318 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 26, NO.

1, MARCH 2011
Investigation of Methods for Reduction of Power
Fluctuations Generated From Large
Grid-Connected Photovoltaic Systems
Walid A. Omran, M. Kazerani, Senior Member, IEEE, and M. M. A. Salama, Fellow, IEEE
AbstractPhotovoltaic (PV) systems are presently allowed to in-
ject into the grid all the power they can generate. However, in the
near future, utilities are expected to impose additional regulations
and restrictions on the power being injected by large centralized
PV systems because of their possible adverse impacts. One of the
main issues associated with large PV systems is the uctuation of
their output power. These uctuations can negatively impact the
performance of the electric networks to which these systems are
connected, especially if the penetration levels of these systems are
high. Moreover, the uctuations in the power of PV systems make
it difcult to predict their output, and thus, to consider them when
scheduling the generating units in the network. The main objec-
tive of this paper is to investigate some methods that can be used
to reduce the uctuations in the power generated from a large
customer-owned PV system, in the order of megawatts. This pa-
per focuses on three methods: 1) the use of battery storage systems;
2) the use of dump loads; and 3) curtailment of the generated power
by operating the power-conditioning unit of the PV system below
the maximum power point. The emphasis in the analysis presented
in this paper is on investigating the impacts of implementing these
methods onthe economical benets that the PVsystemowner gains.
To estimate the maximum revenues gained by the system owner,
an linear programming optimization problem is formulated and
solved. Moreover, the effect of varying different parameters of the
problem is investigated through sensitivity analysis.
Index TermsBattery storage system, dump load, economic,
grid connected, power curtailment, photovoltaic (PV) systems,
storage.
NOMENCLATURE
C
E
Capital cost related to the battery energy capacity
($/kWh).
C
P
Capital cost related to the battery power capacity
($/kW).
d Average annual market discount rate.
E
B
Energy discharged from the battery (kWh).
E
Bmax
Maximum energy storage capacity of the battery
(kWh).
M Number of data points in each pattern.
N Number of photovoltaic (PV) power patterns in
each year.
Manuscript received November 22, 2009; revised February 22, 2010, May 26,
2010, and June 29, 2010; accepted July 21, 2010. Date of publication December
30, 2010; date of current version February 18, 2011. Paper no. TEC-00495-
2009.
W. A. Oman and M. Kazerani are with the Department of Electrical and Com-
puter Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada
(e-mail: waomran@uwaterloo.ca; mkazeran@uwaterloo.ca).
M. M. A. Salama is with the Department of Electrical and Computer En-
gineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada, and
also a Visiting Professor at King Saud University, Saudi Arabia (e-mail:
m.salama@uwaterloo.ca).
Digital Object Identier 10.1109/TEC.2010.2062515
N
B
Number of times the battery will be replaced dur-
ing period T (N
B
= T/T
B
).
O
E
Operating and maintenance costs related to the
battery energy storage capacity ($/kWh).
O
P
Operating and maintenance costs related to the
battery power capability ($/kW).
P
B
Power generated from the battery (kW), +ve:
discharging and ve: charging.
P
Bmax
Maximum power capability of the battery (kW).
P
C
Power curtailed from the PV system (kW).
P
D
Power dumped from the PV system (kW).
P
Dmax
Power rating of the dump load (kW).
P
G
Power injected into the grid (kW).
P
Gmax
Maximum power injected into the grid (kW).
P
PV
Power generated from the PV panels (kW).
R Revenues gained from the system over period
T ($).
S
E
Salvage value related to the battery energy stor-
age capacity ($/kWh).
S
P
Salvage value related to the battery power capa-
bility ($/kW).
SOC
min
Minimum allowable state of charge.
T Number of years studied.
T
B
Lifetime of the battery (years).
x Lower limit for power uctuations (kW).
y Upper limit for power uctuations (kW).
Price of energy sold from the PV system
($/kWh).
Present worth related to the battery power capa-
bility ($/kW).
Present worthcorrespondingtothe batteryenergy
storage capacity ($/kWh).
Present worth corresponding the dump-load
power capacity ($/kW).
R% Percentage change in revenues.
t Minute resolution of data points (min).

PB
Power efciency of the battery charger.

ch
Energy efciency of the battery.
I. INTRODUCTION
G
RID-CONNECTED PV systems are currently being
widely installed in many of the developed countries, such
as Japan, Spain, and U.S. [1]. In addition to their environmental
benets, PV systems have a number of technical and econom-
ical benets. They can be operated to decrease the losses and
0885-8969/$26.00 2010 IEEE
OMRAN et al.: INVESTIGATION OF METHODS FOR REDUCTION OF POWER FLUCTUATIONS GENERATED 319
improve the voltage prole of the feeder to which they are con-
nected [2], [3]. Moreover, the owners of PV systems usually
receive incentives from utilities by setting a high selling price
for the energy generated from these systems. For example, the
Ontario Power Authority (OPA) in Canada offered to pay 42
c
/
/kWh for the power generated from PV systems under the
Ontarios Standard Offer Program that was launched in 2006.
In 2009, the OPA introduced a new program, the Feed in Tar-
iff, which provided customers with higher incentive prices for
the kilowatt-hour generated from PV systems [4]. Recently, the
OPA has announced that 176 PV projects with a total power of
651 MW were awarded contracts under the new program [5],
many of which are in the distribution networks with ratings
in the order of tens of megawatts. In addition, Samsung has
recently announced its plan to build a 100-MW PV plant in On-
tario, which will be the rst phase of a number of plants with a
combined output power of 500 MW [6], [7].
One of the main characteristics of PVsystems is the high vari-
ability of their output power. This variability stems fromthe fact
that these systems are static, and thus, any instantaneous change
in the irradiance reaching the PVarrays leads to a corresponding
change in their output power. The time frame for the short-term
uctuations in irradiance is in the order of seconds to few min-
utes [8], [9]; thus, some studies have considered the uctuations
in the PV power to be within the same time frame [10], [11].
However, other studies have recommended the use of 10-min
irradiance data when studying the power uctuations generated
from PV systems [12], [13]. This is especially suitable for sys-
tems with ratings in the order of tens of megawatts that extend
over a large land area, such as the 10-MW PV plant in Pocking,
Germany [14]. Arecent report published by the North American
Electric Reliability Corporation showed that the output power
of existing large PV systems, with ratings in the order of tens
of megawatts, can change by 70% in a ve- to ten-min time
frame [15]. It should be noted that for a number of small systems
that are distributed over a large land area, the resulting combined
uctuations are much less due to the smoothing effect [16], [17].
Fig. 1 shows an example of the PV power uctuations due to
the uctuations in irradiance. The power displayed in the gure
is based on actual irradiance and temperature measurements at
latitude 39.74

