You are on page 1of 14

Journal Pre-proofs

Original article

Accurate method for PV solar cells and modules parameters extraction using
I-V curves

Ali Kareem Abdulrazzaq, György Bognár, Balázs Plesz

PII: S1018-3639(20)30272-5
DOI: https://doi.org/10.1016/j.jksues.2020.07.008
Reference: JKSUES 418

To appear in: Journal of King Saud University - Engineering Sci‐


ences

Received Date: 4 March 2020


Accepted Date: 17 July 2020

Please cite this article as: Abdulrazzaq, A.K., Bognár, G., Plesz, B., Accurate method for PV solar cells and
modules parameters extraction using I-V curves, Journal of King Saud University - Engineering Sciences (2020),
doi: https://doi.org/10.1016/j.jksues.2020.07.008

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover
page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version
will undergo additional copyediting, typesetting and review before it is published in its final form, but we are
providing this version to give early visibility of the article. Please note that, during the production process, errors
may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2020 Production and hosting by Elsevier B.V. on behalf of King Saud University.
Accurate method for PV solar cells and modules parameters extraction
using I-V curves

ARTICLE INFO ABSTRACT


Keywords: The main contribution of this paper is proposing a new approach for retrieving the five parameters
Parameters extraction of the single diode equivalent model (SDM) of photovoltaic cells/modules including the series and
Single diode model shunt resistances. The least square method is used as an error minimization technique for fitting
Solar cell the non-linear transcendental model equation of the solar panel to the measured I-V characteristics.
Curve fitting Newton Raphson method is applied to solve the system of five non-linear equations which represent
Numerical method the error of each parameter. Initial guess values are calculated with an optimised algorithm depending
I-V characteristics. on information extracted from the same measured data. MATLAB programming script was used in
all implementation steps. This approach is useful for a wide variety of applications where the five
SDM parameters have to be determined from the measured curves, particularly, for self-fabricated
cells/modules or in case of no available datasheet. One of the strengths of this method is the higher
level of accuracy because of the absence of mathematical simplifications and physical assumptions.
The method was validated on different types of PV devices, including a crystalline silicon cell, a
polycrystalline module, and an amorphous module by using measurement data obtained at a wide
range of solar irradiance conditions and temperatures.

1. Introduction glected. Figure 1 shows the single diode equivalent model


circuit.
One of the essential aspects of designing PV system is to
have a mathematical model to simulate the I-V characteris- Rs
tics of the system under different environmental conditions.
It is also required to determine the best operating point of ID Ish I
the PV module and estimate the amount of produced energy
(Menaga and Sankaranarayanan, 2020; Homadi et al., 2020).
For modelling a PV cell/module, researchers use different Iph D Rsh V
approaches each with different objectives including the com-
plexity level and the information to be predicted. Empiri-
cal models were constructed by several research works (Das,
2011, 2013b; Akbaba, 1995). However these models do not
include any environmental dependence nor physically-based ❋✐❣✉r❡ ✶✿ ❚❤❡ s✐♥❣❧❡✲❞✐♦❞❡ ♠♦❞❡❧ ❡q✉✐✈❛❧❡♥t ❝✐r❝✉✐t ♦❢ ❛ P❱

parameters and they are generally used for maximum power ❝❡❧❧✳

point (MPP) or fill factor (FF) calculation. For modelling


the full behaviour of a PV device based on physical parame-
ters typically there are two different concepts. The first one Generally, most solar cells manufacturer provide a panel
is the double diode model (DDM) with two diodes used for datasheet that includes information like the short circuit cur-
simulating the diffusion and the recombination of minority rent, the open circuit voltage, the number of cells and their
carriers phenomena in the PV solar cell separately (Chan and connection configuration, as well as the maximum power
Phang, 1987; Hejri et al., 2014b; Sandrolini et al., 2010; Lun point. However, this information measured at standard test-
et al., 2015). However, to avoid the complexity due to more ing conditions (STC) is not enough to build a five parame-
unknowns in a non-linear implicit equation, researchers are ters based accurate power prediction model. Several tech-
tending to use the second concept incorporating only a single niques have been employed by researchers to extract these
diode (Di Piazza et al., 2017a,b; Ghani et al., 2014; Rhouma parameters, each with different accuracy. Typically, these
et al., 2017; Farivar and Asaei, 2010; Toledo and Blanes, techniques are divided into analytical and numerical models.
2016; Bonkoungou et al., 2013; Ayodele et al., 2016; Az- One aspect of the analytical models is to determine the un-
zouzi et al., 2016; Bastidas-Rodriguez et al., 2014). The sin- known parameters from the data provided in the datasheet by
gle diode model (SDM) meant to simulate both mentioned applying some simplification producing a set of non-linear
physical phenomena with a single ideality factor. Although, equations (Di Piazza et al., 2017a; Farivar and Asaei, 2010;
the DDM is more accurate at low voltages under dark con- Toledo and Blanes, 2016; Jadli et al., 2018; Hosseini et al.,
ditions, where the current is very small, at maximum power 2018; Mehta et al., 2019; Silva et al., 2016; Hejri et al., 2014a;
point the voltage is sufficiently high and the photo-current Silva et al., 2017; Cubas et al., 2014; Das, 2013a; Saetre
dominates. Thus, the effect of the second diode (the one et al., 2011; Et-torabi et al., 2017; Bourdoucen and Gastli,
related to the recombination of minority carriers) can be ne- 2007; Ibrahim and Anani, 2017; Vergura, 2016; Can and
ORCID (s):
Ickilli, 2014; Sera et al., 2007; Brano and Ciulla, 2013; Chikh

❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier Page 1 of 13


❆❝❝✉r❛t❡ ♠❡t❤♦❞ ❢♦r P❱ s♦❧❛r ❝❡❧❧s ❛♥❞ ♠♦❞✉❧❡s ♣❛r❛♠❡t❡rs ❡①tr❛❝t✐♦♥ ✉s✐♥❣ ■✲❱ ❝✉r✈❡s

and Chandra, 2015). proach of fitting the model and extracting the five charac-
In (Di Piazza et al., 2017a), the proposed numerical anal- teristics parameters is considered to be accurate because of
ysis is based on the information provided by the manufac- using a full range I-V characteristics, robust mathematical al-
turers’ datasheets under STC, nominal operating cell tem- gorithm, and optimised steps for the parameters initial guess
perature (NOTC), and low irradiance conditions (LIC). This values.
technique seems promising from the point of using low com-
putational cost; however, it considerably sacrifices the accu- 2. PV Solar Cell Single Diode Model
racy of the final model. Therefore, it is not sufficient for ap-
plication where accurate power prediction is required. Some Under illumination, the solar cell single diode equivalent
other approaches rely on measurements at a limited num- model is described by the implicit form given as (Chegaar
ber of characteristic points which is again a disadvantageous et al., 2001):
from the point of view of overall accuracy (Toledo and Blanes, (𝑉 +𝐼𝑅𝑠 )
𝑉 + 𝐼𝑅𝑠
2016). The authors in (Jadli et al., 2018) propose a two- 𝐼 = 𝐼𝑝ℎ − 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) − . (1)
𝑅𝑠ℎ
step model based on the manufacturer datasheet. First, the
simulated annealing algorithm is used to estimate two of the The five parameters that appear in the SDM model equa-
unknown electrical model parameters, namely, series resis- tion characterize the PV module at a specific meteorological
tance and ideality factor. Second, analytical formulations are condition. These parameters are the photo-generated current
used to evaluate the remaining parameters. Also depending (𝐼𝑝ℎ ), reverse saturation current (𝐼𝑜 ), the ideality factor of the
on the PV module datasheet, the method presented in (Hos- PV cell (𝑛), cell series resistance (𝑅𝑠 ), and shunt resistance
seini et al., 2018) uses analytical formulations to calculate (𝑅𝑠ℎ ). For a panel consisting of multiple cells connected in
the five SDM parameters as a function of the environmen- series and/or in parallel, the number of these cells have to
tal conditions, including the irradiance spectrum. The pro- considered and can be given as (Ma et al., 2014; Vijayaku-
posed set of equations are solved using numerical iterative mari, 2020; Ashfaq et al., 2017):
method. The research work presented in (Mehta et al., 2019)
𝑉 + 𝐼 𝑅 𝑉 𝐼
uses the datasheet information to derive four equations. An- 𝑁𝑠 𝑁𝑝 𝑠 + 𝑅
𝑁𝑠 𝑁𝑝 𝑠
other equation was obtained using the concept of the area 𝐼 = 𝑁𝑝 𝐼𝑝ℎ −𝑁𝑝 𝐼𝑜 (𝑒 𝑛𝑉𝑇 −1)−𝑁𝑝 , (2)
under a measured I-V curve to complete the set of five equa- 𝑅𝑠ℎ
tions. Then, an iterative method was employed to solve these where 𝑁𝑠 and 𝑁𝑝 are the numbers of series and parallel con-
equations and extract the value of the five parameters. An- nected cells, respectively. In the rest of our discussion, we
other approach that uses a two-step method was proposed in will refer to Equation 1 as the SDM model. All the math-
(Silva et al., 2016). The authors use analytical expressions ematical formulations that follow are based on Equation 1.
based on the datasheet information to determine three out However, applying the model for a module rather than a cell
of five parameters. The second step is represented by fitting requires including the number of cells in the corresponding
the model results to the P-V or I-V curves provided by the position. The complete I-V characteristics of a PV cell is
datasheet while scanning the other two parameters within a shown in Figure 2 for a given irradiance and temperature.
predefined range. The best fitting will be considered to ex- At low voltages, the current through the diode is very low,
tract the five parameters.
Numerical models are required to be used because of the
implicit nature of the SDM equation (Sandrolini et al., 2010;
Ghani et al., 2014; Ayodele et al., 2016; Kim et al., 2011;
Ghani and Duke, 2011; Ghani et al., 2013; Appelbaum and
Peled, 2014). In (Ghani and Duke, 2011) the authors pre-
sented a numerical model using the Newton-Raphson method
to solve the SDM equation at the maximum power point to
determine the values of series and shunt resistances. This
technique is considered to solve the model equation repre-
sented in an explicit form using the Lambert W-function.
The authors in (Appelbaum and Peled, 2014) have used a
slightly modified version of the Newton-Raphson method to
solve a model reduced to three parameters instead of five, by
using some algebraic manipulations. ❋✐❣✉r❡ ✷✿ ❚❤❡ P❱ s♦❧❛r ❝❡❧❧ ■✲❱ ❝❤❛r❛❝t❡r✐st✐❝ ❝✉r✈❡✳
In the presented research work, the implicit non-linear
equation has been considered for computing the SDM model
parameters. A computational intelligence approach is pro- thus the output current is high and very close to the photo-
posed to solve this implicit equation. The root mean square generated current. With increasing the voltage, the output
algorithm is used for error minimization and fitting the model current decreases more rapidly. Important points to be high-
equation to the measured I-V characteristic curves. This ap- lighted on this curve are the short circuit current (𝐼𝑠𝑐 ) and
the open circuit voltage (𝑉𝑜𝑐 ). 𝐼𝑠𝑐 is the maximum current

❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier P❛❣❡ ✷ ♦❢ ✶✸


❆❝❝✉r❛t❡ ♠❡t❤♦❞ ❢♦r P❱ s♦❧❛r ❝❡❧❧s ❛♥❞ ♠♦❞✉❧❡s ♣❛r❛♠❡t❡rs ❡①tr❛❝t✐♦♥ ✉s✐♥❣ ■✲❱ ❝✉r✈❡s

that one can get from a PV module which is measured in Because we are interested in the slope of the curve at the
short circuit condition (𝑉 = 0). 𝑉𝑜𝑐 is the maximum voltage short circuit region (when 𝑉 ≃ 0); therefore, we can do the
𝑉 +𝐼𝑅𝑠 𝐼𝑠𝑐 𝑅𝑠
a PV module can produce, measured at zero output current
following approximation: 𝑒 𝑛𝑉𝑇 = 𝑒 𝑛𝑉𝑇 ≃ 0, where the
condition (𝐼 = 0).
term (𝐼𝑠𝑐 𝑅𝑠 ) is minimal compared to (𝑛𝑉𝑇 ). It’s also known
The explicit form of the SDM can be formulated with the
that 𝑅𝑠 is much lower than 𝑅𝑠ℎ (usually different in orders of
help of Lambert W-function as shown in Equation 3 (Abdul- 𝑅
razzaq, 2018). We used this form to plot new I-V character- magnitude), that is, 𝑅 𝑠 ≃ 0. This gives a simple expression
𝑠ℎ
istics curve after substituting the extracted parameters to be to find the initial value of the shunt resistance defined by the
compared with the measured curves. negative slope of the I-V curve near the short circuit region as
shown in Equation 6. This slope could be determined from
𝑅𝑠ℎ (𝐼𝑝ℎ + 𝐼𝑜 ) the measured I-V curve.
𝐼=
𝑅𝑠ℎ + 𝑅𝑠 𝑑𝑉
𝑅𝑠ℎ (𝑉 +𝑅𝑠ℎ (𝐼𝑝ℎ +𝐼𝑜 )) 𝑅𝑠ℎ = − | . (6)
𝑛𝑉 𝑅𝐼 𝑅𝑠ℎ 𝑑𝐼 𝑆.𝐶.
− 𝑇𝑊( 𝑠 𝑜 𝑒 𝑛𝑉𝑇 (𝑅𝑠ℎ +𝑅𝑠 ) )
𝑅𝑠 𝑛𝑉𝑇 𝑅𝑠ℎ + 𝑅𝑠 𝑅𝑠 significantly affect the slope near the open voltage point
𝑉 (𝑉𝑜𝑐 ). Back to Equation 1, by substituting 𝐼 = 0 and 𝑉 =
− , (3)
𝑅𝑠ℎ + 𝑅𝑠 𝑉𝑜𝑐 , we get the following expression:
where, 𝑊 is the principal branch of Lambert W-function.
𝑉𝑜𝑐
𝑉𝑜𝑐
0 = 𝐼𝑝ℎ − 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) − ( ). (7)
3. Algorithm for Estimating the Parameters 𝑅𝑠ℎ
Initial Guess Values 𝑉
The term 𝑅𝑜𝑐 is very small comparing with 𝐼𝑝ℎ and the ex-
𝑠ℎ
Since we restrict ourselves to solve the non-linear model ponential term is much higher than 1. Hence, we have the
shown in Equation 1, the initial guess values require spe- following expression:
cial care because of their direct effect on the convergence of
the numerical solving method and how fast the solution can 𝑉 +𝐼𝑅𝑠 𝑉𝑜𝑐

be achieved. A particular case in this paper is considered by 𝐼𝑜 𝑒 𝑛𝑉𝑇 = 𝐼𝑜 𝑒 𝑛𝑉𝑇 = 𝐼𝑝ℎ . (8)
having only measured full I-V characteristic curves of the PV
Substituting this in Equation 5, we get the following:
modules. Therefore, the following formulation is proposed
to determine the starting values for the five parameters de- 𝑑𝐼 1 𝐼𝑝ℎ 1 𝐼𝑝ℎ
pending on the measured I-V curves. (1 + 𝑅𝑠 ( + )) = −( + ). (9)
𝑑𝑉 𝑅𝑠ℎ 𝑛𝑉𝑇 𝑅𝑠ℎ 𝑛𝑉𝑇
The formulation of this algorithm is based on the solar
cell equivalent circuit shown in Figure 1 and its mathemati-
cal representation shown in Equation 1. At short circuit con- 1 𝐼𝑝ℎ 𝑑𝑉 𝐼𝑝ℎ 1
1 + 𝑅𝑠 ( + )=− ( + ). (10)
dition, for small 𝑅𝑠 values, 𝐼𝑠𝑐 is approximately equal to 𝐼𝑝ℎ . 𝑅𝑠ℎ 𝑛𝑉𝑇 𝑑𝐼 𝑛𝑉𝑇 𝑅𝑠ℎ
Therefore, we will consider that the photo-current is equal to
the measured current value corresponding to the closest pos-
𝐼𝑝ℎ 1 𝑑𝑉 𝐼𝑝ℎ 1
itive voltage to the zero value. 𝑉𝑜𝑐 is the maximum output 𝑅𝑠 ( + )=− ( + ) − 1. (11)
voltage measured with zero external output current (𝐼 = 0) 𝑛𝑉𝑇 𝑅𝑠ℎ 𝑑𝐼 𝑛𝑉𝑇 𝑅𝑠ℎ
of the PV module. The open circuit voltage in this respect is
considered to be corresponding to the lowest positive value 𝑑𝑉 1
of the measured current. Near 𝑉 = 0, the shunt resistance 𝑅𝑠 = − −( 𝐼𝑝ℎ
). (12)
𝑑𝐼 1

can show a direct linear influence on the I-V curve. This 𝑅𝑠ℎ 𝑛𝑉𝑇
concept is used to estimate the value of 𝑅𝑠ℎ by finding the
1
model equation output current derivative with respect to the Another approximation comes from the fact that 𝑅𝑠ℎ
<<
output voltage: 𝐼𝑝ℎ
𝑛𝑉𝑇
. That is, we have an expression for the series resistance
𝑑𝐼 𝐼 ( 𝑉 +𝐼𝑅𝑠 𝑑𝐼
) as a function of the ideality factor given as
= − 𝑜 𝑒 𝑛𝑉𝑇 (1 + 𝑅𝑠 )
𝑑𝑉 𝑛𝑉𝑇 𝑑𝑉
𝑑𝑉 𝑛𝑉
1 𝑅 𝑑𝐼 𝑅𝑠 = − − 𝑇. (13)
− − 𝑠 . (4) 𝑑𝐼 𝐼𝑝ℎ
𝑅𝑠ℎ 𝑅𝑠ℎ 𝑑𝑉
An expression to estimate the value of the reverse satu-
𝑅 𝐼 𝑅 𝑉 +𝐼𝑅𝑠 ration current (𝐼𝑜 ) is deduced from the Equations 7 and 8,
𝑑𝐼
(1 + 𝑠 + 𝑜 𝑠 𝑒 𝑛𝑉𝑇 ) = given as
𝑑𝑉 𝑅𝑠ℎ 𝑛𝑉𝑇
+𝐼𝑅𝑠
𝐼𝑜 𝑉 𝑛𝑉 1 𝐼𝑝ℎ
− 𝑒 𝑇 − . (5) 𝐼0 = − 𝑉𝑜𝑐
. (14)
𝑛𝑉𝑇 𝑅𝑠ℎ 𝑛𝑉𝑇
𝑒
❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier P❛❣❡ ✸ ♦❢ ✶✸
❆❝❝✉r❛t❡ ♠❡t❤♦❞ ❢♦r P❱ s♦❧❛r ❝❡❧❧s ❛♥❞ ♠♦❞✉❧❡s ♣❛r❛♠❡t❡rs ❡①tr❛❝t✐♦♥ ✉s✐♥❣ ■✲❱ ❝✉r✈❡s

The ideality factor generally has a value between 1 and PV solar panel is normally easy to be measured for a given
2. It is mainly affecting the model near the maximum power environmental condition. The data should cover the whole
point (Yordanov et al., 2013). Therefore, we used Equation range of the PV panel normal operation, between the short
1 at the maximum power point for calculating 𝑛 as shown and open circuit conditions. To calculate the five parameters
below: the SDM equation is fitted to these I-V curves. The mea-
(𝑉𝑚 +𝐼𝑚 𝑅𝑠 )
𝑉𝑚 + 𝐼𝑚 𝑅𝑠 sured I-V curve should satisfy the model in the summation
𝐼𝑚 = 𝐼𝑝ℎ − 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) − . (15) form given as follows:
𝑅𝑠ℎ
𝑁 [
The values of 𝑉𝑚 and 𝐼𝑚 are determined by scanning the I-V ∑ 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑉 + 𝐼𝑖 𝑅𝑠 ]
0= 𝐼𝑖 − 𝐼𝑝ℎ + 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) + 𝑖 ,
curve and calculating the power at each I-V point. By simple 𝑅𝑠ℎ
𝑖=1
MATLAB script, this concept is implemented for tracking (16)
for the maximum value of the power that reveals the values
of 𝑉𝑚 and 𝐼𝑚 . To get the initial estimation values of three un- where N is the number of the measured points of the I-V
known parameters (𝑅𝑠 , 𝐼𝑜 , and 𝑛) we have to solve numeri- curve. When solving Equation 16, the results indicate the
cally the three non-linear Equations 13, 14, and 15. Calculat- deviation of the current values from the actual values of the
ing the initial guess values, including solving the three non- five parameters (𝐼𝑝ℎ , 𝐼𝑜 , 𝑅𝑠ℎ , 𝑅𝑠 , and 𝑛).
linear equations, has been implemented by another MAT- Solving the equation and compute the accurate values of
LAB script. The flowchart shown in Figure 3 illustrates the the parameters can only be done if the error term is zero or
code implementation for calculating the initial values for the at acceptable minimum value. The error term can be mini-
SDM parameters. mized using the least square method (Marquardt, 1963). The
sum of the squares of the error (𝐸) is given by
Start 𝑁 [
∑ 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑉𝑖 + 𝐼𝑖 𝑅𝑠 ]2
𝐸= 𝐼𝑖 −𝐼𝑝ℎ +𝐼𝑜 (𝑒 𝑛𝑉𝑇 −1)+ . (17)
𝑖=1
𝑅𝑠ℎ
Enter full measured Performing partial differentiation with respect to 𝐼𝑝ℎ , 𝐼𝑜 , 𝑅𝑠ℎ , 𝑅𝑠 ,
data (I-V Curve)
and 𝑛 and equating to zero, we get the following expressions,
respectively:

