Computers and Electrical Engineering 94 (2021) 107304

Contents lists available at ScienceDirect

Computers and Electrical Engineering

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

An efficient Manta Ray Foraging Optimization algorithm for

parameter extraction of three-diode photovoltaic model✩
Essam H. Houssein a , Gamela Nageh Zaki a , Ahmed A. Zaki Diab b ,∗,
Eman M.G. Younis a
a
Faculty of Computers and Information, Minia University, Egypt
b
Faculty of Engineering, Minia University, Egypt

ARTICLE INFO ABSTRACT

Keywords: The Parameter estimation of solar cell models is considered an important problem in the
Manta Ray Foraging Optimization (MRFO) computational simulation and design of photovoltaic (PV) systems. In this paper, the PV
PV parameter extraction parameters of single, double, and triple-diode models are extracted and tested under different
Single diode
environmental conditions. The parameter estimation of the three models is presented as an
Double diode
optimization process with an objective function to reduce the difference between the measured
Three diode
Solar energy
and calculated data. Moreover, a new optimization algorithm for extracting the PV parameters
of the single, double, and three-diode models called the Manta Ray Foraging Optimization
(MRFO) algorithm has been proposed. The results show that the obtained parameters are
the optimal values and give the least difference between the measured and calculated data
compared with other algorithms.

1. Introduction

Energy shortage has encouraged researchers worldwide to find alternative clean and cost-effective energy solutions. Currently,
40.4% of the world’s energy needs are achieved through coal, which is dangerous and scarce energy source. This has attracted
the attention of the world to renewable energy sources. Generating electricity from renewable energy sources is a huge challenge.
Technologies for producing electricity from renewable sources play a significant role in tackling this challenge. Solar energy is
the most sustainable type of renewable energy. It is an alternative solution to overcome the shortage of fossil fuels Thus, the
manufacturing and use of photovoltaic cells are increasing.
The photovoltaic (PV) system is the most efficient method for converting solar energy to electrical energy [1]. It has attracted
significant attention in the last few decades. The capacity of installed PV systems has increased significantly worldwide. Accurate
modeling of PV modules is essential because it allows designers to optimize system execution and maximize the profitability of
the system. The performance of a PV system depends on several important features such as the latitude of the location, slope and
plate azimuth, and shading obstacles. These properties fundamentally impact the amount of solar energy converted by the panels.
The solar PV system has attracted significant attention because of its advantages, such as cleanliness, fullness, durability, and low
operational and nonoperational costs. However, it has some disadvantages such as high initial investment costs and low conversion
efficiency of PV modules.

✩ This paper is for special section VSI-sgmg. Reviews processed and recommended for publication by Guest Editor Dr. H. H. Alhelou.
∗ Corresponding author.
E-mail addresses: [email protected] (E.H. Houssein), [email protected] (G.N. Zaki), [email protected] (A.A.Z. Diab),
[email protected] (E.M.G. Younis).

https://doi.org/10.1016/j.compeleceng.2021.107304
Received 5 September 2020; Received in revised form 20 April 2021; Accepted 22 June 2021
Available online 1 July 2021
0045-7906/© 2021 Published by Elsevier Ltd.
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

There are various mathematical models of PV solar cells. The major models described in literature are the single-diode (described
by five variables) and double-diode (described by seven variables) models. Recently, other models have been presented to achieve
the optimum output from various PV technologies such as the three-diode models. The simplest mathematical models of PV solar
cells assume the single-diode model (SDM) [2]. The double-diode model (DDM) is more advanced than the single-diode model [3].
The most sophisticated model is the three-diode (TDM) [4]. There are many unknown variables for each of these models.
There are five unknown parameters in the single-diode model: the saturation current (𝐼𝑜 ), photo current (𝐼𝑝ℎ ), ideality factor of
the diode (a), shunt resistance (𝑅𝑠ℎ ), and series resistance (𝑅𝑠 ). The two-diode model has seven unknown parameters. The unknown
parameters are the reverse saturation currents (𝐼𝑜1 ) and (𝐼𝑜2 ), ideality factors 𝑎1 and 𝑎2 for the two diodes and other parameters 𝐼𝑝ℎ ,
𝑅𝑠 , and 𝑅𝑠ℎ . The three-diode model has unknown nine parameters: the reverse saturation currents (𝐼𝑜1 ), (𝐼𝑜2 ) and (𝐼𝑜3 ) and ideal
factors 𝑎1 , 𝑎2 and 𝑎3 for the three diodes as well as other parameters 𝐼𝑝ℎ , 𝑅𝑠 and 𝑅𝑠ℎ .
The datasheet of the manufacturer of the PV module has other parameters, such as the open circuit voltage (𝑉𝑜𝑐 ), voltage,
current at the point of maximum power (𝑉𝑚𝑝𝑝 and 𝐼𝑚𝑝𝑝 ), short-circuit current (𝐼𝑠𝑐 ), maximum power (𝑃𝑚𝑝𝑝 ), current–voltage and the
temperature coefficients (Kv and Ki). These parameters can also be obtained experimentally. The extraction of the unknown variables
is considered a fundamental issue, because the exact determination of these variables plays a significant in the optimization of solar
cells.
There are two methods for estimating unknown parameters of the PV module: the traditional [5] and metaheuristic methods.
These methods were fully explained in [6].
A detailed mathematical model of the PV module has been presented based on three main points: the open circuit, maximum
power, and short-circuit points on the I–V curve [7]. The mathematical methods depend on solving a collection of formulas or
equations to be solved for estimating the unknown parameters of the PV system. This approach focuses on the starting point, which
is the main concern with of the algorithm. These techniques tend to get caught in local solutions. The metaheuristic approach is
better in determining PV variables. Most values in the equations are excluded in the manufacturer’s datasheet. Thus, this method
is not regarded as correct. The metaheuristic method is suitable for estimating PV variables.
This paper presents a methodology for identifying the unknown variables of PV cells using manta ray foraging optimization
(MRFO). The main contributions of this study are given as follows:

1. MRFO, inspired by smart behaviors of manta rays, is used to determine PV variables.

2. MRFO can determine the optimal values of the PV parameters of the single, double, and three-diode models.
3. After comparing the MRFO with other metaheuristic algorithms, the results showed that MRFO provided the least difference
between the calculated and actual values.

The remainder of the paper is organized as follows. Section 2 presents the related work. Section 3 provides descriptions of
the mathematical model of the PV cell/module. Section 4 presents the explanations of the optimization problem of the solar
cell. Section 5 introduces the MRFO technique. In Section, 6 the simulation results, analysis, and comparative study with other
metaheuristic algorithms are reported. Finally, conclusions and future work are presented in Section 7.

2. Related work

Artificial intelligence techniques are premium for treating nonlinear equations. Recently, different optimization methods have
been proposed to determine the variables of PV solar energy called metaheuristic: ABC [8], MSSO [9], CPSO [10], CSO [11],
adaptive DE algorithm [12], GA with convex combination crossover [13], WOA with chaotic maps [14], SCE with a competitive
complex evolution strategy [15], ImCSA [16], TLO [17], ISCA [18], WOA with an opposition-based learning strategy [19], COA [20],
SMA [21], HS [22], and other methods.
From these studies, it is clear that the form of the PV simulations fundamentally focuses on finding the appropriate variables of
the solar cell. However, this research field still requires more attention. The techniques used to perform solar PV variables extraction
have some disadvantages such as, long convergence time, expanded errors, and complexity. Thus, this study presents an alternative
method, a new optimization technique called MRFO, to determine the solar PV variables. It will be explained and evaluated in
the next sections. When using the MRFO technique, the minimum error between the estimated and experimental data, minimum
convergence time for the optimal values, and best curve-fitting (I–V) were obtained. Table 1 presents various techniques used in
estimating the PV parameters with their main results (root mean squared error).

3. Photovoltaic modeling

Many mathematical models discuss the operation and physical behavior of the PV generator. This section explains the
mathematical models of the PV unit [23] that SDM, DDM and TDM can present. Besides, it examines the SDM, DDM, and TDM
models used in the proposed method. Their objective functions are formulated as follows:

3.1. Single diode model

Equivalent representation of a model circuit diagram with a single diode to estimate the PV properties are displayed in Fig. 1. As
shown in Fig. 1, the SDM is located on a parallel current source with a diode, account for a shunt resistor leakage current and series
resistance to show damages with respect to the load current. The SDM is the most widely used electrical model for representing

2
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Table 1
Summary of previous research on the application of MA methods in SCs.
References Year Method Main result
Wei et al. [10] 2011 CPSO SDM: RMSE of 1.3900𝐸 −03 %
Askarzadeh et al. [22] 2012 HS DDM: RMSE of 1.2600 × 10−03 %
Oliva et al. [8] 2014 ABC SDM: RMSE of 9.8615𝐸 −04 %
DDM: RMSE of 9.8387𝐸 −04 %
TDM: RMSE of 9.8452𝐸 −04 %
Chellaswamy and Ramesh [12] 2016 Adaptive DE algorithm SDM: RMSE of 9.3 × 10−4 ,
DDM: RMSE of 0.000924%, PV: RMSE of 0.002131%
Guo et al. [11] 2016 CSO SDM : RMSE of 9.8602 × 10−04 %
DDM: RMSE of 9.8252 × 10−04 %
Lin et al. [9] 2017 MSSO SDM: RMSE of 9.8607𝐸 −04 %
DDM: RMSE of 9.8281𝐸 −04 %
Hamid et all.[13] 2017 GA with convex combination crossover SDM: RMSE of 9.8602 × 10−04 %
DDM: RMSE of 9.8248 × 10−04 %
PV: RMSE of 0.002425%
Oliva et al. [14] 2017 WOA with chaotic maps SDM: RMSE of 9.8602 × 10−04 %
DDM: RMSE of 9.8272 × 10−04 %
Gae et al. [15] 2018 SCE with a competitive complex evolution strategy SDM: RMSE of 9.860219 × 10−04
DDM: RMSE of 9.824849 × 10−04 %
Abd Elaziz and Oliva [19] 2018 WOA with an opposition-based learning strategy SDM: RMSE of 9.8602 × 10−04 %
DDM: RMSE of 9.8251 × 10−04 %
TDM: RMSE of 9.8249 × 10−04 %
Kang et al. [16] 2018 ImCSA SDM: RMSE of 9.8602𝐸 −04 %
DDM: RMSE of 9.8249𝐸 −04 %
Chen et al. [18] 2019 ISCA SDM: RMSE of 9.8602𝐸 −04 %
DDM: RMSE of 9.8237𝐸 −04 %
Li et al. [17] 2019 TLO SDM: RMSE of 9.8609𝐸 −04 %
DDM: RMSE of 9.8612𝐸 −04 %
TDM: RMSE of 9.8613𝐸 −04 %
Diab and Ahmed A Zaki [20] 2020 COA SDM: RMSE of 7.7547𝐸 −04 %
DDM: RMSE of 7.6480𝐸 −04 %
Kumar et al. [21] 2020 SMA SDM: RMSE of 9.8582𝐸 −04 %
DDM: RMSE of 9.8148𝐸 −04 %
TDM: RMSE of 9.80143𝐸 −04 %

Fig. 1. Single diode model of the PV cell.

the behavior of a PV solar module due to its average complexity and precise results. This model is developed by collecting currents
with a nonphysical diode ideality factor. In SDM, we must estimate the five unknown variables: (𝐼𝑝ℎ , 𝐼𝑠𝑑 , 𝑛, 𝑅𝑠 , and 𝑅𝑠ℎ ) of the PV
cell. The output current of the solar cell can be subedited as in Eq. (1).

𝐼𝐿 = 𝐼𝑝ℎ − 𝐼𝑑 − 𝐼𝑠ℎ (1)

where 𝐼𝐿 indicate the output current of cell, 𝐼𝑑 is the current of the diode, 𝐼𝑝ℎ denote the total current produced by the solar cell,
and 𝐼𝑠ℎ is the current of shunt resistor. According to the Shockley equation, the diode current 𝐼𝑑 can be formulated as Eq. (2).

𝑉𝐿 + 𝑅𝑠 .𝐼𝐿
𝐼𝑑 = 𝐼𝑠𝑑 .[𝑒𝑥𝑝( ) − 1] (2)
𝑛.𝑉𝑡

where 𝑉𝐿 indicate the output voltage of the cell, and 𝑉𝑡 is the junction thermal voltage defined by Eq. (3).