N and longitude 105.18

W with a 10-min time

resolution obtained from a weather station [18]. The displayed
gure highlights the fact that most of the severe uctuations in
the output power of PV systems occur at maximum irradiance
level around noon. In some cases, this period might coincide
with the off-peak loading period of the electric network, which
is the case for some utilities in Western and Central Europe [11].
In general, the severity of impacts of the PV power uctuations
on the electric network is governed by several factors, including:
type of clouds, location and penetration level of the PV system,
and topologies of the PVsystemand the electric network. More-
over, the frequency of these uctuations are dictated by factors,
such as wind speed, type, height and size of passing clouds, as
well as the topology of the PV system and the area it covers.
The increasing penetration level of PV systems and other in-
termittent renewable systems is raising the concerns of some
utilities due to the possible negative impacts of the power uc-
tuations generated fromthese systems on the network operation.
Fig. 1. Fluctuations in the output power of a large PV system.
For example, very large PVsystems can affect the scheduling of
utility generating units [19]. Also, the uctuation in the power
of these systems can lead to unstable operation of the electric
network prior to the fault conditions, high power swings in the
feeders [20], and unacceptable voltage uctuations at certain
nodes in the electric network [11]. Moreover, the random uc-
tuations of the power output generated from PV systems does
not allow for considering them in the scheduling process of
electricity generation.
Currently, PVsystems are allowed to inject all their generated
power into the grid unless any of the operating limits of the
network is violated. In these cases, the utility might request the
system owner to take remedial steps to smooth the output of
the PV system or to restrict the generation of the PV system to
a certain limit. However, the high costs of the energy generated
fromthe PVsystem, and some of the methods that can be used to
smooth out the power uctuations, require careful economical
analysis before implementing any technology that achieve this
task.
One of the objectives of this paper is to examine the econom-
ical aspect of installing a battery-based storage (BS) system to
reduce the power uctuations of a large PV system, with rating
in the order of megawatts. This can be achieved by estimating
the maximum revenues that the system owner can obtain after
installing the BS system. To achieve this task, a linear program-
ming (LP) problem that can calculate the revenues of selling
the PV energy to the grid is formulated. The optimization prob-
lem also enables the estimation of the battery power and energy
ratings. Another objective of this paper is to investigate other
methods that can be used to reduce the power uctuations. One
of these methods is to install a dump load to dissipate the excess
PVpower. Another method is to curtail the PVpower by operat-
ing belowthe maximumpower point (MPP). For all investigated
methods, the revenues are compared with those obtained from
the PV system without any reduction in the uctuations, which
is considered the base case.
II. ENERGY STORAGE SYSTEMS
Energy storage devices are usually installed with PV systems
to perform one or more of the following tasks:
320 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 26, NO. 1, MARCH 2011
1) Bridge power uctuations: smoothing the output power of
the PV system can be achieved by storing energy during
high generation periods and delivering energy when the
output of the system falls due to passage of clouds. In this
case, the storage device should have a response time that
matches the speed of the uctuations and should be able to
deliver energy over the range of few minutes to few hours.
2) Shift the peak generation period: This is an important task
whenever the peak loading of the feeder is not around
noon, which corresponds to the peak generation period of
PV systems. In this case, the storage device stores energy
during the peak generation period and injects this energy
during the peak loading period. Thus, the storage device
should be able to operate for up to few hours.
3) Outage protection: In this case, the storage device along
with the PV system can be operated during electricity
outages to feed critical loads in the feeder. This has to be
done after coordinating with the local utility so as not to
violate the islanding protection regulations.
In general, there are a number of technologies that can be
used for storing and delivering electric energy. These technolo-
gies vary in performance and suitability for use with PVsystems.
For example, pumped hydro storage systems require signicant
land area with suitable topography for placing the reservoirs
at different altitudes. Thus, they might not be suitable for use
with large PV systems installed on large at land areas. Also,
superconducting magnetic energy storage systems are not suit-
able for bridging power uctuations in the order of few minutes
up to few hours, as in the case of PV systems, because they
have very low energy densities. These systems are rather used
for pulsed-power and system-stability applications [21]. Ultra-
capacitors also have low energy densities [22], and thus, cannot
be used alone to bridge PVpower uctuations. However, current
trends suggest the use of ultracapacitors with batteries to achieve
high power and energy densities and extend the lifetime of bat-
teries [23]. Conventional low-speed ywheels have high power
and low energy densities, while expensive advanced high-speed
ywheels have higher power and energy densities [24].
Batteries store energy through an electrochemical process,
and thus, have quick response in both charge and discharge pro-
cesses [25]. Moreover, batteries have mediumpower and energy
densities, and thus, they are one of the most suitable technolo-
gies that can be used to bridge the power uctuations gener-
ated from PV systems. During the past few years, a number
of studies focused on analyzing the different aspects of inte-
grating BS systems with PV systems, especially standalone PV
systems [26][28]. Other studies focused on using BS systems
with small grid-connected PV systems to supply power to criti-
cal loads during power outages of the connected grid [29], [30].
Some studies investigated the use of BS systems to shift the
peak generation of the PV system to the periods of peak load-
ing conditions [31][33]. Recently, few studies investigated the
use of battery storage systems to bridge power uctuations gen-
erated from intermittent renewable energy systems, especially
wind systems. For example, in [34], the probabilistic approach
was used to determine the power capacity of a storage system
to bridge power uctuations generated from a wind energy con-
Fig. 2. Grid-connected photovoltaic/battery system.
version system. However, the approach presented in this paper
cannot identify the energy capacity of the storage system. This
is mainly because the temporal information in the data is com-
pletely lost upon using the probabilistic approach. In [35], the
time series of the wind power was used to estimate the power and
energy ratings of a battery storage system. The main drawback
of the method applied in this study is that it assumes unlimited
initial energy in the battery, which is not a realistic assumption.
III. REDUCING POWER FLUCTUATIONS USING
A BATTERY STORAGE SYSTEM
As discussed earlier, the power uctuations generated from
a large PV system can negatively impact the operation of the
electric network. In this section, it is assumed that the utility
requests from the system owner to limit the power uctuations
of a 10-MW PV system within a certain range during a spe-
cic period of time (e.g., 250 kW/10 min from 9:00 A.M. to
3:00 P.M.). Accordingly, the system owner considers installing
a BS system, as shown in Fig. 2, to comply with the power uc-
tuations limit. In this case, it is important to choose the proper
power and energy ratings of the battery and estimate the rev-
enues gained from selling the energy of the PV/BS system. To
achieve this task, an LP optimization problem is formulated (1)
to (13). The input to the optimization problem is the historical
time series of the power generated from the PV system alone
P
P V
calculated over a number of years. In this paper, the DC
output power of the PV system is estimated from the irradiance
and temperature data obtained from [18] and the conversion
model provided in [36]. The AC output power is then obtained
by using the manufacturers efciency curve for a PV inverter.
Objective Function:
R = max

