Energy Conversion and Management: X 19 (2023) 100405

Contents lists available at ScienceDirect

Energy Conversion and Management: X

journal homepage: www.sciencedirect.com/journal/energy-conversion-and-management-x

Dandelion Optimizer algorithm-based method for accurate photovoltaic

model parameter identification
Abdelfattah Elhammoudy *, Mustapha Elyaqouti , El Hanafi Arjdal , Dris Ben Hmamou ,
Souad Lidaighbi , Driss Saadaoui , Imade Choulli , Ismail Abazine
Laboratory of Materials, Signals, Systems and Physical Modelling, Faculty of Science, Ibn Zohr University, Agadir, Morocco

A R T I C L E I N F O A B S T R A C T

Keywords: The utilization of photovoltaic (PV) energy has experienced a significant surge in the last few decades, resulting
Photovoltaic cells in a rise in research endeavours to comprehend its workings better. One of the focal points of this research is the
Parameters extraction electrical modelling of PV cells and modules. Several equivalent circuits have been proposed to model them, such
Optimization
as the single-diode (SDM), double-diode (DDM), and triple-diode model (TDM). The main challenge is identifying
Dandelion Optimizer algorithm
the optimal circuit model parameters. This study introduces a novel method based on a metaheuristic algorithm
named Dandelion Optimizer (DO) coupled with a numerical method Newton-Raphson (NR) to estimate the PV
parameters. Various PV models, including the single-diode model (SDM) and double-diode model (DDM), were
utilized by the proposed method (DONR) to determine the PV parameters of six different cells and modules, such
as RTC France, Photowatt-PWP201, and STP6-120/36. A comparative analysis was conducted with ten other
widely recognized metaheuristic methods to demonstrate the effectiveness of the proposed method. The results
demonstrate that the proposed algorithm is more accurate in estimating the optimal PV parameters than the
other methods. According to the experimental results, the proposed method is superior to other methods in
accurately estimating the optimal PV parameters in terms of accuracy, reliability, and convergence. Specifically,
the root mean squared error values obtained by the proposed method using (SDM, DDM) for RTC France,
PWP201, and STP6-120/36 are (7.73939E-04, 7.56515E-04), (2.08116E-03, 2.07842E-03) and (1.42575E-02,
1.45952E-02), respectively.

1. Introduction sources, such as hydro power, wind, biomass, geothermal and solar. The
latter is the primary indirect source of many renewable energies like
The history of energy on the earth commenced with the sun; it was wind, bioenergy, and ocean current. Solar energy is directly exploited to
the first energy source to provide heat and light. After that, fire’s dis­ obtain thermal energy by concentrating solar irradiation or obtain
covery made a revolution; it was used for various uses, such as cooking electricity by converting solar irradiation into electricity using photo­
and as a source of light and heat. Later, humans exploited the energy of voltaic (PV) panels; they are semiconductor devices that can convert
wind and water for transportation and grinding grain. The discovery of sunlight easily into electricity. PV energy is one of the world’s fastest
fossil fuels made an industrial revolution, where its production was growing and increasingly important renewable energy technologies;
around 93.467 million tonnes of oil equivalent per decade from 1800 to according to the International Energy Agency, the cumulative PV energy
2009 [1]. Coal, oil, and natural gas are the most common fossil fuels; capacity almost triples to over 2350 GW by 2027, in the leading case [3].
their combustion generates a toxic gas, carbon dioxide CO2, whereas the This increase in using PV technology is accompanied by an increase in
emissions of the CO2 in the world is 33884.06 MTCO2 (Metric tons of scientific research about this technology; among this scientific research,
carbon dioxide) from 1965 to 2021 [2]. Carbon dioxide is a greenhouse there is maximum power point tracking [4] and thermal modelling [5].
gas, and its emission into the atmosphere raises the planet’s tempera­ Their goals are always to give a solution to exploit all the power that the
ture, causing climate change. To solve this problem and reduce the PV module can provide. All research is not conducted directly on a PV
emissions of CO2, the world turned its attention to renewable energy module but on an electrical model simulating its function.

* Corresponding author.
E-mail address: [email protected] (A. Elhammoudy).

https://doi.org/10.1016/j.ecmx.2023.100405
Received 2 April 2023; Received in revised form 26 May 2023; Accepted 6 June 2023
Available online 9 June 2023
2590-1745/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

There are many electrical models in the literature, such as the single Table 1
diode model (SDM) [6], the double diode model (DDM) [7], and the PV The latest metaheuristics algorithms for PV parameters extraction.
module model (MM) [8]; they will be presented in detail in the next Authors & Used Cell/ Remarks & Results
section. The main challenge is solving the nonlinear equation given by Year algorithm Module
those models and determining its unknown parameters; in literature, K. Yu et al. JAYA RTC The improved JAYA (IJAYA) can
there are several methods of different natures. There exist analytical 2017 [44] France identify the parameters of single-
approaches that depend on mathematical equations to determine the PWP201 diode (SMD), double-diode (DDM),
unknown parameters [9]. For instance, V. Lo Brano [10] proposed an and PV module models (MM).
• RMSE (RTC France) = 9.8603E-04
analytical approach that uses data sheets and transparent analytical A
procedures to derive equations for SDM. H. El Achouby et al. [11] used • RMSE (PWP201) = 2.4251E-03 A
four equations that were derived from the three notable points of the I-V K. Yu et al. BSA RTC The proposed method in this work
characteristics to establish a system of equations that can determine all 2018 [42] France (MLBSA) is a variant of the
PWP201 backtracking search algorithm (BSA)
parameters except the ideality factor, which is treated as a variable
using multiple learning. MLBSA is
parameter. Numerical methods, on the other hand, rely on an initial comparable to other approaches in
estimate of the unknown parameters and iterative improvements to the terms of accuracy, dependability, and
estimate until convergence is achieved. For example, M.G. Villalva et al. computational efficiency, according
[12] introduced a method whereby the series resistance is initially set to to its results.
• RMSE (RTC France) = 9.8602E-04
zero and subsequently incremented until a satisfactory level of precision A
is achieved between the calculated maximum power values and its • RMSE (PWP201) = 2.425075E-03
corresponding value in the manufacturer’s datasheet. A. Elhammoudy A
et al. [13] proposed a numerical method that uses Dichotomy to solve a S. Li, et al. TLBO RTC The improved teaching–learning-
2019 [41] France based optimization ITLBO improves
nonlinear equation for the series resistance, and an iterative process to
PWP201 teaching and learning phases. ITLBO
determine the diode ideality factor. Furthermore, some methods are a STM6-40/ was used to accurately extract the PV
hybrid between analytical and numerical [14 15 16]. In recent years, 36 parameters for SDM, DDM, and MM.
metaheuristics algorithms have gained popularity among researchers for STP6- • RMSE (RTC France) = 9.8602E-04
solving engineering problems due to their advanced development. These 120/36 A
• RMSE (PWP201) = 2.4251E-03 A
algorithms are widely employed to solve a range of engineering prob­
J. Liang, DE RTC The results indicate that SEDE’s
lems, including PV modelling [17]. Moreover, they have found appli­ et al. 2020 France performance is highly competitive
cations in wind speed prediction [18], forecasting crude oil prices [19], [29] PWP201 with other variants of other
and plant disease detection [20]. Several categories of metaheuristic algorithms, such as PSO and TLBO.
• RMSE (RTC France) = 9.860218E-
algorithms include evolutionary algorithms, swarm intelligence algo­
04 A
rithms, and sociology-based algorithms. • RMSE (PWP201) = 2.425074E-04
Natural evolution principles like selection, mutation, and reproduc­ A
tion inspire evolutionary algorithms. Among these algorithms is the M. Mostafa SMA PWP201 The proposed method uses the slime
genetic algorithm (GA) [21], which uses natural selection and repro­ et al. LSM 20 mould algorithm (SMA) to determine
2020 [45] the PV parameters of SDM and DDM.
duction to search for better solutions to a given problem. Several ap­
The results show that SMA achieves
proaches are employed, including variations of this algorithm or its minimized RMSE compared to other
integration with other algorithms, to estimate the PV parameters [22 methods, such as IJAYA.
23]. Evolutionary strategies (ES) [24] algorithm uses principles of bio­ • RMSE (PWP201) = 2.0574E-04 A
D. Saadaoui GA RTC GAMNU is a variant of the GA
logical evolution to search for the best solutions.
et al. France algorithm based on non-uniform
Differential evolution (DE) [25] algorithm inspired by the principles 2021 [21] PWP201 mutation; it accurately identifies the
of natural evolution and the movement of animals in a population. It has STP6- unknown parameters for SDM and
improved to other novel algorithms, such as the heterogeneous differ­ 120/36 DDM.
ential evolution algorithm (HDE) [26], the directional permutation ESP-160 • RMSE (RTC France) = 9.8618E-04
PPW A
differential evolution (DPDE) [27], the fuzzy adaptive differential evo­
• RMSE (PWP201) = 2.382420E-03
lution algorithm (FADE) [28], and the self-adaptive ensemble-based A
differential evolution (SEDE) [29]. The flower pollination algorithm L. Li et al. TLBO RTC Dynamic Self-Adaptive and Mutual-
(FPA) [30] is based on the pollination process in nature for solving 2021 [40] France Comparison Teaching-Learning-
PWP201 Based Optimization (DMTLBO) is
optimization problems by simulating the movement of pollen grains
STM6-40/ applied to SDM, DDM, and the PV
between the flowers. The honey badger algorithm (HBA) [31] is created 36 modules models (MM). The results
from the intelligent foraging behaviour of the honey badger. STP6- given by this method are comparable
The collective behaviour of animals and insects, such as ants and 120/36 with several variants of the TLBO
bees, inspires swarm intelligence algorithms. For example, the artificial algorithm.
• RMSE (RTC France) = 9.8602E-04
bee swarm optimization (ABSO) [32] algorithm is based on the hy­
A
pothesis that a swarm of bees can find food more effectively than a single • RMSE (PWP201) = 2.4251E-03 A
bee by segmenting the search space into smaller regions and examining S. Gao et al. DE RTC The directional permutation
them concurrently. The particle swarm optimization (PSO) [33] algo­ 2021 [27] France differential evolution algorithm
rithm is inspired by the movement of birds and fish in a flock. It involves PWP201 (DPDE) is a variant of the DE
STM6-40/ algorithm. It is applied for SDM,
a group of particles that move and interact with each other to find the 36 DDM, triple-diode (TDM), and MM
optimal solution to a given problem. Various improvement approaches STP6- and compared with 15 methods. The
have utilized this algorithm, which involves modifications to the PSO or 120/36 results show that DPDE outperforms
combining it with other methods to identify the PV parameters [34 35 its peers in terms of solution
accuracy.
36]. The salp swarm algorithm (SSA) [37] is inspired by the movement
• RMSE (RTC France) = 9.860218E-
patterns of salpas and based on the idea that a group of salpas can move 04 A
and interact with each other to solve complex problems more effectively (continued on next page)
than a single salpa working alone, it has been developed to a novel