Iph = Highest positive current value


with positive voltage 𝜕𝐸 ∑𝑁
Voc = Highest positive voltage value =−2
with positive current
𝜕𝐼𝑝ℎ 𝑖=1
[ 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑉 + 𝐼𝑖 𝑅𝑠 ]
𝐼𝑖 − 𝐼𝑝ℎ + 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) + 𝑖
𝑅𝑠ℎ
Calculate Rs, Io, and n using the =0 → 𝑓1 . (18)
introduced three non-linear equation
using (fsolve) function in MATLAB

𝜕𝐸 ∑ 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑁
=2 (𝑒 𝑛𝑉𝑇 − 1)
𝜕𝐼𝑜 𝑖=1
Export the estimated
initial values to PV model
[ 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑉 + 𝐼𝑖 𝑅𝑠 ]
𝐼𝑖 − 𝐼𝑝ℎ + 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) + 𝑖
𝑅𝑠ℎ
=0 → 𝑓2 . (19)
End

❋✐❣✉r❡ ✸✿ ❋❧♦✇❝❤❛rt s❤♦✇s t❤❡ ❜❛s✐❝ st❡♣s ❢♦r t❤❡ ✐♥✐t✐❛❧ ❣✉❡ss ∑𝑁
𝑉 + 𝐼𝑖 𝑅𝑠
𝜕𝐸
✈❛❧✉❡s ❡st✐♠❛t✐♦♥ ♦❢ t❤❡ ♠♦❞❡❧ ♣❛r❛♠❡t❡rs✳ =−2 ( 𝑖 )
𝜕𝑅𝑠ℎ 𝑖=1 𝑅2𝑠ℎ
[ 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑉 + 𝐼𝑖 𝑅𝑠 ]
𝐼𝑖 − 𝐼𝑝ℎ + 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) + 𝑖
𝑅𝑠ℎ
4. Curve Fitting Modelling
=0 → 𝑓3 . (20)
This section aims to propose a simple and accurate method
for extracting the five SDM parameters from a measured I-
V curve at a given temperature and irradiance. The imple-
mented methodology requires the measured full I-V charac- 𝜕𝐸 ∑ 𝐼 𝐼 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑁
𝐼
=2 ( 𝑜 𝑖 𝑒 𝑛𝑉𝑇 + 𝑖 )
teristics of the PV panel as an input. The I-V curve of any 𝜕𝑅𝑠 𝑖=1
𝑛𝑉 𝑇 𝑅 𝑠ℎ

❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier P❛❣❡ ✹ ♦❢ ✶✸


❆❝❝✉r❛t❡ ♠❡t❤♦❞ ❢♦r P❱ s♦❧❛r ❝❡❧❧s ❛♥❞ ♠♦❞✉❧❡s ♣❛r❛♠❡t❡rs ❡①tr❛❝t✐♦♥ ✉s✐♥❣ ■✲❱ ❝✉r✈❡s

[ 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑉 + 𝐼 𝑖 𝑅𝑠 ]
𝐼𝑖 − 𝐼𝑝ℎ + 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) + 𝑖 Solving a system of equations using the Newton-Raphson
𝑅𝑠ℎ method requires an initial value for each parameter as dis-
=0 → 𝑓4 . (21) cussed in Section 3.
The difficulty of finding the convergence and reaching
the global minimum has been improved by an additional code
to limit the search space around the initial estimated values.
𝜕𝐸 ∑ 𝐼 (𝑉 + 𝐼𝑖 𝑅𝑠 ) 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑁
From experience, the following ideas have been used for se-
=2 ( 𝑜 𝑖 𝑒 𝑛𝑉𝑇 )
𝜕𝑛 𝑛 2𝑉 lecting the search space limits for each one of the five char-
𝑖=1 𝑇
[ 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑉 + 𝐼𝑖 𝑅𝑠 ] acteristic parameters:
𝐼𝑖 − 𝐼𝑝ℎ + 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) + 𝑖
𝑅𝑠ℎ • 𝐼𝑝ℎ : Subtracting or adding value between 1% and 5%
=0 → 𝑓5 . (22) (selected value depends on the I-V curve slope near
the short circuit point) of the highest measure current
Equations 18 to 22 represent a system of five non-linear to its value to be considered as a lower or upper limit,
equations with five unknowns; thus, it can be solved. How- respectively.
ever, the implicit nature of the presented equations imply the
use of a numerical method. Therefore, the Newton-Raphson • 𝐼𝑜 : Two orders of magnitude below and above the
method is considered because of its fast convergence and ro- initial guess value.
bust performance (Ben-Israel, 1966). A general expression • 𝑅𝑠ℎ : One order of magnitude below and above the
for the Newton Raphson method can be written as follows: initial guess value.
𝑋 (𝑡+1) = 𝑋 (𝑡) − [𝐽 (𝑋 (𝑡) )]−1 [𝐹 (𝑋 (𝑡) )], (23) • 𝑅𝑠 : One order of magnitude below and above the ini-
tial guess value.
where 𝑡 is the index for the iteration number. Both of 𝑋 (𝑡+1)
and 𝑋 (𝑡) are five elements vectors that hold the next and the • 𝑛 : between 1 and 2.
current values of the five parameters, respectively. 𝐽 (𝑋 (𝑡) ) is
This strategy has the advantage of always converging to-
the Jacobian matrix that contains the partial derivatives for
ward a good solution as long as having proper initial guesses.
each equation corresponding to each of the five parameters;
In certain cases, if proper initial parameter values are not
therefore, it will be 5𝑋5 matrix computed for the current
available the search space between the lower and upper lim-
value of the five parameters. 𝐹 (𝑋 (𝑡) ) contains the five par-
its around the initial guess can be broadened. However, this
tial differential equations to be calculated for the current val-
may require a higher number of iterations (longer computa-
ues of the parameters. The developed system of non-linear
tional time).
equations can be represented in the following form:
To ensure good solution and finding the global mini-
⎡ 𝑡+1
𝐼𝑝ℎ ⎤ ⎡ 𝑡
𝐼𝑝ℎ ⎤ mum, at each iteration, the calculated five parameters are
⎢ ⎥ ⎢ ⎥ substituted in the explicit form of the single diode model,
⎢ 𝐼𝑜𝑡+1 ⎥ ⎢ 𝐼𝑜𝑡 ⎥ given by equation (3). This allows constructing an I-V curve
⎢ ⎥ ⎢ ⎥ to be compared with the measured one using the mean abso-
⎢ 𝑅𝑡+1
𝑠ℎ ⎥
= ⎢ 𝑅𝑡𝑠ℎ ⎥
⎢ ⎥ ⎢ ⎥ lute percentage error (MAPE), which is defined in Appendix
⎢ 𝑅𝑡+1
𝑠 ⎥ ⎢ 𝑅𝑡𝑠 ⎥ B. The value of the MAPE of each iteration is stored together
⎢ ⎥ ⎢ ⎥ with the corresponding extracted five parameters. After the
⎣ 𝑛 𝑡+1 ⎦ ⎣ 𝑛𝑡 ⎦
−1
maximum number of iterations is reached, the stored errors
𝜕𝑓1 𝜕𝑓1 𝜕𝑓1 𝜕𝑓1 𝜕𝑓1
⎡ ⎤ will be compared and the lowest value will be considered as
⎢ 𝜕𝐼𝑝ℎ 𝜕𝐼𝑜 𝜕𝑅𝑠ℎ 𝜕𝑅𝑠 𝜕𝑛
⎥ the best solution. This procedure highly improves the solu-
⎢ 𝜕𝑓2 𝜕𝑓2 𝜕𝑓2 𝜕𝑓2 𝜕𝑓2 ⎥
⎢ 𝜕𝐼𝑝ℎ 𝜕𝐼𝑜 𝜕𝑅𝑠ℎ 𝜕𝑅𝑠 𝜕𝑛 ⎥ tion and leads to a better fitting compared to the traditional
⎢ ⎥ way of comparing the sum of the squared errors to a specific
−⎢ ⎥
𝜕𝑓3 𝜕𝑓3 𝜕𝑓3 𝜕𝑓3 𝜕𝑓3
limit. This is mainly due to the fact that during the evalua-
⎢ 𝜕𝐼𝑝ℎ 𝜕𝐼𝑜 𝜕𝑅𝑠ℎ 𝜕𝑅𝑠 𝜕𝑛 ⎥
⎢ 𝜕𝑓4 𝜕𝑓4 𝜕𝑓4 𝜕𝑓4 𝜕𝑓4 ⎥ tion of the five equations each equation gives results which
⎢ 𝜕𝐼𝑝ℎ 𝜕𝐼𝑜 𝜕𝑅𝑠ℎ 𝜕𝑅𝑠 𝜕𝑛 ⎥ differ from each other by orders of magnitude. Thus the low-
⎢ ⎥ est error value (the square sum of their solutions) does not
⎢ 𝜕𝑓5 𝜕𝑓5 𝜕𝑓5 𝜕𝑓5 𝜕𝑓5 ⎥
⎣ 𝜕𝐼𝑝ℎ 𝜕𝐼𝑜 𝜕𝑅𝑠ℎ 𝜕𝑅𝑠 𝜕𝑛 ⎦ necessarily mean that it indicates the best fitting. Since the
⎡ 𝑡 , 𝐼 𝑡 , 𝑅𝑡 , 𝑅𝑡 , 𝑛𝑡 ) ⎤ equation with the highest value will have the highest impact
𝑓1 (𝐼𝑝ℎ 𝑜 𝑠ℎ 𝑠
⎢ ⎥ on the summed error. Another important feature provided
⎢ 𝑓2 (𝐼𝑝ℎ , 𝐼𝑜 , 𝑅𝑠ℎ , 𝑅𝑡𝑠 , 𝑛𝑡 ) ⎥
𝑡 𝑡 𝑡 by our proposed method is the possibility of selecting the ac-
⎢ ⎥ ceptable level of accuracy to reduce the computational cost.
𝑋 ⎢ 𝑡 , 𝐼 𝑡 , 𝑅𝑡 , 𝑅𝑡 , 𝑛𝑡 ) ⎥
𝑓3 (𝐼𝑝ℎ . (24) This concept has been implemented by comparing the cal-
⎢ 𝑜 𝑠ℎ 𝑠