k=1

j=1
M

i=1

t
60
P
G
(i, j, k)

1
(1 +d)
k

PB
P
B max

1

ch
E
B max

(1)
OMRAN et al.: INVESTIGATION OF METHODS FOR REDUCTION OF POWER FLUCTUATIONS GENERATED 321
=
N
B

i=1
C
P
1
(1 +d)
(i1)T
B
+O
P
(1 +d)
T
1
d(1 +d)
T

N
B

i=1
S
P
1
(1 +d)
iT
B
(2)
=
N
B

i=1
C
E
1
(1 +d)
(i1)T
B
+O
E
(1 +d)
T
1
d(1 +d)
T

N
B

i=1
S
E
1
(1 +d)
iT
B
(3)
.
Constraints:
P
G
(i, j, k) = P
PV
(i, j, k) +P
B
(i, j, k) i, j, k (4)
P
G
(i 1, j, k) x P
G
(i, j, k) P
G
(i 1, j, k) +y
i > 1, j, k (5)
E
B
(i, j, k) = E
B
(i 1, j, k)
t
60
P
B
(i 1, j, k)
i > 1, j, k (6)
E
B
(1, j, k) = E
B
(M, j 1, k) j > 1, k (7)
E
B
(1, 1, k) = E
B
(M, N, k 1), k > 1 (8)
E
B
(1, 1, 1) = E
Bmax
(9)
P
Bmax
P
B
(i, j, k) P
Bmax
i, j, k (10)
SOC
min
E
Bmax
E
B
(i, j, k) E
Bmax
i, j, k (11)
0 P
G
(i, j, k) P
Gmax
i, j, k (12)
P
Bmax
, E
Bmax
, E
B
, P
G
0. (13)
The objective function (1) calculates the net revenues gained
fromthe PV/BS system. The rst termin this equation calculates
the revenues gained from selling the energy generated from the
system to the grid. The second and third terms account for
the net costs related to the battery power and energy ratings,
respectively. The net costs of the battery are calculated from (2)
and (3), where the rst terms in these equations estimate the
present worth of the capital cost related to the battery power and
energy ratings, respectively [37]. These terms consider the fact
that the battery might have to be changed a number of times
during the period of study. For example, if the battery lifetime
is 2 years, then for a study period of 6 years, the battery has to
be replaced twice after the rst installation. The second terms
in both equations account for the present worth of the annual
operating and maintenance costs, while the third terms consider
the present worth of the salvage value at the end of the battery
lifetime [37].
The problem is subject to several technical constraints. The
rst constraint (4) ensures that the power balance of the systemis
maintained, where it is assumed that the battery can be energized
from the PV system only. The second constraint (5) ensures that
the change in the power output of the PV/BS systemduring each
time step remains within the upper and lower bounds specied
by the utility. This constraint also facilitates the prediction of
the generation of the system during a short period of time.
For example, for a dataset with 10-min time resolution, if the
generation is 5 MW at a certain instant, and the values of x and
y are specied to be 100 kW/10 min, then, the power generated
from the system after 1 h is expected to be within 4.4 and
5.6 MW. Equation (6) displays the constraint that ensures the
energy balance of the battery is maintained at any instant during
the day. On the other hand, (7) ensures that the energy of the
battery at the end of any day is available for the beginning of the
next day in the same year. Similarly, (8) ensures that the energy
of the battery at the end of last day of the year is available
for the beginning of the rst day in the next year, while (9)
assumes that the battery is fully charged at the beginning of the
period of study. Equation (10) ensures that the battery power
during charge or discharge does not exceed the power rating
of the battery. To limit the depth of discharge of the battery
to the recommended level, (11) ensures that the battery energy
at any instant is not less than the minimum allowable state of
charge. This equation also does not permit the energy inside the
battery to exceed the battery rating during the charging period.
In general, the contract between the utility and the systemowner
species the power rating of the PV system installed at a certain
location. Thus, the power generated from the PV/BS system
should not exceed the ratings specied in the contract, as shown
in (12). The last constraint in this optimization problem is the
nonnegativity constraint, where the power and energy ratings
of the battery, the energy of the battery at any instant, and the
power injected into the grid at any instant are all positive. The
problem is modeled in the General Algebraic Modeling System
(GAMS) and solved using the MOSEK solver [38].
IV. CHOICE OF BATTERY TYPE AND SENSITIVITY ANALYSIS
In this section, the optimization problem is solved to nd
the ratings of the battery that satisfy the requirements of the
uctuation limits. The problem is also used to choose between
two types of batteries by comparing the revenues gained from
selling the energy generated from the system in both cases.
The chosen type of battery is further examined by changing
the different parameters of the battery and investigating the
corresponding changes in revenues.
A. Comparison of Battery Types
There are a number of battery technologies that can be used
for storing electric energy with lead-acid (LA) batteries being
the benchmark for storage batteries. This is mainly due to their
low cost, maturity of technology, and availability in the market
[39]. However, these batteries have short lifetime and require
frequent maintenance [21]. The new technology sodiumsulfur
(NaS) batteries have high power and energy densities and are
suitable for large-scale applications. They dont suffer from
self-discharge and their efciency is around 90% [22]. These
batteries operate at high temperatures, 300

