Materials Today: Proceedings 52 (2022) 108–123

Contents lists available at ScienceDirect

Materials Today: Proceedings

journal homepage: www.elsevier.com/locate/matpr

Multiple learning JAYA algorithm for parameters identifying of

photovoltaic models
Driss Saadaoui a,⇑, Mustapha Elyaqouti a, Khalid Assalaou a, Dris Ben hmamou b, Souad Lidaighbi a
a
Laboratory of Materials and Renewable Energies, Faculty of Science, Ibn Zohr University, Agadir, Morocoo
b
Laboratory of Electronics, Signal Processing and Physical, Faculty of Science, Ibn Zohr University, Agadir, Morocoo

a r t i c l e i n f o a b s t r a c t

Article history: Extracting the optimum parameters of the solar photovoltaic (PV) model based on measured current–
Available online 26 November 2021 voltage data accurately, reliably, and quickly is a vital step to simulating, evaluating, and optimizing
the model PV. Although several meta-heuristics have been proposed to extract photovoltaic parameters
Keywords: using experimental data, it suffers from slow convergence and poor efficiency which leads to poor solu-
Parameter extraction tion quality. In this paper, a multiple learning JAYA (MLJAYA) approach has been proposed to extract the
Photovoltaic models PV parameters under different operating conditions. The main advantage of multiple learning strategies
Multiple learning JAYA
introduced in the MLJAYA proposed is its ability to balance exploration and exploitation capabilities. The
Optimization
chaos perturbation mechanism is employed to improve the quality of the population to escape the local
optimum and improve the accuracy of the basic JAYA algorithm. The structure of MLJAYA is very simple
and requires only the size of the population and stop criterion. The efficacy of the proposed MLJAYA has
been verified by including different PV modules models such as single diode, double diode, and PV mod-
ule, and the results obtained was compared with different variants of JAYA and other state-of-the-art
optimization algorithms. The best results of root mean square error (RMSE), and absolute error generated
by MLJAYA are (RMSE = 9.8602E
04, SIAE = 1.781248E
02), and (RMSE = 9.8248E
04, SIAE = 1.76E
02)
for RTC solar cell single diode and double diode respectively, and (RMSE = 2.4250748E
03, SIAE = 4.68
6375E
02) for Photowatt-PWP201. The results show the superior performance in terms of accuracy
and reliability. Since, the MLJAYA show is a potential approach for extracting parameters from photo-
voltaic models.
Ó 2021 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the International Network
of Biomaterials and Engineering Science.

1. Introduction general, photovoltaic systems operate in harsh climatic conditions

and photovoltaic panels degrade easily, which affects the solar
Unluckily, fossil fuel energy use has had serious and rising neg- energy utilization efficiency [9]. Therefore, to optimize PV systems,
ative environmental impacts, such as CO2 emissions, global warm- it is important to assess the actual behavior of PV panels in opera-
ing, air pollution, and overall global environment degradation [4]. tion using an efficient model based on measured current–voltage
Our future energy needs must be met by a combination of sustain- data. Several PV models have been used to describe the nonlinear
able technologies with minimal environmental impact. Potentially, characteristics of PV systems, such as the single diode model, dou-
the majority of these technologies will use solar energy in all its ble diode model, and three diode model [10-14]. Overall, the
forms [5]. However, solar cells are a major contributor to the pro- parameters of PV models affect the performance of PV systems,
duction of environmentally friendly electricity. Solar energy is con- so it is important to extract the parameters of PV models accu-
verted into electrical energy using a system of photovoltaic cells rately to improve the performance of PV models. Up to now, sev-
that convert the absorbed sunlight into direct electrical current eral approach have been proposed to estimate parameters of PV
by utilizing the intrinsic properties of semiconductors [6-8]. In models, can be broadly classified into deterministic and heuristic
methods [15]. The deterministic techniques require a series of
model restrictions, including differentiability and convexity, in
⇑ Corresponding author. order to be properly applied. So, numerous meta-heuristic algo-
E-mail address: [email protected] (D. Saadaoui).

https://doi.org/10.1016/j.matpr.2021.11.106
2214-7853/Ó 2021 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the International Network of Biomaterials and Engineering Science.
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

rithms have already been in the estimation of single diode model rithm is applied in many optimization problems, such as PV system
(SDM) and double diode model (DDM). In heuristic algorithms, maximum power point (MPP) tracking [35], sizing optimization of
exploration and exploitation are two very important features to a micro-channel heat sink [36], solving constrained mechanical
achieve success in solving a particular optimization problem. design problems [37], hydro-static thrust bearing and rolling ele-
Exploration refers to the ability of an algorithm to detect a wide ment bearing [37], and other real-world problems [38,39]. In the
range of solutions, located in different areas of the search space. JAYA algorithm, the solution update is based on the average of
In contrast, exploitation emphasizes the idea of strengthening the best solution and the worst solution simultaneously, even
the research mechanism on solution-rich areas with the aim of though the convergence rate is enhanced, does not guarantee
finding better solutions or improving existing solutions [16]. Each diversity of the population, weakening the exploration ability.
of the preceding strategies has these advantages: (1) Exploitation Therefore, many JAYA modified have been introduced in an
strategies are known to improve the speed of convergence towards attempt to improve the population diversity of the algorithm by
a global optimum, and also to increase the probability of confine- implementing other update strategies. Jian and Weng [30] pro-
ment in local optima. (2) Exploration strategies are known to posed a Logistic Chaotic JAYA (LCJAYA) algorithm for identifying
increase the probability of finding regions in the search space the parameters of PV cell and module models. In this approach, a
where the global optimum is more favorable to be found, this logistic chaotic map strategy is introduced into the solution updat-
may influence the speed of convergence of the algorithm [17]. ing phase of JAYA algorithm, which improves the population diver-
Many meta-heuristic approaches have been used and improved sity of algorithm. An improved JAYA (IJAYA) optimization
to extract the PV models parameters in the past decade. They are algorithm is proposed by Yu et al. [29] to solve the parameters
memetic adaptive differential evolution (MADE) [18], teaching– identification of PV models problem. In this approach, a self-
learning-based artificial bee colony (TLABC) algorithm [19], niche adaptive weight is introduced to adjust the tendency to approach
particle swarm optimization in parallel computing (NPSOPC) algo- the best solution and avoid the worst solution during the different
rithm [20], modified JAYA algorithm [21], genetic algorithm [22], stages of the search, which enables the algorithm to approach the
differential evolution with biogeography based optimization (DE/ promising area at the early stage and implement the local search at
BBO) [23], improved chaotic whale optimization algorithm the later stage. The performance improvement of the IJAYA is lim-
(CWOA) [24]. ited the fact that the elite chaotic learning method used to refine
However, the performance of these meta-heuristic algorithms the quality of the best solution at each generation, it did not take
depends on the determination of the specific parameters of the into account that different perturbations around the best solution
algorithm. Any bad parameter will lead to early convergence, should be attributed to different search stages. Chaotic JAYA
reduced optimization accuracy and increased computational load. (CJAYA) [40], is developed to explore the search space without get-
The competitive swarm optimizer (CSO) [25] is an improved vari- ting trapped in the regions of local optima. The drawback of CJAYA
ant of particle swarm optimization (PSO) that gives good results, is the imbalance between exploration and exploitation. Therefore,
generally for unimodal optimization problems. On the other hand, the performance improvement of CJAYA is bounded and should be
it is quickly trapped in local optima when solving complex multi- improved when solving the parameter estimation problems
modal optimization problems, due to its poor exploration. Winner- parameter estimation problems of PV models. Notwithstanding
leading competitive swarm optimizer with dynamic Gaussian that the previously presented meta-heuristic algorithms can pro-
mutation (WLCSODGM) [26] proposed for extracting the parame- duce accurate results, it is sometimes difficult for them to system-
ters of the PV models. It found that this variant of PSO falls into atically find a more accurate and consistent global optimal
local minima for complex multimodal optimization problems such solution. First of all, the estimation of the PV model parameters
as the PV model parameter extraction problem due to its poor is a multi-modal problem with many local optima. Following this,
exploration capability. Liang, et al. [27] proposed a classified per- these heuristic algorithms, each have their own evolutionary
turbation mutation-based PSO algorithm (CPMPSO), for parame- method that is adapted to different types of problems, including
ters extraction of PV models. In this approach, the authors mono-modal, multi-modal or complex problems [41]. Moreover,
divided the individuals into two categories according to the fitness the algorithms each have their own parameters, which require an
values: The first one has individuals with high quality and is adjustment to adapt to different phases of evolution and to differ-
updated using an effective exploitation operator, whereas, in the ent problems.
second, individuals were refreshed based on an effective explo- Based on the above considerations, this paper proposes a mod-
ration operator. Its convergence speed still needs improvement. ified JAYA algorithm namely Multiple Learning JAYA algorithm
Furthermore, it has a hard time avoiding being trapped in local (MLJAYA) algorithm to extract parameters from PV cell single
minima for the DDM. In addition, many meta-heuristic algorithms and double diodes and module models with greater accuracy and
require additional specific control parameters, such as acceleration reliability. Three main improvements exist in MLJAYA. Firstly, a
factors in PSO, number of mates and mutation control factor in bird self-adaptive weight is applied to adjust the matching tendency
mating optimizer (BMO), mutation probability and crossover prob- towards the best solution and to avoid the worst solution. This
ability in genetic algorithm (GA). The different algorithms intro- weight is intended to guide the algorithm to access the potential
duced, the modeling type, type of solar cell /panel, performance area at an early stage and to implement the local search at a later
criteria and results were listed in Table 1. stage. Secondly, a multiple learning strategy has been introduced
Recently, JAYA method is among the most recent meta-heuristic to improve the global search capacity and the balance between
algorithms, which was first proposed by Rao [28] for solving differ- exploitation and exploration. Thirdly, to improve the quality of
ent types of optimization problems, it does not require any addi- the population, chaos perturbation mechanism is used to escape
tional parameters specific [28]. This algorithm has two the local optimal and improves the precision of the basic JAYA
drawbacks: first, a lack of population diversity leads to poor overall algorithm. Moreover, MLJAYA has a very simple design and doesn’t
search capacity; second, insufficient exploration capacity leads to add any new parameters to Jaya. The MLJAYA algorithm is vali-
an imbalance between exploration and exploitation [29]. In Ref. dated using single diode model, double diode model, and PV mod-
[30], JAYA is used to extract the unknown parameters of PV mod- ule model. The comparison demonstrates the accuracy of the
els. The obtained results demonstrates JAYA’s excellent optimiza- MLJAYA over well-known algorithms.
tion in relation to other well-known methods such as TLBO [31], To summarize, the main contributions of this study are as
CS [32], BLPSO [33], LBSA [34], and NNA [30]. To date, JAYA algo- follows:
109
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Table 1
Meta-heuristic algorithms used for PV parameter estimation.