𝐾.𝑇
𝑉𝑡 = (3)
𝑞

3
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Fig. 2. Double diode model of the PV cell.

Fig. 3. Three-diode model of the PV cell.

where 𝑘 is the constant of the Boltzmann (1.2806503 × 10−23 J∕K), 𝑇 is the temperature of the junction in Kelvin, and 𝑞 is the charge
of the electron (1.60217646 × 10−19 C). According to Ohm’s law, the current of the shunt resistor is calculated using Eq. (4).
𝑉𝐿 + 𝑅𝑠 .𝐼𝐿
𝐼𝑠ℎ = (4)
𝑅𝑠ℎ
Thus, collecting Eqs. (1)–(4), the relationship among the output voltage, output current, and the model parameters for the SDM
is represented by Eq. (5).
𝑉𝐿 + 𝑅𝑠 .𝐼𝐿 𝑉 + 𝑅𝑠 .𝐼𝐿
𝐼𝐿 = 𝐼𝑃 ℎ − 𝐼𝑠𝑑 [𝑒𝑥𝑝( ) − 1] − 𝐿 (5)
𝑛.𝑉𝑡 𝑅𝑠ℎ
The Eq. (5) illustrates that there exist five unknown parameters (𝐼𝑝ℎ , 𝐼𝑠𝑑 , 𝑛, 𝑅𝑠 , and 𝑅𝑠ℎ ), which must be determined.

3.2. Double-diode model

SDM is a suitable approach for the real solar cell. However, it is not entirely correct for various implementations [3]. For this
reason, the DDM model was proposed, which provides more accurate results. This approach consists of two diodes to derive the
current photo-generated source (𝐼𝑝ℎ ). One of these diodes is configured as a rectifier, and the other diode is used to model the
charge recombination current and some nonrealities. The equivalent circuit diagram for DDM is represented in Fig. 2. As shown in
Fig. 2, the electric current flowing through the charge is obtained as in Eq. (7).

𝐼𝐿 = 𝐼𝑝ℎ − 𝐼𝑑1 − 𝐼𝑑2 − 𝐼𝑠ℎ (6)

So,
𝑉𝐿 + 𝑅𝑠 .𝐼𝐿 𝑉 + 𝑅𝑠 .𝐼𝐿 𝑉 + 𝑅𝑠 .𝐼𝐿
𝐼𝐿 = 𝐼𝑃 ℎ − 𝐼𝑠𝑑1 [𝑒𝑥𝑝( ) − 1] − 𝐼𝑠𝑑2 [𝑒𝑥𝑝( 𝐿 ) − 1] − 𝐿 (7)
𝑛1 .𝑉𝑡 𝑛2 .𝑉𝑡 𝑅𝑠ℎ
where 𝐼𝑑1 and 𝐼𝑑2 are the first and second diode currents. 𝑛1 and 𝑛2 are diode ideal factors, 𝐼𝑠𝑑1 and 𝐼𝑠𝑑2 are the saturation currents.
There are seven unknown parameters for DDM (𝐼𝑝ℎ , 𝐼𝑠𝑑1 , 𝐼𝑠𝑑2 , 𝑛1 , 𝑛2 , 𝑅𝑠 , and 𝑅𝑠ℎ ) need to be estimated.

3.3. Three-diode model

In this model, the impact of grain limits and large leakage current is considered [4]. Thus, the third diode is added in parallel
to the two diodes of the DDM as represented in Fig. 3. The current of the output is calculated as follows:

𝐼𝐿 = 𝐼𝑝ℎ − 𝐼𝑑1 − 𝐼𝑑2 − 𝐼𝑑3 − 𝐼𝑠ℎ (8)

𝑉𝐿 + 𝑅𝑠 .𝐼𝐿 𝑉 + 𝑅𝑠 .𝐼𝐿 𝑉 + 𝑅𝑠 .𝐼𝐿 𝑉 + 𝑅𝑠 .𝐼𝐿

𝐼𝐿 = 𝐼𝑃 ℎ − 𝐼𝑠𝑑1 [𝑒𝑥𝑝( ) − 1] − 𝐼𝑠𝑑2 [𝑒𝑥𝑝( 𝐿 ) − 1] − 𝐼𝑠𝑑3 [𝑒𝑥𝑝( 𝐿 ) − 1] − 𝐿 (9)
𝑛1 .𝑉𝑡 𝑛2 .𝑉𝑡 𝑛3 .𝑉𝑡 𝑅𝑠ℎ

4
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Table 2
Parameter ranges for the three models of diode in the 2 case studies.
Parameters Single Double Three
Case study 1 Case study 2 Case study 1 Case study 2 Case study 1 Case study 2
Lower Upper Lower Upper Lower Upper Lower Upper Lower Upper Lower Upper
𝐼𝑝ℎ 0.5 0.8 0 2 0.5 0.8 0 2 0.5 0.8 0 2
𝐼𝑠ℎ1 0.2𝐸 −6 0.4𝐸 −6 0 50𝐸 −6 0.2𝐸 −6 0.4𝐸 −6 0 50𝐸 −6 0.2𝐸 −6 0.4𝐸 −6 0 50𝐸 −6
𝐼𝑠ℎ2 – – – – 0.2𝐸 −6 0.4𝐸 −6 0 50𝐸 −6 0.2𝐸 −6 0.4𝐸 −6 0 50𝐸 −6
𝐼𝑠ℎ3 – – – – – – – – 0.2𝐸 −6 0.4𝐸 −6 0 50𝐸 −6
𝑅𝑠ℎ 50 60 0 200 50 60 0 200 50 60 0 200
𝑅𝑠 0.03 0.05 0 2 0.03 0.05 0 2 0.03 0.05 0 2
𝑛1 1 5 1 50 1 5 1 50 1 5 1 50
𝑛2 – – – – 1 5 1 50 1 5 1 50
𝑛3 – – – – – – – – 1 5 1 50

where 𝐼𝑑1 , 𝐼𝑑2 and 𝐼𝑑3 are the first, second, and third diode currents; 𝑛1 , 𝑛2 and 𝑛3 are the diode ideal factors, 𝐼𝑠𝑑1 , 𝐼𝑠𝑑2 and 𝐼𝑠𝑑3
are the saturation currents. Eq. (9) illustrates that there are nine unknown variables for TDM (𝐼𝑝ℎ , 𝐼𝑠𝑑1 , 𝐼𝑠𝑑2 , 𝐼𝑠𝑑3 , 𝑛1 , 𝑛2 , 𝑛3 , 𝑅𝑠 , and
𝑅𝑠ℎ ) that need to be estimated.

4. Problem formulation

The major goal of modeling the solar cells is to minimize the errors between the experimental and extracted data by estimating
the optimal values of the parameters. The important requirements for the application of any optimization algorithm are obtained
by identifying the solution vector (X), search area, and the objective function. The solution vector X for SDMX = 𝐼𝑝ℎ , 𝐼𝑠𝑑 , 𝑛, 𝑅𝑠 ,
and 𝑅𝑠ℎ , vectors X for DDM X=𝐼𝑝ℎ , 𝐼𝑠𝑑1 , 𝐼𝑠𝑑2 , 𝑛1 , 𝑛2 , 𝑅𝑠 , and 𝑅𝑠ℎ , and for TDM X=𝐼𝑝ℎ , 𝐼𝑠𝑑1 , 𝐼𝑠𝑑2 , 𝐼𝑠𝑑3 , 𝑛1 , 𝑛2 , 𝑛3 , 𝑅𝑠 , and 𝑅𝑠ℎ .
The objective function is obtained through RMSE. The error function (𝐽 (𝑉𝑙 , 𝐼𝐿 , 𝑋)) is the difference between the estimated and
experimental currents. (𝐽 (𝑉𝑙 , 𝐼𝐿 , 𝑋)) is expressed in the equation as follows.

𝐽 (𝑉𝐿 , 𝐼𝐿 , 𝑋) = 𝐼𝐿 − 𝐼𝐿𝑒𝑥𝑝 (10)

where the estimated current 𝐼𝐿 can be calculated from Eqs. (5), (6), and (9)for SDM, DDM, and TDM models. 𝐼𝐿𝑒𝑥𝑝 is the measured
current.
√
√
√1 ∑ 𝐾
( )2
𝑅𝑀𝑆𝐸 = √ (𝑗 𝑉𝑡 , 𝐼𝑡 , 𝑋 ) (11)
𝐾 𝑖=1

To clarify the optimization problem, it can be rewritten as a constrained optimization problem. It is defined as a nonlinear
least-squares regression problem as follows:

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓 (𝑋) (12)

𝑆𝑢𝑏𝑗𝑒𝑐𝑡 𝑡𝑜 𝑋 ∈ Θ (13)

where 𝑓 is the objective function and X is the solution vectors as five, seven, or nine parameters for SSD, DDM, or TDM, respectively.
Θ denotes the bound constraints of X that consider physical reality as the parameter ranges, which are listed in Table 2. We
can observe that the optimization problem of Eq. (12) is subjected to only bound constraints, which may be an unconstrained
optimization problem with an initial interval box defined by the bound constraints in interval analysis. Once the function bounds
over a subregion are available and ready, any unpromising subregion can be detected and discarded, which the lower bound may
be greater than a known minimum at hand; as the upper bound of the global minimum; from the subsequent search for a global
minimizer.

5. Manta ray foraging optimization algorithm (MRFO)

This method is inspired by the smart behavior of manta rays [24]. It imitates three individual strategies for finding foraging,
including searching for the chain foraging, cyclone foraging, and somersault foraging to advance efficiency improvements form to
solve several optimization problems.

5.1. Chain foraging

The manta rays line from head to tail and form a food chain. Individuals except for the first movement toward the food and in
front of it. This mathematical model is described as follows:
{ 𝑑
𝑑
𝑥𝑖 (𝑡) + 𝑟.(𝑥𝑑𝑏𝑒𝑠𝑡 (𝑡) − 𝑥𝑑𝑖 (𝑡)) + 𝑎.(𝑥𝑑𝑏𝑒𝑠𝑡 (𝑡) − 𝑥𝑑𝑖 (𝑡)) 𝑖=1
𝑥𝑖 (𝑡 + 1) = (14)
𝑥𝑑𝑖 (𝑡) + 𝑟.(𝑥𝑑𝑖−1 (𝑡) − 𝑥𝑑𝑖 (𝑡)) + 𝑎.(𝑥𝑑𝑏𝑒𝑠𝑡 (𝑡) − 𝑥𝑑𝑖 (𝑡)) 𝑖 = 2, … , 𝑁

5
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

√
𝑎 = 2.𝑟. |log(𝑟)| (15)

Where 𝑥𝑑𝑖 (𝑡)

is the position of 𝑖𝑡ℎ individual at time 𝑡 in 𝑑𝑡ℎ dimension, 𝑟 is a random vector in [0,1], 𝑎 is a weight coefficient, 𝑥𝑑𝑏𝑒𝑠𝑡 (𝑡)
is the plankton with high concentration. The position update of the 𝑖𝑡ℎ individual is determined by the position 𝑥𝑖−1 (𝑡) of the (𝑖 − 1)th
current individual and the position 𝑥𝑏𝑒𝑠𝑡 (𝑡) of the food.