2
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Table 1 (continued ) improvement using self-adaptive technique-based genetic algorithm

Authors & Used Cell/ Remarks & Results (SSAGA) [38].
Year algorithm Module Sociology-based algorithms use principles from the social sciences,
• RMSE (PWP201) = 2.425074E-03
particularly sociology, to solve problems or make decisions; they are
A designed to mimic individuals’ or groups’ behaviour and decision-
D. Wang DE RTC The heterogeneous differential making processes within society. Out of these algorithms, there is
et al. France evolution algorithm (HDE) improves teaching learning-based optimization (TLBO) [39]. It is based on an
2022 [26] PWP201 the DE algorithm. The authors
education process that involves the exchange of knowledge and skills
STM6-40/ applied it to determine the SDM,
36 DDM, and TDM parameters; they between teachers and students, as well as among students themselves. It
STP6- compared the results with numerous has enhanced other novel algorithms, like dynamic self-adaptive and
120/36 algorithms. mutual-comparison teaching-learning-based optimization (DMTLBO)
• RMSE (RTC France) = 9.860218E- [40], and improved teaching-learning-based optimization (ITLBO) [41].
04 A
• RMSE (PWP201) = 2.425074E-03
The following are additional metaheuristic algorithms, such as the
A multiple learning backtracking search algorithm (MLBSA) [42], which
M. Abdel- GTO RTC The authors propose a new algorithm combines backtracking search principles with machine learning tech­
Basset et France called MIGTO, a variant of the gorilla niques. The memory-based improved gorilla troops optimizer (MIGTO)
al PWP201 troops optimizer (GTO) algorithm.
[43] and improved Jaya algorithm (IJAYA) [44] is a variant of the Jaya
2022 [43] STM6-40/ MIGTO has estimated the PV
36 parameters for SDM, DDM, TDM and algorithm, which the behaviour of birds inspires in flocking and
STP6- MM with precision. swarming.
120/36 • RMSE (RTC France) = 9.860218E- Table 1 details the results and performance of the latest meta­
KC200GT 04 A heuristics algorithm for the extraction of PV parameters. Most of the
SM55 • RMSE (PWP201) = 2.425074E-03
ST40 A
studies in the literature employ the RTC France cell and PWP201 module
T. Düzenli HBA RTC This work presents two variants of as a validation benchmark for their methods. In this regard, Table 1 also
et al. France HBA: Chaotic HBA and HBA.OBL is a provides the corresponding root mean square error (RMSE) values for
2022 [46] PWP201 combination between HBA and SDM yielded by each method, allowing for a thorough evaluation of
STM6-40/ opposition-based learning (OBL).
their accuracy.
36 They have been tested on SDM, DDM,
STP6- and MM. They obtained reliable and A novel method is presented in this paper based on a metaheuristic
120/36 consistent parameter estimates for algorithm named Dandelion Optimizer (DO) [50] and a numerical
SM55 various panels, as evidenced by low method of Newton–Raphson to identify the unknown PV parameters
ST40 RMSE and standard deviation values. using SDM, DDM, and MM. The main contributions of this work can be
KC200GT • RMSE (RTC France) = 9.8602E-04
summarized in the following points:
A
• RMSE (PWP201) = 2.4251E-03 A
M.A. El- NGO PWP201 In this study, the authors use a new • An application of a recently introduced optimization algorithm
Dabah KC200GT algorithm called Northern Goshawk called DO for the purpose of extracting PV parameters.
et al. CS6K-280- Optimization (NGO) to identify the
• A numerical method, specifically Newton-Raphson, is utilized to
2023 [47] M PV parameters using TDM. The
results show that the algorithm’s determine the nonlinear equation of the objective function (RMSE).
convergence is fast compared with • An extensive comparison with ten competitive metaheuristics algo­
other methods mentioned in the rithms verifies that DONR performs the best from the perspective of
work. Notably, the authors utilized a statistical results.
cost function as the objective
function, instead of relying on the
RMSE. The rest of this work is divided into four sections: the mathematical
Z. Garip HHO RTC The proposed methods in this work modelling of PV cells and modules and the demonstration of their
2023 [48] France are variants of the Harris Hawks objective functions are brought in the second section. The DO algorithm
PWP201 optimization (HHO) algorithm using
and the proposed method are introduced in the third section. The fourth
a fractal sequence obtained from
fractal henon chaotic maps. The three
section presents the results and the comparison. Finally, the fifth section
proposed variants show the is the conclusion and some future work.
superiority and accuracy of the
performance in determining the PV 2. PV models and their objective functions
parameters for SDM, DDM, and TDM.
However, a disadvantage of these
methods is that the RMSE values are In literature, the single diode model (SDM), the double diode model
either significantly higher or lower (DDM), and the PV module model (MM) are usually used to describe the
than the mean. current–voltage characteristics of solar cells and PV modules. This sec­
Abd El- DE RTC This combines two algorithms:
tion introduces more details about these models and their objective
Mageed QSO France queuing search optimization (QSO)
et al. PWP201 and differential evolution (DE). The
functions.
2023 [49] STM6-40/ proposed method IQSODE can extract
36 the parameters of PV models, such as 2.1. Single diode model (SDM)
STP6- SDM, DDM, and MM, with reasonable
120/36 accuracy.
Fig. 1(a) shows that the single diode model consists of a current
ST40 • RMSE (RTC France) = 9.860218E-
SM55 04 A source representing the incident flux irradiation. It is parallel with a
• RMSE (PWP201) = 2.425074E-03 diode representing the P-N junction connected with a parallel resistance
A Rsh (shunt resistance) and a series resistance Rs representing the leakage
current and the ohmic contacts, respectively [15]. The output current IL
is calculated in the following manner:
IL = Iph − ID − Ish (1)

3
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Fig. 1. The corresponding electric circuit for: (a) single diode model, (b) double diode model and (c) PV module model.

Where Iph represents the photo-generated current, ID is the diode current 2.3. PV module model (MM)
determined by the Shockley diode equation, and Ish is the shunt resistor
current. The PV module model (MM) consists of many identical solar cells
[ ( ) ] connected in series and parallel [26]. The output current IL for the PV
q(VL + Rs IL )
ID = Isd exp − 1 (2) module model based on a single diode model can be expressed as
akT
follows:
VL + Rs I L [ ( ( )) ]
(3) q VL Np + Rs IL Ns VL Np + Rs IL Ns
Ish =
Rsh IL = Iph Np − Isd Np exp − 1 − (7)
akTNs Np Rsh Ns

Where a represents the diode ideality factor, q = 1.6021764610− 19 C is As the single diode model there is five unknown parameters are Iph ,
the charge of the electron, k = 1.380650310− 23 J/K is the Boltzmann Isd , Rs , Rsh and a.
constant, and T is the cell temperature in Kelvin (K).
After the Eqn 1, 2 and 3, the output current can be written as the
following expression: 2.4. Objective function
[ ( ) ]
q(VL + Rs IL ) VL + Rs I L The problem with all the previous models is that the unknown pa­
IL = Iph − Isd exp − 1 − (4)
akT Rsh rameters need to be identified for the condition that the calculated
There are five unknown parameters are Iph , Isd , Rs , Rsh and a. current is almost identical to the experimental current. Therefore, the
identification problem is converted into an optimization problem; the
2.2. Double diode model (DDM) goal is to optimize an objective function is the root mean squared error
(RMSE). It is the differences between the simulated current Isimi and the
The single diode model neglects the effect of recombination current experimental current values Ii divided by the number of data n.
loss in the space charge region. The double diode model considers this √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√∑
loss; it has another diode representing it, as shown in Fig. 1(b) [51]; √n
√ (Isimi − Ii )2
√
consequently, there are two diodes, current ID1 and ID2 . As a result, the RMSE = i=1 (8)
output current for DDM is determined by: n

IL = Iph − ID1 − ID2 − Ish (5) After the output current from the previous models, the simulated
current is defined for every model by the following equations:
[ ( ) ] [ ( ) ]
q(VL + Rs IL ) q(VL + Rs IL )
IL =Iph − Isd 1 exp
a1 kT
− 1 − Isd 2 exp
a2 kT
− 1 • SDM
VL + Rs I L ⎧ [ ( ) ]
−
Rsh ⎨ Isim = f (IL , VL , X) = Iph − Isd exp q(VL + Rs IL ) − 1 − VL + Rs IL
⎪
akT Rsh
(6) ⎪
⎩ { }
X = Iph , Isd 1 , Rs , Rsh , a1 , Isd 2 , a2
There are seven unknown parameters are Iph , Isd1 , Rs , Rsh , a1 , Isd2 and
(9)
a2 .

• DDM

4
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Fig. 2. The schematic of the rising phase of dandelion seeds for the two cases.