Ref. Year Used algorithms Modeling Type of solar cell / Performance criteria Results
type panel
[42] 2020 grey wolf optimizerand SDM, DDM Photowatt-PWP- Best RMSEMean  It is a promising candidate technique for
cuckoo search(GWOCS) 201 module RMSEWorst extracting the parameters of solar PV cell mod-
modelSTM6-40/36 RMSEMedian els under different operating conditions.
module model RMSEStandard  The accuracy of solution and convergence
deviation (Std. dev.) speed of global optimization problems show
that GWOCS performs better than other algo-
rithms such as GWO, EGWO, and mGWO.
[43] 2020 Transient Search Three-diode KC-200-GT ; MSX- Extraction  The TSO algorithm achieved the best minimum
Optimization (TSO) model (TDM) 60 CS6K280M timeobjective values for the objective function among other
of the PV functionBest fitness algorithms such as GWO, WOA, and SFO
module.  The TSO approach and the objective function
can be exploited to find the TDM model of all
commercial PV cells based on the dataset val-
ues of PVs.
[19] 2018 teaching–learning-based SDM; DDM RTC France silicon Best RMSEMean  Experimental results demonstrate that TLABC
artificial bee colony (TLABC) solar RMSEWorst RMSEStd. has highly competitive performance in terms
cellPhotowatt- dev. of accuracy and reliability different PV parame-
PWP-201 ter estimation problems.
[44] 2017 Evaporation Rate based SDM; DDM KC-200-GTSX RMSE ; AEMAE ; REMRE  In terms of RMSE, MAE and MRE, the ER-WCA is
Water Cycle Algorithm (ER- 200 N1STH-235- ; NFE advantageous to NMMPSO, GOTLBO, MABC,
WCA) WH CSO, BBO-M methods even under changing
irradiation and temperature conditions
 ER-WCA is a promising optimization technique
for PV cell/module identification.
[34] 2018 Multiple learning SDM; DDM RTC solar Best RMSEMean  The experimental and statistical analyses show
backtracking search cellPhotowatt- RMSEWorst RMSEStd. the superiority of MLBSA to solve the parame-
algorithm (MLBSA) PWP201 dev.; IAE ters identification problems of different PV
compared with other methods in terms of
accuracy, reliability, and computational
efficiency.
 The MLBSA can be seen as an efficient method
for solving optimization problems in any
energy and non-energy system.
 The performance of MLBSA can be degraded
when solving larger or smaller systems com-
pared to other methods.
[45] 2020 Grasshopper Optimization TDM KC200GT ; MSX-60 Objective function  The GOA approach has been effective in opti-
Algorithm (GOA) convergence.AE mizing the parameters of the three-diode
model.
 The GOA is capable of solving other optimiza-
tion problems in several research areas such
as wind energy systems, other renewable
energy systems and smart grids.
[46] 2021 Stochastic FractalSearch SDM; DDM R.T.C solar cellSTP6 Residual auto  The obtained results by STS method show the
(SFS) 120/36ESP-160 correlation function superiority, perfectness, and effective modeling
PPW (RCAF)AE ; MAE ; RMSE concerning various performance parameters.
 The SFS method can be used as a high valued
optimization technique for the extraction of
solar PV parameters under any test conditions.
 The SFS method defines a new error metric
(RCAF)
[47] 2018 Differential evolution with SDM; DDM R.T.C solar Mean of IAEsPower  The DE/WOA algorithm is validation with 13
whale optimization cellPhotowatt- errorCurrent classical numerical benchmark functions
algorithm (DE/WOA) PWP201 errorMean ; Std. dev.  The experimental results demonstrate that DE/
WOA can present encouraging results under
different irradiances, temperatures, and
dynamic weatherconditions, comparing with
original DE, WOA, and five advanced variants
of them and is highly competitive with some
recently-proposed parameter extraction
methods.
[32] 2019 biogeography-based hetero- SDM; DDM R.T.C solar Best RMSEMean  The best performance of BHCS is due to the
geneous cuckoo search cellSTM6-40/ RMSEWorst combination of two strategies, namely the
(BHCS) 36STP6-120/36 RMSEMedian RMSEStd. heterogeneous cuckoo search strategy and the
dev. biogeography-based discovery operator.
 In comparison with CS, BBO and several other
meta-heuristic algorithms, BHCS has very com-
petitive performances in terms of accuracy and
reliability.
 The BHCS algorithm is an efficient tool for PV
parameter estimation.

110
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Table 1 (continued)

Ref. Year Used algorithms Modeling Type of solar cell / Performance criteria Results
type panel
[48] 2021 gradient-based optimizer SDM; DDM; R.T.C solar cell AE; RMSE  The GBO algorithm has many advantages,
(GBO) TDM including solution accuracy, balance, and con-
vergence speed between analysis and
exploitation.
 The results obtained by the GBO were more
accurate than those obtained by most of the
ten competitor algorithms.
 GBO is a good candidate to solve solar cell sys-
tem optimization problems.
[49] 2021 Reinforcement learning- SDM; DDM R.T.C solar cell Best RMSE  The RLDE algorithm essentially applies RL to
based differential evolution Photowatt- Mean RMSE the selection and adjustment of operators in
(RLDE) PWP201 Worst RMSE DE.
STM6-40/36 Std. dev.  RLDE can be considered as a very promising
STP6-120/36 approach for parameter extraction of other
complex PV models.
[50] 2017 Shuffled complex evolution DDM Known PV RE; SSE; MAE; MBE ;  Compared with AM, LM, GA, DE, and PSO
(SCE) cellUnknown PV MAEP ; RMSE approach, the results obtained show that SCE
module has a low convergence computation time and
a significant ability to solve all global optimiza-
tion problems.
[51] 2021 Improved gaining-sharing SDM; DDM R.T.C solar Best RMSEMean  The IGSK algorithm will not allow solving
knowledge (IGSK) cellPhotowatt- RMSEWorst multi-objective optimization problems or
PWP201 RMSEMedian RMSEStd. problems under constraints.
STM6-40/36STP6- dev  Compared with other state-of-the-art algo-
120/36ST40 ; rithms, IGSK has excellent performance in
SM55 terms of convergence speed and accuracy and
reliability of parameter values.
[52] 2016 Differential Evolution and SDM KC120-1 multi- MBE ; RMSECorrelation  The achieved results demonstrate the superior-
Electro-magnetism like crystalline 120 Wp coefficient R2CPU ity of the proposed evolutionary method with
algorithms PV execution time integrated mutation per iteration and the eval-
uative algorithm with adaptive mutation per
iteration, compared to the electromagnetic
type algorithm electromagnetic.
 These advantages are tied to high estimation
accuracy, rapid convergence, and fewer control
parameters
[26] 2020 Winner-leading competitive SDM; DDM R.T.C solar cell IAE; SIAE; RMSE  The approach (WLCSODGM) presented per-
swarm optimizer with Photowatt- Best RMSE forms better in terms of the accuracy of param-
dynamic Gaussian Mutation PWP201 Mean RMSE eters, robustness, convergence, and statistics.
(WLCSODGM) KC200GT Worst RMSE  Both coefficients of broadening E and narrow-
MSX-60 Std. dev. ing S in the DGM have a significant effect on
the performance of the WLCSODGM method.


 
 A novel method MLJAYA is developed to extract the unknown qðV þ Rs IÞ V þ Rs I
I ¼ Iph
Io exp
1
ð1Þ
parameters of PV cell and module models. akT Rsh
 In MLJAYA, a multiple learning strategy is used to ensure a good
where Iph denotes the photo current, Io indicate the reverse satura-
balance between exploitation and exploration.
 A chaos perturbation mechanism is proposed to improve the tion current of the diode, a ideality factor , Rs and Rsh denotes series
quality of the best solution in each generation. resistance and shunt resistance respectively,
 The performance of MLJAYA is exhaustively evaluated through q ¼ 1:60217646  10
19 C denotes the electron charge, k is the
parameter identification problems of different PV models. Boltzmann constant (k ¼ 1:3806503  10
23 J=K) , and T is the tem-
perature in Kelvin (K).
The remainder of this paper is structured as follows. PV models
and mathematical background is presented in Section 2. Section 3 2.1.2. Double diode model
description of the LMJAYA algorithm. Results and discussions are The double diode (DDM) model is more precise, accurate, and
presented in Section 4. Finally, conclusions are presented in better than the single diode (SDM) because in this case, it takes
Section 5. into account the effect of recombination current loss in the deple-
tion region [54-60]. In this model, two diodes are connected in par-
allel with a shunt resistance and series resistance. Fig. 1(b) shows
2. Photovoltaic models (PV) and mathematical background
the equivalent circuit of the DDM. The output current I in the dou-
ble diode circuit of Fig. 1(b) is given by Eq. (2):
2.1. Solar cells

  
 
qðV þRs IÞ qðV þRs IÞ V þRs I
2.1.1. Single diode model I¼ Iph
I01 exp
1
I02 exp
1

a1 kT a2 kT Rsh
Due to its simplicity, the single diode model (SDM) is usually
ð2Þ
applied to simulate the characteristic current–voltage of the solar
cell. Fig. 1(a) shows the equivalent circuit of SDM [53]. where I01 and I02 are the reverse saturation current of the diode D1
In SDM, the current I and voltage V at the output terminal of the and D2 respectively, a1 denotes the ideality factor of the diode D1
PV panel module are related by the equation [32,54]: and a2 is the ideality factor of the diode D2.
111
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Fig. 1. The equivalent circuit (a) for the single-diode model and (b) for double-diode model.
     
