Available online at www.sciencedirect.

com

Solar Energy 90 (2013) 29–42

www.elsevier.com/locate/solener

An improved electric circuit model for photovoltaic modules based

on sensitivity analysis
M.U. Siddiqui a, A.F.M. Arif a,⇑, A.M. Bilton b, S. Dubowsky b, M. Elshafei c
a
Department of Mechanical Engineering, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia
b
Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
c
Department of Systems Engineering, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia

Received 3 September 2012; accepted 28 December 2012

Available online 4 February 2013

Communicated by: Associate Editor Nicola Romeo

Abstract

Fast and accurate electrical models of PV systems are crucial for their design. Electrical models of PV devices range from very simple
scaling equations to complex models requiring large amount of experimental data. The purpose of the present work is two-fold. First, an
optimization methodology to estimate the model parameters for the five parameter single-diode equivalent electrical circuit model of PV
devices using only manufacturer supplied data is presented and compared to three electrical models from literature for six different PV
modules. Second, a sensitivity analysis is carried out to identify the relative importance of the five model parameters. Using the results of
sensitivity analysis, the electrical model is improved by including two additional parameters. The suggested model modification results in
marked improvement in the prediction accuracy for all six modules considered.
Ó 2013 Elsevier Ltd. All rights reserved.

Keywords: PV module; Five parameter model; Sensitivity analysis; Parameter estimation; Equivalent circuit

1. Introduction Crystalline silicon cells are further divided into sub-types

The fundamental building block of photovoltaic (PV)
systems is the PV cell. A PV cell is a semiconductor diode
which when exposed to light generates charge carriers. If
the PV cell is connected to an external circuit, current flows
through the circuit. But, a single PV cell generates very low
power, about 0.5 W in case of crystalline silicon cell. There-
fore, series and parallel combinations of cells are made to
get a desired electrical power and voltage. Commercially,
PV manufacturers generally construct PV modules which
consist of 36 or more PV cells connected in series.
The two types of commercially available solar cells are
crystalline silicon solar cells and thin film solar cells.
⇑ Corresponding author. Tel.: +966 38602579.
E-mail addresses: afmarif@kfupm.edu.sa, afmarif1@gmail.com
(A.F.M. Arif).
depending on the crystal size and structure. These sub-
types are mono-crystalline silicon and poly-crystalline sili-
con. The main thin film solar cells technologies include
amorphous silicon (a-Si), Cadmium Telluride (CdTe),
Copper Indium Selinide (CIS) and Copper Indium Gallium
Selinide (CIGS). Thin film technologies have the advantage
of reducing the requirement of light absorbing materials
and thus are less expensive than crystalline silicon
technologies.
Accurate models of the electrical performance of PV sys-
tems are vital for system designers. Over the years, various
electrical models to predict instantaneous as well as long
term performance of PV systems have been developed.
Attempts to develop accurate PV models have included
models based on the analytical knowledge of PV cell
behavior, models based on empirical correlations, as well
as models which combine the two approaches. King et al.

0038-092X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.solener.2012.12.021
30 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42
Nomenclature
A modified diode ideality factor (V)
A diode ideality factor
AM air mass
AOI angle of incidence
E irradiance (W/m2)
Eg band-gap energy of PV cell material (eV)
FF fill factor
fd fraction of diffuse radiation absorbed in the
module
I PV module output current (A)
IL light current (A)
Io diode reverse saturation current (A)
k Boltzmann’s constant, 1.38066E23 (J/K)
m irradiance dependence parameter for IL
n temperature dependence parameter for a
NCS number of cells in series in a module’s cell-string
Np number of cell-strings in parallel in module
Ns number of cells in series in a module’s cell-string
NSC normalized sensitivity coefficient
P electrical power (W)
q elementary charge, 1.60218  1019 (coulomb)
Rs series resistance (X)
Rsh shunt resistance (X)
S plane-of-array absorbed solar radiation at oper-
ating conditions (W/m2)
T temperature (°C)
V voltage (V)
Xi nominal value of model parameters in sensitivity
analysis
Y nominal value of function in sensitivity
analysis
(2004) developed an empirical model for simulation of PV
systems called the Sandia Labs PV model. The model
included its own thermal and radiation models and
required thirty experimentally determined constants to pre-
dict the performance of the PV system at five key points on
the current–voltage (I–V) curve. Hishikawa et al. (2000)
developed a methodology for translating the I–V curves
from one irradiance level to another. Marion et al. (2004)
extended this work to include the effect of varying cell tem-
peratures. Their model, called the bilinear interpolation
model, used four experimentally determined I–V curves
at two different irradiances and two different temperatures.
Townsend (1989) developed a semi-empirical model for
predicting the performance of direct coupled PV systems.
His model, known as the four parameter model, requires
only data available in manufacturer datasheets to deter-
mine parameters for an equivalent electric circuit that can
predict the electrical performance of PV systems. Duffie
and Beckman (1991) presented an improved electric circuit
Greek
aimp temperature coefficient of maximum power
point current
aisc temperature coefficient of short circuit current
bvoc temperature coefficient of open circuit voltage
bvmp temperature coefficient of maximum power
point voltage
DX i change in model parameters in sensitivity analysis
DY change in function value due to DX i change in
model parameters
d(Tc) thermal voltage per cell at temperature Tc
lVoc temperature coefficient of open circuit voltage
lIsc temperature coefficient of short circuit current
c overall diode ideality factor of PV module
Subscripts
0 reference cell condition
amb ambient
b beam radiation
c PV cell
diff diffuse radiation
e effective radiation; experimental
m module back surface; modeled
mp maximum power point
oc open circuit point
ref reference cell condition
sc short circuit point
x IV point at module voltage equal to half of open
circuit voltage
xx IV point at module voltage equal to average of
max. power and open circuit voltages
model for PV devices called the five parameter single-diode
model. The additional parameter in the five parameter
model improved the prediction accuracy of the model for
thin film PV cell types. De Soto et al. (2006) presented a
method of finding the model parameters for the five param-
eter model using only manufacturer data. Valerio et al.
(2010) developed a new variation of the five parameter
model capable of more accurately accounting for changes
in operative temperature and solar irradiance. To find the
model parameters for a given PV module, a system of
I–V equations was solved based on manufacturer’s data
for module performance at standard test conditions (STCs)
with a trial and error process. Almonacid et al. (2009)
adopting a different methodology for predicting PV electri-
cal performance trained an artificial neural network (ANN)
using measured irradiance, ambient temperature and
current–voltage curves. They verified the trained ANN
using a second dataset by comparing the predicting electric
power output with the measured values. The model
 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42
predictions showed an error of 5% for irradiance values
greater than 100 W/m2 which increased to around 8%
for lower irradiance values.
The work presented in this paper is two folds. First, a
methodology to estimate the model parameters for the five
parameter model through an optimization technique using
only manufacturer supplied data is presented. The method
utilizes a standard optimization algorithm and thus simpli-
fies the parameter estimation task for the user. The accu-
racy of the estimation method is compared with the
accuracies of several electrical models from literature for
PV modules of various cell technologies. In the second
part, the results of a sensitivity analysis carried to check
the influence of the model parameters on the model output
are presented. These results can be used as a guideline for
future improvement in the five parameter model. Finally,
one possible improvement to the five parameter model
based on the sensitivity analysis results is proposed. The
parameter estimation methodology for the new model
was also developed and presented and the results of the
model were compared to other models from literature.
2. Electric circuit modeling of PV devices
Any PV element, whether it is a PV cell, module or an
entire array, can be represented by an equivalent electric
circuit (Duffie and Beckman, 1991). The equivalent circuit,
shown in Fig. 1, comprises of a light dependent current
source, a p–n junction diode and two resistances. It should
be noted that the equivalent circuit of Fig. 1 is based on the
physics of crystalline silicon cell.
The current–voltage relationship of the PV device repre-
sented by the equivalent circuit of Fig. 1 is described by Eq.
(1) and is governed by five model parameters (IL, Io, a, Rs
and Rsh). The electrical model that describes the electrical
performance of a PV device using Eq. (1) is called the five
parameter model.
   