⎧ [ ( ) ] [ ( ) ]
⎨ Isim = f (IL , VL , X) = Iph − Isd 1 exp q(VL + Rs IL ) − 1 − Isd 2 exp q(VL + Rs IL ) − 1 − VL + Rs IL
⎪
a1 kT a2 kT Rsh (10)
⎪
⎩ { }
X = Iph , Isd , Rs , Rsh , a

• MM global search and a local search. The second phase involves the descent
of the seeds as they rise to a certain height and then steadily fall toward
the ground. Finally, during the landing phase, the dandelion seeds

⎧ [ ( ( )) ]
⎨ Isim = f (IL , VL , X) = Iph Np − Isd Np exp q VL Np + Rs IL Ns
⎪
− 1 −
VL Np + Rs IL Ns
akTNs Np Rsh Ns (11)
⎪
⎩ { }
X = Iph , Isd , Rs , Rsh , a

As we can see, the equation of the simulated current is nonlinear. So, randomly settle in different locations under the influence of wind and
to calculate the simulated current, it is necessary to use a numerical weather, ultimately growing into new dandelions.
method for solving nonlinear algebraic problems. In the literature, Similar to other metaheuristic algorithms, Dandelion Optimizer (DO)
several numerical methods of solving Nonlinear equations exist, such as is based on the iterative evolution of an initial population. The popu­
the Newton–Raphson method [52], the Regular-Falsi method [53], and lation is represented by the following expression, where P represents the
the secant method, etc. The Newton–Raphson method is the most population size, and D represents the variable dimension.
powerful and able method of solving the nonlinear equation [54]. Its ⎡ ⎤
role consists of computing the simulated current to determine the ⎢
X11 ⋯ X1D
⎥
objective function. However, as we have seen previously, there are un­ population = ⎢⎣ ⋮ ⋱ ⋮ ⎦
⎥ (12)
known parameters that need to estimate, and this is the essential part. To XP1 ⋯ XPD
determine these parameters and to optimize our objective function, we
will use a metaheuristic algorithm named Dandelion Optimizer (DO) In the context of optimization algorithms, the creation of each in­
[50]. dividual, represented as Xi , is subject to the specific constraints of the
given problem. More specifically, the expression of each individual is
3. Dandelion Optimizer generated within the boundaries determined by the upper bound (UB)
and the lower bound (LB) of the problem.
Dandelion Optimizer (DO) is a novel swarm intelligence bioinspired
Xi = rand × (UB − LB) + LB (13)
optimization algorithm proposed by Shijie Zhao [50]. DO algorithm
simulates the proceeding of dandelion seed flight to other places
Where i is an integer range from 1 to P and rand is a random number
depending on the wind; this process is divided into three phases. The
compare between 1 and 2.
first is the rising phase, in which the seeds rise in a spiral manner under
During the process of initialization, DO considers the individual with
the action of drag force; this is achieved with sunny weather. Further­
the optimal fitness value as the initial elite Xelite , as represented by the
more, during rainy weather, the dandelion seeds disperse within a local
following mathematical expression.
area. This difference of rising gives two scenarios of search, namely a

5
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Fig. 3. The schematic of descending phase of dandelion seeds and the Brownian motion model of its descending trajectory.

⎧
⎪
⎪ υx = r × cosθ
⎪
⎪
⎨
υy = r × sinθ (19)
⎪
⎪ 1
⎪
⎪
⎩ r= θ
e
In the previous equations y is a standard normal distribution N(0,1),
tmax is the maximum number of iterations and θ is a random number
between [ − π, π].
In the case of inclement weather, specifically on a rainy day, the
dandelion seeds can’t rise properly because of the air resistance. This
phenomenon can be quantified through the following mathematical
expression:
Xt+1 = Xt × k (20)
Where k is a regulator of the local search domain.
⎧
⎪
⎨q = 1 2 1
t2 − t+1+
tmax 2 − 2tmax + 1 tmax 2 − 2tmax + 1 tmax 2 − 2tmax + 1 (21)
⎪
⎩
k = 1 − rand() × q

Fig. 2 presents the manner how the dandelion seeds rise in the two
previous cases, in a clear day and in a rainy day.
In descending phase, After the dandelion seeds rise to a certain
Fig. 4. The schematic of landing phase of dandelion seeds. altitude, they drop steadily. To simulate the moving trajectory of
dandelion seeds, DO use the Brownian motion. The corresponding
{
Xelite = (find(fbest == f (Xi ) ) ) mathematical expression in this phase is:
(14)
fbest = min(f (Xi ) )
Xt+1 = Xt − α × βt × (Xmean t − α × βt × Xt ) (22)
Where find() indicates two indexes with equal values.
Where βt indicates the Brownian motion [55], and Xmean t is the mean
During the rising phase, the behaviour of dandelion seeds is influ­
position of the population is represented by the following mathematical
enced by weather conditions, with two possible cases. In clear weather
expression:
conditions, dandelion seeds can ascend to significant heights. This
phenomenon can be mathematically expressed as follows: 1∑ P
Xmean t = Xi (23)
Xt+1 = Xt + α × υx × υy × lnY × (Xs − Xt ) (15) P i=1

Where Xt is the position of the dandelion seed in the iteration t, Xs is the Fig. 3 illustrates the schematic descending phase of dandelion seeds and
randomly generated position, α is an adaptive parameter used to adjust the Brownian motion model that describes their trajectory. During the
the search step length, υx and υy represent the lift component coefficients seeds’ descent, individuals’ evolution is determined by the mean
of the dandelion seed due to the separated eddy action, and lnY is a position.
lognormal distribution subject to μ = 0 and σ 2 = 1. After the two previous phases, the dandelion seed will randomly
choose where it will land, and this is the landing phase. Consequently,
Xs = rand(1, D) × (UB − LB) + LB (16) the optimal solution is the landing place where the dandelion seed will
⎧ ( ) easily survive. The corresponding mathematical expression in this phase
⎨ √1̅̅̅̅̅ exp − 1 (lny)2 y ≥ 0
⎪ is:
(17)
2
lnY = y 2π 2σ
⎪
⎩ Xt+1 = Xelite + levy(λ) × α × (Xelite − Xt × δ) (24)
0y < 0
( ) Where Xelite is the optimal position of the dandelion seed, levy(λ) is the
1 2 function of Levy flight, and δ is a linear increasing function between 1
α = rand() × t2 − t+1 (18)
and 2. Fig. 4 presents the dandelion seed’s landing phase and the search
tmax 2 tmax
for the best solution.

6
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Table 2
Upper and lower bounds of all parameters.
Parameter RTC France PWP201 STP6-120/36 S75 SM55 ST36
SDM/DDM SDM/DDM SDM/DDM

LB UB LB UB LB UB LB UB LB UB LB UB

Ipv (A) 0 1 0 2 0 8 0 2 × Isc 0 2 × Isc 0 2 × Isc

I0 , I0 1 , I0 2 (μA) 0 1 0 50 0 50 0 100 0 100 0 100
Rs (Ω) 0 0.5 0 2 0 0.36 0 2 0 2 0 2
Rsh (Ω) 0 100 0 2000 0 1500 0 5000 0 5000 0 5000
a, a1 , a2 1 2 1 50 1 50 1 2 1 2 1 2

Table 3
Estimated parameters and RMSE values for SDM.
Ipv (A) I0 (A) Rs (Ω) Rsh (Ω) a RMSE(A)

RTC France DONR 0.76079 3.03000E-07 3.66538E-02 52.41756 1.47482 7.73939E-04

FPA [30] 0.76077 3.22617E-07 3.63804E-02 53.65319 1.48105 9.86021E-04
HBA [31] 0.76077 3.23020E-07 3.63770E-02 53.71851 1.48118 9.86021E-04
IJAYA [44] 0.76076 3.23622E-07 3.63708E-02 53.83082 1.48137 9.86021E-04
HDE [26] 0.76078 3.23020E − 07 3.63770E-02 53.71852 1.48118 9.86021E-04
MIGTO [43] 0.76077 3.23020E-07 3.63770E-02 53.71852 1.481183 9.86021E-04
DMTLBO [40] 0.7608 3.2302E-06 3.6340E-02 53.7183 1.4812 9.86020E-04
ITLBO [41] 0.7608 3.2310E-07 3.6340E-02 53.7187 1.4812 9.86020E-04
DPDE [27] 0.76077 3.23020E-07 3.63770E-02 53.71852 1.48118 9.86021E-04
SEDE [29] 0.76077 3.23020E-07 3.63770E-02 53.71852 1.48118 9.86021E-04
MLSBA [42] 0.76077 3.23020E-07 3.63770E-02 53.71852 1.48118 9.86021E-04
PWP201 DONR 1.03052 3.09823E-06 1.21791 952.86730 48.19547 2.08116E-03
FPA [30] 1.03048 3.47234E-06 1.20163 985.04957 48.63156 2.42520E-03
HBA [31] 1.03051 3.48226E-06 1.20127 981.98275 48.64283 2.42507E-03
IJAYA [44] 1.03050 3.48544E-06 1.20120 983.27900 48.64631 2.42507E-05
HDE [26] 1.03051 3.48226E − 06 1.20127 981.98224 48.64280 2.42507E-03
MIGTO [43] 1.02462 20.3362E-06 0.02676 1323.48643 56.55348 2.42507E-03
DMTLBO [40] 1.03050 3.48230E-06 1.20130 981.98220 48.64280 2.42510E-03
ITLBO [41] 1.03050 3.48270E-06 1.20130 982.40380 48.64330 2.42510E-03
DPDE [27] 1.03051 3.48220E-06 1.20271 981.98227 48.64283 2.42507E-03
SEDE [29] 1.03051 3.48226E-06 1.20127 981.98246 48.64283 2.42507E-03
MLSBA [42] 1.03051 3.48226E-06 1.20127 981.98222 48.64283 2.42507E-03
STP6-120/36 DONR 7.4759 1.9924E-06 0.1683 869.0440 1.2544 1.42575E-02
FPA [30] 7.4724 2.3491E-06 0.1652 814.0852 1.2606 1.66007E-02
HBA [31] 7.4603 2.6278E-06 0.1638 1500 1.2699 1.67843E-02
IJAYA [44] 7.4672 2.2536E-06 0.1654 771.8252 1.2753 1.66013E-02
HDE [26] 7.4725 2.3349E-06 0.1654 799.9160 1.2601 1.66006E-02
MIGTO [43] 7.4718 3.7256E-06 0.1573 297.3660 1.3004 1.66006E-02
DMTLBO [40] 7.4725 2.3350E-06 0.1656 799.9308 1.2601 1.66010E-02
ITLBO [41] 7.4725 2.3350E-06 0.1656 799.9164 1.2601 1.66010E-02
DPDE [27] 7.4725 2.3349E-06 0.1654 799.9166 1.2601 1.66006E-02
SEDE [29] 7.4725 2.3349E-06 0.1654 799.9160 1.2756 1.66006E-02
MLSBA [42] 7.4715 2.4673E-06 0.1643 902.6820 1.2803 1.66006–02