2.2. Photovoltaic module model ynewi;j ¼ yi;j þ a1  ybest;j
yi;j 
a2  yworst;j
yi;j  ð5Þ

The PV module model contains Ns cells connected in series and where yi;j denotes the j-th variable of the i-th candidate solution, |
Np cells connected in parallel. The equivalent electric circuit of the yi;j | is the absolute value of yi;j , ybest;j and yworst;j respectively denotes
PV panel module is shown in Fig. 2. Furthermore, the output cur- the values of the j-th variable of the best solution and the worst
rent of the PV module is expressed as follows [18]: solution, ynewi;j is the update variable of yi;j , a1 and a2 are two ran-

  dom numbers in [0,1].
qðV=Ns þ Rs I=Np Þ V=Ns þ Rs I=Np
I=Np ¼ Iph
Io exp
1
ð3Þ In the Jaya algorithm, the solution is updated only using one
akT Rsh
solution converges to the best solution and another solution
diverges from the worst solution simultaneously [28]. Thereby, a
2.3. Objective function good intensification and diversification of the search process
attained.
The parameters estimation of the solar PV model is studied as
an optimization problem to minimize the difference between mea- 3.2. The proposed MLJAYA
sured data and the simulated ones. However, the root mean square
error (RMSE) is a widely used objective function to quantify the The MLJAYA algorithm is a modified version of the JAYA algo-
total error, as shown in Eq. (4) [27,32,42,44,57,58] : rithm for solving the parameters identification problems of differ-
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u ent PV models. The solutions in the MLJAYA algorithm are updated
u X N
RMSE ¼ t1=N ðImeasu
Itheor Þ2 ð4Þ in a similar manner as in the JAYA algorithm based on Eq. (5). How-
i¼1
ever, in order to handle complex optimization problems effectively
and efficiently, the MLJAYA algorithm is incorporated with self-
where, Imeasu and Itheor are the measured and simulated current adaptive weight, multiple learning strategy and a chaos perturba-
respectively and N is the number of experimental data. tion mechanism. Firstly, a self-adaptive weight is applied to adjust
the matching tendency towards the best solution and to avoid the
3. Multiple learning JAYA algorithm (MLJAYA) worst solution. Secondly, a multiple learning strategy has been
introduced to achieve a good balance between the exploration
3.1. JAYA algorithm and exploitation abilities. Thirdly, Chaos perturbation mechanism
is used to escape the local optimal and improves the precision of
JAYA algorithm is simple and has proved the most effective in the basic JAYA algorithm [60]
solving the optimizing problems with constrained and uncon-
strained [59]. In this algorithm the initial solutions are randomly 3.2.1. Self-adaptive weight
generated by respecting the upper and lower bounds of the process In the search procedure of JAYA, the population must reach the
variables. The search mechanism of the JAYA algorithm is driven by promising region of the search region at the early stage, and local
the current optimal solution and the current worst solution, which search should be implemented in promising way to refine the qual-
can be described as follows. ity of population. For this purpose, the self-adaptive strategy is

Fig. 2. Equivalent circuit of PV models.

112
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

introduced to adjust the degree of avoiding the worst solution and the function values of the best and worst solution decreases as the
approaching the best solution. For this purpose, Eq. (5) is replaced search process proceeds. Thus, taking into account the position of
by Eq. (6): the best solution, the optimal area can be located at the beginning,
      while the local search in the promising region can be reached at
ynewi;j ¼ yi;j þ a1  ybest;j
yi;j 
W:a2  yworst;j
yi;j  ð6Þ the later stage, because the probability of the best solution being
The weight W presented in Eq. (7) is given by: approached and that of avoiding the worst solution are the same.
8
< g ðybest Þ
2
if yworst –0 3.2.2. Multiple learning strategy
W¼ g ðyworst Þ ð7Þ
: In the conventional JAYA algorithm, the best solution and worst
1 otherwise
solution plays an important role in updating the population, this
where g ðybest Þ denotes the objective function values of the best solu- approach can speed up the convergence rate and increase the
tion and g ðyworst Þ denotes the objective function values of the worst exploitation capacity of the algorithm. As a result, the population
solution. diversity and exploration capability of the JAYA is quickly deterio-
It is evident from Eq. (7) that the added weight is self-adaptive, rating. Therefore, a multiple learning strategy is introduced to have
and that its value increases progressively as the difference between a good balance between exploitation and exploration. The pre-

Fig. 3. Flowchart of the MLJAYA algorithm.

113
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Table 2
Parameters boundary of different PV models.

Parameter RTC solar cell Photowatt- PWP201

Lower bound Upper bound Lower bound Upper bound
Iph ðAÞ 0 1 0 2
Io , Io1 , Io2 (lA) 0 1 0 50
Rs(O) 0 0.5 0 2
Rsh (O) 0 100 0 2000
a , a1 , a2 1 2 1 50

Table 3
Parameters of the algorithms used.

Algorithm Parameters Algorithm Parameters

PGJAYA NP = 20 SATLBO NP = 40
IJAYA NP = 20 IGHS NP = 30, HMCR = 0.95 (ref_askarzadeh2012)
EJADE NP = 50 IBSA NP =
CLJAYA NP = 20 jDE NP = 400 when D = 100, Fl = 0.1, Fu = 0.9, s1 = s2 = 0.1
NP = 100 when D = 30
JAYA NP = 20 MLBSA NP = 50
DE/BBO NP = 100, F = rand(0.1, 1), MADE NP = 20, e = 0.05 for SDM
CR = 0.9, E = I = 1, k = 2, pmax ¼ 0:005 e = 0.01 for DDM
LBSA NP = 50, mix rate = 1. GWOCS NP = 30
GOTLBO NP = 50, JR = 0.3 TLBO NP = 50
CLPSO NP = 40, w = 0.9 ? 0.4c = 1.49445, m = 7 MLJAYA NP = 30, F = 3.randn

sented multiple learning methods allow some individuals to where Z iþ1 ¼ ðz1;iþ1 ; z2;iþ1 ;          ; zD;iþ1 Þ
improve the diversity of the population by learning from both
the historical and current population, while other individuals focus 3.2.4. Pseudo code of MLJAYA
on improving the convergence speed by learning from the best The setups of the proposed MLJAYA is described in algorithm 1,
individual in the current population. The multiple learning meth- where NFT is the number of function tests and (NFT_max) is the
ods are used to generate a new individual, as shown in Eq. (9). maximum number of function tests, NP is the size of the initial
(   population. In addition, the flowchart of MLJAYA is shown in
yij þ F  yworst;j
yi;j þ yk;j
yi;j if ða < b=a  Cð0; 1ÞÞ
y0i;j ¼  Fig. 3. It is clear that MLJAYA will stop the search run if the maxi-
yi;j þ rand: ybest;j
yi;j otherwise mum number of generations is achieved.
ð9Þ
Algorithm 1: Pseudo code of MLJAYA.
where k is an integer randomly selected from {1, 2, . . .,NP} and sat-
isfies k – i, so yk;j is the value of the jth variable for the individual k Input: population size (NP) and maximum number of
of current population. a and b are two uniformly distributed ran- function tests (NFT_max)
dom numbers within [0, 1]. ybest;j is the value of the jth variable Output: optimal member in the population
1 Generate an initial population and evaluate fitness of
for the best individual of current population.
each member in the population
2 NFT = NP;
3.2.3. Chaos perturbation mechanism
3 While NFT < NFT_max do
The chaos perturbation mechanism is used for preventing the
4 Selected the best member YBest and the worst member
local optimum and thus improving the performance of the algo-
YWorst
rithm. However, chaos is a nonlinear phenomenon, characterized
5 For i ¼ 1 to NP do
by its ergodicity, randomness, and high sensitivity to initial condi-
6 If Yi is not best member //Update the weight by self-
tions [61,62]. We use this randomness and the ergodicity of chaos
adaptive strategy// Calculate the weight by using Eq.
to avoid the search colliding with a local optimum and to overcome
(7)
the shortcomings of a traditional optimization algorithm. The
7 Uptate the k-th variable of the solution Yi using Eq.6
chaos system used in this research is the logistic mapping defined
8 Elseif ( a < b/a  C(0,1))//multiple learning
by Eq.10. Subsequently, the best solution is updated using equation
strategy//
(11).
 9 Update the k-th variable of the member Yi using Eq.
zj;iþ1 ¼ lj zj;i 1
zj;i ; zj;i 2 ½0; 1 j ¼ 1    D; i ¼ 1; 2    ð10Þ (9) Else // Chaos perturbation mechanism//
Update the k-th variable of the member using Eq.(11)
where zj;i is the jth chaotic variable in the ith generation, and lj is a
10 End if
chaotic attractor. When lj ¼ 4 , the logistic mapping is considered a 11 Compute the fitness function value of the new member
full mapping in the range of [0, 1], 12 NFT = NFT + 1
The jth learner of the class is expressed by a vector ybest;j and the 13 Receive new member if it’ s better than the worst
0
new learner ybest;j generated by the chaotic search in the neighbor- member ;
hood of ybest;j is obtained by Eq. (11). 14 End for
0
15 End while
ybest;j ¼ ybest;j þ randð2Z iþ1
1Þifrand < ð1
NFT=NFT maxÞ ð11Þ

114
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Table 4
The optimal parameters obtained by MLJAYA compared with other state-of-art algorithms for SDM of the solar cell.

Algorithm Iph ðAÞ Io (lA) Rs(O) Rsh(O) a Best RMSE (10
4 ) Worst RMSE (10
4 ) Mean RMSE (10
4 )
MLJAYA 0.7608 0.32302 0.0364 53.7185 1.4812 9.8602 9.8602 9.8602
PGJAYA [33] 0.7608 0.3230 0.0364 53.7185 1.4812 9.8602 9.8603 9.8602
IJAYA [29] 0.7608 0.3228 0.0364 53.7595 1.4811 9.8603 9.92 9.92
EJADE [63] 0.7608 0.3230 0.0364 53.7185 1.4812 9.8602 9.8602 9.8602
CLJAYA [68] 0.7608 0.3230 0.0364 53.7185 1.4812 9.8602 9.8602 9.8602
JAYA [68] 0.7608 0.3246 0.0363 52.3754 1.4817 9.9786 14.783 11.617
MADE [18] 0.7608 0.3230 0.0364 53.7185 1.4812 9.8602 9.8602 9.8602
GWOCS [42] 0.76077 0.3219 0.0362 53.6320 1.4808 9.8607 9.9095 9.8874
TLBO [68] 0.76078 0.3271 0.0363 53.9466 1.4824 9.8636 12.3579 10.4761
BBO-M [69] 0.76078 0.31874 0.0364 53.3627 1.4798 9.8630 NA NA
IGHS [64] 0.76077 0.34351 0.03613 53.2845 1.4874 9.9306 NA NA
DE/BBO [34] 0.7605 0.32477 0.0364 55.2627 1.4817 9.9922 22.3 12.9
LBSA [34] 0.7606 0.34618 0.0362 59.0978 1.4881 10.143 15.9 12.4
GOTLBO [70] 0.76078 0.33155 0.0362 54.1154 1.4838 9.8744 14.388 10.289
CLPSO [34] 0.7608 0.34302 0.0361 54.1965 1.4873 9.9633 11.8724 10.5871
SATLBO [71] 0.7608 0.3232 0.0363 53.7256 1.4812 9.8602 9.9494 9.878

Fig. 4. Comparison on the (a) I-V and (b) P-V curve between the datasheet and simulated data obtained by MLJAYA algorithm.