C400

C, and thus,
require additional heating [39].
In this study, the rst type of batteries considered is the tradi-
tional LA battery, while the second is the new technology NaS
322 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 26, NO. 1, MARCH 2011
TABLE I
PARAMETERS USED IN THE OPTIMIZATION PROBLEM
battery. The main differences between the two types of batteries
are: 1) the NaS battery has an operating lifetime that is almost
three times of that of the LA battery [22]; 2) the operating and
maintenance costs of NaS battery is much lower than that of
the LA battery [40]; and 3) the capital cost of the LA battery
is lower than that of the NaS battery, especially the cost related
to the energy rating [41]. In this study, the lifetime of the LA
battery is assumed to be 2 years [42], and thus, that of the NaS
battery is assumed to be 6 years. For both batteries, the charging
efciencies and the capital costs, which constitute the main por-
tion of the batteries cost, are obtained from [41]. The operating
cost of the LA battery is assumed to be 10% of the capital cost,
and ten times that of the NaS battery. The salvage values for
both batteries are assumed to be 1% of the capital costs. The
parameters used in the optimization problem are displayed in
Table I.
The selling price of the electricity produced by the PV power
to the grid without reducing the power uctuations, i.e., without
using BS system is considered as the base case and is calculated
by (14). Thus, the revenues gained in the base case during the
6-year period of study are $23,666,808
R
pv
=
T

k=1

j=1
M

i=1

t
60
P
PV
(i, j, k)