Before proceeding to the optimisation process, the DO algorithm (continued )

requires initialising the dandelion seeds and calculating their fitness Algorithm: Dandelion Optimizer (DO)
function. For our problem, the fitness function is the RMSE, and the elite 13: Update dandelion seeds using Eq. (24)
individual is chosen as the dandelion seed with the optimum fitness 14: Arrange dandelion seeds from good to bad according to fitness values
value. The pseudocode for this algorithm is detailed below. 15: Update Xelite
Algorithm: Dandelion Optimizer (DO) 16: if
f(Xelite )〈f(Xbest )
Input: P, tmax and D 17: Xbest = Xelite ,f(Xbest ) = f(Xelite )
Output: the optimal dandelion seed Xbest and its fitness value fbest 18: end if
1: Initialization 19: end While
2: Calculate the fitness value of each dandelion seeds 20: Return Xbest and f(Xbest )
3: Select Xelite according to fitness values
4: While (t < tmax ) do
Rising Phase
5: if randn() < 1.5 do
6: Generate adaptive parameters using Eq. (18)
7: Update dandelion seeds using Eq. (15)
8: Else if do As mentioned in the previous section, the DO algorithm optimizes an
9: Generate adaptive parameters using Eq. (21)
objective function to determine the unknown parameters. In this case,
10: Update dandelion seeds using Eq. (20)
11: end if the objective function is represented by the RMSE, which is a nonlinear
Descending phase equation that can be solved using the numerical algorithm known as the
12: Update dandelion seeds using Eq. (22) Newton-Raphson (NR) method. Therefore, the DONR method combines
landing phase the DO algorithm and the NR method, creating a hybrid approach to
(continued on next column) solving this problem.

7
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Table 4
RMSE statistical results of different algorithms.
RMSE(A)

Best Worst Mean

RTC DONR 7.73939317E- 9.03120134E- 8.30121021E-

France 04 04 04
FPA 9.86021877E- 9.86021877E- 9.86021877E-
04 04 04
IJAYA 9.86021919E- 9.86038050E- 9.86025774E-
04 04 04
DPDE 9.86021877E- 9.86021877E- 9.86021877E-
04 04 04
MLSBA 9.86021877E- 9.86021877E- 9.86021877E-
04 04 04

PWP201 DONR 2.05363160E- 2.09513690E- 2.07959365E-

03 03 03
FPA 2.42520992E- 2.60813358E- 2.51987147E-
03 03 03
IJAYA 2.42507503E- 2.42519518E- 2.42511032E-
Fig. 5. The convergence of DO.NR and other algorithms for RTC France (SDM). 03 03 03
DPDE 2.42507486E- 2.42507486E- 2.42507486E-
03 03 03
MLSBA 2.42507486E- 2.43644827E- 2.42694856E-
03 03 03

STP6- DONR 1.42575394E- 1.44045830E- 1.43313450E-

120/36 02 02 02
FPA 1.66007669E- 2.27256207E- 5.95090817E-
02 01 02
IJAYA 1.66013435E- 1.66071013E- 1.66035969E-
02 02 02
DPDE 1.66006031E- 1.66006031E- 1.66006031E-
02 02 02
MLSBA 1.66006031E- 1.67739108E- 1.66236638E-
02 02 02

4. Results and discussion

In this section, we will demonstrate the accuracy of the proposed

method DONR for identifying the parameters of the single-diode model
SDM and double-diode model DDM. The first demonstration step is
comparing the proposed method and other metaheuristics algorithms.
Fig. 6. The convergence of DO.NR and other algorithms for PWP201 (SDM). Such as. Flower pollination Algorithm (FPA) [30]. Honey Badger Al­
gorithm (HBA) [31], IJAYA [44], HDE[26], MIGTO [43], DMTLBO [40],
ITLBO [41], DPDE[27], SEDE [29] and MLSBA [42] using two PV
modules and one cell are Photowatt-PWP201 which uses 36 poly­
crystalline silicon cells in series; its data has extracted under 1000W/m2
and 45 C. Another polycrystalline module is STP6-120/36 has 36 con­
◦

nected cells in series; it collects 24 pairs of current–voltage data under

1000W/m2 at 55 C and RTC France solar cell under 1000W/m2 at 33 C;
◦ ◦

their experimental data are obtained from other research works such as
[40]. The comparisons are executed on the best results represented by
the RMSE values for SDM and DDM. The second demonstration step is
applying the proposed method on three different PV modules under
STCs and real conditions. The three modules are S75 module contains 36
polycrystalline silicon solar cells. SM55 module contains 36 mono­
crystalline silicon solar cells, and ST36 comprises 42 monolithic struc­
tures of series connected Copper Indium Diselenide (CIS) based solar
cells. Their experimental data come from the technical data sheet of the
manufacturer. Table 2 presents each parameter’s search range by giving
the lower and upper bounds for each PV module and cell. For the sake of
fair comparison, the upper and lower bounds of RTC France, PWP201
and STP6-120/36 are extracted from other research [44 40 42]. It should
Fig. 7. The convergence of DO.NR and other algorithms for STP6-120/ be noted that the dimension of the variable D is the number of param­
36 (SDM). eters, D = 5 for SDM and D = 7 for DDM. The population size is P = 50,
and the maximum number of iterations tmax = 50000.
For the single diode model, the comparison results for the estimated
parameters and RMSE are presented in Table 3 for RTC France, PWP201

8
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Table 5
The absolute individual error of current and power for RTC France (SDM).
Measured data Simulated current data Simulated power data

V(V) I(A) P(W) Isim (A) AIE I(A) Psim (A) AIE P(A) RE

1 − 0.2057 0.764 − 0.1572 0.76392 7.9018E-05 − 0.1571 1.6254E-05 1.0342E-04

2 − 0.1291 0.762 − 0.0984 0.76255 5.4810E-04 − 0.0984 7.0760E-05 7.1929E-04
3 − 0.0588 0.7605 − 0.0447 0.76129 7.8803E-04 − 0.0448 4.6336E-05 1.0361E-03
4 0.0057 0.7605 0.0043 0.76013 3.6862E-04 0.0043 2.1011E-06 4.8470E-04
5 0.0646 0.76 0.0491 0.75907 9.2705E-04 0.0490 5.9887E-05 1.2198E-03
6 0.1185 0.759 0.0899 0.75810 9.0367E-04 0.0898 1.0709E-04 1.1906E-03
7 0.1678 0.757 0.1270 0.75718 1.7712E-04 0.1271 2.9721E-05 2.3397E-04
8 0.2132 0.757 0.1614 0.75626 7.4499E-04 0.1612 1.5883E-04 9.8412E-04
9 0.2545 0.7555 0.1923 0.75522 2.7960E-04 0.1922 7.1159E-05 37009E-04
10 0.2924 0.754 0.2205 0.75381 1.9316E-04 0.2204 5.6481E-05 2.5618E-04
11 0.3269 0.7505 0.2453 0.75152 1.0236E-03 0.2457 3.3461E-04 1.3638E-03
12 0.3585 0.7465 0.2676 0.74746 9.5525E-04 0.2680 3.4246E-04 1.2796E-03
13 0.3873 0.7385 0.2860 0.74015 1.6504E-03 0.2867 6.3921E-04 2.2348E-03
14 0.4137 0.728 0.3012 0.72738 6.2449E-04 0.3009 2.5835E-04 8.5781E-04
15 0.4373 0.7065 0.3090 0.70685 3.5207E-04 0.3091 1.5396E-04 4.9832E-04
16 0.459 0.6755 0.3101 0.67513 3.7005E-04 0.3099 1.6985E-04 5.4781E-04
17 0.4784 0.632 0.3023 0.63070 1.3029E-03 0.3017 6.2330E-04 2.0615E-03
18 0.496 0.573 0.2842 0.57192 1.0757E-03 0.2837 5.3353E-04 1.8772E-04
19 0.5119 0.499 0.2554 0.49941 4.0663E-04 0.2556 2.0815E-04 8.1488E-04
20 0.5265 0.413 0.2174 0.41350 5.0180E-04 0.2177 2.6420E-04 1.2150 E-03
21 0.5398 0.3165 0.1708 0.31731 8.1124E-04 0.1713 4.3791E-04 2.5631E-04
22 0.5521 0.212 0.1170 0.21224 2.4384E-04 0.1172 1.3462E-04 1.1501E-03
23 0.5633 0.1035 0.0583 0.10286 6.4081E-04 0.0579 3.6097E-04 6.1914E-03
24 0.5736 − 0.01 − 0.0057 − 0.00917 8.2857E-04 − 0.0053 4.7527E-04 − 8.2857E-02
25 0.5833 − 0.123 − 0.0717 − 0.12442 1.4242E-03 − 0.0726 8.3071E-04 − 1.1578E-02
26 0.59 − 0.21 − 0.1239 − 0.20936 6.4021E-04 − 0.1235 3.7772E-04 − 3.0485E-03

∑
AIE(A) 1.7861E-02 6.7634E-03
MAE(A) 6.8696E-04
MAPE(%) 4.4016E-06

Table 6
The absolute individual error of current and power for PWP201 (SDM).
Measured data Simulated current data Simulated power data