Table 5
The individual absolute errors (IAE) of current and power of MLJAYA algorithm on single diode model.

Item Benchmark data Simulated current data Simulated power data

V(V) I(A) Isim(A) IAEC(A) Psim(W) IAEp(W)
1
0.2057 0.7640 0,764111773510811 0,000111773510810931
0,15717779 2,29918E-05
2
0.1291 0.762 0,762686766931466 0,000686766931466409
0,09846286 8,86616E-05
3
0.0588 0.7605 0,761378860677928 0,000878860677927529
0,04476908 5,1677E-05
4 0.0057 0.7605 0,760178372817480 0,000321627182520157 0,00433302 1,83327E-06
5 0.0646 0.760 0,759080063397046 0,000919936602953864 0,04903657 5,94279E-05
6 0.1185 0.759 0,758067128339312 0,000932871660687673 0,08983095 0,000110545
7 0.1678 0.757 0,757115696984078 0,000115696984077651 0,12704401 1,9414E-05
8 0.2132 0.757 0,756166136282915 0,000833863717085048 0,16121462 0,00017778
9 0.2545 0.7555 0,755111281845299 0,000388718154700984 0,19217582 9,89288E-05
10 0.2924 0.754 0,753688167288479 0,000311832711521132 0,22037842 9,11799E-05
11 0.3269 0.7505 0,751411183909163 0,000911183909163449 0,24563632 0,000297866
12 0.3585 0.7465 0,747370311139112 0,000870311139112068 0,26793226 0,000312007
13 0.3873 0.7385 0,740116703841217 0,00161670384121682 0,2866472 0,000626149
14 0.4137 0.728 0,727413012461820 0,000586987538179762 0,30093076 0,000242837
15 0.4373 0.7065 0,706964289682593 0,000464289682592489 0,30915548 0,000203034
16 0.4590 0.6755 0,675299187114843 0,000200812885157164 0,30996233 9,21731E-05
17 0.4784 0.632 0,630881679651015 0,00111832034898529 0,3018138 0,000535004
18 0.4960 0.573 0,572074238795868 0,000925761204132125 0,28374882 0,000459178
19 0.5119 0.499 0,499482798110195 0,000482798110194749 0,25568524 0,000247144
20 0.5265 0.413 0,413489887723459 0,000489887723459093 0,21770243 0,000257926
21 0.5398 0.3165 0,317228453575752 0,000728453575752464 0,17123992 0,000393219
22 0.5521 0.212 0,212132620149102 0,000132620149101897 0,11711842 7,32196E-05
23 0.5633 0.1035 0,102778189694791 0,000721810305208528 0,05789495 0,000406596
24 0.5736
0.01
0,00915874229842810 0,000841257701571896
0,00525345 0,000482545
25 0.5833
0.123
0,124252477776826 0,00125247777682634
0,07247647 0,00073057
26 0.5900
0.21
0,209033140961002 0,000966859038997708
0,12332955 0,000570447
Total IAE 0.01781248 0,00665235

115
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Table 6
Calculated values of IAEC obtained by MLJAYA and other advanced algorithms SDM.

IAEC
Item Vm(V) Im(V) MLJAYA PGJAYA IJAYA CLJAYA JAYA GWOCS
1
0.2057 0.7640 0,00011177 0.00003464 0.00003108 0.00008770 0.00020205 0.00009030
2
0.1291 0.762 0,00068676 0.00059413 0.00064510 0.00066309 0.00074090 0.00066300
3
0.0588 0.7605 0,00087886 0.00083513 0.00087259 0.00085531 0.00089959 0.00085299
4 0.0057 0.7605 0,00032162 0.00032210 0.00029691 0.00034601 0.00033252 0.00034934
5 0.0646 0.760 0,00091993 0.00088239 0.00086807 0.00094479 0.00095942 0.00094906
6 0.1185 0.759 0,00093287 0.00086395 0.00085885 0.00095766 0.00099803 0.00096404
7 0.1678 0.757 0,00011569 0.00020632 0.00020437 0.00009165 0.00002770 0.00008313
8 0.2132 0.757 0,00083386 0.00073798 0.00074403 0.00085864 0.00094444 0.00086758
9 0.2545 0.7555 0,00038871 0.00030682 0.00031304 0.00041313 0.00051901 0.00042318
10 0.2924 0.754 0,00031183 0.00026812 0.00026992 0.00033612 0.00046075 0.00034636
11 0.3269 0.7505 0,00091118 0.00089886 0.00090585 0.00089097 0.00074896 0.00087770
12 0.3585 0.7465 0,00087031 0.00078955 0.00080817 0.00085385 0.00069587 0.00083956
13 0.3873 0.7385 0,00161670 0.00148866 0.00151849 0.00161722 0.00144570 0.00159097
14 0.4137 0.728 0,00058698 0.00077985 0.00074326 0.00061778 0.00079846 0.00060496
15 0.4373 0.7065 0,00046428 0.00032714 0.00036230 0.00047265 0.00029052 0.00045663
16 0.4590 0.6755 0,00020081 0.00030107 0.00027606 0.00021985 0.00039164 0.00019848
17 0.4784 0.632 0,00111832 0.00123581 0.00122648 0.00124173 0.00138810 0.00110683
18 0.4960 0.573 0,00092576 0.00098904 0.00099570 0.00107164 0.00117632 0.00090912
19 0.5119 0.499 0,00048279 0.00072822 0.00071050 0.00060702 0.00055795 0.00049840
20 0.5265 0.413 0,00048988 0.00075155 0.00073174 0.00064879 0.00066665 0.00049728
21 0.5398 0.3165 0,00072845 0.00105272 0.00103942 0.00101011 0.00109882 0.00072056
22 0.5521 0.212 0,00013262 0.00011444 0.00011485 0.00015494 0.00031368 0.00010309
23 0.5633 0.1035 0,00072181 0.00135725 0.00134042 0.00124869 0.00102801 0.00077758
24 0.5736
0.01 0,00084125 0.00119127 0.00121803 0.00128246 0.00154955 0.00075595
25 0.5833
0.123 0,00125247 0.00255095 0.00251336 0.00250741 0.00220330 0.00137012
26 0.5900
0.21 0,00096685 0.00165139 0.00168223 0.00152767 0.00184212 0.00082449
Total IAE 0.01781248 0.02125936 0.02129082 0.02152688 0.02228006 0.0177207

Table 7
The optimal parameters obtained by MLJAYA compared with other state-of-art algorithms for DDM of the solar cell.

Algorithm Iph ðAÞ Io1 (lA) Io2 (lA) Rs(O) Rsh(O) a1 a2 Best RMSE (10
4 ) Worst RMSE (10
4 ) Mean RMSE (10
4 )
MLJAYA 0.7608 0.82523 0.22021 0.0367 56.2627 1.9988 1.4491 9.8294 14.2102 10.618
PGJAYA [51] 0.76077 0.29500 0.2102 0.03645 54.3036 1.4735 2 9.8443 9.9499 9.8582
IJAYA [29] 0.7601 0.00504 0.75094 0.0376 77.8519 1.2186 1.6247 9.8293 14.055 10.269
EJADE [63] 0.7608 0.21031 0.88534 0.0368 55.8135 1.4450 2 9.8263 9.8602 9.8363
CLJAYA [68] 0.76078 0.22605 0.74876 0.03674 55.4859 1.4510 1.9999 9.8249 9.8602 9.8308
JAYA [68] 0.76077 0.28663 0.29702 0.03641 53.5136 1.4712 1.9997 9.9307 14.793 11.767
MADE [18] 0.7608 0.7394 0.22460 0.03680 55.4329 1.9963 1.4505 9.8261 9.8786 9.8608
GWOCS [42] 0.76076 0.53772 0.24855 0.03666 54.7331 2 1.4588 9.8334 10.017 9.9411
TLBO [68] 0.7610 0.67834 0.22954 0.03670 55.8211 1.9803 1.4525 9.8290 15.2057 11.5977
BBO-M [69] 0.76077 0.23260 0.69230 0.03670 55.3582 1.4534 1.9995 9.8251 NA NA
IGHS [64] 0.76079 0.97310 0.16791 0.03690 56.8368 1.2126 1.42814 9.8635 NA NA
IBSA [34] 0.7608 0.21507 0.26624 0.0366 51.9008 1.8718 2 9.9663 NA NA
LBSA [34] 0.7606 0.29814 0.27096 0.0363 60.188 1.4760 1.92202 10.016 17.372 12.728
GOTLBO [70] 0.7608 0.8002 0.2205 0.0368 56.0753 2 1.449 9.8318 13.947 11.475
CLPSO [34] 0.7607 0.25843 0.38615 0.0367 57.9422 1.4625 1.9435 9.9894 11.458 15.494
jDE [51] 0.76078 0.2575 0.49160 0.03659 54.7712 1.4620 1.9999 9.8298 NA NA

Fig. 5. Comparison on the (a) I-V and (b) P-V curve between the datasheet and simulated data obtained by MLJAYA algorithm.

116
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Table 8
The individual absolute errors (IAE) of current and power of MLJAYA algorithm on double diode model.