V þ I  Rs V þ I  Rs
I ¼ I L  I o  exp 1  ð1Þ
a Rsh
Before the equation can be used, the unknown model
parameters need to be determined. A variety of techniques
have been reported to estimate the model parameters. De
Soto et al. (2006) and Boyd et al. (2011) used a specialized
non-linear equation solver to get a solution. Villalva et al.
(2009) explicitly defined one parameter, aref, and then
Fig. 1. Equivalent circuit of a PV cell.
31
solved for the remaining parameters. Townsend (1989) sim-
plified the model by assuming the shunt resistance to be
infinite which reduces the non-linearity of the system. Car-
rero et al. (2011) used an iterative procedure to find all five
parameters. Their method only requires the I–V data of
three points i.e. the short circuit, open circuit and maxi-
mum power points. Their method uses simplified forms
of the I–V equation written at the three points and provides
fast convergence.
Once the parameters at the reference condition are
known, Eqs. (2)–(6) are used to calculate the parameters
at operating conditions (solar irradiance, cell temperature)
of the PV device. Using these modified parameters, Eq. (1)
is used to predict the performance of the PV system at any
temperature and irradiance.
a ¼ aref T c =T ref ð2Þ
I L ¼ S=S ref ðI L;ref þ lIsc ðT c  T ref ÞÞ ð3Þ
 
NCS:T ref
3 aref ðEg;ref =T ref E g =T c Þ
I o ¼ I o;ref ðT c =T ref Þ e ð4Þ
S ref
Rsh ¼ Rsh;ref ð5Þ
S
Rs ¼ Rs;ref ð6Þ
The subscript ref represents the parameters at STC, T is the
temperature of the PV panel, S is the absorbed solar radi-
ation, lIsc is the temperature coefficient of short circuit cur-
rent, NCS is the number of cells connected in series and Eg
is the band-gap energy of the PV cell material (Eg = 1.12 -
eV for crystalline silicon, 1.6 eV for amorphous silicon and
1.04 eV for Copper Indium Selenide). Eq. (7) can be used
to calculate Eg at the new temperature. The constant
0.0003174 is for mono-crystalline silicon and De Soto
et al. (2006) have suggested that this value can be used
for all technologies with little error.
Eg ¼ Eg;ref ð1  0:0003174ðT c =T ref ÞÞ ð7Þ
3. Model parameters estimation using manufacturer supplied
data
As shown in Section 2, different methodologies have
been used for calculating the model parameters for the five
parameter single diode model. In the present study, the
parameter estimation task was carried out as an optimiza-
tion problem. Similar techniques based on optimization
algorithms have already been reported in literature (Ike-
gami et al., 2001; Jervase and Bourdoucen, 2001; Ortiz-
Rivera and Peng, 2005; Moldovan et al., 2009; Zagrouba
et al., 2010; Ishaque and Salam, 2011; Ishaque et al.,
2011). In these works, the objective function used was the
current prediction error at multiple points on the I–V curve
which was calculated using experimentally measured volt-
ages. Hence, experimental data was required for the
parameter estimation task.
In the present work, it is shown that only the manufac-
turer supplied data for three I–V points (short circuit, open
32
circuit and maximum power point) is sufficient for the esti-
mation task using optimization methods. A multi-variable
optimization technique, the Nelder–Mead simplex search
algorithm (Nelder and Mead, 1965), was used to minimize
the error in an objective function by varying the five model
parameters. The objective function used in the current
work is defined by Eq. (8). It calculates two errors whose
sum is minimized. First, the errors in current prediction
at short circuit, maximum power point and open circuit
conditions are calculated using manufacturer supplied volt-
ages and the model parameters to be estimated. Second, the
slope of the power–voltage (P–V) curve is calculated at
maximum power point (MPP) using the manufacturer sup-
plied values for maximum power point current and voltage
and the model parameters to be estimated. The slope dP/
dV will become zero when the correct model parameters
are used. Therefore, any non-zero value is considered to
be an error in the model parameters estimation. The three
known I–V points and the slope provide four pieces of
information from which the five parameters are estimated.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P3
i¼1 ðI m ðV e;i ; I L ; I o ; a; Rs ; Rsh Þ  I e ðV e;i ÞÞ
error ¼
3
 