⎢ 𝑓4 (𝐼𝑝ℎ , 𝐼𝑜 , 𝑅𝑠ℎ , 𝑅𝑠 , 𝑛 ) ⎥
𝑡 𝑡 𝑡 𝑡 𝑡 culated MAPE after each iteration with a preselected limit.
⎢ ⎥ If a value of the MAPE below the acceptable limit is de-
⎢ 𝑡 , 𝐼 𝑡 , 𝑅𝑡 , 𝑅𝑡 , 𝑛𝑡 ) ⎥
𝑓5 (𝐼𝑝ℎ
⎣ 𝑜 𝑠ℎ 𝑠 ⎦ tected, the iteration process will be terminated and the five

❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier P❛❣❡ ✺ ♦❢ ✶✸


❆❝❝✉r❛t❡ ♠❡t❤♦❞ ❢♦r P❱ s♦❧❛r ❝❡❧❧s ❛♥❞ ♠♦❞✉❧❡s ♣❛r❛♠❡t❡rs ❡①tr❛❝t✐♦♥ ✉s✐♥❣ ■✲❱ ❝✉r✈❡s

parameters will be considered. Otherwise, the iteration will


Start
continue until the specified maximum number of iterations
is reached. Figure 4 shows the flowchart of the proposed
method.
Importing the measured and
optimized data (I-V Curve)
5. Demonstrative Examples
The proposed method is used to extract the five SDM
parameters of PV cells/modules for a given I-V measured
curve. The code related to the initial guess values of the pa- Calculating the initial parameters values
rameters is executed first. The resulting data is exported to
the main program code for numerically computing the pa-
rameter values as well as calculating some other indicators
to monitor the performance of the algorithm, , according to Use the current values to calculate the
values of the 5 system equations and the
the method described in section 4. Quantifying the error and 25 components of the Jacobian Matrix
evaluating the performance is performed by calculating the
error at the maximum power point (𝛿), mean absolute per-
centage error (MAPE), and the coefficient of determination Calculate the next parameters value by
(𝑅2 ). multiply the inverse of the Jacobian
Three different PV samples have been used for validating matrix by the system function value
the proposed model. The first sample is a crystalline silicon
cell manufactured especially for research purposes (Plesz,
2015). The second sample is a commercial polycrystalline Calculate and store the MAPE
PV module (SPS 10W 12V) that includes 36 cells in series.
The third sample is an amorphous module with 68 cells con-
nected in series (WSK0045). The used measurement sys-
tems record the I-V curves with roughly 50 data points (𝑉 , 𝐼)
between the short and open circuit points. Yes
For extracting the five SDM parameters, the above de- MAPE <= Limit
scribed modified version of the Newton-Raphson method
(by adopting two functional concepts, introduced in Section
4) has been used to solve the system of non-linear equations No
Break the loop
produced by applying the least square method. Figure 5a and
5b (in the beginning of the next page) show how the MAPE
and 𝛿 are changing with increasing the number of iterations, No
for different PV samples, respectively. Table 4 included in Iter = Iter_Max
the Appendix C shows the result of applying the parameter
extraction method on the three PV samples using different
numbers of iterations without terminating the iteration pro- Yes
cess by monitoring the MAPE value, but only until reaching
the maximum number of iterations.
Each of the Figures 6 and 7 show seven fitted curves to- Consider the solution of the lowest error
gether with the measured data points under different envi- value and plot the fitted curve using the
ronmental conditions of a normal winter sunny day for the explicit function form using the
polycrystalline module and cloudy day for the amorphous numerically calculated parameters
modules, respectively. For the monocrystalline cell, previ-
ous data from laboratory measurement were available (Plesz,
2015). These measurements covered a wide range of irradi- Display the numerically
ance (200 𝑊 ∕𝑚2 - 1268 𝑊 ∕𝑚2 ) and temperature (10𝑜 𝐶 - calculated parameters
85𝑜 𝐶) and thus are suitable to test the high-performance ap-
plicability of the presented method. Figure 8a shows fitted
I-V curves for different irradiance conditions, while Figure End
8b reports the results of applying the model to the monocrys-
❋✐❣✉r❡ ✹✿ ❋❧♦✇❝❤❛rt ❢♦r t❤❡ ◆❡✇t♦♥ ❘❛♣❤s♦♥ ♥✉♠❡r✐❝❛❧ ❝❛❧❝✉✲
talline cell at different temperatures.
❧❛t✐♦♥ ♠❡t❤♦❞♦❧♦❣②✳
Table 1 shows the extracted results for the five parame-
ters and the performance monitoring parameters for the three
samples. The parameters shown in Table 1 are extracted for
𝐼 − 𝑉 curves that have been measured under different irradi-

❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier P❛❣❡ ✻ ♦❢ ✶✸


5 0.08
Crystalline cell
Polycrystalline Module
Crystalline cell
Amorphous Module 0.07 Polycrystalline Module
4 Amorphous
0.06

3 0.05
MAPE

0.04
2
0.03

0.02
1
0.01

0 0
0 1000 2000 3000 4000 5000 0 1000 2000 3000 4000 5000
Number of iterations Number of iterations

0.4 0.35
0.35 0.3
0.3
0.25
0.25
I (A)
0.2
I(A)

0.2
0.15
0.15
0.1
0.1
0.05
0.05
0
0 5 10 15 20 0 5 10 15 20 25
V(V) V (V)

ance conditions (𝐺) and temperatures (𝑇 ). Since measure- under a laboratory or sunny day environment; another en-
ment data under a controlled environment is available for the vironmental variant represented by a cloudy day has been
crystalline cell, different curves measured under the same considered in this validation using the amorphous module.
temperature have been selected to highlight the irradiance Figure 9 highlights the model applicability at a wide range
affect one the extracted parameters, but mainly the photo- of temperatures and shows the temperature effect on the re-
generated current. For the polycrystalline module, 𝐼 − 𝑉 verse saturation current (crystalline cell under laboratory en-
curves measured under a sunny day environment have been vironment). The mentioned figure reflects the well-known
used to validate the proposed method. However, not only fact of the substantial temperature effect on the reverse satu-

0.035
0.05
0.03
0.04
0.025
I (A)
I (A)

0.03 0.02

0.015
0.02
0.01
0.01
0.005

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
V (V) V (V)

Preprint submitted to Elsevier


G(𝐖∕𝐦𝟐 ) I(𝐨 𝐂) I𝐩𝐡 (𝐀) I𝐨 (𝐀) R𝐬𝐡 (𝛀) R𝐬 (𝛀) n 𝜹 R𝟐 MAPE%
−5
200 45 0.0091 1.67 ⋅ 10 333.99 1.12 1.84 0.0002 0.99994 0.068
400 45 0.0176 1.93 ⋅ 10−5 261.31 0.87 1.86 0.0014 0.99992 0.078
600 45 0.0247 2.31 ⋅ 10−5 251.20 0.84 1.89 0.0007 0.99998 0.050
800 45 0.0318 1.93 ⋅ 10−5 198.62 0.94 1.84 0.0007 0.99996 0.057
1000 45 0.0381 1.49 ⋅ 10−5 151.11 1.04 1.77 0.0010 0.99982 0.113
1118 45 0.0450 1.78 ⋅ 10−5 180.98 1.06 1.80 0.0016 0.99987 0.124
1268 45 0.0510 2.61 ⋅ 10−5 182.85 1.01 1.88 0.0001 0.99998 0.047
119 16 0.0942 1.43 ⋅ 10−8 2360.7 14.32 1.64 0.0020 0.99552 0.92
191 18 0.1586 1.01 ⋅ 10−6 6835.3 12.41 1.76 0.0087 0.99880 0.64
215 19 0.2014 2.17 ⋅ 10−7 7961.1 13.69 1.78 0.0062 0.99954 0.20
582 26 0.2732 1.00 ⋅ 10−7 3862.1 14.24 1.49 0.0017 0.99978 0.28
638 27 0.3582 1.70 ⋅ 10−8 1978.0 14.12 1.27 0.0031 0.99967 0.46
700 29 0.4043 2.64 ⋅ 10−7 7317.7 13.87 1.44 0.0056 0.99958 0.65
758 29 0.5415 8.04 ⋅ 10−8 2143.2 11.88 1.40 0.0103 0.99951 0.60
66 14 0.094 5.29 ⋅ 10−5 964.9 8.35 1.92 0.011 0.99873 0.889
99 13 0.154 9.98 ⋅ 10−5 675.8 1.48 1.94 0.008 0.99880 0.861
119 15 0.176 7.99 ⋅ 10−5 490.2 0.59 1.90 0.009 0.99901 0.711
134 18 0.208 8.67 ⋅ 10−5 539.9 0.65 1.91 0.014 0.99892 0.768
200 26 0.264 8.24 ⋅ 10−5 393.4 0.43 1.78 0.013 0.99897 0.745
231 28 0.326 16.8 ⋅ 10−5 307.5 0.20 1.87 0.007 0.99930 0.710
256 30 0.350 19.4 ⋅ 10−5 375.7 0.19 1.89 0.0001 0.99937 0.683

ration current represented by an exponential increase in the ronmental conditions in different ways. However, studying
current value. Besides the achieved level of accuracy and the the environmental effects on the SDM parameters are beyond
parameters applicability to be used in the PV electrical mod- the scope of this article.
elling, the introduced results show that the extracted param- The possibility of flexible adjustment between accuracy
eters have a physical meaning, which can be used for further and computational cost is highlighted in Table 2. In which,
investigations. the proposed method compares both situations, with and with-
out including the concept of termination the iteration process
10 -4 after reaching an acceptable (preselected) MAPE value. In
1.5
G=200 W/m2 this comparison, the number of maximum iterations is se-
G=400 W/m2 lected to be 2000, and the acceptable MAPE value is 1%.
G=600 W/m2
G=800 W/m2
The table summarizes the results of 400 I-V curves measured
1
G=1000 W/m2 under a wide range of environmental conditions.
Io (A) G=1118 W/m2
G=1268 W/m2
0.5