1
(1 +d)
k

.
(14)
For each battery type, and for all the other case studies, the
percentage change in revenues is calculated by the following:
R% =
R R
PV
R
PV
100. (15)
The results obtained from using the two types of batteries
to limit the power uctuations to 250 kW/10 min during the
period 9:00 A.M. to 3:00 P.M. are displayed in Table II. The
revenues calculated in this table are those obtained from selling
the energy to the grid during the 6-year period of study.
The results of Table II indicate that for both types of batteries,
there is a loss in revenues as compared to the case when no
battery is installed. However, the loss in revenues is less, if the
TABLE II
REVENUES OBTAINED FROM THE TWO TYPES OF BATTERIES
TABLE III
EFFECT OF CHANGING THE OPERATING PERIOD OF THE BATTERY
expensive NaS battery with the parameters displayed in Table I
is used. The results also show that the required power ratings
of the two batteries are not the same, even with the power-
related efciency being the same. This is due to the fact that
the optimization problem aims to maximize the revenues while
limiting the uctuations to a certain value. Thus, for different
parameters of the batteries, different power and energy patterns
are generated, and accordingly, different ratings for the batteries
are specied. Fig. 3 illustrates this fact as it shows the power
generated from the two PV/BS systems for the same PV power
prole. Figs. 4 and 5 show the corresponding power and energy
proles of the batteries indicating different maximumvalues for
the two battery types.
B. Effect of Changing the Operating Period of the Battery
In this section, the effect of changing the operating period of
the battery is examined. In all cases, the NaS battery is used to
limit the uctuations during each time step to 250 kW/10 min;
however, the time period during which uctuations are limited
is varied. The results displayed in Table III show that the power
rating of the battery is constant for all time periods, indicating
that the maximumpower uctuations occur around noon. On the
other hand, the energy rating of the battery depends on the period
of operation, as it is determined by the power pattern of the
battery. The results also showthat the gained revenues are almost
the same for different operating periods of the battery, with the
2-h operating period producing slightly higher revenues. The
main reason behind the independence of the revenues from the
operating period is that the capital cost per kilowatt is relatively
high compared to the capital cost per kilowatt-hour. Thus, if the
power rating of the battery is constant, the revenues gained from
the system will not be signicantly affected.
OMRAN et al.: INVESTIGATION OF METHODS FOR REDUCTION OF POWER FLUCTUATIONS GENERATED 323
Fig. 3. Power patterns generated from the PV/BS systems for a cloudy day.
Fig. 4. Power pattern for the two types of batteries.
Fig. 5. Energy patterns for the two types of batteries.
C. Effect of Changing the Power Fluctuation Limit
To examine the relation between the revenues gained and
the power uctuation limit, the NaS battery is used to limit the
uctuations to 400 kW/10 min (4% of the rated PV power),
250 kW/10 min (2.5% of the rated PV power), 100 kW/
10 min (1% of the rated PV power), and 0 kW/10 min (constant
power). The revenues displayed in Table IV for all four limits
TABLE IV
EFFECT OF CHANGING THE POWER FLUCTUATION LIMIT
TABLE V
EFFECT OF CHANGING THE BATTERY PARAMETERS
are lower than the case when no battery is used. Moreover,
the revenues decrease as the uctuation limit decreases, with
the lowest revenues corresponding to the case when the PV/BS
system generates constant power. On the other hand, the power
and energy ratings of the battery increase as the uctuation limit
decreases to increase the capability of the battery in bridging
the uctuations.
D. Effect of Changing the Battery Parameters
In this section, the effects of reducing the capital cost and
increasing the efciency of the battery due to the possible tech-
nology advances are investigated. Any change in the capital cost
of the battery is accompanied by a similar change in the salvage
value of the battery.
The results of Table V show that reducing the capital cost
related to the power and energy ratings of the battery by 50%
reduces the loss in revenues by approximately 50%. The results
also indicate that the reduction in the capital cost per kilowatt
has a higher impact on decreasing the loss in revenues as com-
pared to the reduction in the capital cost per kilowatt-hour. This
is because the capital cost per kilowatt used in this paper consti-
tutes the main portion of the battery cost. On the other hand, the
results in Table V show that increasing the power and energy
efciencies of the battery does not have a considerable impact
on the change in revenues of the system.
It is worth mentioning that, for all cases, the power and energy
ratings of the battery are estimated by the optimization problem
to be 3677 kW and 6491 kWh, respectively. These are the
ratings of the battery that can satisfy the 250 kW/10 min power
uctuations constraint. Any increase in the battery ratings will
324 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 26, NO. 1, MARCH 2011
lead to additional loss in the revenues, as it will lead to additional
costs to the system owner.
V. REDUCTION OF POWER FLUCTUATIONS
BY INSTALLING A DUMP LOAD
Another method that can be used to reduce the uctuations in
the power generated from the PV system is to reduce the power
injected into the grid during periods of high uctuations. One
method that can be used for this purpose is to replace the BS
system in Fig. 2 by a dump load. The dump load consists of
a resistance and a controller to control the power ow through
the load. In some cases, the load is cooled to avoid excessive
heating during operation. The main function of the dump load is
to absorb the excess power generated from the PV system, and
thus, to smooth the power injected into the grid. In the analysis
presented here, the dump load is assumed to operate during
the 6-year period of study without replacement. The capital
cost of the dump load including the controller is assumed to
be $200/kW while the operating costs and the salvage value of
the dump load are considered to be zero. To estimate the power
rating of the dump load and the revenues obtained from selling
the energy generated fromthis system, the optimization problem
is modied to the following:
R = max

k=1

j=1
M

i=1

t
60
P
G
(i, j, k)