 dP 
þ  
dV ðI e;MP ;V e;MP ;I L ;I o ;a;Rs ;Rsh Þ
4. Validation of parameter estimation methodology
In order to validate the proposed parameter estimation
methodology and to compare its accuracy with the accu-
racy of models available in literature, a total of six PV
modules, three crystalline silicon modules and three thin
film modules, were selected. The I–V curves available in
the datasheets of the modules were digitized and five key
I–V points were extracted from each curve. These points
are short circuit (SC), open circuit (OC), maximum power
point (MPP), point with voltage equal to half the voltage
at maximum power point (X) and point with voltage equal
to the average of maximum power point voltage and open
circuit voltage (XX).
4.1. Selected PV modules
The PV modules selected for model validation are listed
below. The manufacturer supplied electrical performance
characteristics of the modules at STC are listed in Table
1. The I–V points for crystalline silicon and thin film mod-
ules which are extracted from the I–V curves on the module
datasheets are listed in Tables 2 and 3 respectively.
1. Astro power AP110 (mc-Si).
2. Shell solar S36 (pc-Si).
3. Kyocera KC-40T (pc-Si).
4. Shell solar ST36 (CIS).
5. Solarex MST43LV (2-a-Si).
M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42
6. Unisolar PVL-124 (3-a-Si).
4.2. Error statistics
Two metrics have been selected to compare the PV mod-
els. The metrics selected are the root mean square error and
the mean bias error, given by Eqs. (9) and (10) respectively.
In these equations, y is the modeled value, x is the mea-
sured value and n is the number of values.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
1
Pn 2
n i¼1 ðy i  xi Þ
RMSE ¼ 1
Pn ð9Þ
n i¼1 xi
1
Pn
n i¼1 ðy i  xi Þ
MBE ¼ 1
Pn ð10Þ
n i¼1 xi
The root mean square error is used to check the accuracy
of the model in predicting actual measured values. The
mean bias error, on the other hand, does not represent
the accuracy of the model. Rather it is used to check
2 whether the model over-predicts or under-predicts the
performance.
ð8Þ 4.3. Electrical models selected for comparison
Three electrical performance models were selected for
comparison with the five parameter model using the pro-
posed parameters estimation methodology.
The selected models are:
1. Four parameter electric circuit model (Townsend, 1989).
2. Sandia Labs model (King et al., 2004).
3. Villalva et al. electric circuit model (Villalva et al., 2009).
It is important to note here that two of the models
selected for comparison, four parameter electric circuit
model and Villalva et al. electric circuit model, are varia-
tions of the five parameter model and differ from the pro-
posed modeling methodology in either the assumptions
involved or the parameter estimation process.
4.3.1. Four parameter PV model
Four parameter model is based on the assumption that
the shunt resistance is high enough to be considered infinite
and neglected. This results in only four unknown parame-
ters and the simplified model is called the four parameter
model (Townsend, 1989). The advantage of this model is
that with the removal of the fifth parameter, Rsh, the non-
linearity in the system of equations used to determine the
parameters reduces significantly and its solution is greatly
simplified.
The I–V characteristics of the PV system in the four
parameter model are described by the following equation:
  
qðV þ IRS Þ
I ¼ I L  I 0 exp 1 ð11Þ
ckT C
 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42 33

Table 1
Electrical characteristics of the selected modules at STC.
Astro power AP- Shell solar S- Kyocera KC- Shell solar ST- Solarex MST43- Unisolar PVL-
110 36 40T 36 LV 124
Short circuit current (Isc) (A) 7.5 2.3 2.65 2.68 3.3 5.1
Open circuit voltage (Voc) (V) 20.7 21.4 21.7 22.9 22.7 42
MPP current (Imp) (A) 6.6 2.182 2.48 2.28 2.6 4.13
MPP voltage (Vmp) (V) 16.7 16.5 17.4 15.8 16.5 30
Number of cells in series (NCS) 36 36 36 42 16 20
Isc temperature coefficient (lIsc) 0.0034 0.001 0.00106 0.00032 0.00066 0.001
(A/°C)
Voc temperature coefficient (lVoc) 0.08 0.076 0.0821 0.1 0.1 0.0038
(V/°C)

Table 2
Extracted I–V points for crystalline silicon modules.
Point AP-110 S36 KC-40T
V (V) I (A) V (V) I (A) V (V) I (A)
2 2 2
1000 W/m and 25 °C 1000 W/m and 25 °C 1000 W/m and 25 °C
SC 0.00 7.50 0.00 2.30 0.00 2.65
X 8.35 7.30 8.25 2.29 8.70 2.61
MPP 16.70 6.60 16.50 2.18 17.40 2.48
XX 18.70 4.60 18.95 1.53 19.55 1.82
OC 20.70 0.00 21.40 0.00 21.70 0.00
1000 W/m2 and 60 °C 1000 W/m2 and 60 °C 1000 W/m2 and 50 °C
SC 0.00 7.70 0.00 2.33 0.00 2.74
X 6.46 7.43 7.26 2.33 7.69 2.70
MPP 12.92 6.70 14.52 2.09 15.38 2.55
XX 15.13 3.88 16.64 1.44 17.49 1.78
OC 17.34 0.00 18.75 0.00 19.6 0.00
800 W/m2 and 45 °C 400 W/m2 and 25 °C 400 W/m2 and 25 °C
SC 0.00 6.00 0.00 0.92 0.00 1.22
X 7.33 5.88 8.25 0.89 8.32 1.21
MPP 14.65 4.98 16.54 0.84 16.64 1.15
XX 16.41 3.54 18.41 0.64 18.82 0.82
OC 18.18 0.00 20.27 0.00 20.99 0.00

The parameters IL, Io and Rs are the same as in the five I sc ¼ I sc0 :f1 ðAM a ÞfðEb f2 ðAOIÞ þ fd Ediff Þ=E0 gf1
parameter model. The parameter c is related to the param-
þ aisc ðT c  T 0 Þg ð12Þ
eter a in the 5 parameter model by the relation a = ckTc/q.