5.2. Cyclone foraging

Manta ray on a line developing a spiral search for food. The mathematical equation to model the spiral shape movement of
manta rays in n-dimensional (n-D) space is given as follows:
{ 𝑑
𝑥𝑏𝑒𝑠𝑡 + 𝑟.(𝑥𝑑𝑏𝑒𝑠𝑡 (𝑡) − 𝑥𝑑𝑖 (𝑡)) + 𝛽.(𝑥𝑑𝑏𝑒𝑠𝑡 (𝑡) − 𝑥𝑑𝑖 (𝑡)) 𝑖 = 1
𝑥𝑑𝑖 (𝑡 + 1) = (16)
𝑥𝑑𝑏𝑒𝑠𝑡 + 𝑟.(𝑥𝑑𝑖−1 (𝑡) − 𝑥𝑑𝑖 (𝑡)) + 𝛽.(𝑥𝑑𝑏𝑒𝑠𝑡 (𝑡) − 𝑥𝑑𝑖 (𝑡)) 𝑖 = 2, … , 𝑁
𝑇 −𝑡+1
𝛽 = 2𝑒𝑟1 𝑇 . sin(2𝜋𝑟1 ) (17)

where 𝛽 is the coefficient of weight, 𝑇 is the number of maximum iterations, and 𝑟1 is a random number in [0,1].
The mechanism converges on exploration and compeers MRFO to obtain a big goal in the global search. The equation is given
as follows:

𝑥𝑑𝑟𝑎𝑛𝑑 = 𝐿𝑏𝑑 + 𝑟.(𝑈 𝑏𝑑 − 𝐿𝑏𝑑 ) (18)

{
𝑥𝑑𝑟𝑎𝑛𝑑 + 𝑟.(𝑥𝑑𝑟𝑎𝑛𝑑 − 𝑥𝑑𝑖 (𝑡)) + 𝛽.(𝑥𝑑𝑟𝑎𝑛𝑑 − 𝑥𝑑𝑖 (𝑡)) 𝑖=1
𝑥𝑑𝑖 (𝑡 + 1) = (19)
𝑥𝑑𝑟𝑎𝑛𝑑 + 𝑟.(𝑥𝑑𝑖−1 (𝑡) − 𝑥𝑑𝑖 (𝑡)) + 𝛽.(𝑥𝑑𝑟𝑎𝑛𝑑 − 𝑥𝑑𝑖 (𝑡)) 𝑖 = 2, … , 𝑁

Where, 𝑥𝑑𝑟𝑎𝑛𝑑 is a random position search, 𝐿𝑏𝑑 and 𝑈 𝑏𝑑 are the lower and upper boundary of the 𝑑𝑡ℎ dimension.

5.3. Somersault foraging

With this behavior, the food position is considered the fulcrum. Thus, they always change their positions by the best position.
This model is formulated as follows:

𝑥𝑑𝑖 (𝑡 + 1) = 𝑥𝑑𝑖 (𝑡) + 𝑆.(𝑟2 .𝑥𝑑𝑏𝑒𝑠𝑡 − 𝑟3 .𝑥𝑑𝑖 (𝑡)) 𝑖 = 1, … ., 𝑁 (20)

𝑆 is the factor of somersault and 𝑆 = 2, 𝑟2 and 𝑟3 are two random numbers between [0, 1].

5.4. Time complexity of the MRFO

The time complexity of the MRFO depends on the number of parameters, number of individuals, and maximum number of
iterations. The three strategies of manta ray are implemented for every iteration. Thus, the total time complexity of this algorithm
is identified as follows:

𝑂(𝑀𝑅𝐹 𝑂) = 𝑂(𝑇 (𝑂(𝑐𝑦𝑐𝑙𝑜𝑛𝑒𝑓 𝑜𝑟𝑎𝑔𝑖𝑛𝑔 + 𝑐ℎ𝑎𝑖𝑛𝑓 𝑜𝑟𝑎𝑔𝑖𝑛𝑔) + 𝑂(𝑠𝑜𝑚𝑒𝑟𝑠𝑎𝑢𝑙𝑡𝑓 𝑜𝑟𝑎𝑔𝑖𝑛𝑔))) (21)

𝑂(𝑀𝑅𝐹 𝑂) = 𝑂(𝑇 (𝑛𝑑 + 𝑛𝑑)) = 𝑂(𝑇 𝑛𝑑) (22)

where 𝑇 is the number of maximum iterations, 𝑛 is the number of individuals, and 𝑑 is the number of variables.

6. Simulation and experimental results

To evaluate the performance of the PV cell/module estimation, a PV cell by RTC France and a PV module by Photowatt (PWP201)
are considered reference problems in the identification of PV systems. The cell is a commercial silicon solar cell with a diameter of
57 mm, operating at 33 ◦ C below (1000 W∕m2 ). The PV module considered has 36 polycrystalline PV cells connected in series and
operating at 45 ◦ C, always under (1000 W∕m2 ) [5]. It is worth noting that many studies utilized these datasets. To examine the
precision, robustness and reliability of the proposed MRFO for determining the variables of different PV models, i.e., SDM, DDM,
and TDM two recordings of experiences, benchmark datasets of a standard PV cell and a standard PV. The modules are described
in the next subsections. The benchmark data of a standard PV cell and module are selected to prove the efficiency of the MRFO
and compare it with existing results. All programs were performed with MATLAB R2016b using a computer with an Intel (R) Core
(TM) i3 processor, 2.3 GHz, 4 GB RAM, and Windows 7 system. To estimate the robustness of the proposed algorithm, RMSE was
analyzed statistically. In addition, the best solution quality and modeling performance of the proposed algorithm are rated based
on the mean absolute error (MAE) and the mean relative errors (MRE) given as follows:

𝐴𝑏𝑠𝑜𝑙𝑢𝑡𝑒𝐸𝑟𝑟𝑜𝑟(𝐴𝐸) = ||𝐼𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑 − 𝐼𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 || (23)

1 ∑
𝑁
𝑀𝐴𝐸 = 𝐴𝐸𝑖 (24)
𝑁 𝑖=1

6
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Fig. 4. Comparison between the experimental data and simulated data of RTC France solar cell (SDM) (a) I–V characteristics; (b) P–V characteristics.

|𝐼𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑 − 𝐼𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 |
𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐸𝑟𝑟𝑜𝑟(𝑅𝐸) = | | (25)
𝐼𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑

1 ∑
𝑁
𝑀𝑅𝐸 = 𝑅𝐸𝑖 (26)
𝑁 𝑖=1

𝐼𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 indicated the current calculated using of Newton’s method, 𝐼𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑 indicated the experimentally acquired value of current,
and 𝑁 is the number of data points on the I–V characteristics.

6.1. Case study 1: RTC France PV cell

The proposed algorithm used to estimate the values of the parameters of SDM, DDM and TDM of a 57 mm diameter commercial
(RTC France) silicon solar cell below 1000 W∕m2 at 33 ◦ C. depends on the experimental data from the RTC France PV cell 1000
(W∕m2 at 33 ◦ C) the five unknown variables for SDM, seven unknown variables for DDM and nine parameters for TDM have been
estimated. The experimental curve had 26 pairs of I–V data. The results considered one hundred races for all models. However, the
results of the best run and objective function values (RMSE) have been registered.

6.1.1. Results of the single-diode model

For the SDM, the values of the parameters identified for the best and worst run of 30 runs are shown in Table 3. Five unknown
variables that need to be extracted for SDM. The range of every parameter used is listed in Table 2. Besides, Table 4 offer the
determined variable values after running the algorithm 30 times for the SDM and values that correspond to the objective function
variation and calculation costs. The minimum value of the objective function that provided the best solution is 7.7301𝐸 −04 .
Other newer algorithms from the literature are selected for comparison since their performance is relatively better in identifying
variables for SDM, such as gray wolf optimizer (GWO) [25], whale optimization algorithm (WOA) [14], Sine cosine algorithm
(SCA) [26], moth-flame optimization (MFO) [27], ant lion optimizer (ALO) [28], multiverse optimizer (MVO) [29]. The results of
the comparison include the determined parameters, RMSE, MAE, and MRE.
From Table 4, one can see that the MRFO reaches the lowest RMSE value under all selected algorithms. Another notable result
is the decrease or stability of MAE values compared to these contemporaries, and the lower MRE compared among all methods.
Thus, the proposed method was very precise when determining the PV parameters with OF distribution (RMSE) values close to the
minimum value.
Besides, the I–V and P–V characteristics are the best models evaluated using the MRFO and the experimental data are shown in
Fig. 4. From Fig. 4, the two curves of the tested and the calculated data have been shown in the same curve that the results obtained
by MRFO are optimal. The individual absolute error (IAE) and relative error (RE) between the experimental and calculated data are
shown in Fig. 5 and Table 9. The calculated data from the MRFO closely match the experimental data for SDM, indicating that the
parameters identified by MRFO are accurate.
Fig. 6 shows the convergence curves of different optimizations algorithms for estimating the parameters of SDM of the RTC
france cell. From Fig. 6, the proposed MRFO algorithm achieves the best value while the MVO algorithm is the worst. The remaining
algorithms do not reach the minimum best value of the RMSE as the proposed algorithm (MRFO). Thus, the best one that reaches
the coordinates.

7
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Table 3
Optimal parameters for single-diode model of RTC France PV cell.
Run 𝐼𝑝ℎ 𝐼𝑠𝑑 𝑅𝑠 𝑅𝑠ℎ 𝑛 RMSE
The best run 0.7608 3.1077𝐸 −07 0.03655 52.9026 1.4773 7.7301𝐸 −04
The worst run 0.7607 3.2321𝐸 −07 0.0364 53.4836 1.4812 7.8071𝐸 −04
Min 0.7607 3.0071𝐸 −07 0.0364 51.7103 1.4740 7.7301𝐸 −04
Mean 0.7608 3.1087𝐸 −07 0.0365 52.8863 1.4773 7.7414𝐸 −04
Max 0.7609 3.2321𝐸 −07 0.0367 53.8600 1.4812 7.8071𝐸 −04
STD 3.1610𝐸 −05 5.7616𝐸 −09 8.0644𝐸 −05 0.4816 0.0019 1.6000𝐸 −06

Table 4
Comparison results for single-diode model of RTC France PV cell.
Parameters WOA GWO MVO SCA MFO ALO MRFO
𝐼ℎ𝑝 (A) 0.76405 0.7606 0.7630 0.7515 0.7607 0.7601 0.7608
𝐼𝑠ℎ (μA) 2.8588𝐸 −07 2.2496𝐸 −07 3.9989𝐸 −07 2.5606𝐸 −07 3.9953𝐸 −07 2.4432𝐸 −07 3.0908𝐸 −07
𝑅𝑠ℎ (ohm) 59.9940 54.6069 56.3258 54.2298 60 57.2379 52.7129
𝑅𝑠 (ohm) 0.0484 0.0385 0.0377 0.0372 0.0355 0.0375 0.0366
𝑛 1.4702 1.4455 1.5027 1.4593 1.5029 1.4534 1.4767
RMSE 0.0132 0.0011 0.0031 0.0072 0.0009 0.0010 7.7307𝐸 −04
MAE 0.0095 0.0008 0.0027 0.0064 0.0007 0.0009 6.7784𝐸 −04
MRE 0.0124 0.0011 0.0036 0.0084 0.0009 0.0012 8.8723𝐸 −04
Elepsed-time 0.006208 s 0.000267 s 0.000138 s 0.000141 s 0.000218 s 0.000138 s 0.000074 s

Fig. 5. Error index values of the simulated current data and the experimental current data of RTC France solar cell (SDM) (a) IAE values; (b) RE values.

6.1.2. Results of the double diode model

The results and analysis of the implementation of MRFO to estimate the parameters of DDM with the RTC France the PV cell are
presented in this subsection. In DDM, the number of decision variables or the dimension of the problem during optimization is seven.
The range used for the various parameters of PV cells of the diode model is presented in Table 2. In this model, there are seven
unknown parameters. The values of these parameters for the best and worst run are listed in Table 5. It shows the objective function
(RMSE) with the minimum, maximum, mean, and standard deviation values for each parameter. After the analysis of these results,
we obtain that the proposed method has the minimum (best) error value, which is 7.6046𝐸 −04 , the mean value of the objective
function is 7.8863𝐸 −04 , and the standard deviation of the objective function is 3.0052𝐸 −05 .
In this case, the optimal values for the parameters of DDM, RMSE, MAE, and MRE of the proposed and six tested algorithms are
presented in Table 6 after running and testing these algorithms. Moreover, the I–V and the P–V characteristics of the perfect model
evaluated by the MRFO and the experimental data are shown in Fig. 7. The IAE and RE between the test and calculated data are
presented in Fig. 8 and Table 9.
Table 6 shows that MRFO has the best statistical value compared with other methods. MRFO achieves the best RMSE values
(9.824849𝐸 −04 ) than that achieved by other methods. Besides, it achieves lower or equal MAE values than these contemporaries and
the lowest ERM among all compared methods.

8
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Fig. 6. The Best cost (RMSE) with respect to iteration numbers of proposed algorithm and comparison algorithms for RTC France PV cell of single diode model.