1%
0
10 20 30 40 50 60 70 80
T (oC) 𝜹
0.00526 0.450 34.33
0.00915 0.842 9.76
0.00996 0.893 37.24
0.01530 1.090 10.50
Based on the introduced results in Table 1, we see that 0.01015 1.159 37.26
the shunt resistant varies to a relatively large extent within a 0.01152 1.188 7.30
specific range for each PV sample. This indicates that 𝑅𝑠ℎ
has a limited effect on the 𝐼 − 𝑉 characteristics compared to From the results shown in this section we can conclude
the effect of the reverse saturation current. Similar investi- the following points:
gations have been carried out for the other SDM parameters
which show that these parameters are affected by the envi- • Table 1 and Figures 6, 7, 8a and 8b, prove the effec-

Preprint submitted to Elsevier


❆❝❝✉r❛t❡ ♠❡t❤♦❞ ❢♦r P❱ s♦❧❛r ❝❡❧❧s ❛♥❞ ♠♦❞✉❧❡s ♣❛r❛♠❡t❡rs ❡①tr❛❝t✐♦♥ ✉s✐♥❣ ■✲❱ ❝✉r✈❡s

tiveness and validate the proposed method and its ap- for this comparison are based on the single diode model for
plicability at a wide range of environmental conditions extracting the parameters of a monocrystalline cell. Table
and different PV technologies. 3 compares the best results reported by each reference with
the result of our presented method represented by best and
• Figures 5a and 5b, and Table 4 show that the accu- the average (of 200 𝐼 − 𝑉 curves measured under different
racy of the results are clearly improved with increas- environmental conditions) values of the corresponding pa-
ing the number of iterations up to value between 1000 rameters by running the model for extracting the parameters
and 2000. Hence, such information provides the pos- over the whole available environmental range (mentioned in
sibility to select the optimum number of the maximum section 5). New parameters are shown in the table are root
iterations for the PV module under investigation. mean square error (RMSE), normalized version of RMSE
• From the figures and table mentioned above, we see (RMSEn), and mean absolute error (MAE). Appendix B lists
that even with a low number of iterations we are still the formulation of these parameters.
getting good solutions. The main reason for that is the
good initial guess values calculated by the methodol- 7. Conclusions
ogy introduced in Section 3.
In this paper, an accurate numerical method for retriev-
• Despite the good results obtained for the amorphous ing the single diode model parameters is proposed. This ap-
PV module at early iterations, with increasing the num- proach is based on curve fitting using the least square method
ber of iterations we don’t see a significant improve- to form a system of equations and solving them by a modified
ment on the solution. This is because the SDM model version of the Newton-Raphson method. The only required
does not fit perfectly to the tandem pin diode structure information are the measured I-V characteristic curves of
of the thin-film module. But still, the recorded errors the solar panel. The model is implemented using MATLAB
at the maximum power point and the MAPE are very script. An optimised method was implemented for estimat-
acceptable. ing the parameters’ initial guess values. The method’s flexi-
bility is enhanced by a selectable trade-off between the com-
• Also shown in Tables 1 and 4 and Figures 5a and 5b, putational cost and the acceptable error value. Limiting the
more accurate results were realized for the monocrys- search space of the five parameters ensures the robustness of
talline cell. We state three reasons for this difference. the algorithm by overcoming the potential of the numerical
First, due to the better quality of the measured I-V solution method divergence. This work was validated using
curves since these were obtained under laboratory con- three PV samples: a monocrystalline cell, a polycrystalline
ditions. The other two samples were measured with module, and an amorphous silicon module. The modelling
less accurate measurement systems under outdoor op- results reflect the proposed method’s ability for extracting
erational conditions. Second, due to the physical fact the parameters with very low deviation between the fitted
that the single exponential representation of the SDM model and the measured I-V curves. Comparisons with the
fits single PN-junction much better than a solar mod- recent and the most accurate literature reveal the importance
ule consisting of a number of cells connected in series. of the proposed method in terms of results accuracy. Al-
Third, modelling a PV module assumes that all cells though the best value of the error quantifying parameters
are identical, whereas practically, not all cells have the reported by different references are compared to the aver-
same characteristics because of technology processes age values calculated from fitting 400 I-V curves at various
or as a result of partial environmental effects which environmental conditions, still our proposed parameters ex-
could be different among cells in the module. traction method shows lower error in most cases.
• The result shown in Table 2 reveals that selecting 1%
as an acceptable MAPE will significantly improve the A. Jacobian Matrix Equations
average convergence time (ACT). It will be reduced to The components of the Jacobian matrix used in the im-
one-quarter compared to the case of not comparing to plemented algorithm for solving the system of non-linear
a preselected error limit and terminating the iteration equations are shown below.
process. By monitoring different performance indica-
tors, we see that the accuracy is only slightly lower. 𝜕𝑓1 ∑ 𝑁
= −1 (25)
𝜕𝐼𝑝ℎ 𝑖=1
6. Results of comparison to the literature
𝜕𝑓1 ∑ [ 𝑉𝑖 𝑛𝑉 ]
𝑁
This section is dedicated to briefly highlight the scientific +𝐼𝑖 𝑅𝑠

improvement that has been introduced in this work. Table 3 = 𝑒 𝑇 −1 (26)


𝜕𝐼𝑜 𝑖=1
compares the proposed method (2000 as a maximum num-
ber of iterations has been selected), with different accurate
models from the recent and most accurate literature using 𝜕𝑓1 ∑ [ 𝑉𝑖 + 𝐼𝑖 𝑅𝑠 ]
𝑁

various error quantifying parameters. The selected models = − (27)


𝜕𝑅𝑠ℎ 𝑖=1 𝑅2𝑠ℎ

❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier P❛❣❡ ✾ ♦❢ ✶✸


❆❝❝✉r❛t❡ ♠❡t❤♦❞ ❢♦r P❱ s♦❧❛r ❝❡❧❧s ❛♥❞ ♠♦❞✉❧❡s ♣❛r❛♠❡t❡rs ❡①tr❛❝t✐♦♥ ✉s✐♥❣ ■✲❱ ❝✉r✈❡s

❚❛❜❧❡ ✸
❈♦♠♣❛r✐s♦♥ ♦❢ t❤❡ ♣r♦♣♦s❡❞ ♠❡t❤♦❞ ✇✐t❤ ❞✐✛❡r❡♥t ❡①✐st✐♥❣ ♠❡t❤♦❞s
❘❡❢❡r❡♥❝❡ ❉❡s❝r✐♣t✐♦♥ ▼♦❞✉❧❡ ❊rr♦r ❘❡❢❡r❡♥❝❡ Pr❡s❡♥t❡❞ Pr❡s❡♥t❡❞
t②♣❡ q✉❛♥t✐❢②✐♥❣ ❜❡st ✇♦r❦ ✇♦r❦
♣❛r❛♠❡t❡r ❜❡st ❛✈❡r❛❣❡
❈♦♠♣❛r✐s♦♥ ♦❢ ❢♦✉r ❞✐✛❡r❡♥t ▼❆P❊ 5.5376% 0.026% 0.273%
▼♦♥♦❝r②st❛❧❧✐♥❡
✭P✳ ❛r❛♥❞❛ ❈❤❡♥❝❤❡ ❡t ❛❧✳✱♠❡t❤♦❞s
✷✵✶✽ ✮ ❜❛s❡❞ ♦♥ ❛♥❛❧②t✐❝❛❧ ❘2 0.9934 0.9999991 0.999803
❛♥❞ ♦♣t✐♠✐s❛t✐♦♥ ❛❧❣♦r✐t❤♠s✳
▼❡t❤♦❞ ❜❛s❡❞ ♦♥ ✐♠♣❡r✐❛❧✐st ▼♦♥♦❝r②st❛❧❧✐♥❡ ▼❆❊ 4 ⋅ 10−3 1.54 ⋅ 10−6 5.88 ⋅ 10−5
✭❋❛t❤② ❛♥❞ ❘❡③❦✱ ✷✵✶✼ ✮ ❝♦♠♣❡t✐t✐✈❡ ❛❧❣♦r✐t❤♠ ✭■❈❆✮✳ P♦❧②❝r②st❛❧❧✐♥❡ ▼❆❊ 5 ⋅ 10−4 2.58 ⋅ 10−4 1.82 ⋅ 10−3
❆♠♦r♣❤♦✉s ▼❆❊ 8.3 ⋅ 10−3 5.12 ⋅ 10−4 1.18 ⋅ 10−3
❇❛s❡❞ ♦♥ t❤❡ ❋❧♦✇❡r P♦❧❧✐♥❛t✐♦♥ ▼♦♥♦❝r②st❛❧❧✐♥❡ ❘▼❙❊ 7.27 ⋅ 10−4 1.98 ⋅ 10−6 6.77 ⋅ 10−5
✭❘❛♠ ❡t ❛❧✳✱ ✷✵✶✼ ✮ ❆❧❣♦r✐t❤♠ ✭❇P❋P❆✮✳ P♦❧②❝r②st❛❧❧✐♥❡ ❘▼❙❊ 5.87 ⋅ 10−4 3.25 ⋅ 10−4 2.34 ⋅ 10−3
❆♠♦r♣❤♦✉s ❘▼❙❊ 3.65 ⋅ 10−4 5.98 ⋅ 10−4 1.38 ⋅ 10−3
❆♥❛❧②t✐❝❛❧ ❛❧❣♦r✐t❤♠✱ ❝❛❧❝✉❧❛t❡ ▼♦♥♦❝r②st❛❧❧✐♥❡ ❘▼❙❊ 2.26 ⋅ 10−2 1.98 ⋅ 10−6 6.77 ⋅ 10−5
✭❘❤♦✉♠❛ ❡t ❛❧✳✱ ✷✵✶✼ ✮
♠❛♥② s♦❧✉t✐♦♥s ❛♥❞ s❡❧❡❝t
t❤❡ ♦♥❡ ✇✐t❤ t❤❡ ❧♦✇❡st ❘▼❙❊✳
✭❊t✲t♦r❛❜✐ ❡t ❛❧✳✱ ✷✵✶✼ ✮ ■t❡r❛t✐✈❡ ♠❡t❤♦❞ ✲ ●❛✉ss ❙❡✐❞❡❧ ▼♦♥♦❝r②st❛❧❧✐♥❡ ❘▼❙❊♥ 2.3 ⋅ 10−3 2.4 ⋅ 10−4 2.4 ⋅ 10−3

𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑉𝑖 + 𝐼𝑖 𝑅𝑠 )]
+ 𝐼𝑝ℎ − 2𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) −
𝑅𝑠ℎ
𝜕𝑓1 ∑ [ 𝐼𝑖 𝑉𝑖 𝑛𝑉 𝐼 ]
𝑁 +𝐼𝑖 𝑅𝑠
(34)
= 𝐼𝑜 𝑒 𝑇 + 𝑖 (28)
𝜕𝑅𝑠 𝑖=1 𝑛𝑉𝑇 𝑅𝑠ℎ

𝜕𝑓3 ∑ [ 𝑉𝑖 + 𝐼𝑖 𝑅𝑠 ]
𝑁
= − (35)
𝜕𝑓1 ∑ [ +𝐼𝑖 𝑅𝑠 ]
𝑁
𝑉 + 𝐼𝑖 𝑅𝑠 𝑉𝑖 𝑛𝑉 𝜕𝐼𝑝ℎ 𝑖=1 𝑅2𝑠ℎ
= − 𝐼𝑜 𝑖 𝑒 𝑇 (29)
𝜕𝑛 𝑖=1 𝑛2 𝑉𝑇

𝜕𝑓3 ∑ [ 𝑉𝑖 + 𝐼𝑖 𝑅𝑠 𝑉𝑖 𝑛𝑉 ]
𝑁 +𝐼𝑖 𝑅𝑠
= (𝑒 𝑇 − 1) (36)
𝜕𝑓2 ∑ [𝑁 𝑉𝑖 +𝐼𝑖 𝑅𝑠 ] 𝜕𝐼𝑜 𝑖=1 𝑅2 𝑠ℎ
= − 1 − 𝑒 𝑛𝑉𝑇 (30)
𝜕𝐼𝑝ℎ 𝑖=1

𝜕𝑓3 ∑ 𝑉𝑖 + 𝐼𝑖 𝑅𝑠 [ −(𝑉𝑖 + 𝐼𝑖 𝑅𝑠 )
𝑁
= − 2𝐼𝑖
𝜕𝑓2 ∑ [ 𝑉𝑖 +𝐼𝑖 𝑅𝑠 ]2 𝑅3
𝑁 𝜕𝑅𝑠ℎ 𝑖=1 𝑅𝑠ℎ
𝑠ℎ
= 1 − 𝑒 𝑛𝑉𝑇 (31) 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝜕𝐼𝑜 𝑖=1 + 2𝐼𝑝ℎ − 2𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1)
𝑉 + 𝐼𝑖 𝑅 𝑠 ]
−2 𝑖 (37)
𝜕𝑓2 ∑ [ 𝑉𝑖 + 𝐼𝑖 𝑅𝑠
𝑁 𝑉𝑖 +𝐼𝑖 𝑅𝑠 ] 𝑅𝑠ℎ
= (1 − 𝑒 𝑛𝑉𝑇 ) (32)
𝜕𝑅𝑠ℎ 𝑖=1 𝑅2 𝑠ℎ
𝜕𝑓3 ∑ 𝐼𝑖 [ 𝐼𝑜 (𝑉𝑖 + 𝐼𝑖 𝑅𝑠 ) 𝑉𝑖 𝑛𝑉
𝑁 +𝐼𝑖 𝑅𝑠
= 𝑒 𝑇
𝜕𝑅𝑠 𝑖=1 𝑅𝑠ℎ2
𝜕𝑓2 ∑ 𝐼𝑖 [ 𝑉𝑖 𝑛𝑉
+𝐼𝑖 𝑅𝑠 (
𝑁 𝑛𝑉𝑇
= 𝑒 𝑇 𝐼𝑖 − 𝐼𝑝ℎ − 2𝐼𝑜 𝑉𝑖 + 𝐼𝑖 𝑅𝑠
𝜕𝑅𝑠 𝑖=1 𝑛𝑉𝑇 +2 + 𝐼𝑖 − 𝐼𝑝ℎ
𝑉𝑖 +𝐼𝑖 𝑅𝑠 )
𝑅𝑠ℎ
𝑛𝑉𝑇 𝑉 + 𝐼𝑖 𝑅𝑠 ]
+ + 𝑖 + 2𝐼𝑜 𝑒 𝑛𝑉𝑇 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑅𝑠ℎ 𝑅𝑠ℎ + 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) (38)
𝑛𝑉𝑇 ]
− (33)
𝑅𝑠ℎ
𝜕𝑓3 ∑ [ 𝑉𝑖 + 𝐼𝑖 𝑅𝑠
𝑁
𝑉 + 𝐼𝑖 𝑅𝑠
= ( )(− 𝑖 )
𝜕𝑛 2 2𝑉
𝜕𝑓2 ∑ [ 𝑉𝑖 + 𝐼𝑖 𝑅𝑠 𝑉𝑖 𝑛𝑉
+𝐼𝑖 𝑅𝑠 (
𝑁 𝑖=1 𝑅 𝑠ℎ
𝑛 𝑇
= 𝑒 𝑇 − 𝐼𝑖 𝑉𝑖 +𝐼𝑖 𝑅𝑠 ]
𝜕𝑛 𝑖=1 𝑛2 𝑉𝑇 ⋅ (𝐼𝑜 𝑒 𝑇 )𝑛𝑉 (39)

❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier P❛❣❡ ✶✵ ♦❢ ✶✸


❆❝❝✉r❛t❡ ♠❡t❤♦❞ ❢♦r P❱ s♦❧❛r ❝❡❧❧s ❛♥❞ ♠♦❞✉❧❡s ♣❛r❛♠❡t❡rs ❡①tr❛❝t✐♦♥ ✉s✐♥❣ ■✲❱ ❝✉r✈❡s

𝜕𝑓4 ∑[ 𝑁
𝐼 𝐼 𝐼 𝑉𝑖 +𝐼𝑖 𝑅𝑠 ]
𝑁 +𝐼𝑖 𝑅𝑠 [
𝜕𝑓5 ∑ 𝐼𝑖 𝐼𝑜 𝑉𝑖 𝑛𝑉 𝑉 + 𝐼𝑖 𝑅𝑠 𝑉𝑖 +𝐼𝑖 𝑅𝑠
= − 𝑖 − 𝑖 𝑜 𝑒 𝑛𝑉𝑇 (40) = 𝑒 𝑇 ( 𝑖 𝐼𝑜 𝑒 𝑛𝑉𝑇
𝜕𝐼𝑝ℎ 𝑖=1 𝑅𝑠ℎ 𝑛𝑉𝑇 𝜕𝑅𝑠 𝑖=1 𝑛2 𝑉𝑇 𝑛𝑉𝑇
𝑉𝑖 + 𝐼𝑖 𝑅𝑠 𝑉 + 𝐼𝑖 𝑅𝑠
+ )+( 𝑖 + 1)
+𝐼𝑖 𝑅𝑠 [
𝑅𝑠ℎ 𝑅𝑠ℎ
𝜕𝑓4 ∑ 𝐼𝑖 𝑉𝑖 𝑛𝑉
𝑁 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑉𝑖 +𝐼𝑖 𝑅𝑠
= 𝑒 𝑇 2𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1)
𝜕𝐼𝑜 𝑖=1 𝑛𝑉𝑇 + (𝐼𝑖 − 𝐼𝑝ℎ + 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1)
𝑉𝑖 + 𝐼𝑖 𝑅𝑠 𝑉 + 𝐼𝑖 𝑅𝑠 ]
+ 𝐼𝑖 − 𝐼𝑝ℎ + + 𝑖 ) (48)
𝑅𝑠ℎ 𝑅𝑠ℎ
𝐼𝑖 𝑉𝑖 +𝐼𝑖 𝑅𝑠 ]
+ (𝑒 𝑛𝑉𝑇 − 1) (41)
𝑅𝑠ℎ +𝐼𝑖 𝑅𝑠 [
𝜕𝑓5 ∑ −𝐼𝑜 (𝑉𝑖 + 𝐼𝑖 𝑅𝑠 )2 𝑉𝑖 𝑛𝑉
𝑁 𝑉𝑖 +𝐼𝑖 𝑅𝑠
= ( 𝑒 𝑇 ) (𝐼𝑜 𝑒 𝑛𝑉𝑇 )
𝜕𝑛 𝑖=1 𝑛4 𝑉 2 𝑇
𝜕𝑓4 ∑ 𝐼𝑖 [ 𝑉𝑖 + 𝐼𝑖 𝑅𝑠 𝐼𝑜 (𝑉𝑖 + 𝐼𝑖 𝑅𝑠 ) 𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑁
2𝑛𝑉𝑇 𝑉𝑖 +𝐼𝑖 𝑅𝑠
= − − 𝑒 𝑛𝑉𝑇 +( + 1)(𝐼𝑖 − 𝐼𝑝ℎ + 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1)
𝜕𝑅𝑠ℎ 𝑖=1 𝑅𝑠ℎ2 𝑅𝑠ℎ 𝑛𝑉𝑇 𝑉𝑖 + 𝐼𝑖 𝑅𝑠
𝑉𝑖 +𝐼𝑖 𝑅𝑠
𝑉 + 𝐼𝑖 𝑅𝑠 ] 𝑉𝑖 + 𝐼𝑖 𝑅𝑠 ]
− 𝐼𝑖 + 𝐼𝑝ℎ − 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) − 𝑖 + ) . (49)
𝑅𝑠ℎ 𝑅𝑠ℎ
(42)
B. Equations of the error quantifying
parameters
+𝐼𝑖 𝑅𝑠 (
𝜕𝑓4 ∑ [ 𝐼𝑜 𝐼𝑖2 𝑉𝑖 𝑛𝑉
𝑁 𝑉𝑖 +𝐼𝑖 𝑅𝑠
2𝑛𝑉𝑇
= ( 𝑒 𝑇 ) 𝐼𝑜 𝑒 𝑛𝑉𝑇 +
𝜕𝑅𝑠 𝑖=1 𝑛2 𝑉 2 𝑅𝑠ℎ
𝑇 ∑𝑛
𝑉 + 𝐼𝑖 𝑅𝑠 )
𝑖 𝑖 2
𝑖 (𝐼𝑚𝑒𝑎𝑠 − 𝐼𝑐𝑎𝑙𝑐 )
𝑉𝑖 +𝐼𝑖 𝑅𝑠
2
+ 𝐼𝑖 + 𝐼𝑝ℎ − 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) + 𝑖 𝑅 =1− ∑ ∑ 𝑖 . (50)
𝑅𝑠ℎ 𝑖
(𝐼𝑚𝑒𝑎𝑠 − 1𝑛 𝑛𝑖 (𝐼𝑚𝑒𝑎𝑠 ))2
𝐼𝑖2 ]
+ (43)
𝑅𝑠ℎ2
100% ∑ (𝐼𝑚𝑒𝑎𝑠 − 𝐼𝑐𝑎𝑙𝑐 )
𝑛 𝑖 𝑖
𝑀𝐴𝑃 𝐸 = 𝑖
. (51)
𝑛 𝑖 𝐼𝑚𝑒𝑎𝑠
𝜕𝑓4 ∑ −𝐼𝑜 𝐼𝑖 (𝑉𝑖 + 𝐼𝑖 𝑅𝑠 ) 𝑉𝑖 𝑛𝑉
𝑁 +𝐼𝑖 𝑅𝑠
= ( 𝑒 𝑇 )
𝜕𝑛 𝑛2 𝑉𝑇
1∑ 𝑖
𝑛
𝑖=1
[ 𝐼 𝑉𝑖 +𝐼𝑖 𝑅𝑠 𝑀𝐴𝐸 = |(𝐼𝑚𝑒𝑎𝑠 − 𝐼𝑐𝑎𝑙𝑐
𝑖
)|. (52)
1 1 1 𝑛 𝑖
⋅ + 𝑜 𝑒 𝑛𝑉𝑇 + ( + )
𝑅𝑠ℎ 𝑛𝑉𝑇 𝑛𝑉𝑇 𝑉 𝑖 + 𝐼𝑖 𝑅𝑠
𝑉 +𝐼𝑅𝑠
𝑉 + 𝐼𝑖 𝑅𝑠 ] √
(𝐼𝑖 − 𝐼𝑝ℎ + 𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) + 𝑖 ) ∑𝑛 𝑖 𝑖 )2
𝑛𝑉𝑇 𝑖 (𝐼𝑚𝑒𝑎𝑠 − 𝐼𝑐𝑎𝑙𝑐
(44) 𝑅𝑀𝑆𝐸 = . (53)
𝑛