Item Benchmark data Simulated current data Simulated power data

V(V) I(A) Isim(A) IAEC(A) Psim(W) IAEp(W)
1
0.2057 0.7640 0,763958750706341 4,12E-05
0,15714631502029 8,48498E-06
2
0.1291 0.762 0,762598063210525 0,00059806
0,09845140996047 7,721E-05
3
0.0588 0.7605 0,761348968505920 0,00084897
0,04476731934814 4,99193E-05
4 0.0057 0.7605 0,760201803311741 0,0002982 0,004333150278876 1,69972E-06
5 0.0646 0.760 0,759150430695006 0,00084957 0,049041117822897 5,48822E-05
6 0.1185 0.759 0,758176945963568 0,00082305 0,089843968096682 9,75319E-05
7 0.1678 0.757 0,757254260919216 0,00025426 0,127067264982244 4,2665E-05
8 0.2132 0.757 0,756318452213770 0,00068155 0,161247094011976 0,000145306
9 0.2545 0.7555 0,755256979078226 0,00024302 0,192212901175408 6,18488E-05
10 0.2924 0.754 0,753803078522197 0,00019692 0,220412020159890 5,75798E-05
11 0.3269 0.7505 0,751472411078683 0,00097241 0,245656331181622 0,000317881
12 0.3585 0.7465 0,747363961175409 0,00086396 0,267929980081384 0,00030973
13 0.3873 0.7385 0,740047956558985 0,00154796 0,286620573575295 0,000599524
14 0.4137 0.728 0,727311279278660 0,00068872 0,300888676237582 0,000284924
15 0.4373 0.7065 0,706878484766503 0,00037848 0,309117961388392 0,000165511
16 0.4590 0.6755 0,675280195501096 0,0002198 0,309953609735003 0,00010089
17 0.4784 0.632 0,630957787471239 0,00104221 0,301850205526241 0,000498594
18 0.4960 0.573 0,572240318087602 0,00075968 0,283831197771450 0,000376802
19 0.5119 0.499 0,499703519654239 0,00070352 0,255798231711005 0,000360132
20 0.5265 0.413 0,413717652861094 0,00071765 0,217822344231366 0,000377844
21 0.5398 0.3165 0,317420496480098 0,0009205 0,171343583999957 0,000496884
22 0.5521 0.212 0,212262012953803 0,00026201 0,117189857351795 0,000144657
23 0.5633 0.1035 0,102838205681634 0,00066179 0,057928761260464 0,000372789
24 0.5736
0.01
0,009159327751035 0,00084067
0,005253790397993 0,00048221
25 0.5833
0.123
0,124294949923916 0,00129495
0,072501244290620 0,000755344
26 0.5900
0.21
0,209089946913424 0,00091005
0,123363068678920 0,000536931
Sum of IAE 0.0176 0,00677778

Table 9
The calculated values of IAEC obtained by MLJAYA and other advanced algorithms DDM.

IAEC
Item Vm(V) Im(V) MLJAYA PGJAYA IJAYA CLJAYA JAYA GWOCS
1
0.2057 0.7640 4,12E-05 0.00003464 0.00003108 0.00001670 0.00010219 0.00001696
2
0.1291 0.762 0,00059806 0.00059413 0.00064510 0.00060400 0.00067209 0.00059756
3
0.0588 0.7605 0,00084897 0.00083513 0.00087259 0.00083761 0.00085919 0.00079481
4 0.0057 0.7605 0,0002982 0.00032210 0.00029691 0.00032628 0.00034705 0.00040087
5 0.0646 0.760 0,00084957 0.00088239 0.00086807 0.00089238 0.00095093 0.00099450
6 0.1185 0.759 0,00082305 0.00086395 0.00085885 0.00087863 0.00096978 0.00100387
7 0.1678 0.757 0,00025426 0.00020632 0.00020437 0.00018858 0.00007141 0.00004849
8 0.2132 0.757 0,00068155 0.00073798 0.00074403 0.00075640 0.00089108 0.00089725
9 0.2545 0.7555 0,00024302 0.00030682 0.00031304 0.00032267 0.00046384 0.00044793
10 0.2924 0.754 0,00019692 0.00026812 0.00026992 0.00027758 0.00041319 0.00036586
11 0.3269 0.7505 0,00097241 0.00089886 0.00090585 0.00089924 0.00077950 0.00086416
12 0.3585 0.7465 0,00086396 0.00078955 0.00080817 0.00080159 0.00070232 0.00083315
13 0.3873 0.7385 0,00154796 0.00148866 0.00151849 0.00151082 0.00142673 0.00159294
14 0.4137 0.728 0,00068872 0.00077985 0.00074326 0.00075292 0.00083681 0.00059386
15 0.4373 0.7065 0,00037848 0.00032714 0.00036230 0.00035036 0.00024621 0.00047596
16 0.4590 0.6755 0,0002198 0.00030107 0.00027606 0.00028950 0.00042731 0.00017357
17 0.4784 0.632 0,00104221 0.00123581 0.00122648 0.00123937 0.00140627 0.00108065
18 0.4960 0.573 0,00075968 0.00098904 0.00099570 0.00100544 0.00117943 0.00088583
19 0.5119 0.499 0,00070352 0.00072822 0.00071050 0.00070598 0.00055478 0.00051674
20 0.5265 0.413 0,00071765 0.00075155 0.00073174 0.00073360 0.00063683 0.00051184
21 0.5398 0.3165 0,0009205 0.00105272 0.00103942 0.00104625 0.00101178 0.00073592
22 0.5521 0.212 0,00026201 0.00011444 0.00011485 0.00012315 0.00013709 0.00012636
23 0.5633 0.1035 0,00066179 0.00135725 0.00134042 0.00133651 0.00132002 0.00073824
24 0.5736
0.01 0,00084067 0.00119127 0.00121803 0.00120838 0.00113028 0.00081954
25 0.5833
0.123 0,00129495 0.00255095 0.00251336 0.00254346 0.00277537 0.00127408
26 0.5900
0.21 0,00091005 0.00165139 0.00168223 0.00162809 0.00116770 0.00094888
Total IAE 0.0176 0.02125936 0,02125974 0,02127549 0,02147918 0,01773982

4. Results and discussions Photowatt-PWP201 contains 36 polycrystalline silicon cells in ser-

ies operating at an irradiance of 1000 W/m2 and a temperature of
The main goal of this section is tested to appraise the effective- 45 °C. For a fair judgment, the lower and upper bounds for each
ness of the proposed algorithm by conducting using the bench- parameter are adopted as the literature and the detailed descrip-
marks experimental current–voltage data of single diode, double tions are given in Table 2. The parameters of the algorithms are
diode, and PV module. This data are a 57 mm diameter commercial set in Table 3 [18,35,63,64]. In order to demonstrate the competi-
R.T.C. France silicon solar cell operating at an irradiance of 1000 W/ tive performance of the MLJAYA algorithm in the parameters iden-
m2 and a temperature of 33 °C, and a solar module named tification, it was compared with other state-of-the-art algorithms.

117
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Table 10
Comparison MLJAYA with different others algorithms on PV module model.

Algorithm Iph ðAÞ Io (lA) Rs(O) Rsh(O) A ¼ a  Ns Best RMSE (10
3 ) Worst RMSE (10
3 ) Mean RMSE (10
3 )
MLJAYA 1.0305 3.4823 1.2013 981.9822 48.6428 2.4250748 2.49419 2.44395
PGJAYA [33] 1.0305 3.4818 1.2013 981.8545 48.6424 2.425075 2.426764 2.425144
IJAYA [29] 1.0305 3.4703 1.2016 977.3752 48.6298 2.4251 2.4389 2.4289
GWO [72] 1.02982 4.3863 1.17573 1186.5926 49.54686 2.526088 NA NA
CLJAYA [68] 1.030514 3.4822628 1.201271 981.982279 48.64283 2.425075 NA NA
JAYA [68] 1.03038 3.410219 1.204298 972.58408 48.5631 2.4309 2.595873 2.45371
MADE [18] 1.0305 3.4823 1.2013 981.9823 48.6428 2.4250 2.4251 2.4251
GWOCS [42] 1.03049 3.4650 1.2019 982.7566 48.62367 2.4251 2.4275 2.4261
MLBSA [72] 1.030514 3.4823 1.20127 981.9822 48.64283 2.5250748 NA NA
DE/BBO [34] 1.0303 3.6172 1.1969 1015.1 48.7894 2.428255 2.5256 2.4616
LBSA [34] 1.0304 3.5233 1.2014 1020.4 48.6866 2.4296 2.5344 2.4674
GOTLBO [70] 1.0307 3.5124 1.1995 969.9313 48.8214 2.426583 2.5638 2.4754
CLPSO [34] 1.0304 3.6131 1.1978 1017.0 48.7847 2.4281 2.5433 2.4549
SATLBO [71] 1.0306 3.4715 1.2017 972.9357 48.6313 2.4251 NA NA

Fig. 6. Comparison on the (a) I-V and (b) P-V curve between the datasheet and simulated data obtained by MLJAYA algorithm.

These algorithms are PGJAYA, IJAYA, EJADE, CLJAYA, JAYA, MADE, I-V and P-V curves simulated data are in very good agreement with
GWOCS, TLBO, BBO-M, IGHS, DE/BBO, LBSA, GOTLBO, CLPSO, the benchmark data, which confirm the accuracy of estimated
SATLBO, IBSA, jDE, GWO, and MLBSA. The comparisons are con- parameters. On the other hand, to show the quality of the results,
ducted on the root mean square error (RMSE) and the individual the individual absolute error (IAE) between the measured data and
absolute error (IAE) values to illustrate the accuracy of each algo- simulated data on different voltages are recorded, as shown in
rithm. Then, the convergence curves are analyzed and presented Table 5 and 6. The current IAEC and power IAEP values are not
to evaluate the robustness of each algorithm. The simulations are greater than 0.00227392110 and 0.001326378respectively. Addi-
performed using the MATLAB platform under Windows7 64-bit. tionally, as can be seen in Table. . . .., the sum of the individual
The MATLAB code runs on an Intel(R) Core (TM) i7
4600 M CPU absolute current errors (IAEC) obtained by MLJAYA, PGJAYA, IJAYA,
@ 2.90 GHz 2.90 GHz HP ZBook 15, 16 GB RAM. CLJAYA, JAYA, and GWOCS are 0.01781248, 0.02125936,
0.02129082, 0.02152688, 0.02228006, and 0.0177207, respec-
tively. It is quite clear that MLJAYA and GWOCS provide the small-
4.1. Case 1: single diode model
est. Although the differences are small, any decrease in the sum IAE
value is significant because it indicates an improvement in the
The outcomes of the application the MLJAYA and fifteen other
accuracy of the parameters. The Fig. 7 illustrates the boxplot of
methods for the parameter estimation of RTC PV cell with the
fourteen algorithms for the SDM for the solar cell RTC, which
SDM are recorded in Table 4. It can be noted that MLJAYA, PGJAYA,
shows the distribution of results obtained by different algorithms
EJADE, CLJAYA, MADE, and SATLBO obtained the best RMSE value
in 30 runs. It can be seen that the proposed approach presents
(9.8602E
04), followed by IJAYA (9.8603E
04), GWOCS
excellent performances compared with other compared algorithms
(9.8607E
04), BBO-M (9.8630E
04), TLBO (9.8636E
04), GOTLBO
in terms of robustness. The error measurements confirm that the
(9.8744E-04), IGHS (9.9306E-04), CLPSO (9.9633), DE/BBO
estimated currents are in perfect agreement with the measured
(9.9922E-04), and LBSA (10.143–04). Even though the RMSE values
currents. The individual absolute error values shown in Table 1
obtained by the IJAYA, GWOCS, BBO-M, and TLBO algorithms are
are expressed as follows [65-71]:
very close to the value (9.8602E-05), any reduction of the objective
function is significant since it leads to an improvement of the  
IAEC ¼ Imeas;i
Isimu;i  ð12Þ
knowledge of the real values of the parameters.
As a further analysis, based on the identification parameters of  
IAEP ¼ Pmeas;i
Psimu;i  ð13Þ
the MLJAYA algorithm, the I-V output curves, and P-V output
curves are shown in Fig. 4. As can be clearly seen from Fig. 4, the
118
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Table 11
The individual absolute errors (IAE) of current and power of MLJAYA algorithm on PV module model.