V(V) I(A) P(W) Isim (A) AIE I(A) Psim (A) AIE P(A) RE

1 0.1248 1.0315 0.1287 1.0291 4.0931E-03 6.1875 2.4779E-02 2.3519E-03

2 1.8093 1.03 1.8636 1.0273 4.2356E-03 7.3792 3.0650E-02 2.6351E-03
3 3.3511 1.026 3.4382 1.0256 2.2127E-03 8.4538 1.8407E-02 3.8932E-04
4 4.7622 1.022 4.8670 1.0239 4.4108E-04 9.4069 4.1063E-03 1.8883E-03
5 6.0538 1.018 6.1628 1.0221 2.7387E-03 10.2241 2.7979E-02 4.0207E-03
6 7.2364 1.0155 7.3486 1.0197 3.1009E-03 10.8781 3.4249E-02 4.1709E-03
7 8.3189 1.014 8.4354 1.0162 2.8980E-03 11.3309 3.4202E-02 2.1821E-03
8 9.3097 1.01 9.4028 1.0104 1.9410E-03 11.5379 2.4249E-02 4.3671E-04
9 10.2163 1.0035 10.2521 1.0008 6.0704E-04 11.4579 7.9663E-03 2.7291E-03
10 11.0449 0.988 10.9124 0.9849 2.2429E-04 11.0644 3.0724E-03 3.1385E-03
11 11.8018 0.963 11.3651 0.9601 1.6757E-03 10.3562 2.3832E-02 3.0093E-03
12 12.4929 0.9255 11.5622 0.9236 1.9456E-03 9.3554 2.8599E-02 2.0972E-03
13 13.1231 0.8725 11.4499 0.8731 9.5073E-04 8.1038 1.4389E-02 6.9575E-04
14 13.6983 0.8075 11.0614 0.8077 9.0437E-04 6.6536 1.4046E-02 2.7775E-04
15 14.2221 0.7265 10.3324 0.7282 3.2558E-04 5.0567 5.1743E-03 2.3065E-03
16 14.6995 0.6345 9.3268 0.6364 1.1272E-03 3.3642 1.8286E-02 3.0663E-03
17 15.1346 0.5345 8.0894 0.5355 3.0320E-03 1.6188 5.0100E-02 1.7787E-03
18 15.5311 0.4275 6.6395 0.4284 3.8072E-04 − 0.1408 6.3956E-03 2.1154E-03
19 15.8929 0.3185 5.0619 0.3182 5.7865E-05 − 1.8916 9.8659E-04 1.0222E-03
20 16.2229 0.2085 3.3825 0.2074 1.9435E-04 − 3.6080 3.3583E-03 5.4062E-04
21 16.5241 0.101 1.6689 0.0980 1.6090E-03 − 5.2709 2.8139E-02 3.0019E-02
22 16.7987 − 0.008 − 0.1344 − 0.0084 4.0931E-03 6.1875 2.4779E-02 − 4.7590E-02
23 17.0499 − 0.111 − 1.8925 − 0.1109 4.2356E-03 7.3792 3.0650E-02 − 5.2130E-04
24 17.2793 − 0.209 − 3.6114 − 0.2088 2.2127E-03 8.4538 1.8407E-02 − 9.2992E-04
25 17.4885 − 0.303 − 5.2990 − 0.3014 4.4108E-04 9.4069 4.1063E-03 − 5.3101E-03

∑
AIE(A) 4.2165E-02 4.1871E-01
MAE(A) 1.6866E-03
MAPE(%) 2.5419e-05

and STP6-120/36. Note that the best RMSE is given by the proposed among all the chosen algorithms are in the order of 7.73939E-04A for
method comparing with other algorithms for all the modules. As shown RTC France, 2.08116E-03 A for PWP201 and 1.42575E-02A for STP6-
in tables, the proposed method gives the minimum values of RMSE 120/36. The RMSE values of other algorithms are very close to each

9
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Table 7
The absolute individual error of current and power for STP6-120/36 (SDM).
Measured data Simulated current data Simulated power data

V(V) I(A) P(W) Isim (A) AIE I(A) Psim (A) AIE P(A) RE

1 19.21 0.00 0.0000 0.0031 3.0976E-03 0.0595 5.9505E-02 ————

2 17.65 3.83 67.5995 3.8302 2.2615E-04 67.6035 3.9915E-03 5.9046E-05
3 17.41 4.29 74.6889 4.2722 1.7850E-02 74.3781 3.1076E-01 4.1607E-03
4 17.25 4.56 78.6600 4.5448 1.5238E-02 78.3971 2.6286E-01 3.3417E-03
5 17.10 4.79 81.9090 4.7846 5.4264E-03 81.8162 9.2792E-02 1.1328E-03
6 16.90 5.07 85.6830 5.0811 1.1063E-02 85.8700 1.8697E-01 2.1821E-03
7 16.76 5.27 88.3252 5.2732 3.1766E-03 88.3784 5.3240E-02 6.0277E-04
8 16.34 5.75 93.9550 5.7770 2.7043E-02 94.3969 4.4188E-01 4.7031E-03
9 16.08 6.00 96.4800 6.0382 3.8160E-02 97.0936 6.1361E-01 6.3600E-03
10 15.71 6.36 99.9156 6.3499 1.0120E-02 99.7566 1.5898E-01 1.5911E-03
11 15.39 6.58 101.2662 6.5694 1.0645E-02 101.1024 1.6382E-01 1.6177E-03
12 14.93 6.83 101.9719 6.8165 1.3547E-02 101.7696 2.0226E-01 1.9834E-03
13 14.58 6.97 101.6226 6.9600 9.9796E-03 101.4771 1.4550E-01 1.4317E-03
14 14.17 7.10 100.6070 7.0896 1.0433E-02 100.4592 1.4784E-01 1.4694E-03
15 13.59 7.23 98.2557 7.2189 1.1136E-02 98.1044 1.5134E-01 1.5402E-03
16 13.16 7.29 95.9364 7.2850 5.0469E-03 95.8700 6.6417E-02 6.9229E-04
17 12.74 7.34 93.5116 7.3320 7.9676E-03 93.4101 1.0151E-01 1.0854E-03
18 12.36 7.37 91.0932 7.3636 6.4201E-03 91.0138 7.9353E-02 8.7111E-03
19 11.81 7.38 87.1578 7.3959 1.5884E-02 87.3454 1.8760E-01 2.1523E-03
20 11.17 7.41 82.7697 7.4200 9.9812E-03 82.8812 1.1149E-01 1.3469E-03
21 10.32 7.44 76.7808 7.4385 1.4946E-03 76.7654 1.5425E-02 2.0089E-04
22 9.740 7.42 72.2708 7.4460 2.5968E-02 72.5237 2.5293E-01 3.4997E-03
23 9.060 7.45 67.4970 7.4516 1.6384E-03 67.5118 1.4844E-02 2.1992E-04
24 0.00 7.48 0.0000 7.4690 1.0980E-02 0.0000 0.0000Eþ00 1.4678E-03

∑
AIE(A) 2.7252E-01 3.8249E+00
MAE(A) 1.1355E-02
MAPE(%) 2,0258E-05

Fig. 8. Comparisons between experimental and simulated data for RTC France (SDM) (a) I-V characteristics; (b) P-V characteristics.

other. To enhance the prominence of the proposed method’s superiority, ence between simulated and experimental data for current (AIE I) and
we have included Figs. 5, 6, and 7, which depict the RMSE average power (AIE P), respectively. Additionally, alternative metrics, including
evolution of the proposed method compared with other algorithms. It is the relative error (RE), the mean absolute error (MAE), and the mean
evident from the convergence curves that the proposed algorithm ex­ absolute percentage error (MAPE) are utilized to facilitate a compre­
hibits a remarkable convergence performance in contrast to other al­ hensive evaluation of the simulated and experimental current data. The
gorithms such as DPDE, IJAYA, and MLBSA. results presented in Tables 5, 6, and 7 indicate a statistically insignifi­
Additionally, Table 4 summarizes the Root Mean Square Error cant deviation between the experimental and simulated currents. Spe­
(RMSE) statistical results, including the minimum (Best), the maximum cifically, for the RTC France cell, the absolute individual error values fall
(Worst), and the mean. These results are obtained from 10 independent within the range of 7.9018E-05A to 1.6504E-03A, with a corresponding
executions of the proposed and other methods. Compared to other MAPE of 4.4016E-06%. These results provide strong evidence that the
methods, the proposed method exhibits lower values for the best, worst, proposed method is capable of producing simulated data that exhibits
and mean RMSE across all three tested modules, as shown in the table. high precision and accuracy, which yields a more optimal fit to the
The estimated parameters by the proposed method are laid into our experimental data. This is corroborated by the I-V and P-V characteristic
objective function to calculate the simulated current and power data, curves presented in Figs. 8, 9, and 10, which demonstrate a strong
which is compared with the experimental data using the absolute indi­ correlation between the simulated and experimental data.
vidual error. The AIE metric is calculated as the absolute of the differ­

10
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Simulated data
1 10 Experimental data
0.8

Current (A)

Power (W)
0.6
5
Simulated data
0.4
Experimental data
0.2
0
0
-0.2
-5
-0.4
0 5 10 15 0 5 10 15
Voltage (V) Voltage (V)

Fig. 9. Comparisons between experimental and simulated data for PWP201 (SDM) (a) I-V characteristics; (b) P-V characteristics.

8
7 100 Simulated data
Experimental data
6 80
5
Current (A)

Power (W)
4 60
Simulated data
3 Experimental data
40
2
1 20

0 0
-1
0 5 10 15 20 0 5 10 15 20
Voltage (V) Voltage (V)

Fig. 10. Comparisons between experimental and simulated data for STP6-120/36 (SDM) (a) I-V characteristics; (b) P-V characteristics.