Table 5
Optimal parameters for double-diode model of RTC France PV cell.
Run 𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝑅𝑠 𝑅𝑠ℎ 𝑛1 𝑛2 RMSE
Best run 0.7608 2.2619𝐸 −07 3.9314𝐸 −07 0.0368 54.2188 1.4515 1.8760 7.6046𝐸 −04
Worst run 0.7607 2.0370𝐸 −07 2.0393𝐸 −07 0.0354 57.8522 1.4869 1.5276 8.7606𝐸 −04
Min 0.7605 2.0370𝐸 −07 2.0016𝐸 −07 0.0354 52.9295 1.4515 1.4455 7.6046𝐸 −04
Mean 0.7608 2.6025𝐸 −07 2.5529𝐸 −07 0.0364 55.1621 1.6805 1.6424 7.8863𝐸 −04
Max 0.7613 3.5046𝐸 −07 3.9314𝐸 −07 0.0369 58.5191 1.9989 1.9999 8.7606𝐸 −04
STD 1.3887𝐸 −04 3.8767𝐸 −08 4.6922𝐸 −08 3.6307𝐸 −04 1.6801 0.2068 0.2245 3.0052𝐸 −05

Table 6
Comparison results for double-diode model of RTC France PV cell.
Parameters WOA GWO MVO SCA MFO ALO MRFO
𝐼ℎ𝑝 (A) 0.7658 0.7614 0.7629 0.7586 0.7609 0.7617 0.7607
𝐼𝑠ℎ1 (μA) 2.9957𝐸 −07 3.3771𝐸 −07 3.9698𝐸 −07 3.9637𝐸 −07 2.8619𝐸 −07 3.5427𝐸 −07 2.2685𝐸 −07
𝐼𝑠ℎ2 (μA) 3.9438𝐸 −07 2.4649𝐸 −07 3.4509𝐸 −07 2.0000𝐸 −07 3.0667𝐸 −07 3.5047𝐸 −07 2.5403𝐸 −07
𝑅𝑠ℎ (ohm) 59.0196 56.7361 50.6069 60 60 54.2679 54.9462
𝑅𝑠 (ohm) 0.0493 0.0335 0.0332 0.0300 0.0345 0.0325 0.0366
𝑛1 1.4795 1.5431 1.6007 1.8613 1.4906 1.5492 1.4618
𝑛2 1.9201 1.5442 1.5443 1.4413 1.6617 1.5830 1.8208
𝐑𝐌𝐒𝐄 0.0154 0.0014 0.0023 0.0102 0.0011 0.0018 7.6842𝐸 −04
𝐌𝐀𝐄 0.0115 0.0014 0.0019 0.0072 0.0009 0.0015 6.6931𝐸 −04
𝐌𝐑𝐄 0.0151 0.0015 0.0025 0.0095 0.0012 0.0019 8.7606𝐸 −04
𝐄𝐥𝐚𝐩𝐬𝐞𝐝 − 𝐓𝐢𝐦𝐞 0.000141 s 0.000251 s 0.000275 s 0.000145 s 0.000166 s 0.000222 s 0.000053 s

Fig. 9 shows the convergence curves of various optimization algorithms for estimating the parameters of DDM of the RTC France
cell. From the Figure, it is shown that the proposed MRFO algorithm reaches the coordinates, indicating MRFO is the best algorithm
while the remaining algorithms do not reach the minimum best value of RMSE as the proposed algorithm.

6.1.3. Results of the three-diode model

For the TDM of RTC France solar cell at 33 ◦ C, the parameters and RMSE values estimated using the proposed MRFO algorithm
are summarized in Table 7. The limit for the nine unknown parameters of TDM is shown in Table 2. In Table 7, the value of the
best error(RMSE) is 7.6083𝐸 −04 , and the standard deviation is 1.2099𝐸 −04 .
Table 8 shows that the proposed MRFO algorithm achieves the best accurate RMSE value among all compared methods and
Fig. 10 for observing the correspondence between the experimental and calculated data. From Figs. 10(a) and (b), the simulated
data for the TDM are perfect with the measured data at almost all data points. It means that this algorithm also has the highest
computing efficiency in estimating the most accurate parameter values from TDM.
Table 9 shows the comparison between the simulated and measured data, which consist of IAE and RE between the experiment
and calculated data, and Fig. 11 for the observation of IAE and RE.

9
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Fig. 7. Comparison between the experimental data and simulated data of RTC France solar cell (DDM) (a) I–V characteristics; (b) P–V characteristics.

Fig. 8. Error index values of the simulated current data and the experimental current data of RTC France solar cell (DDM) (a) IAE values; (b) RE values.

Fig. 12 displays the convergence curves of various optimization algorithms for estimating the parameters of TDM ofRTC France
cell. From the Figure, it is shown that the proposed MRFO algorithm is the best because it achieves the best (minimum) value of
RMSE.
For more validation, a comparison between the proposal results with those of the existing algorithms for the same case study has
been presented. Table 10 presents Various algorithms that estimate PV parameters for SDM. The proposed algorithm results have
been compared with various existing algorithms, such slime mould algorithm (SMA) in 2020 with an RMSE of 9.8582𝐸 −04 , Coyote
optimization algorithm (COA) in 2020 with an RMSE of 7.7547𝐸 −04 , teaching-learning optimization algorithm (TLO) in 2019 with
an RMSE of 9.8609𝐸 −04 , improved sine cosine algorithm (ISCA) in 2019 with an RMSE of 9.8602𝐸 −04 , improved cuckoo search
algorithm (ImCSA) in 2018 with an RMSE of 9.8602𝐸 −04 , modified simplified swarm optimization (MSSO) in 2017 with an RMSE
of 9.8607𝐸 −04 , artificial bee colony (ABC) in 2014 with an RMSE of 9.8615𝐸 −04 and chaos particle swarm optimization (CPSO) in
2011 with an RMSE of 1.3900𝐸 −03 . As shown in Table 10, the proposed MRFO achieved the best result of 7.7301𝐸 −04 . Table 11
presents various techniques used for estimating the PV parameters of DDM. After comparing the results of the existing algorithms
reported in different years (2020, 2019, 2018,2017, 2015, and 2012) and under the same conditions. It can be concluded that the
proposed MRFO obtained the optimal solution with the smallest error of 7.7301𝐸 −04 compared to these algorithms. Table 12 listed
algorithms for TDM parameters used recently. As shown in Table 12, the proposed MRFO achieves the best RMSE compared with
those of the existing algorithms.

10
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Fig. 9. The Best Cost (RMSE) with respect to iteration numbers of proposed algorithm and comparison algorithms for RTC France PV cell of double diode model.

Table 7
Optimal parameters for three-diode model of RTC France PV cell.
Run 𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝐼𝑠𝑑3 𝑅𝑠 𝑅𝑠ℎ 𝑛1 𝑛2 𝑛3 RMSE
Best run 0.7607 2.6318𝐸 −07 2.3678𝐸 −07 2.5070𝐸 −07 0.0367 55.0167 1.9131 1.4551 1.9965 7.6083𝐸 −04
Worst run 0.7615 2.5716𝐸 −07 2.2482𝐸 −07 2.3106𝐸 −07 0.0338 55.9953 1.5729 1.8881 1.4939 1.3418𝐸 −03
Min 0.7605 2.0022𝐸 −07 2.0007𝐸 −07 2.0031𝐸 −07 0.0338 51.2170 1.4508 1.4544 1.4451 7.6083𝐸 −04
Mean 0.7609 2.5382𝐸 −07 2.3309𝐸 −07 2.5027𝐸 −07 0.0358 56.9082 1.8107 1.7395 1.6963 8.8205𝐸 −04
Max 0.7615 3.5304𝐸 −07 3.3080𝐸 −07 3.5874𝐸 −07 0.0367 59.9961 2.0000 1.9997 1.9965 1.3418𝐸 −03
STD 2.3113𝐸 −04 4.9429𝐸 −08 3.2316𝐸 −08 4.9775𝐸 −08 6.7410𝐸 −04 2.1476 0.2183 0.2159 0.2379 1.2099𝐸 −04

Table 8
Comparison of the results for three-diode model of RTC France PV cell.
Parameters WOA GWO MVO SCA MFO ALO MRFO
𝐼ℎ𝑝 (A) 0.7667 0.7611 0.7614 0.7560 0.7605 0.7615 0.76077
𝐼𝑠ℎ1 (μA) 2.2309𝐸 −07 3.0297𝐸 −07 3.3273𝐸 −07 2.0000𝐸 −07 2.0000𝐸 −07 3.8979𝐸 −07 2.3539𝐸 −07
𝐼𝑠ℎ2 (μA) 3.9027𝐸−07 2.2386𝐸 −07 3.9977𝐸 −07 3.5918𝐸 −07 4.0000𝐸 −07 3.7385𝐸 −07 2.4289𝐸 −07
𝐼𝑠ℎ3 (μA) 3.9029𝐸−07 3.0436𝐸 −07 2.2175−07 4.0000𝐸 −07 3.9999𝐸 −07 3.7425𝐸 −07 2.1963𝐸 −07
𝑅𝑠ℎ (ohm) 58.6704 58.9166 59.1165 60 50 54.8311 53.5857
𝑅𝑠 (ohm) 0.0488 0.0346 0.0351 0.0302 0.0300 0.0338 0.03689
𝑛1 1.9539 1.9892 1.9463 2 1.5863 1.8705 1.9062
𝑛2 1.7596 1.6676 1.5218 2 2 1.5059 1.9198
𝑛3 1.5185 1.4925 1.6737 1.5081 1.5358 1.9555 1.4485
𝐑𝐌𝐒𝐄 0.0179 0.0012 0.0030 0.0059 0.0038 0.0015 7.5936𝐸 −04
𝐌𝐀𝐄 0.0135 0.0009 0.0022 0.0059 0.0026 0.0012 6.7021𝐸 −04
𝐌𝐑𝐄 0.0176 0.0012 0.0028 0.0068 0.0034 0.0016 8.7724𝐸 −04
Elapsed-Time 0.000204 s 0.000202 s 0.000139 s 0.000144 s 0.000207 s 0.000142 s 0.000052 s

6.2. Case study 2: Photowatt-PWP201 PV module

Based on the experimental data from the Photowatt-PWP201 PV module (1000 W∕m2 at 45 ◦ C), the five ignored variables for
SDM, seven unknown variables for DDM and nine ignored variables of TDM have been estimated. The experimental curve had
twenty six pairs of I–V data. The solutions considered the 100 races for three models. The parameter values for the best execution,
objective function values (RMSE), MAE, and MRE have been presented.

6.2.1. Results of the single diode mode

Table 13 presents the optimal values of the unknown parameters for SDM. As shown in Table 13, the proposed method achieves
the best value for each parameter and the objective function (RMSE). The standard deviation value of RMSE is 5.9016𝐸 −06 . The
minimum (best) and maximum values of RMSE are the same value of 2.0521𝐸 0−3 , indicating that the proposed MRFO achieved the
best value for the objective function in all cases.
The extracted parameters and RMSE value of MRFO, MAE, and MRE were compared with other algorithms: WOA, GWO, MVO,
SCA. MFO, ALO. After running each algorithm of these comparison algorithms in MATLAB, the experimental results are shown
in Table 14. The general best RMSE values among the compared methods are obtained using the proposed MRFO. As shown in
Table 14, the RMSE value is 2.0419𝐸 −03 . The proposed MRFO has the lowest MAE and MRE among all other algorithms, indicating

11
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Table 9
Error metrics for each measurement in the three models (SDM, DDM and TDM) of the solar cell obtained by MRFO.
Item Single diode Double diode Three diode
𝑉𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑 𝐼𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑 𝐼𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 𝐼𝐴𝐸 𝑅𝐸 𝐼𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 𝐼𝐴𝐸 𝑅𝐸 𝐼𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 𝐼𝐴𝐸 𝑅𝐸