Item Benchmark data Simulated current data Simulated power data

V(V) I(A) Isim(A) IAEC(A) Psim(W) IAEp(W)
1 0.1248 1.0315 1.02910483147811 0.00239516852 0.12843228296 0.00029892
2 1.8093 1.0300 1.02736674239367 0.00263325760 1.85881464701 0.00476435
3 3.3511 1.0260 1.02572746353177 0.00027253646 3.43731530304 0.0009133
4 4.7622 1.0220 1.02409281404295 0.00209281404 4.87693479903 0.0099664
5 6.0538 1.0180 1.02227744513878 0.00427744513 6.18866319738 0.0258948
6 7.2364 1.0155 1.01991627774018 0.00441627774 7.38052215223 0.03195795
7 8.3189 1.0140 1.01634860811052 0.00234860811 8.45490243601 0.01953784
8 9.3097 1.0100 1.01048146575136 0.00048146575 9.40727930170 0.0044823
9 10.2163 1.0035 1.00061393548725 0.00288606451 10.2225721491 0.0294849
10 11.0449 0.9880 0.984532752378583 0.00346724762 10.8740657967 0.0382954
11 11.8018 0.9630 0.959505119503797 0.00349488049 11.3238875193 0.04124588
12 12.4929 0.9255 0.922820917609443 0.00267908239 11.5287094416 0.03346951
13 13.1231 0.8725 0.872579994281680 7.9994281e-05 11.4509545229 0.00104977
14 13.6983 0.8075 0.807252422243689 0.00024757775 11.0579858556 0.00339139
15 14.2221 0.7265 0.728312245219096 0.00181224521 10.3581295827 0.02577393
16 14.6995 0.6345 0.637111312228311 0.00261131222 9.36521773410 0.03838498
17 15.1346 0.5345 0.536184041928488 0.00168404192 8.11493100097 0.0254873
18 15.5311 0.4275 0.429480310131995 0.00198031013 6.67030164469 0.03075639
19 15.8929 0.3185 0.318741875604404 0.00024187560 5.06573275479 0.0038441
20 16.2229 0.2085 0.207355818800211 0.00114418119 3.36391271281 0.01856194
21 16.5241 0.1010 0.096132877051555 0.00486712294 1.58850927368 0.08042483
22 16.7987
0.008
0.00835960930196 0.00035960930
0.1404305687 0.00604097
23 17.0499
0.111
0.11097026089431 2.9739105e-05
1.8920318512 0.00050705
24 17.2793
0.209
0.20928021189372 0.00028021189
3.6162155653 0.00484187
25 17.4885
0.302
0.30208067936876 8.0679368e-05
5.2829379611 0.00141096
Total IAE 0.04686375 0.48078704

Table 12
Calculated values of IAEC obtained by MLJAYA and other advanced algorithms DDM.

IAEC
Item Vm(V) Im(V) MLJAYA PGJAYA IJAYA CLJAYA JAYA GWOCS
1 0.1248 1.0315 0.00239516852 0.00238036 0.00245607 0.00238084 0.00245607 0.00240597
2 1.8093 1.0300 0.00263325760 0.00261867 0.00262813 0.00261893 0.00262813 0.00264227
3 3.3511 1.0260 0.00027253646 0.00025814 0.00020772 0.00025820 0.00020772 0.00028302
4 4.7622 1.0220 0.00209281404 0.00210704 0.00221094 0.00210715 0.00221094 0.00208051
5 6.0538 1.0180 0.00427744513 0.00429155 0.00444181 0.00429180 0.00444181 0.00426211
6 7.2364 1.0155 0.00441627774 0.00443032 0.00461867 0.00443068 0.00461867 0.00439919
7 8.3189 1.0140 0.00234860811 0.00236269 0.00257921 0.00236311 0.00257921 0.00233705
8 9.3097 1.0100 0.00048146575 0.00049575 0.00072904 0.00049615 0.00072904 0.00048363
9 10.2163 1.0035 0.00288606451 0.00287134 0.00263257 0.00287103 0.00263257 0.00282156
10 11.0449 0.9880 0.00346724762 0.00345177 0.00321631 0.00345162 0.00321631 0.00333714
11 11.8018 0.9630 0.00349488049 0.00347827 0.00324876 0.00347832 0.00324876 0.00328487
12 12.4929 0.9255 0.00267908239 0.00266092 0.00243167 0.00266118 0.00243167 0.00242690
13 13.1231 0.8725 7.9994281e-05 0.00010009 0.00034179 0.00009966 0.00034179 0.00011492
14 13.6983 0.8075 0.00024757775 0.00022523 0.00004717 0.00022574 0.00004717 0.00016575
15 14.2221 0.7265 0.00181224521 0.00183695 0.00215112 0.00183648 0.00215112 0.00147504
16 14.6995 0.6345 0.00261131222 0.00263835 0.00299715 0.00263800 0.00299715 0.00197421
17 15.1346 0.5345 0.00168404192 0.00171321 0.00210668 0.00171306 0.00210668 0.00119493
18 15.5311 0.4275 0.00198031013 0.00201127 0.00241115 0.00201133 0.00241115 0.00130780
19 15.8929 0.3185 0.00024187560 0.00027424 0.00064662 0.00027448 0.00064662 0.00015622
20 16.2229 0.2085 0.00114418119 0.00111086 0.00081061 0.00111049 0.00081061 0.00065583
21 16.5241 0.1010 0.00486712294 0.00483326 0.00464639 0.00483283 0.00464639 0.00265575
22 16.7987
0.008 0.00035960930 0.00032571 0.00031522 0.00032539 0.00031522 0.00017322
23 17.0499
0.111 2.9739105e-05 0.00006337 0.00013491 0.00006352 0.00013491 0.00003631
24 17.2793
0.209 0.00028021189 0.00024715 0.00068793 0.00024727 0.00068793 0.00010220
25 17.4885
0.302 8.0679368e-05 0.00213691 0.00141908 0.00213641 0.00141908 0.00100513
Total IAE 0,04686375 0,04892342 0,05011672 0,04892367 0,05011672 0,04178153

4.2. Case2: double diode model jDE (9.8298E-0-04), JAYA (9.8307E-0-04), GWOCS (9.8334E-0-04),
PGJAYA (9.8443E-0-04), IGHS (9.8635E-0-04), JAYA (9.9307E-0-
In this case, seven parameters are needed to be identified. The 04), IBSA (9.9663E-0-04), CLPSO (9.9894E-04), and LBSA
estimated parameters and the RMSE of several methods are listed (10.016E-0-04). Compared to the RMSE values given in Tables 4
in Table 7. The results of the application of the MLJAYA and and 7, it can be seen that the two-diode model is generally more
many other methods for the parameter extraction of RTC PV cell accurate than its single-diode counterpart, regardless of the
with the DDM show that CLJAYA (9.8249E-0-04), BBO-M choice of the optimization algorithm.
(9.8251E-0-04), MADE (9.8261E-0-04), EJADE (9.8263E-0-04), Furthermore, the extracted parameters of MLJAYA algorithm
TLBO (9.8290E-0-04), IJAYA (9.8293E-04), MLJAYA (9.8294E-05), were used to plot the I-V and P-V curves, which are illustrated in

119
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Fig. 7. Best boxplot of different algorithms for the single diode model.

Fig. 8. Best boxplot of different algorithms for the double diode model.

Fig. 5. It is clear that the measured and simulated data obtained by parameters and their corresponding RMSE results obtained by
MLJAYA are very coherent for the I-V and P-V curves. In addition, MLJAYA and the reported algorithms are given in Table 10. It can
the values of the individual absolute current error (IAEC) and indi- be noted that the MLJAYA and MLBSA algorithms obtained the best
vidual absolute power error (IAEP) shown in Table 8, it can be seen RMSE value (2.425074E-03), while MADE, PGJAYA and CLJAYA pro-
that all IAEC values are less than 0,00104221, and all IAEP are less vided the second best RMSE value (2.425075E-03), followed by
than 0,000755344. According to Table 9, the sum of absolute cur- IJAYA, GWOCS, and SATLBO (2.4251E-03), GOTLBO (2.426583E-
rent errors obtained by MLJAYA, PG JAYA, JAYA, CL JAYA, JAYA, 03), CLPSO (2.4281E-03), DE/BBO (2.428255E-03), LBSA (2.4296E-
and GWOCS are 0.0176, 0.02125936, 0.02125974, 0.02127549, 03), JAYA (2.4309E-03), MLBSA (2.5250748E-03), and GWO
0.02127549, and 0.01773982, respectively. Clearly, MLJAYA can (2.526088E-03). We see that the difference between the first best
achieve the smallest sum of AEI, which means that MLJAYA can RMSE value and second-best RMSE values is of the order of 10E-
provide a solution with higher accuracy. 06, it is meaningful for any reduction in the objective function.
As per results are shown in boxplots of Fig. 8, it can be observed To conduct an additional survey on the quality of the parame-
that the proposed MLJAYA exhibits the best performance com- ters obtained by MLJAYA, these parameters were put into the PV
pared with CLPSO, LBSA, DE/BBO, TLBO, and JAYA in terms of solu- model to reconstruct the I-V and P-V characteristics. According
tion distribution. to the I–V and P-V characteristics plotted in Fig. 6, we can see that
the simulated data obtained by MLJAYA fit the experimental data
4.3. Case 3: PV module very well. In addition, the individual absolute error (IAE) between
the experimental data and simulated data for MLJAYA and the sum
In order to validate the high performance of the MLJAYA algo- IAEC values of MLJAYA, PGJAYA, IJAYA, CLJAYA, JAYA, and GWOCS
rithm, it was compared to four JAYA variants and ten state-of- are summarized in Table 11 and Table 12. The obtained results
the-art meta-heuristics algorithms. The best statistical extracted are 0.04686375, 0.04892342, 0.05011672, 0.04892367,

120
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

Fig. 9. Best boxplot of different algorithms for PV model.