1
(i +d)
k

P
Dmax
. (16)
Subject to:
P
G
(i, j, k) = P
PV
(i, j, k) P
D
(i, j, k) i, j, k (17)
P
G
(i 1, j, k) x P
G
(i, j, k) P
G
(i 1, j, k) +y
i > 1, j, k (18)
0 P
D
(i, j, k) P
Dmax
i, j, k (19)
0 P
G
(i, j, k) P
Gmax
i, j, k (20)
P
G
, P
D
0. (21)
The problem is solved to estimate the revenues gained
from the system with the same uctuations limits displayed in
Table IV. The percentage change in revenues and the power
rating of the dump load are shown in Table VI for different
uctuation limits. The base case for calculating the percentage
change in revenues is the case when no power is dumped.
Comparing the results displayed in Table VI with those of
Table IV leads to the following observations:
1) Dumping part of the generated power to limit the power
uctuations will lead to loss of revenues. This loss depends
on the uctuation limit imposed by the network operator.
2) For high uctuation limits, dumping the power leads to
less loss in revenues as compared to the case when a BS
system with parameters given in Table I is installed. On
the other hand, for lower uctuation limits, the use of a
BS system leads to lower loss in revenues.
TABLE VI
REDUCTION OF POWER FLUCTUATIONS USING A DUMP LOAD
TABLE VII
REDUCTION OF FLUCTUATIONS BY OPERATING BELOW THE MPP
3) For the case when it is required to generate constant
power from the system (uctuation limit = 0), dumping
the power leads to 51% loss in the revenues. This large
loss is due to the fact that during periods of high uctua-
tion, the PV dump-load system has to generate a constant
power equal the lowest power during this period, while
any additional generation is lost in the dump load.
VI. REDUCTION OF POWER FLUCTUATIONS
BY OPERATING BELOW THE MPP
A third method that can be used to reduce the uctuations in
the output power of the PV system is to curtail the power by
operating belowthe MPP rather than dumping the power using a
dump load. This method does not require any additional instal-
lations, as only the control strategy of the power-conditioning
unit (PCU) needs to be modied. In this case, the curtailed
power can be considered the same as the dumped power and,
thus, (16)(21) can be still used but with setting = 0 and
replacing P
D
with P
C
. The results in Table VII show that the
loss in revenues for all limits of power uctuations is less than
in the case of using a dump load. This is due to the savings in
the capital cost of the dump load. The loss in revenues is also
less than in the case of using a BS system except for the case
when the generated power from the PV system is required to be
constant.
Based on the aforementioned discussions, it is logical to in-
vestigate the combination of a BS system with the curtailment
of power by operating the PCU below the MPP. In this case,
the original optimization problem (1)(13) can be applied with
minor modications. The rst modication is to account for the
curtailed power in the power-balance constraint (4), which is
OMRAN et al.: INVESTIGATION OF METHODS FOR REDUCTION OF POWER FLUCTUATIONS GENERATED 325
TABLE VIII
EFFECT OF CHANGING THE POWER FLUCTUATION LIMIT
Fig. 6. Percentage change in revenues for different methods.
replaced by the following:
P
G
(i, j, k)=P
PV
(i, j, k) +P
B
(i, j, k) P
C
(i, j, k) i, j, k.
(22)
The second modication is in the nonnegativity constraint
(13), which becomes:
P
Bmax
, E
Bmax
, E
B
, P
G
, P
C
0. (23)
The results displayed in Table VIII and Fig. 6 show that
installing a BS system and curtailing the power by operating
below the MPP leads to the least loss in revenues for all power
uctuations limits. Moreover, the power and energy ratings of
the battery are highly reduced as compared to the case when
power is not curtailed. Thus, combining both methods leads
to the most economical solution for the system owner. In this
case, the uctuations are reduced by charging and discharging
the battery. When the power limit of the battery is reached, the
excess power is curtailed by operating the PCU below the MPP.
Fig. 7 shows the power proles for different systemcomponents
in a cloudy day when the power uctuations are limited to
250 kW/10 min. It can be seen that the curtailed power is zero
except when the power discharged from the battery reaches the
capacity limit (1099 kW) from 10:00 to 10:10 A.M. and from
11:00 to 11:10 A.M.
VII. INCENTIVES FOR THE PV SYSTEM OWNER
The aforementioned results show that all the methods investi-
gated in this paper can effectively reduce the power uctuations
of the large PV systems. However, the revenues of selling the
Fig. 7. Power proles for different system components.
TABLE IX
PRICE OF SELLING THE PV ENERGY FOR THE DIFFERENT METHODS
PV energy to the grid obtained from any of the methods is less
than that obtained in the case when no limit on the power uctu-
ations is imposed. Thus, the utilities have to cooperate with the
system owner to compensate for the revenue losses. A possible
way of cooperation is by providing additional incentives to the
system owner. These incentives can be in the form of increas-
ing the selling price of PV energy during the periods when the
uctuations are reduced. This increase in price depends mainly
on the amount of revenues lost, and thus, on the method ap-
plied to reduce the power uctuations. The lowest increase in
the selling price is expected to correspond to the case when a
BS system is operated while curtailing the output power of the
PV arrays by operating below the MPP. Table IX shows the sell-
ing prices required to overcome the loss in revenues based on a
nominal selling price of $0.42/kWh and a uctuation limit of
250 kW/10 min.
The results of Table IX show that using batteries to limit the
uctuation during each time step to 250 kW/10 min, may lead
the utility to pay an additional $0.12 (28.6%) for each kilowatt-
hour during the period of operation of the battery. In case of
using a dump load, the utility would have to pay an additional
$0.13/kWh (30%), which is slightly higher than that in the case