1 −0.2057 0.7640 0.76416878 0.00016877 0.00022091 0.763910342 8.9657𝐸 −05 0.000117353 0.764091257 9.1257𝐸 −05 0.000119447
2 −0.1291 0.7620 0.76271661 0.00071661 0.00093797 0.762517143 0.000517143 0.000676889 0.762662693 0.000662693 0.0008674
3 −0.0588 0.7605 0.76138378 0.00088378 0.00115678 0.761238377 0.000738377 0.000966461 0.761351403 0.000851403 0.001114402
4 0.0057 0.7605 0.76016043 0.00033956 0.00044446 0.760064455 0.000435545 0.000570085 0.760147449 0.000352551 0.000461454
5 0.0646 0.7600 0.75904125 0.00095874 0.00125489 0.758989935 0.001010065 0.001322075 0.759045027 0.000954973 0.001249965
6 0.1185 0.7590 0.75800957 0.00099042 0.00129636 0.757997941 0.001002059 0.001311595 0.758026485 0.000973515 0.001274234
7 0.1678 0.7570 0.75704149 4.1492𝐸 −05 5.4309𝐸 −05 0.757063631 6.3631𝐸 −05 8.3278𝐸 −05 0.757066046 6.6046𝐸 −05 8.6447𝐸 −05
8 0.2132 0.7570 0.75607799 0.00092200 0.00120681 0.756126429 0.000873571 0.001143417 0.756101946 0.000898054 0.001175463
9 0.2545 0.7555 0.7550135 0.00048653 0.00063682 0.755077879 0.000422121 0.000552515 0.755025581 0.000474419 0.000620968
10 0.2924 0.7540 0.75358725 0.00041274 0.00054023 0.753654752 0.000345248 0.000451896 0.753574576 0.000425424 0.000556838
11 0.3269 0.7505 0.75131716 0.00081715 0.00106958 0.751373081 0.000873081 0.001142777 0.751269038 0.000769038 0.001006595
12 0.3585 0.7465 0.74729686 0.00079686 0.00104301 0.747327353 0.000827353 0.001082922 0.747210525 0.000710525 0.000930006
13 0.3873 0.7385 0.74007961 0.00157961 0.00206756 0.740075711 0.001575711 0.002062449 0.739966608 0.001466608 0.001919644
14 0.4137 0.7280 0.72742624 0.00057376 0.00075099 0.727387545 0.000612455 0.000801643 0.727314335 0.000685665 0.000897467
15 0.4373 0.7065 0.70703144 0.00053143 0.00069559 0.706968592 0.000468592 0.000613341 0.706957297 0.000457297 0.000598556
16 0.4590 0.6755 0.67541008 8.9918𝐸 −05 0.00011769 0.675340266 0.000159734 0.000209075 0.675405106 9:4893𝐸 −05 0.000124207
17 0.4784 0.6320 0.63100905 0.00099094 0.00129704 0.630949964 0.001050036 0.001374393 0.631080316 0.000919684 0.001203775
18 0.4960 0.5730 0.57218289 0.00081710 0.00106950 0.572145404 0.000854596 0.001118582 0.572310197 0.000689803 0.000902884
19 0.5119 0.4990 0.49954121 0.00054120 0.00070837 0.499526486 0.000526486 0.000689118 0.499685587 0.000685587 0.000897365
20 0.5265 0.4130 0.41347952 0.00047952 0.00062765 0.413481302 0.000481302 0.000629976 0.413600953 0.000600953 0.000786588
21 0.5398 0.3165 0.31714929 0.00064928 0.00084985 0.317157522 0.000657522 0.00086063 0.317221449 0.000721449 0.000944305
22 0.5521 0.2120 0.21200005 4.7598𝐸 −08 6.2301𝐸 −08 0.212005096 5.0964𝐸 −06 6.6708𝐸 −03 0.212016349 1.6348𝐸 −09 2.1398𝐸 −05
23 0.5633 0.1035 0.10261938 0.00088061 0.00115264 0.102614335 0.000885665 0.001159247 0.10259324 0.00090676 0.001186858
24 0.5736 −0.0100 −0.0093122 0.00068782 0.00090028 −0.0093309 0.00066908 0.000875759 −0.0093544 0.000645529 0.000844933
25 0.5833 −0.1230 −0.1243674 0.00136739 0.00178977 −0.1244008 0.001400854 0.001833579 −0.1243917 0.001391753 0.001821666
26 0.5900 −0.2100 −0.2090994 0.00090056 0.00117874 −0.2091429 0.00085706 0.001121806 −0.2090867 0.000913294 0.001195411

Table 10
Estimated model parameters for SDM RTC France solar cell, obtained by various optimization methods.
Techniques Year 𝐼𝑝ℎ 𝐼𝑠𝑑 𝑅𝑠 𝑅𝑠ℎ 𝑛 RMSE

MRFO 2020 0.7608 3.1077𝐸 −07 0.03655 52.9026 1.4773 7.7301𝐸 −04
SMA [21] 2020 0.7607 0.3231 0.0363 53.7149 1.4811 9.8582𝐸 −04
COA [20] 2020 0.7608 0.3084 0.0366 52.8267 1.4765 7.7547𝐸 −04
TLO [17] 2019 0.7609 0.3329 0.0354 56.0305 1.4847 9.8609𝐸 −04
ISCA [18] 2019 0.7608 0.3230 0.0363 53.7185 1.4811 9.8602𝐸 −04
ImCSA [16] 2018 0.7608 0.3230 0.0364 53.7185 1.4817 9.8602𝐸 −04
MSSO [9] 2017 0.7607 0.3235 0.0364 53.7423 1.4812 9.8607𝐸 −04
ABC [8] 2014 0.7605 0.3599 0.0360 52.1479 1.4959 9.8615𝐸 −04
CPSO [10] 2011 0.7607 0.4 0.0354 59.012 1.5033 1.3900𝐸 −03

Fig. 10. Comparison between the experimental data and simulated data of RTC France solar cell (TDM) (a) I–V characteristics; (b) P–V characteristics.

that the proposed MRFO performs the best extraction in obtaining the optimal value of the variables. The I–V and P–V characteristics
between the experimental and calculated data acquired from MRFO are shown in Fig. 13.
The I–V characteristic of MRFO, the IAE and RE between the tested and calculated data are presented. The results showed that the
calculated data obtained from MRFO significantly agree with the experimental data in SDM, meaning that the parameters extracted
by the MRFOs are very accurate.

12
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Fig. 11. Error index values of the simulated current data and the experimental current data of RTC France solar cell (TDM) (a) IAE values; (b) RE values.

Fig. 12. The Best cost (RMSE) with respect to iteration numbers of proposed algorithm and comparison algorithms for RTC France PV cell of three diode model.

Table 11
Estimated model parameters for DDM RTC France solar cell, obtained by various optimization methods.
Techniques Year 𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝑅𝑠 𝑅𝑠ℎ 𝑛1 𝑛2 RMSE

MRFO 2020 0.7608 2.2619𝐸 −07 3.9314𝐸 −07 0.0368 54.2188 1.4515 1.8760 7.6046𝐸 −04
HS [22] 2012 0.7618 0.1255 0.2547 0.0354 46.8269 1.4944 1.4998 1.2600𝐸 −03
ABC [8] 2015 0.7608 0.1927 0.9999 0.0369 55.9335 1.4380 1.9837 9.8387𝐸 −04
MSSO [9] 2017 0.7607 0.2349 0.6716 0.03668 55.7147 1.4543 1.9953 9.8281𝐸 −04
ImCSA [16] 2018 0.7607 0.2259 0.7473 0.0367 55.4827 1.4515 2.0000 9.8249𝐸 −04
TLO [17] 2019 0.7615 0.2975 0.4699 0.0348 53.4254 1.6811 1.4826 9.8612𝐸 −04
ISCA [18] 2019 0.7607 0.7493 0.2259 0.0367 55.4854 2.0000 1.4510 9.8237𝐸 −04
COA [20] 2020 0.7607 0.2447 0.3802 0.0369 53.5129 1.4563 1.9899 7.6480𝐸 −04
SMA [21] 2020 0.7607 0.7487 0.2265 0.0367 55.7145 2.0000 1.4546 9.8148𝐸 −04

Fig. 14 shows the convergence curves of different optimization algorithms for determining the parameters of the Photowatt-
PWP201 PV module for SDM. As shown in Fig. 14, the proposed MRFO algorithm achieves the best value while the remaining
algorithms do not reach the minimum best value of RMSE as the proposed MRFO algorithm.

13
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Table 12
Estimated model parameters for TDM RTC France solar cell, obtained by various optimization methods.
Run Year 𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝐼𝑠𝑑3 𝑅𝑠 𝑅𝑠ℎ 𝑛1 𝑛2 𝑛3 RMSE

MRFO 2020 0.0.7607 2.6318𝐸 −07 2.3678𝐸 −07 2.5070𝐸 −07 0.0367 55.0167 1.9131 1.4551 1.9965 7.6083𝐸 −04
SMA [21] 2020 0.7608 2.64𝐸 −07 5.51𝐸 −12 5.81𝐸 −07 0.0368 54.9152 1.4609 1.1467 2.0209 9.80143𝐸 −04
TLO [17] 2019 0.7608 2.54𝐸 −07 4.56𝐸 −08 1.48𝐸 −06 0.03671 55.3144 1.4602 1.7408 2.2514 9.8613𝐸 −04
OBWOA [30] 2015 0.7608 0.2353 0.2213 0.4573 0.0366 55.4448 1.4543 2 2 9.8249𝐸 −04
ABC [8] 2015 0.7608 1.66𝐸 −07 1.98𝐸 −07 3.48𝐸 −08 0.0364 53.7220 1.4873 1.4775 4.4665 9.8452𝐸 −04

Table 13
Optimal parameters for single-diode model of Photowatt-PWP201 PV module.
Run 𝐼𝑝ℎ 𝐼𝑠𝑑 𝑅𝑠 𝑅𝑠ℎ 𝑛 RMSE

The best run 1.0324 2.5139𝐸 −06 1.2395 748.8128 47.4240 2.0401𝐸 −03
The worst run 1.0332 2.5633𝐸 −06 1.2359 708.3053 47.4977 2.0633𝐸 −03
Min 1.0315 2.3798𝐸 −06 1.2284 708.3053 47.2249 2.0401𝐸 −03
Mean 1.0323 2.5419𝐸 −06 1.2385 756.4881 47.4625 2.0441𝐸 −03
Max 1.0332 2.7584𝐸 −06 1.2463 822.0781 47.7656 2.0633𝐸 −03
STD 2.9795𝐸 −04 8.8812𝐸 −08 4.0531𝐸 −03 25.1304 0.1264 5.9016𝐸 −06

Table 14
Comparison results for single-diode model of Photowatt-PWP201 PV module.
Parameters WOA GWO MVO SCA MFO ALO MRFO
𝐼ℎ𝑝 (A) 1.3135 1.0412 1.3666 1.0547 1.0384 1.0282 1.0323
𝐼𝑠ℎ (μA) 0 2.9876𝐸 −06 1.5427𝐸 −06 3.1606𝐸 −06 2.0000𝐸 −07 2.1239𝐸 −07 2.5050𝐸 −06
𝑅𝑠ℎ (ohm) 16.9229 401.4866 16.1197 1111.4314 361.3808 735.7503 751.4096
𝑅𝑠 (ohm) 0.0622 1.1807 0.6021 1.1730 1.4950 1.5313 1.2402
𝑛 49.4231 50 15.8530 50 41.2287 41.3515 47.4107
𝐑𝐌𝐒𝐄 0.2839 0.0042 0.2839 0.0201 0.0052 0.0066 2.0419𝐸 −03
𝐌𝐀𝐄 0.2458 0.0021 0.2465 0.0186 0.0052 0.0054 1.6740𝐸 −03
𝐌𝐑𝐄 0.2376 0.0020 0.2383 0.0180 0.0041 0.0052 1.6182𝐸 −03
Elapsed-Time 0.000199 s 0.000438 s 0.000281 s 0.000159 s 0.000147 s 0.000197 s 0.000048 s

Table 15
Optimal parameters for double-diode model of Photowatt-PWP201 PV module.
Run 𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝑅𝑠 𝑅𝑠ℎ 𝑛1 𝑛2 RMSE

Best run 1.0325 1.0218𝐸 −06 1.4658𝐸 −06 1.2406 736.2174 48.9581 49.4584 2.0412𝐸 −03
Worst run 1.0337 1.0733𝐸 −06 1.4776𝐸 −07 1.3239 571.5418 46.5046 48.4363 2.5420𝐸 −03
Min 1.0310 1.0424𝐸 −07 1.4776𝐸 −07 1.2291 551.5075 46.3353 45.7095 2.0412𝐸 −03
Mean 1.0328 1.0971𝐸 −06 1.0182𝐸 −06 1.2657 693.7174 48.6671 48.3724 2.1469𝐸 −03
Max 1.0353 2.1622𝐸 −06 2.3897𝐸 −06 1.3239 860.5923 49.9810 49.8987 2.5421𝐸 −03
STD 7.7305𝐸 −04 4.5214𝐸 −07 5.0548𝐸 −07 0.0217 66.1582 1.0588 1.1408 1.2257𝐸 −04

Fig. 13. Comparison between the experimental data and simulated data of Photowatt-PWP201 PV module (SDM) (a) I–V characteristics; (b) P–V characteristics.