Table 8
Estimated parameters and RMSE values for DDM.
Ipv (A) I0 1 (A) Rs (Ω) Rsh (Ω) a1 I0 2 (A) a2 RMSE(A)

RTC France DONR 0.76077 2.1535E-07 3.6847E-02 54.9301 1.4473 5.7817E-07 1.9242 7.56515E-04
FPA 0.76073 1.0000E-06 3.6710E-02 58.2248 1.9917 2.0632E-07 1.4442 9.90582E-04
HBA 0.76078 2.2345E-07 3.6751E-02 55.5443 1.4500 7.7121E-07 1.9999 9.82488E-04
IJAYA 0.76010 5.0445E-09 3.7600E-02 77.8519 1.2185 7.5094E-07 1.6247 9.82930E-03
HDE 0.76078 7.4934E-07 3.6740E-02 55.4854 2.000 2.2597E-07 1.451 9.82484E-04
MIGTO 0.76078 2.2597E-07 3.6740E-02 55.4854 1.4510 7.4934E-07 1.9999 9.82484E-04
DPDE 0.76078 2.2596E-07 3.6740E-02 55.4869 1.4510 7.4949E-07 2.0000 9.82484E-04
SEDE 0.76077 2.7410E-07 3.6510E-02 54.2881 1.9999 2.8473E-07 1.4703 9.83996E-04
MLSBA 0.76080 2.8411E-07 3.6486E-02 54.2657 1.4703 2.7277E-07 1.9824 9.84184E-04

PWP201 DONR 1.0305 5.4424E-09 1.22000 949.3939 42.8319 3.0633E-06 48.1854 2.07842E-03
FPA 1.0292 9.6636E-07 1.20165 1161.7881 50 2.6346E-06 48.4082 2.45263E-3
HBA 1.0305 3.1176E-07 1.20164 981.0549 48.6388 3.1590E-06 48.6293 2.42510E-03
IJAYA 1.0306 1.7940E-08 1.20676 925.9600 47.5070 3.2690E-06 48.4311 2.44562E-03
PDPE 1.0305 1.7389E-06 1.20127 981.9822 48.6428 1.7433E-06 48.6428 2.425074E-03
SEDE 1.0305 1.8570E-06 1.20134 980.8042 48.6436 1.6223E-06 48.6351 2.425076E-03
MLSBA 1.0304 7.6463E-08 1.20118 987.2186 48.5828 3.4107E-06 48.6496 24.25099E-03

STP6-120/36 DONR 7.4707 1.6347E-05 0.1697 1029.9171 1.9999 1.5139E-06 1.2269 1.45952E-02
IJAYA 7.4759 2.4601E-06 0.1635 837.2170 1.2650 3.6638E-06 1.9990 1.69522E-02
PDPE 7.4725 2.1127E-18 0.1654 799.9163 1.8827 2.3349E-06 1.2601 1.66006E-02
SEDE 7.4765 2.4125E-06 0.1688 582.4523 1.2754 1.9140E-08 1.0607 1.68465E-02
MLSBA 7.4768 2.2461E-06 0.1654 633.7942 1.2584 6.8850E-08 1.3008 1.66258E-02

AIE I = |Isim − I| (25)

|Isim − I|
RE = (27)
AIE P = |Psim − P| (26) I

11
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Table 9
The individual absolute error of current and power for RTC France, PWP201 and STP6-120/36 (DDM).
AIE I(A) AIE P(A)
∑ ∑
min max mean AIE I min max mean AIE P

RTC France 5.8712E-06 1.5130E-03 6.6192E-04 1.7210E-02 1.9477E-06 8.0006E-04 2.4845E-04 6.4599E-03
PWP201 5.6311E-05 4.2061E-03 1.6937E-03 4.2344E-02 3.0410E-04 5.0770E-02 1.6891E-02 4.2229E-01
STP6-120/36 1.2634E-03 3.9293E-02 1.1611E-02 2.7867E-01 1.4529E-02 3.2773E-01 1.6367E-1 3.9283E+00

0.4
0.8 Simulated data
0.3 Experimental data
0.6
0.2
Current (A)

Power (W)
0.4
Simulated data 0.1
Experimental data
0.2
0
0
-0.1
-0.2
-0.2
-0.2 0 0.2 0.4 0.6 -0.2 0 0.2 0.4 0.6
Voltage (V) Voltage (V)

Fig. 11. Comparisons between experimental and simulated data for RTC France (DDM) (a) I-V characteristics; (b) P-V characteristics.

Simulated data
1 Experimental data
10
0.8
Current (A)

Power (W)

0.6
5
Simulated data
0.4
Experimental data
0.2
0
0
-0.2
-5
-0.4
0 5 10 15 0 5 10 15
Voltage (V) Voltage (V)

Fig. 12. Comparisons between experimental and simulated data for PWP201 (DDM) (a) I-V characteristics; (b) P-V characteristics.

8
100 Simulated data
Experimental data
6 80
Current (A)

Power (W)

60
4 Simulated data
Experimental data
40
2
20

0 0

0 5 10 15 20 0 5 10 15 20
Voltage (V) Voltage (V)

Fig. 13. Comparisons between experimental and simulated data for STP6-120/36 (DDM) (a) I-V characteristics; (b) P-V characteristics.

12
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Table 10
Estimated parameters and RMSE values.
Module G(W/m2) T(◦ C) Ipv (A) I0 (A) Rs (Ω) Rsh (Ω) a RMSE(A)

S75 1000 25 4.69604 3.46270E-09 0.29610 265.80261 1.11266 2.10321E-02

800 3.76768 3.42222E-10 0.37297 229.24998 1 1.96894E-02
600 2.82454 5.75053E-10 0.36125 249.17081 1.02111 1.43517E-02
1000 20 4.66524 6.89282E-08 0.24558 9999.99929 1.33459 2.02255E-02
40 4.71009 4.66170E-07 0.26897 4992.59065 1.30141 1.49538E-02
60 4.75909 4.29980E-06 0.28090 4999.99519 1.31604 3.15111E-02

SM55 1000 25 3.44017 3.05679E-09 0.45223 730.17893 1.12596 1.59231E-02

800 2.75574 9.27591E-09 0.46501 446.78118 1.19005 1.03913E-02
600 2.07242 8.21283E-10 0.57769 359.31712 1.05417 7.17021E-03
1000 20 3.43311 8.42543E-09 0.43270 8511.07179 1.22617 1.59449E-02
40 3.45965 1.55314E-07 0.43540 2315.47929 1.25450 1.37874E-02
60 3.49435 3.09004E-06 0.40270 2697.09018 1.32594 1.17385E-02

ST36 1000 25 2.70188 2.90981E-07 1.20780 251.13770 1.57215 1.23625E-02

800 2.15634 9.31874E-06 0.89482 330.91325 2 1.01836E-02
600 1.61495 9.06910E-06 0.97965 486.29810 2 7.59098E-03
1000 20 2.67335 1.26364E-05 0.86481 2218.16783 1.82656 9.38347E-03
40 2.71419 1.62820E-05 1.02491 369.90382 1.87068 9.93862E-03
60 2.75255 4.21228E-05 1.13518 245.92137 1.73779 1.33903E-02

Fig. 14. I-V characteristics for S75. (a) For different irradiation and T = 25◦ . (b) For different temperature and G = 1000 W/m2.

Fig. 15. I-V characteristics for SM55. (a) For different irradiation and T = 25◦ . (b) For different temperature and G = 1000 W/m2.

1∑ n To further compare, the double diode model was utilized. Table 8 dis­
MAE = |Isimi − Ii | (28) plays various methods’ extracted parameters and RMSE values, with the
n i=1
proposed method yielding the most favourable results. It is worth noting
1∑ n ⃒ ⃒
⃒Isimi − Ii ⃒ that even with the DDM, the proposed method still demonstrated the
MAPE = ⃒
⃒
⃒ × 100 (29) best RMSE, with values of approximately 7.56515E-04A, 2.07842E-03A,
n i=1 Ii ⃒
and 1.45952E-02A for RTC France, PWP201, and STP6-120/36,
respectively, thereby confirming the previous results and the

13
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

Fig. 16. I-V characteristics for ST36. (a) For different irradiation and T = 25◦ . (b) For different temperature and G = 1000 W/m2.

Table 11
Comparison of RMSE values under different levels of irradiation and temperature for S75, SM55 and ST36.
T(C ) 25 20 40 60
◦

2
G(W/m ) 1000 800 600 1000

RMSE(A)

S75 DONR 2.1032E-02 1.9689E-02 1.4351E-02 2.0225E-02 1.4953E-02 3.1511E-02

IJAYA 5.4881E-02 5.5413E-02 3.0920E-02 2.9658E-02 2.4174E-02 4.3337E-02
PDPE 3.7357E-02 3.2690E-02 2.3396E-02 2.9570E-02 2.1694E-02 4.3327E-02
SEDE 3.7826E-02 4.0490E-02 2.6927E-02 2.9570E-02 2.1694E-02 4.3327E-02
MLSBA 3.9850E-02 4.7661E-02 2.8946E-02 2.9570E-02 2.1836E-02 4.3327E-02

SM55 DONR 1.5923E-02 1.0391E-02 7.1702E-03 1.5944E-02 1.3787E-02 1.1738E-02

IJAYA 3.7207E-02 2.4772E-02 1.9883E-02 2.5641E-02 2.4231E-02 1.6242E-02
PDPE 2.6748E-02 1.6052E-02 9.9458E-03 2.4793E-02 1.8254E-02 1.5569E-02
SEDE 2.6748E-02 1.6052E-02 9.9589E-03 2.4793E-02 1.8342E-02 1.5570E-02
MLSBA 3.4983E-02 2.0215E-02 1.6472E-02 2.5138E-02 1.9823E-02 1.5659E-02