The methodology to find the model parameters for the four I mp ¼ I mp0 C 0 Ee þ C 1 E2e f1 þ aimp ðT c  T 0 Þg ð13Þ
parameter model is presented in Appendix A.
V oc ¼ V oc0 þ N s dðT c Þ lnðEe Þ þ bvoc ðEe ÞðT c  T 0 Þ ð14Þ

4.3.2. Sandia Labs PV model V mp ¼ V mp0 þ C 2 N s dðT c Þ lnðEe Þ þ C 3 N s fdðT c Þ

The Sandia Labs PV model was developed by King et al. 2
 lnðEe Þg þ bvmp ðEe ÞðT c  T 0 Þ ð15Þ
(2004) and uses empirical correlations and experimentally
determined coefficients to predict the performance of the P mp ¼ I mp V mp ð16Þ
PV system. The accuracy of the model has been demon-

I x ¼ I x0 C 4 Ee þ C 5 E2e f1 þ aisc ðT c  T 0 Þg ð17Þ
strated for modules of all technologies as well as for con-

centrator modules and for large arrays of modules. I xx ¼ I xx0 C 6 Ee þ C 7 E2e f1 þ aimp ðT c  T 0 Þg ð18Þ
Electrical, thermal, solar spectral and optical effects for
photovoltaic modules are all included in the model. The
currents and voltages at five points on the I–V curve are 4.3.3. Villalva et al. electric circuit model
calculated by the Sandia Labs model using Eqs. (12)– Villalva et al. (2009) developed a model to predict the
(18). The remaining equations used in the model are listed electrical performance of PV systems based on the equiva-
in Appendix A. lent electric circuit of Fig. 1. The difference in their model
from other models based on the single-diode equivalent cir-
34 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42

Table 3
Extracted I–V points for thin film modules.
Point PVL-124 ST36 MST43LV
V (V) I (A) V (V) I (A) V (V) I (A)
1000 W/m2 and 25 °C 1000 W/m2 and 25 °C 1000 W/m2 and 25 °C
SC 0.00 5.10 0.00 2.68 0.00 3.30
X 15.00 4.98 7.90 2.65 8.25 3.10
MPP 30.00 4.13 15.80 2.28 16.50 2.60
XX 36.00 2.62 19.35 1.49 19.60 1.75
OC 42.00 0.00 22.90 0.00 22.70 0.00
600 W/m2 and 25 °C 400 W/m2 and 25 °C 250 W/m2 and 25 °C
SC 0.00 3.07 0.00 1.07 0.00 0.88
X 14.7 3.03 7.77 1.06 8.15 0.77
MPP 29.4 2.62 15.53 0.89 16.30 0.67
XX 35.24 1.71 18.25 0.61 18.46 0.52
OC 41.09 0.00 20.96 0.00 20.61 0.00
200 W/m2 and 25 °C 1000 W/m2 and 60 °C
SC 0.00 1.06 0.00 2.72
X 14.74 1.04 6.36 2.68
MPP 29.47 0.91 12.71 2.18
XX 34.04 0.66 16.10 1.35
OC 38.60 0.00 19.50 0.00

Table 4
Parameters for the five parameter model at STC using proposed methodology.
Parameter AP-110 S-36 KC-40T ST-36 MST-43LV PVL-124
IL,ref (A) 7.5084 2.3199 2.6540 2.7148 3.5075 5.3001
Io,ref (A) 3.4686e6 9.8279e7 1.3503e8 9.2532e7 2.33e16 1.14e23
aref (V) 1.4249 1.4587 1.1366 1.5451 0.6141 0.7726
Rs,ref (X) 0.0527 0.4235 0.4747 1.6187 1.6270 2.5023
Rsh,ref (X) 46.8713 2002.273 2117.7476 124.7165 25.8786 63.7659

cuit is in the methodology for calculating the model param-
eters. During the parameter estimation process, one of the
five parameters, aref, is explicitly specified. The remaining
parameters are calculated using performance data of the
PV module available in module datasheet. The procedure
to estimate the model parameters is presented in Appendix
A.
4.4. Parameters calculation for the selected modules
Before the electrical performance of the modules can be
simulated, the model parameters for the modules were
determined. The procedures to determine the parameters
for the five parameters model using the proposed method-
ology is described in Section 3. The procedure to determine
the parameters for the four parameter model and Villalva
et al. model are described in Appendix A. The parameters
determined using the proposed method for the five param-
eter model at STC are listed in Table 4, the estimated
parameters for the four parameter model at STC are listed
in Table B.1 and the parameters for the Villalva et al.
model are listed in Table B.2 The parameters for the Sandia
Labs PV model are experimentally determined and were
obtained from the parameters library in the System Advi-
sor Model (SAM) software available from the National
Renewable Energy Labs website. The parameters have
been listed in Table B.3.
4.5. Comparison with other electrical models – results
The electrical performance of the six selected PV mod-
ules as predicted by the five parameter model using the pro-
posed parameter estimation methodology and the three
selected models from literature was calculated at the
absorbed radiation and cell temperature conditions listed
in Table 2. From the results of these calculations, the root
mean square error and mean bias error were calculated for
the currents and voltages at the five key points and for
maximum power. The model errors in currents and volt-
ages at the five key points are shown in Figs. 2 and 3.
Fig. 4 shows the error in the prediction of maximum power
for each module.
The results in Figs. 2–4 show that the five parameter
model using the proposed parameter estimation methodol-
ogy performs satisfactorily compared to the four parameter
model and the Sandia Labs model with some instances in
which it shows better accuracy than the two and some
instances in which it shows larger errors than them. The
five parameter model provided better results when the
parameters estimated using the new methodology were
 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42 35

Fig. 2. Crystalline silicon modules current and voltage errors.