14
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Fig. 14. The Best cost (RMSE) with respect to iteration numbers of proposed algorithm and comparison algorithms for Photowatt-PWP201 PV module of single
diode model.

Table 16
Comparison results for double-diode model of Photowatt-PWP201 PV module.
Parameters WOA GWO MVO SCA MFO ALO MRFO
𝐼ℎ𝑝 (A) 1.0439 1.0408 1.0485 1.0459 1.0237 1.0267 1.0321
𝐼𝑠ℎ1 (μA) 2.1386𝐸 −06 2.8435𝐸 −09 0 3.0721𝐸 −06 6.3687𝐸 −07 2.7924𝐸 −06 9.1128𝐸 −07
𝐼𝑠ℎ2 (μA) 0 2.9225𝐸 −06 6.3312𝐸 −07 0 2.0000𝐸 −07 2.1348𝐸 −07 1.6226𝐸 −06
𝑅𝑠ℎ (ohm) 639.0296 386.5405 363.3248 654.1636 2000 863.0756 764.8141
𝑅𝑠 (ohm) 1.7256 1.1832 1.5382 1.3961 1.4121 1.0805 1.2394
𝑛1 48.7719 40.3283 11.7393 50 44.7228 50 49.4954
𝑛2 48.4604 50 44.5812 1 50 49.5812 49.2153
𝐑𝐌𝐒𝐄 0.0287 0.0045 0.0114 0.0132 0.0056 0.0097 2.0429𝐸 −03
𝐌𝐀𝐄 0.0227 0.0036 0.0092 0.0120 0.0047 0.0076 1.6991𝐸 −03
𝐌𝐑𝐄 0.0220 0.0035 0.0089 0.0116 0.0046 0.0074 1.6425𝐸 −03
Elapsed-Time 0.000179 s 0.000336 s 0.000236 s 0.000148 s 0.000207 s 0.000328 0.000052 s

Table 17
Optimal parameters for three-diode model of Photowatt-PWP201 PV module.
Run 𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝐼𝑠𝑑3 𝑅𝑠 𝑅𝑠ℎ 𝑛1 𝑛2 𝑛3 RMSE

Best run 1.0323 1.1787𝐸 −06 6.4164𝐸 −07 7.3192𝐸 −07 1.2384 752.5965 49.4518 48.7769 49.7246 2.0411𝐸 −03
Worst run 1.0331 5.0177𝐸 −07 6.5189𝐸 −08 1.2399𝐸 −06 1.2851 639.7905 48.2474 44.3495 48.3924 2.1915𝐸 −03
Min 1.0314 1.4123𝐸 −07 6.5189𝐸 −08 3.1708𝐸 −08 1.2329 629.4852 47.6564 44.3495 44.8762 2.0411𝐸 −03
Mean 1.0327 7.4257𝐸 −07 7.9082𝐸 −07 7.1796𝐸 −07 1.2552 706.7433 48.8661 48.6125 48.7647 2.0800𝐸 −03
Max 1.0336 1.7144𝐸 −06 1.8961𝐸 −06 1.4133𝐸 −06 1.2851 794.0224 49.9243 49.9804 49.9113 2.1915𝐸 −03
STD 5.2496𝐸 −04 4.0310𝐸 −07 4.8283𝐸 −07 3.5452𝐸 −07 0.0125 43.1385 0.6160 1.2976 1.0161 3.8507𝐸 −05

6.2.2. Results of the double-diode model

There are seven unknown parameters in DDM that must be estimated. The range of these parameters is listed in Table 2. The
optimization results obtained from running the proposed algorithm 30 times with the best and worst run are presented in Table 15.
As shown in Table 16, the proposed MRFO achieves the optimal value for each unknown parameter of the DDM and the objective
function (RMSE) in the minimum with a value of 2.0412𝐸 −03 that is the best for all.
The extracted parameters for PV modules are presented in Table 16 with comparable results of methods after running each
algorithm under the same conditions. Table 16 shows that the proposed MRFO achieves the best RMSE value and lower MAE and
MRE values. It means that the proposed MRFO is the best in estimating the optimal value of the parameters.
A curve-fitting procedure similar to that of the case study, IAE, and RE for each measurement is calculated. The comparison
results showed that the proposed MRFO results outperform other existing algorithms’ results. Methods related to IAE and the sum
of RE. The I–V and P–V characteristics for the PV module depend on the estimated parameters are displayed in Fig. 15.
Fig. 16 shows the convergence curves of all optimization algorithms for determining the parameters of the Photowatt-PWP201
PV module for DDM. It shows that the proposed MRFO algorithm is the best since it reaches to the minimum (best) value of RMSE,
while the remaining algorithms do not reach the best value of RMSE.

15
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Fig. 15. Comparison between the experimental data and simulated data of Photowatt-PWP201 PV module (DDM) (a) I–V characteristics; (b) P–V characteristics.

Fig. 16. The Best cost (RMSE) with respect to iteration numbers of proposed algorithm and comparison algorithms for Photowatt-PWP201 PV module of double
diode model.

6.2.3. Results of the three-diode model

This section examines the validity of the proposed method for TDM. Table 17 shows the values of 𝑛1 , 𝑛2 , 𝑛3 , 𝑅𝑠 , 𝑅𝑆𝐻 , 𝐼𝑝ℎ , 𝐼𝑠ℎ1 ,
𝐼𝑠ℎ2 , 𝐼𝑠ℎ3 , RMSE, Min, Mean, Max and standard deviation. After the analysis of these results, the proposed algorithm shows the best
performance in solving the optimization problem. Thus, it provides the least error values.
Table 18 presents nine unknown parameters of TDM, RMSE, MAE, and MRE obtained from WOA, GWO, MVO, SCA, MFO, ALO,
and MRFO optimization algorithms after running these algorithms under standard test conditions. It shows that the proposed MRFO
provides a much lower RMSE than other techniques. Thus, in terms of precision, MRFO is the best technique.
Additionally, the I–V and the P–V characteristics between the tested and calculated data obtained from MRFO are displayed in
Fig. 17. From Fig. 17, we observe that the tested and calculated curves are the same. The comparison of the results of the proposed
MRFO with other algorithms showed that the MRFO algorithm is significantly better than other competitive methods based on IAE
and RMSE.

16
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Table 18
Comparison results for three-diode model of Photowatt-PWP201 PV module.
Parameters WOA GWO MVO SCA MFO ALO MRFO
𝐼ℎ𝑝 (A) 1.0254 1.0359 1.0285 1.0506 1.0374 1.0288 1.0326
𝐼𝑠ℎ1 (μA) 0 1.6649𝐸 −06 1.2276𝐸 −06 2.0001𝐸 −07 2.1291𝐸 −06 2.0021𝐸 −07 1.4319𝐸 −06
𝐼𝑠ℎ2 (μA) 2.2828𝐸 −07 3.8189𝐸 −07 1.2276𝐸 −07 6.5057𝐸 −07 2.0012𝐸 −07 3.0757𝐸 −07 9.1012𝐸 −08
𝐼𝑠ℎ3 (μA) 0 6.7447𝐸 −09 0 2.0000𝐸 −07 2.0011𝐸 −07 2.5439𝐸 −07 5.1877𝐸 −07
𝑅𝑠ℎ (ohm) 1912.0007 454.1251 1994.7253 528.3903 2000 836.8726 694.4557
𝑅𝑠 (ohm) 1.6753 1.2448 1.5533 1.1733 2 1.4543 1.2674
𝑛1 23.8170 49.9451 49.4054 50 50 42.1396 48.8703
𝑛2 41.5406 45.6563 43.0818 50 41.6479 49.2224 49.9221
𝑛3 41.8669 45.2898 21.1331 50 50 47.3679 47.5214
𝐑𝐌𝐒𝐄 0.0118 0.0033 0.0121 0.0143 0.0305 0.0049 2.0993𝐸 −03
𝐌𝐀𝐄 0.0091 0.0024 0.0091 0.0139 0.0232 0.0040 1.7842𝐸 −03
𝐌𝐑𝐄 0.0088 0.0024 0.0088 0.0134 0.0225 0.0039 1.7247𝐸 −03
Elapsed-Time 0.000162 s 0.000165 s 0.000145 s 0.000218 s 0.000148 s 0.000204 s 0.000053 s

Fig. 17. Comparison between the experimental data and simulated data of Photowatt-PWP201 PV module (TDM) (a) I–V characteristics; (b) P–V characteristics.

Fig. 18. The Best cost (RMSE) with respect to iteration numbers of the proposed algorithm and comparison to other algorithms for Photowatt-PWP201 PV
module of three diode model.

Fig. 18 shows the convergence curves of different optimization algorithms for determining the parameters of the Photowatt-
PWP201 PV module for TDM. From Table 18 and Fig. 18, the proposed MRFO algorithm is the best one that reaches to coordinates,
while the remaining algorithms do not reach the minimum best value of RMSE as the proposed MRFO algorithm.

17
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Fig. 19. Scatter plot all 30 run and show the optimal and the worst results for Mono-crystalline_ SM55_1000W at 25 ◦ C of SDM.

Fig. 20. Scatter plot all 30 run and show the optimal and the worst results for Mono-crystalline_ SM55_800W at 25 ◦ C of SDM.

6.3. Case study 3

In this case, mono-crystalline solar cell SM55 is selected as a sample for investigating the problem of parameter extraction
determined based on the experimental data measured at the laboratory. Four levels of irradiance: 400, 600, 800, and 1000 W∕m2
at different temperatures (25 ◦ C, 40 ◦ C and 60 ◦ C) are investigated. SDM, DDM and TDM models are used to mimic this cell. MRFO,
MVO, SCA, WOA, and GWO algorithms applied to the three models to estimate their unknown variables are presented in Tables 19,
20, 21, 22, 23, and 24. From these tables, the implementation of the MRFO can obtain the global optimal value for the unknown
parameters of the PV in all three models (single, double, and three-diodes). Thus, MRFO achieves the least error for all models.
A detailed comparison of the proposed algorithm and other algorithms is listed in Table 25. Table 25 contains the least RMSE,
MAE, and MRE. From Table 25, we obtain that the proposed algorithm achieves the best RMSE in all cases of different irradiance
and temperatures in the three models. It means that the MRFO is the best among all the models. Obtaining 1000 W∕m2 and 25 ◦ C
in the SDM, the RMSE is 2.9𝐸 −03 , but in WOA is 0.0728, SCA is 0.0500, GWO is 0.0136, and MVO is 0.0268.
Figs. 19, 20 and 21 show the best (optimal) and worst solutions of SDM using the proposed MRFO at different levels of irradiance
and temperatures (1000 W∕m2 at 25 ◦ C, 800 W∕m2 at 25 ◦ C and 1000 W∕m2 at 40 ◦ C). From these figures, we obtain that the
most solution of 30 runs near the optimal solution (best), and there are solutions of 30 runs close to the worst. It means that the
proposed MRFO achieves the optimal value in every run under different irradiance and temperatures.

18
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Fig. 21. Scatter plot all 30 run and show the optimal and the worst results for Mono-crystalline_ SM55_1000W at 40 ◦ C of SDM.