ST36 DONR 1.2362E-02 1.0183E-02 7.5909E-03 9.3834E-03 9.9386E-03 1.3390E-02

IJAYA 2.5837E-02 1.5125E-02 9.3979E-03 1.8037E-02 1.7129E-02 2.1856E-02
PDPE 2.1256E-02 1.3833E-02 8.9608E-03 1.8037E-02 1.5999E-02 2.1061E-02
SEDE 2.3554E-02 1.3833E-02 8.9608E-03 1.8037E-02 1.5999E-02 2.1061E-02
MLSBA 2.5004E-02 1.4240E-02 8.9611E-03 1.8037E-02 1.6017E-02 2.1188E-02

superiority of the proposed method. parameters and RMSE values for the three PV modules at six different
Furthermore, to demonstrate the superior performance of the pro­ levels of irradiation and temperature. The RMSE values are in order of
posed method for the double diode model, Table 9 summarises the 10− 2 A and 10− 3 A. Figures 14, 15 and 16 present a comparison between
statistical results of the absolute errors for both current and power. To the simulated and experimental data for the three PV modules under
validate these results, the average values of the AIE_I for RTC France, different irradiation levels with constant temperature T=25◦ C and
PWP201, and STP6-120/36 were found to be 6.6192E-04A, 1.6937E- under different temperature values with constant irradiation
03A, and 1.1611E-02A, respectively. These values indicate that the G=1000W/m2. The figures show that the simulated current agrees well
difference between the simulated and experimental current is very with the experimental current.
small, evident from the I-V and P-V characteristics presented in Figs. 11, To further evaluate the superiority of the proposed method DONR,
12, and 13. The superior performance of the proposed method and the Table 11 summarizes the values of RMSE given by DONR compared with
achieved results can be attributed to the effective combination of the other methods. The three modules have been presented in the previous
Newton-Raphson method and the Dandelion Optimizer algorithm. By paragraph; the comparison is carried out under different levels of irra­
employing the Newton-Raphson method to solve the nonlinear equation diation and temperature. The table clearly indicates that DONR out­
of the objective function, we were able to significantly reduce optimi­ performs the other methods in all cases, providing better RMSE values.
zation errors and identify the most optimal parameters. These results To ensure fairness, the comparison was carried out using the same
highlight the effectiveness of the proposed approach and its potential for experimental data, maximum iteration numbers, and population sizes.
enhancing the accuracy and efficiency of optimization techniques, The key takeaways from this section can be summarized as follows:
especially in complex systems with nonlinear objective functions.
In this subsection, we will test the accuracy of the proposed method • The proposed method exhibited clear superiority in determining the
under various levels of irradiance (G) and temperature (T) using three parameters of various modules and cells, utilizing both SDM and
PV modules from different technologies. They are the S75 module DDM, under varying levels of temperature and irradiation;
(polycrystalline) [56], the SM55 module (monocrystalline) [57], and • A novel contribution to the literature by using a numerical method
the ST36 module (thin film) [58]. Table 10 presents the estimated for calculating the objective function. This methodology enhances

14
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

the accuracy of the proposed method and provides a more precise References
determination of the parameters;
• The results indicate that the proposed method outperforms all other [1] Black BC. Burning up: A Global history of fossil fuel consumption by Simon Pirani.
Technol Cult 2020;61(2):700–2. https://doi.org/10.1353/tech.2020.0042.
compared methods. Moreover, they imply that the proposed algo­ [2] bp energy charting tool | Energy economics | Home. [Online]. Available: https
rithm holds promise for substantial advancements in diverse ://www.bp.com/en/global/corporate/energy-economics/statistical-review-of-wo
domains. rld-energy/energy-charting-tool-desktop.html.html#/results/et/carb-emis/unit/M
TCO2/regions/tWORLD/view/area.
[3] Renewable electricity – Renewables 2022 – Analysis - IEA. [Online]. Available: htt
5. Conclusion and future work ps://www.iea.org/reports/renewables-2022/renewable-electricity.
[4] Anurag A, Bal S, Sourav S, Nanda M. A review of maximum power-point tracking
techniques for photovoltaic systems. Int J Sustain Energy 2016;35(5):478–501.
This study introduces a novel method called DONR for identifying https://doi.org/10.1080/14786451.2014.918979.
the PV parameters of various models, such as SDM and DDM. DONR [5] Elyaqouti M, Boulfaf N, Hamid N, Izbaim D, Chaoufi J, Bouhouch L. Thermal and
utilizes Dandelion Optimizer (DO) algorithm, in conjunction with a electrical modelling of photovoltaic modules. Int J Ambient Energy 2022;43(1):
5793–809. https://doi.org/10.1080/01430750.2021.1994464.
numerical approach known as the Newton-Raphson (NR) method, to
[6] Piliougine M, Guejia-Burbano RA, Petrone G, Sánchez-Pacheco FJ, Mora-López L,
estimate unknown parameters and optimize the objective function, a Sidrach-de-Cardona M. Parameters extraction of single diode model for degraded
non-linear equation solved using NR. The Dandelion Optimizer is an photovoltaic modules. Renew Energy 2021;164:674–86. https://doi.org/10.1016/
innovative metaheuristic algorithm that operates in three stages: rising, j.renene.2020.09.035.
[7] Ganesh Pardhu BSS, Kota VR, Radial movement optimization based parameter
descending, and landing. It employs both local and global search stra­ extraction of double diode model of solar photovoltaic cell, Sol Energy, 2021;213
tegies to optimize the objective function. The proposed method’s effi­ (December 2020):312–327, doi: 10.1016/j.solener.2020.11.046.
cacy is validated using six modules and cells, including RTC France solar [8] Yu K, Qu B, Yue C, Ge S, Chen X, Liang J. A performance-guided JAYA algorithm
for parameters identification of photovoltaic cell and module. Appl Energy 2019;
cell, Photowatt-PWP201, STP6-120/36, S75, SM55, and ST36. DONR’s 237(January):241–57. https://doi.org/10.1016/j.apenergy.2019.01.008.
performance is evaluated using the root mean squared error (RMSE) and [9] Khan F, Al-Ahmed A, Al-Sulaiman FA. Critical analysis of the limitations and
the absolute individual error and is compared to other methods and validity of the assumptions with the analytical methods commonly used to
determine the photovoltaic cell parameters. Renew Sustain Energy Rev 2021;140
algorithms. As a result, the following results of RMSE are obtained by (February):110753. https://doi.org/10.1016/j.rser.2021.110753.
using the proposed method DNOR (7.73939E-04, 7.56515E-04), [10] Lo Brano V, Ciulla G. An efficient analytical approach for obtaining a five
(2.08116E-03, 2.07842E-03), and (1.42575E-02, 1.45952E-02) using parameters model of photovoltaic modules using only reference data. Appl Energy
2013;111:894–903. https://doi.org/10.1016/j.apenergy.2013.06.046.
(SDM, DDM) for respectively RTC France, PWP201, and STP6-120/36. [11] El Achouby H, Zaimi M, Ibral A, Assaid EM. New analytical approach for modelling
The results demonstrate that DONR outperforms all compared effects of temperature and irradiance on physical parameters of photovoltaic solar
methods in various conditions. module. Energy Convers Manag 2018;177(September):258–71. https://doi.org/
10.1016/j.enconman.2018.09.054.
This study suggests potential for future research, such as extending
[12] Villalva MG, Gazoli JR, Filho ER. Comprehensive approach to modeling and
the applicability of DONR to maximum power point tracking, and away simulation of photovoltaic arrays. IEEE Trans Power Electron 2009;24(5):
from photovoltaic energy, we will grow it to be applicable for solving the 1198–208. https://doi.org/10.1109/TPEL.2009.2013862.
optimal power flow problem [59]. [13] Elhammoudy A, et al., A novel numerical method for estimation the photovoltaic
cells/modules parameters based on dichotomy method, Results Opt, 2023;12
(February):100445, doi: 10.1016/j.rio.2023.100445.
CRediT authorship contribution statement [14] Choulli I, et al., Energy conversion and management a novel hybrid analytical/
iterative method that exacts the single-diode model’s parameters using Lambert’s
W-function, Energy Convers Manag X, 2023;18(January): 100362, doi: 10.1016/j.
Abdelfattah Elhammoudy: Software, Validation, Formal analysis, ecmx.2023.100362.
Conceptualization, Methodology, Data curation, Visualization, Investi­ [15] Lidaighbi S, Elyaqouti M, Ben Hmamou D, Saadaoui D, Assalaou K, Arjdal E. A new
gation, Writing – original draft, Writing – review & editing. Mustapha hybrid method to estimate the single-diode model parameters of solar photovoltaic
panel. Energy Convers Manage: X 2022;15:100234.
Elyaqouti: Software, Validation, Formal analysis, Conceptualization, [16] Ben hmamou D, Elyaqouti M, Arjdal EH, Saadaoui D, Lidaighbi S, Chaoufi J, et al.
Methodology, Data curation, Visualization, Investigation, Writing – A novel hybrid numerical with analytical approach for parameter extraction of
original draft, Writing – review & editing. El Hanafi Arjdal: Concep­ photovoltaic modules. Energy Convers Manage: X 2022;14:100219.
[17] Jian X, Weng Z, A logistic chaotic JAYA algorithm for parameters identification of
tualization, Methodology, Data curation, Writing – original draft, photovoltaic cell and module models, Optik (Stuttg), 2020;203(December 2019):
Writing – review & editing. Dris Ben Hmamou: Software, Validation, 164041, doi: 10.1016/j.ijleo.2019.164041.
Formal analysis, Conceptualization, Methodology, Data curation, Visu­ [18] Altan A, Karasu S, Zio E, A new hybrid model for wind speed forecasting combining
long short-term memory neural network, decomposition methods and grey wolf
alization, Investigation, Writing – original draft, Writing – review &
optimizer, Appl Soft Comput, 2021; 100(December 2019):106996, doi: 10.1016/j.
editing. Souad Lidaighbi: Conceptualization, Methodology, Data asoc.2020.106996.
curation, Writing – original draft, Writing – review & editing. Driss [19] Karasu S, Altan A, Bekiros S, Ahmad W. A new forecasting model with wrapper-
Saadaoui: Conceptualization, Methodology, Data curation, Writing – based feature selection approach using multi-objective optimization technique for
chaotic crude oil time series. Energy 2020;212:118750. https://doi.org/10.1016/j.
original draft, Writing – review & editing. Imade Choulli: Conceptu­ energy.2020.118750.
alization, Methodology, Data curation, Writing – original draft, Writing [20] Yağ İ, Altan A, Artificial intelligence-based robust hybrid algorithm design and
– review & editing. Ismail Abazine: Conceptualization, Methodology, implementation for real-time detection of plant diseases in agricultural
environments, Biology (Basel), 2022;11(12), doi: 10.3390/biology11121732.
Data curation, Writing – original draft, Writing – review & editing. [21] Saadaoui D, Elyaqouti M, Assalaou K, Ben hmamou D, Lidaighbi S, Parameters
optimization of solar PV cell/module using genetic algorithm based on non-
Declaration of Competing Interest uniform mutation, Energy Convers Manag X, 2021;12(July):100129, doi: 10.1016/j.
ecmx.2021.100129.
[22] Bendaoud R, Amiry H, Benhmida M, Zohal B, Yadir S, Bounouar S, et al. New
The authors declare that they have no known competing financial method for extracting physical parameters of PV generators combining an
interests or personal relationships that could have appeared to influence implemented genetic algorithm and the simulated annealing algorithm. Sol Energy
2019;194:239–47.
the work reported in this paper. [23] Bastidas-Rodriguez JD, Petrone G, Ramos-Paja CA, Spagnuolo G. A genetic
algorithm for identifying the single diode model parameters of a photovoltaic
Data availability panel. Math Comput Simul 2017;131:38–54. https://doi.org/10.1016/j.
matcom.2015.10.008.
[24] Barukčić M, Hederić Ž, Špoljarić Ž. The estimation of I-V curves of PV panel using
Data will be made available on request. manufacturers’ I-V curves and evolutionary strategy. Energy Convers Manag 2014;
88:447–58. https://doi.org/10.1016/j.enconman.2014.08.052.
[25] Bilal, Pant M, Zaheer H, Garcia-Hernandez L, Abraham A, Differential Evolution: A
review of more than two decades of research, Eng Appl Artif Intell, 2020;90
(February):103479, doi: 10.1016/j.engappai.2020.103479.