of using a BS system. If the power is curtailed by operating
below the MPP, the utility should pay an additional $0.09/kWh
(21%), which is a better option. On the other hand, if the system
owner installs a battery and curtails the power by operating
below the MPP, the utility will pay only an additional $0.06/
kWh (14%), which is the best option for the utility.
326 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 26, NO. 1, MARCH 2011
VIII. CONCLUSION
This paper examined the economic aspects of using differ-
ent methods for smoothing the output power of large PV sys-
tems. The results presented show that irrespective of the method
adopted to smooth the PV output power, there is a loss of rev-
enues as compared to the case when the power is not smoothed.
When a battery storage system is used, the expensive NaS bat-
teries prove to be superior to the cheap LA batteries from a
long-term economic point of view. This is mainly due to the
higher efciency and longer lifetime of NaS batteries. With re-
gard to the operating period of the battery, extending the period
of operation of the battery on both sides of the noon hour does
not affect the power rating of the battery. This indicates that the
most severe power uctuations occur around noon. However,
upon changing the uctuations limit, it is observed that the loss
in revenues increases as the power uctuations limit decreases.
This is mainly because setting a low uctuations limit increases
both the power and energy ratings of the battery.
Dumping the excess power to satisfy the power uctuations
constraint, and power curtailment by operating below MPP,
prove to result in lower revenue loss than using a battery when
larger power uctuations are allowed, with power curtailment
being the superior option. Combining a battery storage system
and power curtailment is found to be the most economical so-
lution. In all cases, offering incentives in the form of higher PV
energy sale price seems to be a reasonable approach to encour-
age the PV system owner providing smoother power.
REFERENCES
[1] Trends in photovoltaic applications: Survey report of selected IEA coun-
tries between 1992 and 2007. (2008). [Online]. Available: http://www.
iea-pvps.org/.
[2] B. H. Chowdhury and A. W. Sawab, Evaluating the value of distributed
photovo1aic generations in radial distribution systems, IEEE Trans.
Energy Convers., vol. 11, no. 3, pp. 595600, Sep. 1996.
[3] N. Srisaen and A. Sangswang, Effects of PV grid-connected system
location on a distribution system, in Proc. IEEE Asia Pac. Conf. Circuits
Syst. 2006, 47 Dec., pp. 852855.
[4] Ontario Power Authority, Feed-In Tariff Prices for Renewable
Energy Projects in Ontario, (2010, Aug.). [Online]. Available:
http://www.powerauthority.on.ca/t/
[5] Ontario Power Authority, Ontarios Feed-In Tariff Program: Back-
grounder, (2010, Apr.). [Online]. Available: http://powerauthority.on.ca/
Storage/117/16512_Apr_8_Backgrounder_FINAL.pdf
[6] Samsung website, Samsung C&T, Korea Electric Power Company
to Build Worlds Largest Wind, Solar Panel Cluster in Ontario,
(2010, Jan.). [Online]. Available: http://www.samsung.com/ca/news/
newsRead.do?news_seq=17081&page=1
[7] Renewable Energy World, Samsung Invests $7B in Ontario Wind & So-
lar, (2010, Jan.). [Online]. Available: http://www.renewableenergyworld.
com/rea/news/article/2010/01/samsung-invests-can-7b-in-ontario-wind-
and-solar
[8] W. Jewell and T. Unruh, Limits on cloud-induced uctuation in photo-
voltaic generation, IEEE Trans. Energy Convers., vol. 5, no. 1, pp. 814,
Mar. 1990.
[9] A. Woyte, R. Belmans, and J. Nijs, Fluctuations in instantaneous clear-
ness index: Analysis and statistics, Solar Energy, vol. 81, no. 2, pp. 195
206, Feb. 2007.
[10] W. T. Jewell and R. Ramakumar, The effects of moving clouds on electric
utilities with dispersed PV generation, IEEE Trans. Energy Convers.,
vol. EC-2, no. 4, pp. 570576, Dec. 1987.
[11] A. Woyte, V. Van Thong, R. Belmans, and J. Nijs, Voltage uctuations
on distribution level introduced by photovoltaic systems, IEEE Trans.
Energy Convers., vol. 21, no. 1, pp. 202209, Mar. 2006.
[12] G. Vijayakumar, M. Kummert, S. Klein, and W. Beckman, Analysis of
short-termsolar radiation data, Solar Energy, vol. 79, pp. 495504, 2005.
[13] C. Craggs, E. M. Conway, and N. M. Pearsall, Statistical investigation of
the optimal averaging time for solar irradiance on horizontal and vertical
surfaces in the U.K, Solar Energy, vol. 68, pp. 179187, 2000.
[14] Solar Server, The worlds largest photovoltaic solar power plant
is in Pocking, [Online]. Available: http://www.solarserver.com/
solarmagazin/anlage_0606_e.html
[15] North American Electric Reliability Corp. (Apr. 2009). Accom-
modating High Levels of Variable Generation. [Online]. Available:
http://www.nerc.com/les/IVGTF_Report_041609.pdf
[16] E. Wiemken, H. G. Beyer, W. Heydenreich, and K. Kiefer, Power char-
acteristics of PV ensembles: Experiences from the combined power pro-
duction of 100 grid connected pv systems distributed over the area of
germany, Solar Energy, vol. 70, no. 6, pp. 513518, 2001.
[17] K. Otani, J. Minowa, and K. Kurokawa, Study on areal solar irradiance
for analyzing areally-totalized PV systems, Solar Energy Mater. Solar
Cells, vol. 47, pp. 281288, 1997.
[18] NREL Solar Radiation Research Laboratory (BMS). [Online]. Available:
http://www.nrel.gov/midc/srrl_bms/
[19] S. Rahman and M. Bouzguenda, A model to determine the degree of
penetration and energy cost of large-scale utility interactive photovoltaic
systems, IEEE Trans. Energy Convers., vol. 9, no. 2, pp. 224230, Jun.
1994.
[20] Y. T. Tan and D. S. Kirschen, Impact on the power system of a large
penetration of photovoltaic generation, in Proc. IEEE Power Eng. Soc.
Gen. Meet., Jun. 2428, 2007, pp. 18.
[21] EPRI-DOE handbook of energy storage for transmission and distribution
applications, Electric Power Res. Inst., Palo Alto, CA, Tech. Rep. EPRI,
2003.