Table 19
The extracted parameters for SDM, DDM and TDM for Mono-cystalline solar module by MRFO and other algorithms at irradiance of 400 W/m2 and temperature
of 25 ◦ C.
Parameters
𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝐼𝑠𝑑3 𝑅𝑠 𝑅𝑠ℎ 𝑎1 𝑎2 𝑎3
SMD
MRFO 1.3849 2.9023𝐸 −07 – – 0.2189 498.8072 1.4381 – –
WOA 3.6197 4.5014𝐸 −06 – – 0.3556 334.1200 1.7724 – –
SCA 3.5681 7.5689𝐸 −07 – – 0.1707 500 1.5460 – –
GWO 3.5481 2.2179𝐸 −06 – – 0.1008 498.8594 1.6635 – –
MVO 3.5468 2.6107𝐸 −06 – – 0.1024 499.7114 1.6851 – –
DDM
MRFO 3.5179 2.3047𝐸 −06 3.4982𝐸 −08 – 0.3099 498.1876 2.7292 1.2662 –
WOA 3.7091 1.2109𝐸 −06 1.7384𝐸 −06 – 0.1049 203.1969 2.9912 1.6450 –
SCA 3.5073 1.1710𝐸 −06 1.0911𝐸 −06 – 0.1214 493.2795 1.5868 3.4815 –
GWO 3.5247 2.0618𝐸 −06 3.2414𝐸 −07 – 0.1539 499.51805 3.5737 1.4511 –
MVO 3.5199 4.2198𝐸 −06 1.6099𝐸 −07 – 0.1543 500 3.1852 1.3869 –
TDM
MRFO 3.5303 1.0330𝐸 −07 2.6171𝐸 −06 1.7511𝐸 −06 0.3249 499.4888 1.2385 3.9867 3.2617
WOA 3.5968 5.8571𝐸 −06 2.2932𝐸 −06 2.1417𝐸 −06 0.1581 373.0926 1.6772 1.7690 3.0351
SCA 3.6859 1.9354𝐸 −07 5.4314𝐸 −08 5.0000𝐸 −12 0.1062 233.9192 1.2906 3.1115 1
GWO 3.5440 4.1263𝐸 −06 2.4370𝐸 −07 1.3956𝐸 −06 0.1388 500 2.5724 1.9145 1.6676 1.4743
MVO 3.5423 2.7995𝐸 −06 5.4201𝐸 −07 1.4031𝐸 −06 0.1001 499.9585 2.3922 1.9060 1.4753

Table 20
The extracted parameters for SDM, DDM and TDM for Mono-cystalline solar module by MRFO and other algorithms at irradiance of 600 W/m2 and temperature
of 25 ◦ C.
Parameters
𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝐼𝑠𝑑3 𝑅𝑠 𝑅𝑠ℎ 𝑎1 𝑎2 𝑎3
SMD
MRFO 3.4523 1.5148𝑒−07 – – 0.3530 462.0645 1.3865 – –
WOA 3.6573 3.4119𝐸 −06 – – 0.2498 124.6874 1.7217 – –
SCA 3.5290 1.7195𝐸 −06 – – 0.2139 406.8945 1.6328 – –
GWO 3.4897 3.3436𝐸 −06 – – 0.1268 478.5276 1.7042 – –
MVO 3.5172 1.3472𝐸 −06 – – 0.1000 298.8463 1.5948 – –
DDM
MRFO 3.4721 1.8321𝐸 −07 2.4685𝐸 −07 – 0.2619 499.9532 3.4563 1.4248 –
WOA 3.5849 4.8901𝐸 −06 1.4319𝐸 −06 – 0.3175 193.9336 2.0285 1.6375 –
SCA 3.5520 5.0000𝐸 −12 3.6411𝐸 −06 – 0.2841 500.0000 2.9122 1.7012 –
GWO 3.5057 1.5235𝐸 −06 1.5932𝐸 −06 – 0.1999 365.1534 1.6108 2.8748 –
MVO 3.4925 3.2871𝐸 −09 5.8940𝐸 −06 – 0.1000 499.9165 3.9989 1.7804 –
TDM
MRFO 3.4845 2.3292𝐸 −06 2.9013𝐸 −06 6.2273𝐸 −07 0.2567 498.1860 3.5310 3.6551 1.3896
WOA 3.6386 2.73506𝐸 −06 2.0525𝐸 −06 1.9478𝐸 −07 0.1978 130.6967 1.5513 2.5187 3.8821
SCA 3.5569 3.3862𝐸 −06 1.5403𝐸 −06 6.3256𝐸 −10 0.2907 309.4875 2.3776 1.5200 1
GWO 3.4917 1.6430𝐸 −06 1.1342𝐸 −06 1.9616𝐸 −06 0.2088 497.7203 2.6080 1.9479 1.5078
MVO 3.4906 2.1855𝐸 −06 2.0986𝐸 −06 2.3992𝐸 −06 0.1269 474.8412 2.2409 1.5157 2.0141

19
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Table 21
The extracted parameters for SDM, DDM and TDM for Mono-cystalline solar module by MRFO and other algorithms at irradiance of 800 W/m2 and temperature
of 25 ◦ C.
Parameters
𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝐼𝑠𝑑3 𝑅𝑠 𝑅𝑠ℎ 𝑎1 𝑎2 𝑎3
SDM
MRFO 3.4509 1.6526𝐸 −07 – – 0.3496 494.3648 1.3933 – –
WOA 3.5174 4.8375𝐸 −06 – – 0.3226 236.1436 1.7563 –
SCA 3.4453 4.3385𝐸 −07 – – 0.10000 198.8733 1.4734 – –
GWO 3.4843 8.2134𝐸 −07 – – 0.2290 277.7882 1.5390 – –
MVO 3.5035 5.5002𝐸 −06 – – 0.4723 458.6703 1.7827 – –
DDM
MRFO 3.4584 9.7555𝐸 −07 5.0516𝐸 −07 – 0.2494 491.0692 3.3736 1.4892 –
WOA 3.4682 4.6678𝐸 −06 8.8282𝐸 −07 – 0.2129 411.7262 2.5784 1.5463 –
SCA 3.4297 1.6633𝐸 −06 2.9995𝐸 −11 – 0.1797 500 1.6072 4 –
GWO 3.4790 3.6288𝐸 −06 1.2147𝐸 −06 – 0.1107 332.3092 1.7053 3.2808 –
MVO 3.4767 5.9936𝐸 −06 4.7322𝐸 −06 – 0.1000 499.9869 1.8078 1.9600 –
TDM
MRFO 3.4693 1.0739𝐸 −08 4.5573𝐸 −07 9.7437𝐸 −07 0.2630 499.9999 3.2087 3.6030 1.4300
WOA 3.5087 4.3381𝐸 −06 3.2144𝐸 −06 5.6021𝐸 −06 0.2084 249.7644 1.5921 3.4671 3.0919
SCA 3.5173 1.4046𝐸 −06 8.0368𝐸 −12 5.0000𝐸 −12 0.1764 387.3931 1.4616 1.9848 2.1352
GWO 3.4781 2.8944𝐸 −06 2.5583𝐸 −06 4.4223𝐸 −06 0.1268 385.6046 3.5207 3.7332 1.5898
MVO 3.4813 5.0000𝐸 −12 1.4089𝐸 −07 2.8699𝐸 −06 0.1601 329.9890 3.9890 3.7996 1.5407

Table 22
The extracted parameters for SDM, DDM and TDM for Mono-cystalline solar module by MRFO and other algorithms at irradiance of 1000 W/m2 and temperature
of 25 ◦ C.
Parameters
𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝐼𝑠𝑑3 𝑅𝑠 𝑅𝑠ℎ 𝑎1 𝑎2 𝑎3
SDM
MRFO 3.4525 1.8041𝐸 −07 – – 0.3455 453.2439 1.4008 – –
WOA 3.4811 3.9143𝐸 −06 – – 0.3958 405.0698 1.7306 – –
SCA 3.4935 4.3819𝐸 −06 – – 0.2874 390.6811 1.7388 – –
MVO 3.4828 3.2618𝐸 −06 – – 0.2163 240.4361 1.6933 – –
DDM
MRFO 3.4518 4.1468𝐸 −06 8.8498𝐸 −08 – 0.3660 458.7830 2.6328 1.3446 –
WOA 3.4451 2.0975𝐸 −06 2.6805𝐸 −06 – 0.1263 447.2563 2.5599 1.7890 –
SCA 3.4032 5.9680𝐸 −07 2.36311𝐸 −10 – 0.4699 292.1491 4 1 –
MVO 3.4561 5.0000𝐸 −12 1.5926𝐸 −06 – 0.2823 497.3038 1.0146 1.6129 –
TDM
MRFO 3.4525 4.1905𝐸 −06 1.8044𝐸 −07 2.2118𝐸 −06 0.3421 484.0723 2.9438 1.3339 2.6032
WOA 3.4617 3.0526𝐸 −06 2.5675𝐸 −06 2.9025𝐸−06 0.1890 461.8038 1.8463 1.6025 4
SCA 3.4539 2.0984𝐸 −10 9.4911𝐸 −12 7.3008𝐸 −10 0.5000 270.2985 1.8284 4 1
MVO 3.4762 1.6684𝐸 −06 2.4822𝐸 −06 8.3209𝐸 −07 0.1613 204.8531 1.9741 1.5833 3.6046

Table 23
The extracted parameters for SDM, DDM and TDM for Mono-cystalline solar module by MRFO and other algorithms at temperature of 40 ◦ C and irradiance of
1000 W/m2 .
Parameters
𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝐼𝑠𝑑3 𝑅𝑠 𝑅𝑠ℎ 𝑎1 𝑎2 𝑎3
SDM
MRFO 3.4612 2.5772𝐸 −07 – – 0.3022 379.4510 1.4273 – –
WOA 3.5061 1.6597𝐸 −06 – – 0.2233 144.4335 1.6142 – –
GWO 3.4696 5.1617𝐸 −06 – – 0.1532 369.9749 1.7490 – –
MVO 3.4695 3.7602𝐸 −06 – – 0.1682 484.3936 1.7069 – –
DDM
MRFO 3.4596 3.3939𝐸 −06 3.3354𝐸 −07 – 0.2938 444.2354 3.8519 1.4501 –
WOA 3.4735 4.5663𝐸 −06 1.1908𝐸 −07 – 0.2292 390.7771 1.7373 2.9535
GWO 3.4656 4.2327𝐸 −06 1.6286𝐸 −08 – 0.1844 482.8334 1.7244 1.7397 –
MVO 3.4711 8.1231𝐸 −07 3.9618𝐸 −06 – 0.1660 401.7482 1.8498 1.7247 –
TDM
MRFO 3.4704 3.5633𝐸 −06 4.7653𝐸 −07 4.6713𝐸 −06 0.3089 441.7564 3.9821 1.4055 2.3280
WOA 3.4790 3.8502𝐸 −06 2.1271𝐸 −06 1.7892𝐸 −06 0.2786 455.8353 2.6727 1.5585 1.9630
GWO 3.4978 8.2350𝐸 −07 4.6704𝐸 −06 5.5562𝐸 −06 0.1512 208.4792 1.8477 2.1660 1.6728
MVO 3.5239 3.8898𝐸 −07 5.8571𝐸 −06 1.6471𝐸 −06 0.2453 110.6757 1.3930 2.8608 1.8687

7. Conclusions and future work

Exact modeling of solar PV is necessary before designing the entire system. In this study, the problem of solar cell parameter
identification is addressed using the MRFO algorithm. The optimized parameters of single, two, and three-diode models play a
significant role in accurate modeling. In this paper, a new optimization algorithm called MRFO has been proposed to effectively
extract variables for the solar cell models. The estimation of three PV modules confirms the robustness of the proposed method.
The proposed MRFO extracted the optimal value of the unknown variables of SDM, DDM, and TDM with the best RMSE, MAE, and
MRE values. This algorithm was compared with six other algorithms. Besides, MRFO can produce more accurate and robust results,
achieve a higher success rate, and converge faster than other algorithms. The presented statistical analysis explains the high-quality
of the estimated results. Solutions illustrate that there exists a good balance between the calculated and experimental I–V curves
using the proposed MRFO. The obtained results are used to mimic the performance of single, double, and three-diode solar cell

20
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

Table 24
The extracted parameters for SDM, DDM and TDM for Mono-cystalline solar module by MRFO and other algorithms at temperature of 60 ◦ C and irradiance of
1000 W/m2 .
Parameters
𝐼𝑝ℎ 𝐼𝑠𝑑1 𝐼𝑠𝑑2 𝐼𝑠𝑑3 𝑅𝑠 𝑅𝑠ℎ 𝑎1 𝑎2 𝑎3
SDM
MRFO 3.4527 5.9802𝐸 −07 – – 0.2516 499.1255 1.5061 – –
WOA 3.4908 2.0838𝐸 −06 – – 0.3830 202.8827 1.6531 – –
GWO 3.4797 5.5022𝐸 −06 – – 0.1713 420.7752 1.76263 – –
MVO 3.4532 3.6052𝐸 −07 – – 0.2854 469.9150 1.4584 – –
DDM
MRFO 3.4546 2.2135𝐸 −06 6.0316𝐸 −07 – 0.2507 465.1915 3.5266 1.5069 –
WOA 3.4714 5.5721𝐸 −06 4.4049𝐸 −06 – 0.1362 372.8683 1.7657 2.6349 –
GWO 3.4867 5.2406𝐸 −06 3.2859𝐸 −06 – 0.1462 217.8358 1.7580 2.8931 –
MVO 3.4819 3.0978𝐸 −07 5.9800𝐸 −06 – 0.2586 165.1623 1.4452 3.1584 –
TDM
MRFO 3.4643 9.1133𝐸 −10 1.4166𝐸 −06 8.7488𝐸 −08 0.2499 499.9922 3.4505 1.5100 3.9969
WOA 3.5451 4.0844𝐸 −06 5.4084𝐸 −06 1.7595𝐸 −06 0.3804 113.2699 2.6861 1.6807 3.6578
GWO 3.4731 5.3466𝐸 −07 1.0735𝐸 −06 4.3894𝐸 −06 0.2573 344.2823 2.1599 1.4841 2.3565
MVO 3.4896 3.5238𝐸 −06 4.9209𝐸 −06 1.7388𝐸 −06 0.2576 252.0685 2.2259 4 1.5359