Fig. 3. Thin film modules current and voltage errors.

used, especially for thin film module. For thin film mod- although it differs from the proposed model in the method-
ules, Villalva et al. model showed much larger errors ology for estimating the model parameters. This signifies
36 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42

the importance of the parameter estimation process for
accurate model prediction.
5. Sensitivity analysis of the five parameter model
To identify the relative importance of the five model
parameters, a sensitivity analysis was carried out in which
normalized sensitivity coefficients were calculated for each
model parameter. This allows a one-to-one comparison
between the model parameters. Masi et al. (1999) explained
the physical significance of the normalized sensitivity coef-
ficients as the order of magnitude change in the analyzed
function that will result from one order of magnitude
change in the concerned parameter. Thus, normalized sen-
sitivity coefficients are used to identify the model parame-
ters to which the model is most sensitive. The normalized
sensitivity coefficients (NSCs) are given by Eq. (19), where
Y is nominal value of the function at nominal model
parameters X i and DY is the change in the function value
to due to a change of DX i in the model parameter X i .
For the five parameter model, the maximum power output
of the PV panel is considered to be the function Y and Xi
are the five model parameters (IL, Io, a, Rs and Rsh). Sensi-
tivity analysis was carried out using the module data and
parameters for the AP-110 PV module and the results of
the analysis are shown in Table 5. In the table, the values
X+ and X show 10% increase and decrease in the param-
eter X respectively while Y+ and Y show the correspond-
ing values of the function Y. It is clear from the table that
the electrical power prediction using the model is clearly
more sensitive to two model parameters, IL and a. In order
to check whether the normalized sensitivity coefficients
vary with irradiance and cell temperature, the analysis
was repeated by changing the absorbed radiation and the
cell temperature. The results are shown in Fig. 5 which
shows that within the ranges of absorbed radiation and cell
temperatures considered, the relative importance of the five
parameters does not change.
DY X i
NSC i ¼  ð19Þ
Y DX i
6. Proposed seven parameter electrical model
In this section, improvements are suggested to the trans-
lation Eqs. (2)–(6) based on results of the sensitivity analy-
sis. The constraint imposed in modifying the equations was
that no data not available in the module datasheets could
be used for estimating any additional parameters.
Boyd et al. (2011) attempted to improve the five param-
eter model by introducing two new parameters m and d
into the translation equations for Io and Rs. The modifica-
tions did not provide satisfactory results with errors
remaining the same or sometimes even increasing when
the new parameters were introduced, especially in the max-
imum power prediction. These observations agree with the
results of the sensitivity analysis carried out in the current
work which show that the maximum power prediction is
not very sensitive to Io and Rs.
In the present work, modifications to the model based
on the results of the sensitivity analysis were tried to
improve the model accuracy. Using an approach similar
to Boyd et al. (2011), modifications were made in the trans-
lation equations of parameters IL and a to which the model

Fig. 4. Maximum power prediction errors. (Module 1 = AP-110, Module 2 = S-36, Module 3 = KC-40T, Module 4 = MST-43LV, Module 5 = ST-35,
Module 6 = PVL-124.)

Table 5
Normalized sensitivity analysis results for the five parameter model.
Parameter X X+ X Y+ Y DXi DYi NSCi
IL (A) 7.5084 8.25924 6.75756 122.5 98.02 1.50168 24.48 1.222553
Io (A) 3.47E06 3.82E06 3.12E06 109.4 111.1 6.94E07 1.7 0.005896
a (V) 1.4249 1.56739 1.28241 120.8 99.48 0.28498 21.32 0.927298
Rs (X) 0.0527 0.05797 0.04743 109.9 110.5 0.01054 0.6 0.000734
Rsh (X) 46.8713 51.55843 42.18417 110.8 109.5 9.37426 1.3 0.003448
 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42 37

Fig. 5. Variation of normalized sensitivity coefficients with (a) absorbed radiation and (b) PV cell temperature.

Table 6
Modifications to translation equations for IL and a.
Model Translation equation for IL Translation equation for a
m
7 Parameters (IL and a) IL = (S/Sref) (IL,ref + lIsc(Tc  Tref)) a = aref(Tc/Tref)n
6 Parameters (m) IL = (S/Sref)m(IL,ref + lIsc(Tc  Tref)) a = aref(Tc/Tref)
6 Parameters (n) IL = (S/Sref)(IL,ref + lIsc(Tc  Tref)) a = aref(Tc/Tref)n
7 Parameters (a) IL = (S/Sref)(IL,ref + lIsc(Tc  Tref)) a = aref(G/Gref)m(Tc/Tref)n
7 Parameters (IL) IL = (S/Sref)m(IL,ref + (lIsc)n(Tc  Tref)) a = aref(Tc/Tref)

Table 7
Module information required for parameters m and n.
Module Low irradiance data Temperature coefficient of IMP (A/°C) Temperature coefficient of VMP (V/°C)
Input condition IMP (A) VMP (V)
AP-110 800 W/m2, 45 °C 4.975 14.653 0.00042 0.084
S-36 400 W/m2, 25 °C 0.835 16.54 0.00014 0.076
KC-40T 400 W/m2, 25 °C 1.15 16.64 0.00014 0.084
MST-43LV 250 W/m2, 25 °C 0.67 16.30 0.00095 0.087
ST-36 400 W/m2, 25 °C 0.894 15.53 0.00045 0.074
PVL-124 200 W/m2, 25 °C 0.909 29.472 0.001 0.093

Table 8
Effect of the translation equations modification on maximum power prediction accuracy (RMSE).
Model Module 1 Module 2 Module 3 Module 4 Module 5 Module 6
Original 1.755 2.054 2.746 2.751 1.760 5.456
7 Parameters (IL and a) 1.426 0.378 1.284 0.003 0.942 3.037
6 Parameters (m) 0.889 2.041 1.654 0.051 0.247 2.983
6 Parameters (n) 2.315 0.466 2.542 2.751 3.160 5.486
7 Parameters (a) 1.419 0.407 1.287 0.048 3.150 2.396
7 Parameters (IL) 1.425 0.404 1.285 0.051 2.583 2.988

prediction is most sensitive to. In total, five possible Table 9

improvements to the translation equations were attempted. Parameters m and n for the selected PV modules.
Table 6 shows the five modifications to the translation Module m n
equations of IL and a considered here.
AP-110 1.0959 1.1368
In all the modifications considered, the new parameters S-36 0.9865 1.4144
(m and n) were estimated using a secondary optimization KC-40T 0.8706 1.1157
routine that used the sum of the root mean square errors MST-43LV 1.0736 0.0887
in maximum power prediction at a low irradiance condi- ST-36 1.0959 1.1368
PVL-124 1.1137 0.0010
tion and at a high temperature as the objective function
38 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42