[22] E. Spahic, G. Balzer, B. Hellmich, and W. Munch, Wind energy
storagesPossibilities, in Proc. IEEE Power Tech, 2007, pp. 615620.
[23] M. Glavin and W. Hurley, Ultracapacitor/ battery hybrid for solar energy
storage, in Proc. 42nd Int. Univ. Power Eng. Conf., 2007, pp. 791795.
[24] J. Carrasco, L. Franquelo, J. Bialasiewicz, E. Galv an, R. Guisado, M. Prats,
J. Le on, and N. Moreno-Alfonso, Power-electronic systems for the grid
integration of renewable energy sources: A survey, IEEE Trans. Ind.
Electron., vol. 53, no. 4, pp. 10021016, Aug. 2006.
[25] S. Choi, K. Tseng, D. Vilathgamuwa, and T. Nguyen, Energy stor-
age systems in distributed generation schemes, in Proc. 2008 IEEE
Power Energy Soc. Gen. Meet.Convers. Del. Electr. Energy 21st Cen-
tury, pp. 18.
[26] H. A. M. Maghraby, M. H. Shwehdi, and G. K. Al-Bassam, Probabilistic
assessment of photovoltaic (PV) generation systems, IEEE Trans. Power
Syst., vol. 17, no. 1, pp. 205208, Feb. 2002.
[27] Bagen and R. Billinton, Incorporating well-being considerations in gen-
erating systems using energy storage, IEEE Trans. Energy Convers.,
vol. 20, no. 1, pp. 225230, Mar. 2005.
[28] M. Kolhe, Techno-economic optimumsizing of a stand-alone solar photo-
voltaic system, IEEETrans. Energy Convers., vol. 24, no. 2, pp. 511519,
Jun. 2009.
[29] C. V. Nayar, M. Ashari, and W. W. L. Keerthipala, A grid-interactive
photovoltaic uninterruptible power supply system using battery storage
and a back up diesel generator, IEEE Trans. Energy Convers., vol. 15,
no. 3, pp. 348353, Sep. 2000.
[30] T. E. Hoff, R. Perez, and R. M. Margolis, Maximizing the value of
customer-sited PV systems using storage and controls, Solar Energy,
vol. 81, pp. 940945, 2007.
[31] W.-F. Su, S.-J. Huang, and C.-E. Lin, Economic analysis for demand-side
hybrid photovoltaic and battery energy storage system, IEEE Trans. Ind.
Appl., vol. 37, no. 1, pp. 171177, Jan./Feb. 2001.
[32] P. Poonpun and W. T. Jewell, Analysis of the cost per kilowatt hour to
store electricity, IEEE Trans. Energy Convers., vol. 23, no. 2, pp. 529
534, Jun. 2008.
[33] B. Lu and M. Shahidehpour, Short-term scheduling of battery in a grid-
connected PV/battery system, IEEE Trans. Power Syst., vol. 20, no. 2,
pp. 10531061, May 2005.
[34] J. P. Barton and D. G. Ineld, Energy storage and its use with intermittent
renewable energy, IEEE Trans. Energy Convers., vol. 19, no. 2, pp. 441
448, Jun. 2004.
[35] X. Y. Wang, D. M. Vilathgamuwa, and S. S. Choi, Determina-
tion of battery storage capacity in energy buffer for wind farm,
IEEE Trans. Energy Convers., vol. 23, no. 3, pp. 868878, Sep.
2008.
[36] M. E. Ropp, M. Begovic, and A. Rohatgi, Determination of the curvature
derating factor for the georgia tech aquatic center photovoltaic array, in
OMRAN et al.: INVESTIGATION OF METHODS FOR REDUCTION OF POWER FLUCTUATIONS GENERATED 327
Proc. Conf. Rec. 26th IEEE Photovoltaic Spec. Conf., 1997, pp. 1297
1300.
[37] G. J. Thuesen and W. J. Fabrycky, Engineering Economy (International
Series in Industrial and Systems Engineering), 8th ed. Upper Saddle
River, NJ: Prentice-Hall, 1993.
[38] GAMS Distribution 23.0: A users guide, GAMS Develop. Corp.,
Washington, DC, 2009.
[39] D. Ton, C. Hanley, G. Peek, and J. Boyes, Solar energy grid integration
systemsEnergy storage (SEGIS-ES), Sandia Nat. Lab., Albuquerque,
NM, SANDIA Rep. SAND2008-4247, Unlimited Release, Jul. 2008.
[40] Ali Nourai, Installation of the rst distributed energy storage system
(dess) at American electric power (AEP): A study for the DOE energy
storage systems program, Sandia Nat. Lab., Albuquerque, NM, SANDIA
Rep. SAND2007-3580, Unlimited Release, 2007.
[41] Electricity Storage Association. Storage Technologies, [Online]. Avail-
able: http://www.electricitystorage.org/site/home/
[42] D. P. Jenkins, J. Fletcher, and D. Kane, Lifetime prediction and sizing
of lead-acid batteries for micro-generation storage applications, IET,
Renewable Power Gen., vol. 2, no. 3, pp. 191200, Sep. 2008.
Walid A. Omran was born in Cairo, Egypt. He
received the B.Sc. and M.Sc. degrees in electrical en-
gineering from Ain-Shams University, Cairo, Egypt,
in 1998 and 2005, respectively, and the Ph.D. de-
gree fromthe Department of Electrical and Computer
Engineering, University of Waterloo, Waterloo, ON,
Canada, in 2010.
His research interests include power quality, dis-
tributed power generation, and integration of renew-
able systems in distribution networks.
M. Kazerani (S88M96SM02) received the
B.Sc. degree from Shiraz University, Iran, the
M. Eng. degree from Concordia University, Canada,
and the Ph.D. degree from McGill University,
Canada, in 1980, 1990, and 1995, respectively.
From 1982 to 1987, he was with the Energy Min-
istry of Iran. He is currently a Professor at the De-
partment of Electrical and Computer Engineering,
University of Waterloo, Waterloo, ON, Canada. His
research interests include power electronic circuits
and systems design, power quality/active power l-
ters, matrix converters, distributed power generation, utility interface of alterna-
tive energy sources, battery electric, hybrid electric and fuel cell vehicles, and
FACTS.
Dr. Kazerani is a Registered Professional Engineer in the province of
Ontario.
M. M. A. Salama (F02) received the B.Sc. and
M.Sc. degrees in electrical engineering from Cairo
University, Cairo, Egypt, in 1971 and 1973, respec-
tively, and the Ph.D. degree in electrical engineer-
ing from the University of Waterloo, Waterloo, ON,
Canada, in 1977.
He is currently a Professor in the Department of
Electrical and Computer Engineering, University of
Waterloo. He is also a Visiting Professor at King
Saud University, Saudi Arabia. His research interests
include the operation and control of distribution sys-
tems, cables, insulation systems, power-quality monitoring and mitigation, and
electromagnetics. He has consulted widely with government agencies and the
electrical industry.
Prof. Salama is a Registered Professional Engineer in the Province of
Ontario.