0.05011672, and 0.04178153. It is obvious that GWOCS can CRediT authorship contribution statement
achieve the minimum sum absolute error, followed by MLJAYA
which means MLJAYA can offer a higher accuracy solution than Driss Saadaoui: Writing – original draft, Writing – review &
PG JAYA, JAYA, CL JAYA, and JAYA. editing, Software, Visualization, Investigation, Formal analysis,
Fig. 9 presents the boxplot of ten algorithms for Photowatt- Methodology, Conceptualization. Mustapha Elyaqouti: Formal
PWP201. Based on the comparisons on the solution distribution, analysis, Visualization, Writing – review & editing, Software, Visu-
it can be achieved that the proposed MLJAYA offers superior per- alization, Investigation, Formal analysis, Methodology, Conceptual-
formance compared with CLPSO, GOTLBO, LBSA, BE/BBO, and JAYA ization, Supervision. Khalid Assalaou: Formal analysis,
in terms accuracy and robustness. Visualization, Writing – review & editing, Software, Visualization,
Supervision. Driss Ben hmamou: Formal analysis, Visualization,
5. Conclusion Writing – review & editing. Souad Lidaighbi: Formal analysis,
Visualization, Writing – review & editing.
This paper proposes a multiple learning JAYA (MLJAYA) algo-
rithm for parameter identification different photovoltaic models. Declaration of Competing Interest
In MLJAYA approach three strategy has been used with the aim is
avoid the drawback of JAYA algorithm, and therefore achieve good The authors declare that they have no known competing finan-
accuracy, convergence and reliability. In the MLJAYA the adaptive cial interests or personal relationships that could have appeared
weighting consists of helping the algorithm reach the potential to influence the work reported in this paper.
search space quickly and perform a local search afterward. The
multiple learning strategies are employed to improve the diversity References
of the population, and a chaos perturbation mechanism is pro-
posed to reinforce the quality of the best solution in each genera- [4] Y.u. Wang, S. Zhou, H. Huo, Cost and CO2 reductions of solar photovoltaic
power generation in China: perspectives for 2020, Renew. Sustain. Energy Rev.
tion. The proposed algorithm is applied to solve three solar PV
39 (2014) 370–380.
models, such as the single and double diode models of the bench- [5] R. Foster, M. Ghassemi, A. Cota. Renewable Energy and the Environment 2010.
mark commercial R.T.C France silicon solar cell and the single [6] A. Yahya-Khotbehsara, A. Shahhoseini, A fast modeling of the double-diode
diode of the benchmark Photowatt-PWP 201 PV module. Based model for PV modules using combined analytical and numerical approach, Sol.
Energy 162 (2018) 403–409, https://doi.org/10.1016/j.solener.2018.01.047.
on the experimental results, the conclusions are summarized as [7] H. Attia, High performance PV system based on artificial neural network MPPT
follows: with PI controller for direct current water pump applications, Int. J. Power
Electron. Drive Syst. 10 (3) (2019) 1329.
[8] R. Tamrakar, A. Gupta, A review: extraction of solar cell modelling parameters,
 The accuracy of the double diode model is slightly better than Int. J. Innov. Res. Electr. Electron. Instrum. Control Eng. 3 (2015) 55–60.
the single diode model. [9] Z. Chen, L. Wu, P. Lin, Y. Wu, S. Cheng, Parameters identification of photovoltaic
 The proposed MLJAYA demonstrates the superior performance models using hybrid adaptive Nelder-Mead simplex algorithm based on eagle
strategy, Appl. Energy 182 (2016) 47–57.
to that of the original JAYA. [10] D.F. Alam, D.A. Yousri, M.B. Eteiba, Flower Pollination Algorithm based solar
 The experiment results achieved show that the MLJAYA has bet- PV parameter estimation, Energy Convers Manag 101 (2015) 410–422,
ter performance compared with some variants of JAYA, and https://doi.org/10.1016/j.enconman.2015.05.074.
[11] V. Khanna, B.K. Das, D. Bisht, Vandana, P.K. Singh, A three diode model for
other state-of-the-art algorithms in PV models parameters industrial solar cells and estimation of solar cell parameters using PSO
identification. algorithm, Renew. Energy 78 (2015) 105–113, https://doi.org/10.1016/j.
 As per the solution accuracy results, the I-V and P-V curves, the renene.2014.12.072.
[12] V. Stornelli, M. Muttillo, T. de Rubeis, I. Nardi, A new simplified five-
individual absolute errors (IAE) of current and power, MLJAYA
parameter estimation method for single-diode model of photovoltaic panels,
provides superior or similar performance on single diode model, Energies 12 (2019) 4271.
double diode model and PV module. [13] A. Orioli, A. Di Gangi, A procedure to evaluate the seven parameters of the two-
diode model for photovoltaic modules, Renew. Energy 139 (2019) 582–599.
[14] M.H. Qais, H.M. Hasanien, S. Alghuwainem, Identification of electrical
Therefore, MLJAYA algorithm can be efficiently used and reli- parameters for three-diode photovoltaic model using analytical and
able alternative to parameter identification of PV models. sunflower optimization algorithm, Appl. Energy 250 (2019) 109–117.