the temperature coefficients of either IMP and VMP or

Fig. 6. Comparison of maximum power prediction accuracy (RMSE) of
five parameter model and seven parameter model. (Module 1 = AP-110,
Module 2 = S-36, Module 3 = KC-40T, Module 4 = MST-43LV, Module
5 = ST-35, Module 6 = PVL-124.)
to be minimized. The objective function for the secondary
optimization is given by the following equation:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
error ¼ ðP m  P e Þ 
2
Low Irradiance
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þ ðP m  P e Þ 
2
ð20Þ
High Temeprature
where P is the electrical power output of the PV panel. Sub-
scripts m and e stand for modeled and experimental values
respectively. The new information about the PV module re-
quired for the estimation process includes the maximum
power of the PV module at a low irradiance value and
PMP. Although in the future all manufacturers will have
to provide data at low irradiance condition (200 W/m2)
(State of California Energy Commision, 2008), in the pres-
ent work it was obtained by digitizing a low irradiance I–V
curve from the module datasheet. The additional informa-
tion for the six modules considered here is given in Table 7.
It is important to note that values of the original five
parameters remain unchanged when the new parameters
are introduced. The effect of the five modifications on the
maximum power prediction accuracy of the six modules
considered here is shown in Table 8.
Only the seven parameters model (IL and a), henceforth
referred to as only the seven parameter model, provided
consistent and significant improvement for all PV modules.
Therefore, the best modifications to the translation equa-
tions of IL and a are given by Eqs. (21) and (22). The values
of the two new parameters (m and n) required in the new
seven parameters model for the PV modules considered
in the present study are listed in Table 9.
m
I L ¼ ðS=S ref Þ ðI L;ref þ lIsc ðT c  T ref ÞÞ ð21Þ
n
a ¼ aref ðT c =T ref Þ ð22Þ
6.1. Comparison of proposed seven parameters model with
other models
A comparison of the maximum power prediction accu-
racy between the five parameter model using the parame-
ters estimated using the new proposed methodology and
the seven parameter model is shown in Fig. 6.

Fig. 7. Comparison of accuracy of seven parameter model for crystalline silicon technology.
 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42 39

Fig. 8. Comparison of accuracy of seven parameter model for thin film technology.

Fig. 9. Comparison of maximum power prediction accuracy for seven parameter model. (Module 1 = AP-110, Module 2 = S-36, Module 3 = KC-40T,
Module 4 = MST-43LV, Module 5 = ST-35, Module 6 = PVL-124.)

The introduction of the two new parameters in the
model resulted in marked improvement in the maximum
power prediction accuracy of all the considered modules.
A comparison of the new seven parameter model with
the three models from literature considered here is shown
in Figs. 7–9. Figs. 6–9 show that the introduction of the
new parameters results in much more accurate results than
the five parameter model.
In order to confirm the accuracy of the seven parameter
model, the model was used to predict the maximum power
of two new PV modules. These included Sanyo’s propriety
HIT technology module, HIT-195, which combines mono-
crystalline silicon and amorphous silicon in a layered
arrangement and Suntech’s STP090T amorphous silicon
module. The RMSE in the maximum power prediction
for HIT-195 and STP090T modules using the seven
parameter model was found to be 0.91% and 2.32%
respectively.
7. Conclusions
This paper presents a novel methodology for estimating
the model parameters for the five parameter PV model and
a new seven parameter PV model based on the results of a
sensitivity analysis.
 Comparison of the five parameter model using the new
parameters estimation methodology with other electrical
models shows that the new methodology provides com-
parable results to other models and better results than
40 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42

the five parameter model using the parameters estimated determined by setting the ideality factor A equal to 1 in the
by the methodology of Villalva et al. (2009) for crystal- following equation:
line silicon as well as thin film PV cell types.   
 The results of the sensitivity analysis show that the five 1 kT c;ref NCS I mp;ref
Rs;max ¼ ln 1  þ V oc;ref  V mp;ref
parameter model is several orders of magnitude more I mp;ref q I sc;ref
sensitive to the parameters, IL and a, than the remaining ðA:6Þ
three parameters.
 The proposed seven parameter model shows improve- In the four parameter model, the parameters c and Rs
ment in the prediction accuracy for all modules consid- are assumed to be independent of operating conditions
ered in the study. There is a need to test the proposed and are assumed constant while IL and Io are scaled to
seven parameter model with experimental data to fur- the operating conditions using Eqs. (3) and (4) respectively.
ther validate its accuracy.
A.2. Remaining equations of Sandia Labs model

Acknowledgements In the Sandia Labs model (King et al., 2004), Eqs.