Table 25
The extracted parameters for SDM, DDM and TDM for Mono-cystalline solar module by MRFO and other algorithms at different levels of irradiance and
temperature.
MRFO WOA GWO MVO
RMSE MAE MRE RMSE MAE MRE RMSE MAE MRE RMSE MAE MRE

400 W/m2 SDM 0.0113 0.0094 0.0068 0.0521 0.0398 0.0288 0.0244 0.0197 0.0143 0.0257 0.0205 0.0148
25 ◦ C DDM 0.0114 0.0095 0.0069 0.0566 0.0454 0.0339 0.0145 0.0118 0.0086 0.0134 0.0114 0.0082
TDM 0.0116 0.0096 0.0069 0.0388 0.0305 0.0221 0.0173 0.0143 0.0104 0.0166 0.0136 0.0098

600 W/m2 SDM 0.0102 0.0061 0.0030 0.0761 0.0625 0.0302 0.0205 0.0161 0.0078 0.0207 0.0169 0.0082
25 ◦ C DDM 0.0102 0.0062 0.0030 0.0610 0.0480 0.0231 0.0232 0.0188 0.0092 0.0246 0.0193 0.0093
TDM 0.0103 0.0060 0.0029 0.0586 0.0482 0.0232 0.0145 0.0107 0.0052 0.0127 0.0097 0.0047

800 W/m2 SDM 0.0062 0.0046 0.0017 0.0643 0.0458 0.0166 0.0156 0.0125 0.0045 0.0239 0.0187 0.0068
25 ◦ C DDM 0.0063 0.0047 0.0017 0.0098 0.0079 0.0029 0.0156 0.0125 0.0045 0.0876 0.0581 0.0211
TDM 0.0066 0.0052 0.0019 0.0261 0.0211 0.0077 0.0173 0.0143 0.0104 0.0142 0.0811 0.0043

1000 W/m2 SDM 0.0029 0.0023 6.5604𝐸 −04 0.0728 0.0476 0.0138 0.0136 0.0094 0.0027 0.0268 0.0211 0.0061
25 ◦ C DDM 0.0031 0.0023 6.5908𝐸 −04 0.0391 0.0217 0.0063 0.0286 0.0197 0.0057 0.0187 0.0122 0.0035
TDM 0.0044 0.0030 8.7227𝐸 −04 0.0210 0.0159 0.0046 0.0217 0.0157 0.0046 0.0301 0.0225 0.0065

1000 W/m2 SDM 0.0049 0.0035 0.0010 0.0302 0.0235 0.0067 0.0221 0.0158 0.0045 0.0179 0.0149 0.0043
40 ◦ C DDM 0.0050 0.0035 0.0010 0.0361 0.0496 0.0072 0.0210 0.0146 0.0042 0.0207 0.0207 0.0046
TDM 0.0049 0.0041 0.0012 0.0211 0.0151 0.0044 0.0251 0.0193 0.0055 0.0280 0.0226 0.0065

1000 W/m2 SDM 0.0113 0.0065 0.0018 0.0843 0.0545 0.0156 0.0292 0.0234 0.0067 0.0123 0.0067 0.0019
60 ◦ C DDM 0.0114 0.0064 0.0018 0.0237 0.0184 0.0052 0.0297 0.0225 0.0064 0.0166 0.0124 0.0035
TDM 0.0116 0.0069 0.0020 0.0957 0.0694 0.0198 0.0124 0.0081 0.0023 0.0213 0.0160 0.0046

models. Thus, the proposed MRFO can be a good candidate for solving the parameter extraction problem in solar cell systems.
Future work may extend the use of the MRFO algorithm to solve various optimization problems in other fields of applications.

CRediT authorship contribution statement

Essam H. Houssein: Conceptualization, Investigation, Methodology, Supervision, Writing - review & editing. Gamela Nageh
Zaki: Data curation, Formal analysis, Investigation, Software, Validation, Visualization, Writing - review & editing. Ahmed A. Zaki
Diab: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing -
review & editing. Eman M.G. Younis: Data curation, Formal analysis, Investigation, Supervision, Visualization, Writing - review &
editing.

Declaration of competing interest

No author associated with this paper has disclosed any potential or pertinent conflicts which may be perceived to have impending
conflict with this work. For full disclosure statements refer to https://doi.org/10.1016/j.compeleceng.2021.107304.

Appendix A. Supplementary data

Supplementary material related to this article can be found online at https://doi.org/10.1016/j.compeleceng.2021.107304.

References

[1] Lun S-x, Du C-j, Sang J-s, Guo T-t, Wang S, Yang G-h. An improved explicit I–v model of a solar cell based on symbolic function and manufacturer’s
datasheet. Sol Energy 2014;110:603–14.
[2] Chan DS, Phang JC. Analytical methods for the extraction of solar-cell single-and double-diode model parameters from IV characteristics. IEEE Trans
Electron Devices 1987;34(2):286–93.

21
E.H. Houssein et al. Computers and Electrical Engineering 94 (2021) 107304

[3] Saleem H, Karmalkar S. An analytical method to extract the physical parameters of a solar cell from four points on the illuminated 𝑗{−}𝑣 curve. IEEE
Electron Device Lett 2009;30(4):349–52.
[4] Nishioka K, Sakitani N, Uraoka Y, Fuyuki T. Analysis of multicrystalline silicon solar cells by modified 3-diode equivalent circuit model taking leakage
current through periphery into consideration. Sol Energy Mater Sol Cells 2007;91(13):1222–7.
[5] Easwarakhanthan T, Bottin J, Bouhouch I, Boutrit C. Nonlinear minimization algorithm for determining the solar cell parameters with microcomputers.
Int J Sol Energy 1986;4(1):1–12.
[6] Soon JJ, Low K-S. Photovoltaic model identification using particle swarm optimization with inverse barrier constraint. IEEE Trans Power Electron
2012;27(9):3975–83.
[7] Villalva MG, Gazoli JR, Ruppert Filho E. Comprehensive approach to modeling and simulation of photovoltaic arrays. IEEE Trans Power Electron
2009;24(5):1198–208.
[8] Oliva D, Cuevas E, Pajares G. Parameter identification of solar cells using artificial bee colony optimization. Energy 2014;72:93–102.
[9] Lin P, Cheng S, Yeh W, Chen Z, Wu L. Parameters extraction of solar cell models using a modified simplified swarm optimization algorithm. Sol Energy
2017;144:594–603.
[10] Wei H, Cong J, Lingyun X, Deyun S. Extracting solar cell model parameters based on chaos particle swarm algorithm. In: 2011 international conference
on electric information and control engineering. IEEE; 2011, p. 398–402.
[11] Guo L, Meng Z, Sun Y, Wang L. Parameter identification and sensitivity analysis of solar cell models with cat swarm optimization algorithm. Energy
Convers Manage 2016;108:520–8.
[12] Chellaswamy C, Ramesh R. Parameter extraction of solar cell models based on adaptive differential evolution algorithm. Renew Energy 2016;97:823–37.
[13] Oliva D, Abd Elaziz M, Elsheikh AH, Ewees AA. A review on meta-heuristics methods for estimating parameters of solar cells. J Power Sources
2019;435:126683.
[14] Oliva D, Abd El Aziz M, Hassanien AE. Parameter estimation of photovoltaic cells using an improved chaotic whale optimization algorithm. Appl Energy
2017;200:141–54.
[15] Gao X, Cui Y, Hu J, Xu G, Wang Z, Qu J, Wang H. Parameter extraction of solar cell models using improved shuffled complex evolution algorithm. Energy
Convers Manage 2018;157:460–79.
[16] Kang T, Yao J, Jin M, Yang S, Duong T. A novel improved cuckoo search algorithm for parameter estimation of photovoltaic (PV) models. Energies
2018;11(5):1060.
[17] Li S, Gong W, Yan X, Hu C, Bai D, Wang L, Gao L. Parameter extraction of photovoltaic models using an improved teaching-learning-based optimization.
Energy Convers Manage 2019;186:293–305.
[18] Chen H, Jiao S, Heidari AA, Wang M, Chen X, Zhao X. An opposition-based sine cosine approach with local search for parameter estimation of photovoltaic
models. Energy Convers Manage 2019;195:927–42.
[19] Abd Elaziz M, Oliva D. Parameter estimation of solar cells diode models by an improved opposition-based whale optimization algorithm. Energy Convers
Manage 2018;171:1843–59.
[20] Diab AAZ, Sultan HM, Do TD, Kamel OM, Mossa MA. Coyote optimization algorithm for parameters estimation of various models of solar cells and PV
modules. IEEE Access 2020;8:111102–40.
[21] Kumar C, Raj TD, Premkumar M, Raj TD. A new stochastic slime mould optimization algorithm for the estimation of solar photovoltaic cell parameters.
Optik 2020;223:165277.
[22] Askarzadeh A, Rezazadeh A. Parameter identification for solar cell models using harmony search-based algorithms. Sol Energy 2012;86(11):3241–9.
[23] Biswas PP, Suganthan P, Wu G, Amaratunga GA. Parameter estimation of solar cells using datasheet information with the application of an adaptive
differential evolution algorithm. Renew Energy 2019;132:425–38.
[24] Zhao W, Zhang Z, Wang L. Manta ray foraging optimization: An effective bio-inspired optimizer for engineering applications. Eng Appl Artif Intell
2020;87:103300.
[25] Mirjalili S, Mirjalili SM, Lewis A. Grey wolf optimizer. Adv Eng Softw 2014;69:46–61.
[26] Mirjalili S. SCA: a sine cosine algorithm for solving optimization problems. Knowl-Based Syst 2016;96:120–33.
[27] Mirjalili S. Moth-flame optimization algorithm: A novel nature-inspired heuristic paradigm. Knowl-Based Syst 2015;89:228–49.
[28] Mirjalili S. The ant lion optimizer. Adv Eng Softw 2015;83:80–98.
[29] Mirjalili S, Mirjalili SM, Hatamlou A. Multi-verse optimizer: a nature-inspired algorithm for global optimization. Neural Comput Appl 2016;27(2):495–513.
[30] Batzelis EI, Papathanassiou SA. A method for the analytical extraction of the single-diode PV model parameters. IEEE Trans Sustain Energy
2015;7(2):504–12.

Essam H. Houssein is working as an associate professor at the Faculty of Computers and Information, Minia University, Egypt. Essam H. Houssein serves as a
reviewer of more than 30 journals (Elsevier, Springer, and IEEE). His research interests include wireless sensor networks, IoT, Bioinformatics and Biomedical,
Cloud computing, Soft computing, Image processing, Artificial intelligence, Data mining, Optimization, and Meta-heuristics techniques.

Gamela Nageh Zaki is currently working as an Assistant lecturer at Minia University, Faculty of Computers and Information, computer sciences Department.
Her research interests include Optimization, and Meta-heuristics techniques.

Ahmed A. Zaki Diab is working as an Associate Professor at Faculty of Engineering, Minia University, Egypt. Ahmed A. Zaki Diab serves as a reviewer of
international journals (Elsevier, Springer, and IEEE). Ahmed Diab is an Associate Editor of IEEE ACCESS. He joined Kyushu University, Japan as a Visitor
Professor. His research interests include Renewable Energy, Optimization Algorithms, Machine Drives, Artificial intelligence.

Eman M.G. Younis is currently working as an Associate Professor at Minia University, Faculty of Computers and Information, Information Systems Department.
She got her B.Sc. degree from Zagazig University, Egypt, 2002. She obtained her M.Sc. degree from Menoufia University, Egypt in 2007. She received her Ph.D.
degree from Cardiff University, UK in 2014. She joined NTU, computing Department, as a researcher.

22