121
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

[15] D. Oliva, E. Cuevas, G. Pajares, Parameter identification of solar cells using values, Energy Convers. Manage. 214 (2020) 112904, https://doi.org/10.1016/j.
artificial bee colony optimization, Energy 72 (2014) 93–102. enconman.2020.112904.
[16] M. Črepinšek, S.-H. Liu, M. Mernik, Exploration and exploitation in [44] D. Kler, P. Sharma, A. Banerjee, K.P.S. Rana, V. Kumar, PV cell and module
evolutionary algorithms: A survey, ACM Comput. Surv. 45 (3) (2013) 1–33. efficient parameters estimation using Evaporation Rate based Water Cycle
[17] X.-S. Yang, S. Deb, S. Fong, Metaheuristic algorithms: optimal balance of Algorithm, Swarm Evol. Comput. 35 (2017) 93–110.
intensification and diversification, Appl. Math. Inf. Sci. 8 (3) (2014) 977–983. [45] O.S. Elazab, H.M. Hasanien, I. Alsaidan, A.Y. Abdelaziz, S.M. Muyeen, Parameter
[18] S. Li, W. Gong, X. Yan, C. Hu, D. Bai, L. Wang, Parameter estimation of estimation of three diode photovoltaic model using grasshopper optimization
photovoltaic models with memetic adaptive differential evolution, Sol. Energy algorithm, Energies 13 (2) (2020) 497, https://doi.org/10.3390/en13020497.
190 (2019) 465–474. [46] H. Rezk, T.S. Babu, M. Al-Dhaifallah, H.A. Ziedan, A robust parameter
[19] X. Chen, B. Xu, C. Mei, Y. Ding, K. Li, Teaching–learning–based artificial bee estimation approach based on stochastic fractal search optimization
colony for solar photovoltaic parameter estimation, Appl. Energy 212 (2018) algorithm applied to solar PV parameters, Energy Rep. 7 (2021) 620–640,
1578–1588, https://doi.org/10.1016/j.apenergy.2017.12.115. https://doi.org/10.1016/j.egyr.2021.01.024.
[20] X. Lin, Y. Wu, Parameters identification of photovoltaic models using niche- [47] G. Xiong, J. Zhang, X. Yuan, D. Shi, Y. He, G. Yao, Parameter extraction of solar
based particle swarm optimization in parallel computing architecture, Energy photovoltaic models by means of a hybrid differential evolution with whale
196 (2020) 117054, https://doi.org/10.1016/j.energy.2020.117054. optimization algorithm, Sol. Energy 176 (2018) 742–761, https://doi.org/
[21] T.V. Luu, N.S. Nguyen, Parameters extraction of solar cells using modified JAYA 10.1016/j.solener.2018.10.050.
algorithm, Optik (Stuttg) 203 (2020) 164034, https://doi.org/10.1016/j. [48] A.A.K. Ismaeel, E.H. Houssein, D. Oliva, M. Said, Gradient-based optimizer for
ijleo.2019.164034. parameter extraction in photovoltaic models, IEEE Access 9 (2021) 13403–
[22] J.D. Bastidas-Rodriguez, G. Petrone, C.A. Ramos-Paja, G. Spagnuolo, A genetic 13416, https://doi.org/10.1109/ACCESS.2021.3052153.
algorithm for identifying the single diode model parameters of a photovoltaic [49] Z. Hu, W. Gong, S. Li, Reinforcement learning-based differential evolution for
panel, Math. Comput. Simul. 131 (2017) 38–54, https://doi.org/10.1016/ parameters extraction of photovoltaic models, Energy Rep. 7 (2021) 916–928,
j.matcom.2015.10.008. https://doi.org/10.1016/j.egyr.2021.01.096.
[23] W. Gong, Z. Cai, C.X. Ling, DE/BBO: a hybrid differential evolution with [50] R.C.M. Gomes, M.A. Vitorino, M.B. de Rossiter Correa, D.A. Fernandes, R. Wang,
biogeography-based optimization for global numerical optimization, Soft. Shuffled complex evolution on photovoltaic parameter extraction: a
Comput. 15 (4) (2010) 645–665. comparative analysis, IEEE Trans. Sustain. Energy 8 (2) (2017) 805–815.
[24] D. Oliva, M. Abd El Aziz, H.A. Ella, Parameter estimation of photovoltaic cells [51] K.M. Sallam, M.A. Hossain, R.K. Chakrabortty, M.J. Ryan, An improved gaining-
using an improved chaotic whale optimization algorithm, Appl. Energy 200 sharing knowledge algorithm for parameter extraction of photovoltaic models,
(2017) 141–154, https://doi.org/10.1016/j.apenergy.2017.05.029. Energy Convers. Manage. 237 (2021) 114030, https://doi.org/10.1016/j.
[25] R. Cheng, Y. Jin, A competitive swarm optimizer for large scale optimization, enconman.2021.114030.
IEEE Trans. Cybern. 45 (2014) 191–204. [52] D.H. Muhsen, A.B. Ghazali, T. Khatib, I.A. Abed, A comparative study of
[26] G. Xiong, J. Zhang, D. Shi, L. Zhu, X. Yuan, Z. Tan, Winner-leading competitive evolutionary algorithms and adapting control parameters for estimating the
swarm optimizer with dynamic Gaussian mutation for parameter extraction of parameters of a single-diode photovoltaic module’s model, Renew. Energy 96
solar photovoltaic models, Energy Convers. Manage. 206 (2020) 112450. (2016) 377–389.
[27] J. Liang, S. Ge, B. Qu, K. Yu, F. Liu, H. Yang, P. Wei, Z. Li, Classified perturbation [53] V. Lo Brano, A. Orioli, G. Ciulla, A. Di Gangi, An improved five-parameter model
mutation based particle swarm optimization algorithm for parameters for photovoltaic modules, Sol. Energy Mater. Sol. Cells 94 (8) (2010) 1358–
extraction of photovoltaic models, Energy Convers. Manage. 203 (2020) 1370.
112138. [54] W. Gong, Z. Cai, Parameter extraction of solar cell models using repaired
[28] R. Rao, Jaya: a simple and new optimization algorithm for solving constrained adaptive differential evolution, Sol. Energy 94 (2013) 209–220.
and unconstrained optimization problems, Int. J. Ind. Eng. Comput. 7 (2016) [55] P. Sharma, T.S. Bhatti, A review on electrochemical double-layer capacitors,
19–34. Energy Convers. Manage. 51 (12) (2010) 2901–2912.
[29] K. Yu, J.J. Liang, B.Y. Qu, X. Chen, H. Wang, Parameters identification of [56] A.A. Elbaset, H. Ali, M. Abd El Sattar, New seven parameters model for
photovoltaic models using an improved JAYA optimization algorithm, Energy amorphous silicon and thin film PV modules based on solar irradiance, Sol.
Convers. Manage. 150 (2017) 742–753, https://doi.org/10.1016/j. Energy 138 (2016) 26–35.
enconman.2017.08.063. [57] Messaoud R Ben. Extraction of uncertain parameters of single and double
[30] X. Jian, Z. Weng, A logistic chaotic JAYA algorithm for parameters identification diode model of a photovoltaic panel using Salp Swarm algorithm.
of photovoltaic cell and module models, Optik (Stuttg) 203 (2020) 164041, Measurement 2020;154:107446.
https://doi.org/10.1016/j.ijleo.2019.164041. [58] D.H. Muhsen, A.B. Ghazali, T. Khatib, I.A. Abed, Parameters extraction of double
[31] S. Li, W. Gong, X. Yan, C. Hu, D. Bai, L. Wang, L. Gao, Parameter extraction of diode photovoltaic module’s model based on hybrid evolutionary algorithm,
photovoltaic models using an improved teaching-learning-based optimization, Energy Convers. Manage. 105 (2015) 552–561.
Energy Convers. Manage. 186 (2019) 293–305. [59] R.V. Rao, D.P. Rai, J. Balic, Multi-objective optimization of machining and
[32] X. Chen, K. Yu, Hybridizing cuckoo search algorithm with biogeography-based micro-machining processes using non-dominated sorting teaching–learning-
optimization for estimating photovoltaic model parameters, Sol. Energy 180 based optimization algorithm, J. Intell. Manuf. 29 (8) (2018) 1715–1737.
(2019) 192–206, https://doi.org/10.1016/j.solener.2019.01.025. [60] K. Yu, X. Wang, Z. Wang, An improved teaching-learning-based optimization
[33] K. Yu, B. Qu, C. Yue, S. Ge, X. Chen, J. Liang, A performance-guided JAYA algorithm for numerical and engineering optimization problems, J. Intell.
algorithm for parameters identification of photovoltaic cell and module, Appl. Manuf. 27 (4) (2016) 831–843.
Energy 237 (2019) 241–257. [61] L.dos.S. Coelho, T.C. Bora, L. Lebensztajn, A chaotic approach of differential
[34] K. Yu, J.J. Liang, B.Y. Qu, Z. Cheng, H. Wang, Multiple learning backtracking evolution optimization applied to loudspeaker design problem, IEEE Trans.
search algorithm for estimating parameters of photovoltaic models, Appl. Magn. 48 (2) (2012) 751–754.
Energy 226 (2018) 408–422. [62] B. Alatas, Chaotic bee colony algorithms for global numerical optimization,
[35] C. Huang, L. Wang, R.-C. Yeung, Z. Zhang, H.-H. Chung, A. Bensoussan, A Expert Syst. Appl. 37 (8) (2010) 5682–5687.
prediction model-guided Jaya algorithm for the PV system maximum power [63] S. Li, Q. Gu, W. Gong, B. Ning, An enhanced adaptive differential evolution
point tracking, IEEE Trans. Sustain. Energy 9 (1) (2018) 45–55. algorithm for parameter extraction of photovoltaic models, Energy Convers.
[36] R.V. Rao, K.C. More, J. Taler, P. Ocłoń, Dimensional optimization of a micro- Manage. 205 (2020) 112443.
channel heat sink using Jaya algorithm, Appl. Therm. Eng. 103 (2016) 572–582. [64] A. Askarzadeh, A. Rezazadeh, Parameter identification for solar cell models
[37] R.V. Rao, G.G. Waghmare, A new optimization algorithm for solving complex using harmony search-based algorithms, Sol. Energy 86 (11) (2012) 3241–
constrained design optimization problems, Eng. Optim. 49 (1) (2017) 60–83. 3249.
[38] S.-H. Wang, P. Phillips, Z.-C. Dong, Y.-D. Zhang, Intelligent facial emotion [65] J.P. Ram, T.S. Babu, T. Dragicevic, N. Rajasekar, A new hybrid bee pollinator
recognition based on stationary wavelet entropy and Jaya algorithm, flower pollination algorithm for solar PV parameter estimation, Energy
Neurocomputing 272 (2018) 668–676. Convers. Manage. 135 (2017) 463–476, https://doi.org/10.1016/j.
[39] N. Kumar, I. Hussain, B. Singh, B.K. Panigrahi, Rapid MPPT for uniformly and enconman.2016.12.082.
partial shaded PV system by using JayaDE algorithm in highly fluctuating [66] A. Babu, A.K. Sao. Unit Selection Using Acoustic Supra-Segmental Cues to
atmospheric conditions, IEEE Trans. Ind. Inf. 13 (5) (2017) 2406–2416. Improve Prosody. Recent Adv. Nonlinear Speech Process., Springer; 2016, p.
[40] J.L. Ravipudi, M. Neebha, Synthesis of linear antenna arrays using jaya, self- 265–73.
adaptive jaya and chaotic jaya algorithms, AEU-Int. J; Electron. Commun. 92 [67] C. Chellaswamy, R. Ramesh, Parameter extraction of solar cell models based on
(2018) 54–63. adaptive differential evolution algorithm, Renew. Energy 97 (2016) 823–837.
[41] J. Liang, K. Qiao, M. Yuan, K. Yu, B. Qu, S. Ge, Y. Li, G. Chen, Evolutionary multi- [68] Y. Zhang, M. Ma, Z. Jin, Comprehensive learning Jaya algorithm for parameter
task optimization for parameters extraction of photovoltaic models, Energy extraction of photovoltaic models, Energy 211 (2020) 118644.
Convers. Manage. 207 (2020) 112509, https://doi.org/10.1016/j. [69] Q. Niu, L. Zhang, K. Li, A biogeography-based optimization algorithm with
enconman.2020.112509. mutation strategies for model parameter estimation of solar and fuel cells,
[42] W. Long, S. Cai, J. Jiao, M. Xu, T. Wu, A new hybrid algorithm based on grey wolf Energy Convers. Manage. 86 (2014) 1173–1185.
optimizer and cuckoo search for parameter extraction of solar photovoltaic [70] X. Chen, K. Yu, W. Du, W. Zhao, G. Liu, Parameters identification of solar cell
models, Energy Convers. Manage. 203 (2020) 112243, https://doi.org/10.1016/ models using generalized oppositional teaching learning based optimization,
j.enconman.2019.112243. Energy 99 (2016) 170–180, https://doi.org/10.1016/j.energy.2016.01.052.
[43] M.H. Qais, H.M. Hasanien, S. Alghuwainem, Transient search optimization for [71] K. Yu, X. Chen, X. Wang, Z. Wang, Parameters identification of photovoltaic
electrical parameters estimation of photovoltaic module based on datasheet models using self-adaptive teaching-learning-based optimization, Energy
Convers. Manage. 145 (2017) 233–246.

122
D. Saadaoui, M. Elyaqouti, K. Assalaou et al. Materials Today: Proceedings 52 (2022) 108–123

[72] J. Liang, K. Qiao, K. Yu, S. Ge, B. Qu, R. Xu, K. Li, Parameters estimation of solar solar projects in Iran: new application of a hybrid multi-criteria decision
photovoltaic models via a self-adaptive ensemble-based differential evolution, making approach, Energy Convers. Manag. 86 (2014) 653–663.
Sol. Energy 207 (2020) 336–346, https://doi.org/10.1016/j.solener.2020.06.100. [2] S. Mekhilef, R. Saidur, A. Safari, A review on solar energy use in industries,
Renew. Sustain. Energy Rev. 15 (4) (2011) 1777–1790.
[3] M. Kumar, A. Kumar, An efficient parameters extraction technique of
Further reading photovoltaic models for performance assessment, Sol. Energy 158 (2017)
192–206.
[1] M. Vafaeipour, S. Hashemkhani Zolfani, M.H. Morshed Varzandeh, A. Derakhti,
M. Keshavarz Eshkalag, Assessment of regions priority for implementation of

123