(A.7)–(A.14) are used together with Eqs. (12)–(18) to calcu-
The authors would like to acknowledge the support of late the electrical performance of a PV device.
King Fahd University of Petroleum and Minerals through
the Center for Clean Water and Clean Energy at KFUPM FF ¼ P mp =ðI sc V oc Þ ðA:7Þ
(DSR Project # R6-DMN-08) and MIT. The authors Ee ¼ I sc =½I sc0 f1 þ aisc ðT c  T 0 Þg ðA:8Þ
would also like to thank Prof. S.M. Zubair and Prof. An- dðT c Þ ¼ nkðT c þ 273:15Þ=q ðA:9Þ
war K. Sheikh for their helpful comments during the devel-
opment of this work. Eb ¼ Edni cosðAOIÞ ðA:10Þ
2 3
f1 ðAM a Þ ¼ a0 þ a1 ðAM a Þ þ a2 ðAM a Þ þ a3 ðAM a Þ
Appendix A þ a4 ðAM a Þ
4
ðA:11Þ
2 3
A.1. Parameter estimation methodology for the four f2 ðAOIÞ ¼ b0 þ b1 ðAOIÞ þ b2 ðAOIÞ þ b3 ðAOIÞ
parameter model 4 5
þ b4 ðAOIÞ þ b5 ðAOIÞ ðA:12Þ
aþb:WS
Eqs. (A.1)–(A.6) are used to determine the parameters T m ¼ E:fe g þ T amb ðA:13Þ
for the four parameter model (Townsend, 1989). Writing E
Eq. (11) for the short circuit, open circuit and maximum Tc ¼ Tm þ DT ðA:14Þ
E0
power conditions results in the following equations:
I L;ref ¼ I sc;ref ðA:1Þ A.3. Parameters estimation methodology for Villalva et al.
  model
qV oc;ref
I 0;ref ¼ I sc;ref exp ðA:2Þ
kcT c;ref
In the Villalva model (Villalva et al., 2009), an iterative
qðV mp;ref þ I mp;ref Rs  V oc;ref Þ scheme is used to calculate the model parameters at the ref-
c ¼ A:NCS ¼ ðA:3Þ
kT c;ref lnð1  ðI mp;ref =I sc;ref ÞÞ erence condition. To ease the process of finding the param-
To obtain the fourth equation, Eq. (A.2) is first written eters, the parameter aref is assumed at the beginning of the
in the form of Eq. (A.4) and then the derivative with process and remains unchanged during the algorithm. The
respect to temperature is taken resulting in Eq. (A.5). parameter Io is found using Eq. (A.15). Next, in the itera-
tive algorithm, the resistance Rs is set to an initial values
  of 0 and Rsh is set to an initial value of Rsh,min given by
kT c;ref c I sc;ref
V oc;ref ¼ ln ðA:4Þ Eq. (A.16).
q I 0;ref
     I sc;ref
ck I sc;ref T c;ref lisc qEg I o;ref ¼ ðA:15Þ
lvoc ¼ ln þ  3þ ðA:5Þ expðV oc;ref =aref ÞÞ  1
q I 0;ref I sc;ref AkT c;ref
V mp;ref V oc;ref  V mp
Rsh;min ¼  ðA:16Þ
Here, Eg is the material band-gap energy, lVoc and lIsc are I sc;ref  I mp I mp
the temperature coefficients to Voc and Isc and A = c/NCS
is the diode ideality factor. The current IL is calculated using Eq. (A.17) and the resis-
Eq. (A.5) is solved using a bisection method to deter- tance Rsh is calculated using Eq. (A.18).
mine Rs where the effect of Rs is introduced in the equation
through Eq. (A.3) for c. The minimum value of Rs for the Rsh þ Rs
I L;ref ¼ I sc;ref ðA:17Þ
bisection method is set to be 0 while its maximum value is Rsh
 M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42 41

Table B.1
Parameters for the four parameter model at STC.
Parameter AP-110 S-36 KC-40T ST-36 MST-43LV PVL-124
IL,ref (A) 7.5 2.3 2.65 2.68 3.3 5.1
Io,ref (A) 3.0176e6 1.3873e7 7.1599e8 1.9691e4 1.1114e4 9.9250e5
cref 54.7192 50.1118 48.4726 93.6517 85.8243 150.72
Rs,ref (X) 0.1545 0.4934 0.3548 1.1070 1.0697 1.35

Table B.2
Parameters for Villalva et al. model at STC.
Parameter AP-110 S-36 KC-40T ST-36 MST-43LV PVL-124
IL,ref (A) 7.516 2.3015 2.6511 2.711 3.3714 5.1838
Io,ref (A) 2.50335e7 4.289e8 3.8505e8 2.181e7 2.0643e18 2.5232
aref (V) 1.2018 1.2018 1.2018 1.4021 0.5341 0.6677
Rs,ref (X) 0.0707 0.5 0.3368 0.5 0.5 0.5
Rsh,ref (X) 33.079 769.4 807.9666 43.1282 23.092 30.43