∑ [ −𝐼𝑜 (𝑉𝑖 + 𝐼𝑖 𝑅𝑠 ) 𝑉𝑖 +𝐼𝑖 𝑅𝑠 ]


𝑁 √
𝜕𝑓5 √ 1 ∑𝑛
= 𝑒 𝑛𝑉𝑇 (45) √ 𝑖 𝑖 2
√ 𝑖 (𝐼𝑚𝑒𝑎𝑠 − 𝐼𝑐𝑎𝑙𝑐 )
𝜕𝐼𝑝ℎ 𝑖=1 𝑛2 𝑉𝑇 𝑅𝑀𝑆𝐸𝑛 = √ 𝑛 . (54)
1 ∑𝑛 𝑖 2
𝑛 𝑖 (𝐼 𝑚𝑒𝑎𝑠 )

+𝐼𝑖 𝑅𝑠 [
𝜕𝑓5 ∑ 𝑉𝑖 + 𝐼𝑖 𝑅𝑠 𝑉𝑖 𝑛𝑉
𝑁
= ( 𝑒 𝑇 ) 𝐼𝑖
𝜕𝐼𝑜 𝑖=1 𝑛2 𝑉𝑇
𝑉 +𝐼𝑅𝑠
𝑉𝑖 + 𝐼 𝑖 𝑅 𝑠 ]
− 𝐼𝑝ℎ + 2𝐼𝑜 (𝑒 𝑛𝑉𝑇 − 1) + (46)
𝑛𝑉𝑇

𝜕𝑓5 ∑ [ (𝑉𝑖 + 𝐼𝑖 𝑅𝑠 )2
𝑁 𝑉𝑖 +𝐼𝑖 𝑅𝑠 ]
= − 𝐼𝑜 𝑒 𝑛𝑉𝑇 (47)
𝜕𝑅𝑠ℎ 𝑖=1 𝑛2 𝑉𝑇 𝑅𝑠ℎ2

❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier P❛❣❡ ✶✶ ♦❢ ✶✸


❆❝❝✉r❛t❡ ♠❡t❤♦❞ ❢♦r P❱ s♦❧❛r ❝❡❧❧s ❛♥❞ ♠♦❞✉❧❡s ♣❛r❛♠❡t❡rs ❡①tr❛❝t✐♦♥ ✉s✐♥❣ ■✲❱ ❝✉r✈❡s

C. Result of applying the proposed method Haroon Ashfaq, Ikhlaq Hussain, and Ajay Giri. Comparative analysis of
old, recycled and new PV modules. Journal of King Saud University
using different numbers of iterations - Engineering Sciences, 29(1):22–28, jan 2017. ISSN 10183639. doi:
10.1016/j.jksues.2014.08.004. URL https://linkinghub.elsevier.com/
retrieve/pii/S1018363914000440.
❚❛❜❧❡ ✹ TR Ayodele, ASO Ogunjuyigbe, and EE Ekoh. Evaluation of numerical
❘❡s✉❧t ♦❢ ❛♣♣❧②✐♥❣ t❤❡ ♠❡t❤♦❞ ❢♦r ✹✵✵ ■✲❱ ❝✉r✈❡s ✭♠❡❛s✉r❡❞ algorithms used in extracting the parameters of a single-diode photo-
✉♥❞❡r ❞✐✛❡r❡♥t ❡♥✈✐r♦♥♠❡♥t❛❧ ❝♦♥❞✐t✐♦♥s✮ ❢♦r ❡❛❝❤ P❱ s❛♠♣❧❡ voltaic model. Sustainable Energy Technologies and Assessments, 13:
✉s✐♥❣ ❞✐✛❡r❡♥t ♠❛①✐♠✉♠ ✐t❡r❛t✐♦♥ ♥✉♠❜❡rs 51–59, 2016.
M Azzouzi, D Popescu, and M Bouchahdane. Modeling of electrical char-
P❱ ❙❛♠♣❧❡ ◆✉♠❜❡r ♦❢ ❆✈❡r❛❣❡ ❆✈❡r❛❣❡ ❆✈❡r❛❣❡ acteristics of photovoltaic cell considering single-diode model. Journal
✐t❡r❛t✐♦♥s 𝜹 ❘𝟐 ▼❆P❊ of Clean Energy Technologies, 4(6):414–20, 2016.
J. D. Bastidas-Rodriguez, G. Petrone, C. A. Ramos-Paja, and G. Spagnuolo.
50 0.03715 0.94353 4.921
100 0.03522 0.95027 4.474 A genetic algorithm for identifying the single diode model parameters of
250 0.01202 0.99005 1.429 a photovoltaic panel. Mathematics and Computers in Simulation, 131:
500 0.00595 0.99908 0.598 38–54, 2014. ISSN 03784754. doi: 10.1016/j.matcom.2015.10.008.
❈r②st❛❧❧✐♥❡ ❝❡❧❧ 1000 0.00561 0.99938 0.505 Adi Ben-Israel. A newton-raphson method for the solution of systems of
2000 0.00526 0.99948 0.450 equations. Journal of Mathematical analysis and applications, 15(2):
3000 0.00426 0.99973 0.324 243–252, 1966.
4000 0.00388 0.99979 0.290 Dominique Bonkoungou, Zacharie Koalaga, and Donatien Njomo. Mod-
5000 0.00376 0.99980 0.273 elling and simulation of photovoltaic module considering single-diode
equivalent circuit model in matlab. International Journal of Emerging
50 0.04006 0.91473 3.477
Technology and Advanced Engineering, 3(3):493–502, 2013.
100 0.03634 0.94073 3.159
250 0.01308 0.98579 1.464 H Bourdoucen and A Gastli. Analytical modelling and simulation of photo-
500 0.01108 0.99231 0.968 voltaic panels and arrays. The Journal of Engineering Research [TJER],
P♦❧②✲♠♦❞✉❧❡ 1000 0.01041 0.99272 0.927 4(1):75–81, 2007.
2000 0.00996 0.99407 0.893 Valerio Lo Brano and Giuseppina Ciulla. An efficient analytical approach
3000 0.00897 0.99450 0.825 for obtaining a five parameters model of photovoltaic modules using only
4000 0.00910 0.99470 0.815 reference data. Applied Energy, 111:894–903, 2013.
5000 0.00887 0.99580 0.809 Hayrettin Can and Damla Ickilli. Parameter estimation in modeling of pho-
tovoltaic panels based on datasheet values. Journal of Solar Energy
50 0.01847 0.95696 3.444
Engineering, 136(2):021002, 2014.
100 0.01789 0.95824 3.393
Daniel SH Chan and Jacob CH Phang. Analytical methods for the extraction
250 0.01783 0.97973 2.326
500 0.01387 0.99090 1.671 of solar-cell single-and double-diode model parameters from iv charac-
❆♠♦r✳✲♠♦❞✉❧❡ 1000 0.01141 0.99361 1.474 teristics. IEEE Transactions on Electron devices, 34(2):286–293, 1987.
2000 0.01015 0.99677 1.159 M. Chegaar, Z. Ouennoughi, and A. Hoffmann. A new method for evalu-
3000 0.01041 0.99693 1.148 ating illuminated solar cell parameters. Solid-State Electronics, 45(2):
4000 0.00991 0.99748 1.103 293–296, feb 2001. ISSN 00381101.
5000 0.00983 0.99759 1.092 Ali Chikh and Ambrish Chandra. An optimal maximum power point track-
ing algorithm for pv systems with climatic parameters estimation. IEEE
Transactions on Sustainable Energy, 6(2):644–652, 2015.
Declaration of Conflict of Interest Javier Cubas, Santiago Pindado, and Carlos De Manuel. Explicit expres-
sions for solar panel equivalent circuit parameters based on analytical
The authors declare that they have no known competing formulation and the lambert w-function. Energies, 7(7):4098–4115,
financial interests or personal relationships that could have 2014.
appeared to influence the work reported in this paper. Abhik Kumar Das. An explicit j–v model of a solar cell for simple fill factor
calculation. Solar Energy, 85(9):1906–1909, 2011.
Abhik Kumar Das. Analytical expression of the physical parameters of an
Acknowledgments illuminated solar cell using explicit j–v model. Renewable energy, 52:
95–98, 2013a.
The research reported in this paper was partially sup- Abhik Kumar Das. An explicit j–v model of a solar cell using equivalent
ported by the Stipendium Hungaricum Scholarship Programme, rational function form for simple estimation of maximum power point
the BME-Nanotechnology FIKP grant (BME FIKP-NAT), voltage. Solar Energy, 98:400–403, 2013b.
the grant EFOP-3.6.1-16-2016-00014 and by the Science Ex- Maria Carmela Di Piazza, Massimiliano Luna, Giovanni Petrone, and Gio-
vanni Spagnuolo. Translation of the single-diode pv model parameters
cellence Program at BME under the grant agreement NKFIH-
identified by using explicit formulas. IEEE Journal of Photovoltaics, 7
849-8/2019 of the Hungarian National Research, Develop- (4):1009–1016, 2017a.
ment and Innovation Office. MC Di Piazza, M Luna, G Petrone, and G Spagnuolo. Parameter translation
for single-diode pv models based on explicit identification. In Environ-
ment and Electrical Engineering and 2017 IEEE Industrial and Com-
References mercial Power Systems Europe (EEEIC/I&CPS Europe), 2017 IEEE In-
Ali Kareem Abdulrazzaq. Electro-Thermal Modeling of Photovoltaic (PV) ternational Conference on, pages 1–5. IEEE, 2017b.
Systems. In Spring Wind 2018, pages 17–29, Györ, 2018. Association K Et-torabi, I Nassar-eddine, A Obbadi, Y Errami, R Rmaily, S Sahnoun,
of Hungarian PhD and DLA students. M Agunaou, et al. Parameters estimation of the single and double
Mehmet Akbaba. Performance analysis of solar cell arrays loaded with diode photovoltaic models using a gauss–seidel algorithm and analyti-
passive loads. Applied energy, 52(2-3):209–218, 1995. cal method: A comparative study. Energy Conversion and Management,
J Appelbaum and A Peled. Parameters extraction of solar cells–a compar- 148:1041–1054, 2017.
ative examination of three methods. Solar Energy Materials and Solar Ghias Farivar and Behzad Asaei. Photovoltaic module single diode model
Cells, 122:164–173, 2014.

❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier P❛❣❡ ✶✷ ♦❢ ✶✸


❆❝❝✉r❛t❡ ♠❡t❤♦❞ ❢♦r P❱ s♦❧❛r ❝❡❧❧s ❛♥❞ ♠♦❞✉❧❡s ♣❛r❛♠❡t❡rs ❡①tr❛❝t✐♦♥ ✉s✐♥❣ ■✲❱ ❝✉r✈❡s

parameters extraction based on manufacturer datasheet parameters. In toactive Layer Crystalline Silicon Solar Cells. PhD thesis, Budapest
Power and energy (PECon), 2010 IEEE international conference on, University of Technology and Economics, 2015.
pages 929–934. IEEE, 2010. J. Prasanth Ram, T. Sudhakar Babu, Tomislav Dragicevic, and N. Rajasekar.
Ahmed Fathy and Hegazy Rezk. Parameter estimation of photovoltaic sys- A new hybrid bee pollinator flower pollination algorithm for solar PV
tem using imperialist competitive algorithm. Renewable Energy, 111: parameter estimation. Energy Conversion and Management, 135:463–
307–320, 2017. ISSN 18790682. doi: 10.1016/j.renene.2017.04.014. 476, 2017. ISSN 01968904. doi: 10.1016/j.enconman.2016.12.082.
F Ghani and Mike Duke. Numerical determination of parasitic resistances Mohamed BH Rhouma, Adel Gastli, Lazhar Ben Brahim, Farid Touati, and
of a solar cell using the lambert w-function. Solar Energy, 85(9):2386– Mohieddine Benammar. A simple method for extracting the parameters
2394, 2011. of the pv cell single-diode model. Renewable Energy, 113:885–894,
F Ghani, Mike Duke, and J Carson. Numerical calculation of series and 2017.
shunt resistances and diode quality factor of a photovoltaic cell using Tor Oskar Saetre, Ole-Morten Midtgård, and Georgi Hristov Yordanov. A
the lambert w-function. Solar Energy, 91:422–431, 2013. new analytical solar cell i–v curve model. Renewable Energy, 36(8):
F Ghani, G Rosengarten, M Duke, and JK Carson. The numerical calcula- 2171–2176, 2011.
tion of single-diode solar-cell modelling parameters. Renewable Energy, Leonardo Sandrolini, Marcello Artioli, and Ugo Reggiani. Numerical
72:105–112, 2014. method for the extraction of photovoltaic module double-diode model
Mohammad Hejri, Hossein Mokhtari, Mohammad Reza Azizian, Mehrdad parameters through cluster analysis. Applied Energy, 87(2):442–451,
Ghandhari, and Lennart Soder. On the Parameter Extraction of a Five- 2010.
Parameter Double-Diode Model of Photovoltaic Cells and Modules. Dezso Sera, Remus Teodorescu, and Pedro Rodriguez. Pv panel model
IEEE Journal of Photovoltaics, 4(3):915–923, may 2014a. ISSN 2156- based on datasheet values. In Industrial Electronics, 2007. ISIE 2007.
3381. doi: 10.1109/JPHOTOV.2014.2307161. IEEE International Symposium on, pages 2392–2396. IEEE, 2007.
Mohammad Hejri, Hossein Mokhtari, Mohammad Reza Azizian, Mehrdad Emerson A. Silva, Fabricio Bradaschia, Marcelo C. Cavalcanti, and
Ghandhari, and Lennart Soder. On the parameter extraction of a five- Aguinaldo J. Nascimento. Parameter Estimation Method to Improve
parameter double-diode model of photovoltaic cells and modules. IEEE the Accuracy of Photovoltaic Electrical Model. IEEE Journal of Pho-
Journal of Photovoltaics, 4(3):915–923, 2014b. tovoltaics, 6(1):278–285, jan 2016. ISSN 2156-3381. doi: 10.1109/
Abdulrahman Homadi, Tony Hall, and Lawrence Whitman. Using solar JPHOTOV.2015.2483369.
energy to generate power through a solar wall. Journal of King Saud Emerson A. Silva, Fabricio Bradaschia, Marcelo C. Cavalcanti,
University - Engineering Sciences, mar 2020. ISSN 10183639. doi: Aguinaldo Jose Nascimento, Leandro Michels, and Luiz Paulo
10.1016/j.jksues.2020.03.003. URL https://linkinghub.elsevier.com/ Pietta. An Eight-Parameter Adaptive Model for the Single Diode
retrieve/pii/S1018363919303459. Equivalent Circuit Based on the Photovoltaic Module’s Physics. IEEE
Seyedkazem Hosseini, Shamsodin Taheri, Masoud Farzaneh, Hamed Journal of Photovoltaics, 7(4):1115–1123, jul 2017. ISSN 2156-3381.
Taheri, and Mehdi Narimani. Determination of Photovoltaic Charac- doi: 10.1109/JPHOTOV.2017.2703778.
teristics in Real Field Conditions. IEEE Journal of Photovoltaics, 8(2): FJ Toledo and José M Blanes. Analytical and quasi-explicit four arbitrary
572–580, mar 2018. ISSN 2156-3381. doi: 10.1109/JPHOTOV.2018. point method for extraction of solar cell single-diode model parameters.
2797974. Renewable Energy, 92:346–356, 2016.
Haider Ibrahim and Nader Anani. Evaluation of analytical methods for Silvano Vergura. A complete and simplified datasheet-based model of pv
parameter extraction of pv modules. Energy Procedia, 134:69–78, 2017. cells in variable environmental conditions for circuit simulation. Ener-
Utkarsh Jadli, Padmanabh Thakur, and Rishabh Dev Shukla. A New gies, 9(5):326, 2016.
Parameter Estimation Method of Solar Photovoltaic. IEEE Journal A. Vijayakumari. A non-iterative MPPT of PV array with online measured
of Photovoltaics, 8(1):239–247, jan 2018. ISSN 2156-3381. doi: short circuit and open circuit quantities. Journal of King Saud University
10.1109/JPHOTOV.2017.2767602. - Engineering Sciences, apr 2020. ISSN 10183639. doi: 10.1016/j.
Jong Pil Kim, Ho Lim, Ju Hun Song, Young June Chang, and Chung Hwan jksues.2020.04.007. URL https://linkinghub.elsevier.com/retrieve/
Jeon. Numerical analysis on the thermal characteristics of photovoltaic pii/S1018363920302269.
module with ambient temperature variation. Solar Energy Materials and Georgi Hristov Yordanov, Ole-Morten Midtgard, and Tor Oskar Saetre. Ide-
Solar Cells, 95(1):404–407, 2011. ality factor behavior between the maximum power point and open circuit.
Shu-xian Lun, Shuo Wang, Gui-hong Yang, and Ting-ting Guo. A new In Photovoltaic Specialists Conference (PVSC), 2013 IEEE 39th, pages
explicit double-diode modeling method based on lambert w-function for 0729–0733. IEEE, 2013.
photovoltaic arrays. Solar Energy, 116:69–82, 2015.
Tao Ma, Hongxing Yang, and Lin Lu. Solar photovoltaic system mod-
eling and performance prediction. Renewable and Sustainable Energy
Reviews, 36:304–315, 2014. ISSN 13640321.
Donald W Marquardt. An algorithm for least-squares estimation of nonlin-
ear parameters. Journal of the society for Industrial and Applied Math-
ematics, 11(2):431–441, 1963.
Hitesh K. Mehta, Himanshu Warke, Kaushik Kukadiya, and Ashish K. Pan-
chal. Accurate Expressions for Single-Diode-Model Solar Cell Param-
eterization. IEEE Journal of Photovoltaics, 9(3):803–810, may 2019.
ISSN 2156-3381. doi: 10.1109/JPHOTOV.2019.2896264.
D. Menaga and V. Sankaranarayanan. Performance comparison for grid
connected photovoltaic system using sliding mode control. Jour-
nal of King Saud University - Engineering Sciences, may 2020.
ISSN 10183639. doi: 10.1016/j.jksues.2020.04.012. URL https:
//linkinghub.elsevier.com/retrieve/pii/S1018363920302312.
Luz Elena P. aranda Chenche, Oscar Saul Hernandez Mendoza, and
Enio Pedone Bandarra Filho. Comparison of four methods for parameter
estimation of mono- and multi-junction photovoltaic devices using ex-
perimental data. Renewable and Sustainable Energy Reviews, 81(June
2017):2823–2838, jan 2018. ISSN 13640321.
Balázs Plesz. Investigation of the Temperature Dependence of Thin Pho-

❋✐rst ❆✉t❤♦r ❡t ❛❧✳✿ Preprint submitted to Elsevier P❛❣❡ ✶✸ ♦❢ ✶✸

You might also like