15
A. Elhammoudy et al. Energy Conversion and Management: X 19 (2023) 100405

[26] Wang D, Sun X, Kang H, Shen Y, Chen Q. Heterogeneous differential evolution [43] Abdel-Basset M, El-Shahat D, Sallam KM, Munasinghe K. Parameter extraction of
algorithm for parameter estimation of solar photovoltaic models. Energy Rep 2022; photovoltaic models using a memory-based improved gorilla troops optimizer.
8:4724–46. https://doi.org/10.1016/j.egyr.2022.03.144. Energy Convers Manage 2022;252:115134. https://doi.org/10.1016/j.
[27] Gao S, Wang K, Tao S, Jin T, Dai H, Cheng J. A state-of-the-art differential enconman.2021.115134.
evolution algorithm for parameter estimation of solar photovoltaic models. Energy [44] Yu K, Liang JJ, Qu BY, Chen X, Wang H. Parameters identification of photovoltaic
Convers Manag 2021;230(January):113784. https://doi.org/10.1016/j. models using an improved JAYA optimization algorithm. Energy Convers Manag
enconman.2020.113784. 2017;150(July):742–53. https://doi.org/10.1016/j.enconman.2017.08.063.
[28] Dang J, Wang G, Xia C, Jia R, Li P. Research on the parameter identification of PV [45] Mostafa M, Rezk H, Aly M, Ahmed EM. A new strategy based on slime mould
module based on fuzzy adaptive differential evolution algorithm. Energy Rep 2022; algorithm to extract the optimal model parameters of solar PV panel. Sustain
8:12081–91. https://doi.org/10.1016/j.egyr.2022.09.057. Energy Technol Assessments 2020;42(June):100849. https://doi.org/10.1016/j.
[29] Liang J, Qiao K, Yu K, Ge S, Qu B, Xu R, et al. Parameters estimation of solar seta.2020.100849.
photovoltaic models via a self-adaptive ensemble-based differential evolution. Sol [46] Düzenli˙ T, Kutlu Onay F, Aydemi˙r SB. Improved honey badger algorithms for
Energy 2020;207:336–46. https://doi.org/10.1016/j.solener.2020.06.100. parameter extraction in photovoltaic models. Optik 2022;268:169731. https://doi.
[30] Yang X-S, Karamanoglu M, He X. Flower pollination algorithm: A novel approach org/10.1016/j.ijleo.2022.169731.
for multiobjective optimization. Eng Optim 2014;46(9):1222–37. https://doi.org/ [47] El-Dabah MA, El-Sehiemy RA, Hasanien HM, Saad B. Photovoltaic model
10.1080/0305215X.2013.832237. parameters identification using Northern Goshawk Optimization algorithm. Energy
[31] Hashim FA, Houssein EH, Hussain K, Mabrouk MS, Al-Atabany W. Honey Badger 2023;262(PB):125522. https://doi.org/10.1016/j.energy.2022.125522.
Algorithm: New metaheuristic algorithm for solving optimization problems. Math [48] Garip Z. Parameters estimation of three-diode photovoltaic model using fractional-
Comput Simul 2022;192:84–110. https://doi.org/10.1016/j.matcom.2021.08.013. order Harris Hawks optimization algorithm. Optik 2023;272:170391. https://doi.
[32] Askarzadeh A, Rezazadeh A. Artificial bee swarm optimization algorithm for org/10.1016/j.ijleo.2022.170391.
parameters identification of solar cell models. Appl Energy 2013;102:943–9. [49] Abd El-Mageed AA, Abohany AA, Saad HMH, Sallam KM. Parameter extraction of
https://doi.org/10.1016/j.apenergy.2012.09.052. solar photovoltaic models using queuing search optimization and differential
[33] Wang D, Tan D, Liu L. Particle swarm optimization algorithm: an overview. Soft evolution. Appl Soft Comput 2023;134:110032. https://doi.org/10.1016/j.
Comput 2017. https://doi.org/10.1007/s00500-016-2474-6. asoc.2023.110032.
[34] Ben Hmamou D, Elyaqouti M, Arjdal E, Chaoufi J, Saadaoui D, Lidaighbi S, et al. [50] Zhao S, Zhang T, Ma S, Chen M. Dandelion Optimizer: A nature-inspired
Particle swarm optimization approach to determine all parameters of the metaheuristic algorithm for engineering applications. Eng Appl Artif Intell 2022;
photovoltaic cell. Mater Today Proc 2022;52:7–12. https://doi.org/10.1016/j. 114(May):105075. https://doi.org/10.1016/j.engappai.2022.105075.
matpr.2021.10.083. [51] Gao X, Cui Y, Hu J, Xu G, Yu Y. Lambert W-function based exact representation for
[35] Fan Y, Wang P, Asghar A, Chen H, Random reselection particle swarm optimization double diode model of solar cells: Comparison on fitness and parameter extraction.
for optimal design of solar photovoltaic modules, Energy, 2022;239:121865, doi: Energy Convers Manage 2016;127:443–60. https://doi.org/10.1016/j.
10.1016/j.energy.2021.121865. enconman.2016.09.005.
[36] Abdellatif H, Hossain MI, Abido MA. Parameters estimation of PV models using [52] Verbeke J, Cools R. The Newton-Raphson method. Int J Math Educ Sci Technol
artificial neural network. Arab J Sci Eng 2022;47(11):14947–56. https://doi.org/ 1995;26(2):177–93. https://doi.org/10.1080/0020739950260202.
10.1007/s13369-022-06901-7. [53] Chen J, Li W. An exponential Regula Falsi method for solving nonlinear equations.
[37] Abbassi R, Abbassi A, Heidari AA, Mirjalili S. An efficient salp swarm-inspired Numer Algorithms 2006;41(4):327–38. https://doi.org/10.1007/s11075-006-
algorithm for parameters identification of photovoltaic cell models. Energy 9015-9.
Convers Manage 2019;179:362–72. https://doi.org/10.1016/j. [54] Charity Ebelechukwu O. Comparison of some iterative methods of solving
enconman.2018.10.069. nonlinear equations. Int J Theor Appl Math 2018;4(2):22. https://doi.org/
[38] Kassaymeh S, Abdullah S, Al-Betar MA, Alweshah M, Al-Laham M, Othman Z. Self- 10.11648/j.ijtam.20180402.11.
adaptive salp swarm algorithm for optimization problems. Soft Comput 2022;26 [55] Einstein A. Investigations on the theory of the Brownian movement. Courier
(18):9349–68. https://doi.org/10.1007/s00500-022-07280-9. Corporation; 1956.
[39] Rao RV, Savsani VJ, Vakharia DP. Teaching-learning-based optimization: A novel [56] Mandal S. Modeling of photovoltaic systems using Modified Elephant Swarm Water
method for constrained mechanical design optimization problems. CAD Comput Search Algorithm. Int J Model Simul 2020;40(6):436–55. https://doi.org/
Aided Des 2011;43(3):303–15. https://doi.org/10.1016/j.cad.2010.12.015. 10.1080/02286203.2019.1650488.
[40] Li L, Xiong G, Yuan X, Zhang J, Chen J. Parameter extraction of photovoltaic [57] Meng Z, Zhao Y, Tang S, Sun Y. An efficient datasheet-based parameters extraction
models using a dynamic self-adaptive and mutual- comparison teaching-learning- method for two-diode photovoltaic cell and cells model. Renew Energy 2020;153:
based optimization. IEEE Access 2021;9:52425–41. https://doi.org/10.1109/ 1174–82. https://doi.org/10.1016/j.renene.2020.02.084.
ACCESS.2021.3069748. [58] Abido MA, Khalid MS. Seven-parameter PV model estimation using Differential
[41] Li S, Gong W, Yan X, Hu C, Bai D, Wang L, et al. Parameter extraction of Evolution. Electr Eng 2018;100(2):971–81. https://doi.org/10.1007/s00202-017-
photovoltaic models using an improved teaching-learning-based optimization. 0542-2.
Energy Convers Manag 2019;186:293–305. https://doi.org/10.1016/j. [59] Maheshwari A, Sood YR, Jaiswal S. Flow direction algorithm-based optimal power
enconman.2019.02.048. flow analysis in the presence of stochastic renewable energy sources. Electr Pow
[42] Yu K, Liang JJ, Qu BY, Cheng Z, Wang H. Multiple learning backtracking search Syst Res 2023;216:109087. https://doi.org/10.1016/j.epsr.2022.109087.
algorithm for estimating parameters of photovoltaic models. Appl Energy 2018;
226(February):408–22. https://doi.org/10.1016/j.apenergy.2018.06.010.

16