Table B.3
Parameters for the Sandia Labs PV model.
Parameter AP-110 S-36 KC-40T ST-36 MST-43LV PVL-124
NCS 36 36 36 42 16 20
NP 1 1 1 1 4 1
Isc0 (A) 7.5000 2.3000 2.6500 2.6800 3.5300 5.1000
Voc0 (V) 20.7000 21.4000 21.7000 22.9000 22.2700 42.0000
Imp0 (A) 6.6000 2.1800 2.4800 2.2800 2.5700 4.1300
Vmp0 (V) 16.7000 16.5000 17.4000 15.8000 16.6700 30.0000
aisc (A/°C) 3.3000E04 4.50E04 4.00E04 1.3000E05 6.600E04 1.00E03
aimp (A/°C) 4.2000E04 1.40E04 1.40E04 4.5000E04 9.500E04 1.00E03
C0 0.9970 0.9890 1.0060 0.9720 1.0227 1.0960
C1 0.0030 0.0110 0.0060 0.0280 0.0227 0.0960
bvoc (V/°C) 8.4000E02 7.60E02 8.21E02 9.0600E02 1.050E01 1.60E01
mbvoc 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
bvmp (V/°C) 8.4000E02 7.60E02 8.40E02 7.4400E02 8.700E02 9.30E02
mbvmp 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
N 1.4040 1.3700 1.3700 1.7520 3.5480 3.7700
C2 0.1775 0.1170 0.1170 0.5088 0.2203 1.1416
C3 11.0652 11.0820 11.0820 2.9540 4.0595 2.8912
a0 9.1800E01 9.22E01 9.22E01 9.2100E01 8.937E01 1.05E+00
a1 6.8713E02 7.09E02 7.09E02 7.1815E02 1.416E01 8.21E04
a2 1.0438E02 1.43E02 1.43E02 1.4619E02 5.539E02 2.59E02
a3 7.2504E04 1.17E03 1.17E03 1.2500E03 5.613E03 3.17E03
a4 2.0182E05 3.37E05 3.37E05 3.7406E05 1.770E04 1.10E04
b0 1.0000E+00 1.00E+00 1.00E+00 1.0000E+00 1.000E+00 1.00E+00
b1 2.4380E03 2.47E03 2.44E03 2.4380E03 2.438E03 5.02E03
b2 3.1030E04 3.15E04 3.10E04 3.1030E04 3.103E04 5.84E04
b3 1.2460E05 1.26E05 1.25E05 1.2460E05 1.246E05 2.30E05
b4 2.1120E07 2.14E07 2.11E07 2.1120E07 2.112E07 3.83E07
b5 1.3590E09 1.37E09 1.36E09 1.3590E09 1.359E09 2.31E09
DT (°C) 3.0000 3.0000 3.0000 3.0000 3.0000 1.0000
Fd 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
A 3.5600 3.5370 3.5600 3.4700 3.4700 2.8100
B 0.0750 0.0721 0.0750 0.0594 0.0594 0.0455
C4 0.9840 0.9866 0.9866 0.9829 1.0124 1.0440
C5 0.0160 0.0134 0.0134 0.0171 0.0124 0.0440
Ix0 (A) 7.3000 2.2600 2.6100 2.5900 3.2100 4.7200
Ixx0 (A) 4.6000 1.5300 1.7400 1.5800 1.7700 2.9000
C6 1.1230 1.1183 1.1183 1.0450 1.1185 1.1300
C7 0.1230 0.1183 0.1183 0.0450 0.1185 0.1300
42
V mp;ref ðV mp;ref þ I mp;ref Rs;ref Þ
Rsh;ref ¼ n h
ðV mp;ref þI mp;ref Rs;ref Þ
V mp;ref I L;ref  V mp;ref I o;ref exp
aref
To calculate the maximum power, Eq. (1) is solved for
current for the entire range of voltages from 0 to the open
circuit voltage Voc and the maximum power is found by
multiplying the currents and voltages and searching for
the maximum value. If the error of the predicted power
from the experimental value is within the specified toler-
ance, the solution terminates otherwise the value for Rs is
incremented and the process is repeated. Once the param-
eters at the reference condition are known, they can be
translated on to the operating condition using Eqs. (2)–
(6). Villalva et al. suggested that using Eq. (A.19) to trans-
late Io can improve the accuracy of the model.
I sc;ref þ lisc ðT c  T ref Þ
Io ¼
expððV oc;ref þ lvoc ðT c  T ref ÞÞ=aÞ  1
Appendix B
See Tables B.1–B.3.
References
Almonacid, F., Rus, C., Pérez, P.J., Hontoria, L., 2009. Estimation of the
energy of a PV generator using artificial neural network. Renewable
Energy 34, 2743–2750.
Boyd, M.T., Klein, S.a., Reindl, D.T., Dougherty, B.P., 2011. Evaluation
and validation of equivalent circuit photovoltaic solar cell performance
models. Journal of Solar Energy Engineering 133, 021005.
Carrero, C., Ramı́rez, D., Rodrı́guez, J., Platero, C.a., 2011. Accurate and
fast convergence method for parameter estimation of PV generators
based on three main points of the I–V curve. Renewable Energy 36,
2972–2977.
De Soto, W., Klein, S.A., Beckman, W.A., 2006. Improvement and
validation of a model for photovoltaic array performance. Solar
Energy 80, 78–88.
Duffie, J.A., Beckman, W.A., 1991. Solar Engineering of Thermal
Processes, second ed. John Wiley & Sons, Inc., New York.
Hishikawa, Y., Imura, Y., Oshiro, T., 2000. Irradiance-dependence and
translation of the I–V characteristics of crystalline silicon solar cells.
M.U. Siddiqui et al. / Solar Energy 90 (2013) 29–42
In: Photovoltaic Specialists Conference (PVSC), 2000 28th IEEE, pp.
i o
þ V mp;ref I o;ref  P max;e 1464–1467.
Ikegami, T., Maezono, T., Nakanishi, F., Yamagata, Y., Ebihara, K.,
ðA:18Þ 2001. Estimation of equivalent circuit parameters of PV module and its
application to optimal operation of PV system. Solar Energy Materials
and Solar Cells 67, 389–395.
Ishaque, K., Salam, Z., 2011. An improved modeling method to determine
the model parameters of photovoltaic (PV) modules using differential
evolution (DE). Solar Energy 85, 2349–2359.
Ishaque, K., Salam, Z., Taheri, H., Shamsudin, A., 2011. A critical
evaluation of EA computational methods for Photovoltaic cell
parameter extraction based on two diode model. Solar Energy 85,
1768–1779.
Jervase, J., Bourdoucen, H., 2001. Solar cell parameter extraction using
genetic algorithms. Measurement Science and Technology Science and
Technology 12, 1922–1925.
King, D.L., Boyson, W.E., Kratochvil, J.A., 2004. Photovoltaic Array
Performance Model, System. Sandia National Laboratories, Albu-
querque, New Mexico.
Marion, B., Rummel, S., Anderberg, A., 2004. Current–voltage curve
translation by bilinear interpolation. Progress in Photovoltaics:
ðA:19Þ
Research and Applications 12, 593–607.
Masi, M., Fogliani, S., Carra, S., 1999. Sensitivity analysis on indium
phosphide liquid encapsulated Czochralski growth. Crystal Research
and Technology 34, 1157–1167.
Moldovan, N., Picos, R., Garcia-Moreno, E., 2009. Parameter extraction
of a solar cell compact model using genetic algorithms. In: Spanish
Conference On Electron Devices, CDE 2009. IEEE, pp. 379–382.
Nelder, J.A., Mead, R., 1965. A simplex method for function minimiza-
tion. The Computer Journal 7, 308–311.
Ortiz-Rivera, E.I., Peng, F.Z., 2005. Analytical model for a photovoltaic
module using the electrical characteristics provided by the manufac-
turer data sheet. In: Power Electronics Specialists Conference, 2005.
PESC ’05. IEEE 36th, pp. 2087–2091.
State of California Energy Commision, 2008. Guidelines for California’s
Solar Electric Incentive Programs (Senate Bill 1), second ed..
Townsend, T.U., 1989. MS thesis, A Method for Predicting the Long-
Term Performance of Directly-Coupled Photovoltaic Systems. Uni-
versity of Wisconsin, Madison.
Valerio, L.B., Orioli, A., Ciulla, G., Gangi, A.D., 2010. An improved five-
parameter model for photovoltaic modules. Solar Energy Materials
and Solar Cells 94, 1358–1370.
Villalva, M.G., Gazoli, J.R., Filho, E.R., 2009. Comprehensive approach
to modeling and simulation of photovoltaic arrays. IEEE Transactions
on Power Electronics 24, 1198–1208.
Zagrouba, M., Sellami, A., Bouaı̈cha, M., Ksouri, M., 2010. Identification
of PV solar cells and modules parameters using the genetic algorithms:
application to maximum power extraction. Solar Energy 84, 860–866.