.

REFERENCES
1. Jennings C, Farmer B, Townsend T, Hutchinson P, Gough J, Shipman D. PVUSA—The first decade of experience.
Proceedings of the 25th IEEE PVSC, Washington, DC, 1996; 1513–1516.
2. Lorenzo E, Maquedano C. Operational results of the 100 kWp tracking PV plant at Toledo-PV project. Proceedings of
the 13th European PV Solar Energy Conference, Nice, 1995; 2433–2436.
3. Stern MJ, West RT. Development of a low cost integrated 15 kW A.C. solar tracking sub-array or grid connected PV
power system applications. Proceedings of the 25th IEEE PVSC, Washington, DC, 1996; 1369–1372.
4. Illiceto A, Previ A, et al. Progress report on ENEL’s 33 MWp PV plant. Proceedings of the 12th PV Solar Energy
Conference, Amsterdam, 1994; 1167–1170.
5. De Laquil III P, Kearney D, Geyer M, Diver R. Solar-thermal electric technology. In Renewable Energy: Sources from
Fuels and Electricity. Island Press: Washington, 1993; 213–296.
6. H-World database. Mean Values of Solar Irradiation on Horizontal Surface. Distributed by CENSOLAR, Sevilla, 1993.
7. Duffie JA, Beckman WA. Solar Engineering of Thermal Processes, 2nd edn. Wiley: New York, 1991.
8. Panico D, Garvison P, Wenger H, Shugar D. Backtracking: a novel strategy for tracking PV systems. Proceedings of the
22nd IEEE Photovoltaic Specialist Conference, Las Vegas, 1991; 668–673.
9. Lorenzo E, Macagnan M. Considerations in the design of a one-axis tracking photovoltaic system. Progress in
Photovoltaics: Research and Applications 1994; 2: 45–55.
10. Iqbal M. An Introduction to Solar Radiation. Academic: Ontario, 1983.
11. Collares Pereira M, Rabl. The average distribution of the solar radiation—correlations between diffuse and hemi-
spherical and between daily and hourly insolation values. Solar Energy 1979; 22: 155–164.
12. Liu BYH, Jordan RC. The interrelationship and characteristic distribution of direct, diffuse and total solar radiation.
Solar Energy 1960; 4: 1–19.
13. Hay JE, McKay DC. Estimating solar irradiance on inclined surfaces. Solar Energy 1979; 23: 111–114.
14. Green MA. Solar Cells. Prentice-Hall: Kensington, 1982.
15. Martin N, Ruiz JM. Calculation of the PV modules angular losses under field conditions by means of an analytical
model. Solar Energy Materials and Solar Cells 2001; 70: 25–38.
16. Schmid J, Schmidt H. Inverters for photovoltaic systems. Proceedings of the 5th Contractor’s Meeting of the
Photovoltaic Demonstration Projects, Ispra, Italy, 1991; 122–132.
17. Guidelines for the assessment of PV plants. Document B: analysis and presentation of monitoring data, JRC of the
European Union. Report EUR 16339 EN, Ispra, Italy, 1995.

Copyright # 2002 John Wiley & Sons, Ltd. Prog. Photovolt: Res. Appl. 2002; 10:533–543

You might also like