Genetic Algorithm - 2021
Genetic Algorithm - 2021
A R T I C L E I N F O A B S T R A C T
Keywords: Extracting the optimum parameters of solar photovoltaic (PV) model using the experimental data of cur
Algorithm genetic rent–voltage is very critical in simulating, controlling, and optimizing the PV systems. One of the important
Single diode model problems encountered in modeling and simulating is to find a model that can extract parameters from PV models
Double diode model
quickly, accurately, and reliably. Based on this motivation, the goal of this study is to suggest an improved al
Photovoltaic
gorithm, namely genetic algorithm based on non-uniform mutation (GAMNU), in order to approximate effi
Method optimization
ciently the parameters of solar cells and PV modules. In GAMNU, non-uniform mutation operator is used to
maintain diversity in the explored solutions and the crossover operator follows an adaptive search strategy,
which consists of searching the entire space at the beginning while maintaining a focused search when the
population tends to converge in a certain region of the search space. The performance of the method is
comprehensively evaluated on different solar cell models, including single and double diode, and single diode PV
modules, of a R.T.C France silicon solar cell, ESP-160 PPW PV, STP6-120/36 and Photowatt-PWP201 module.
The results obtained from single and double diode models for R.T.C France are respectively 9.8618×10− 4 and
9.8683×10− 4 , and for Photowatt-PWP201, STP6-120/36 and ESP-160 PPW are 2.3824×10− 3 2.382420230900 ×
10− 3 , 1.6735×10− 2 1.6735786505085 × 10− 2 and 8.2942×10− 2 8.2942 × 10− 2 . The statistical obtained results
show that the proposed method has very competitive performance in terms of accuracy and reliability when
compared to other advanced algorithms. Therefore, the proposed algorithm is highly useful to extract the pa
rameters of solar PV models.
* Corresponding author at: Laboratory of Materials and Renewable Energies, Faculty of Science, Ibn Zohr University, B. P. 8106 Agadir, Morocco.
E-mail address: [email protected] (D. Saadaoui).
https://doi.org/10.1016/j.ecmx.2021.100129
Received 18 July 2021; Received in revised form 28 October 2021; Accepted 30 October 2021
Available online 6 November 2021
2590-1745/© 2021 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/).
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
disadvantages of some of these algorithms is their dependence on the balance between exploration and exploitation properties. A modified
initial values of the parameters, since these can reduce the accuracy and Artificial Bee Colony (ABC) proposed by Ketkar et al. [23–25] is used for
the convergence of the algorithm. In order to tackle the limitation estimating the parameters of double diode model of solar cell solar. The
problems of analytical and iterative methods, meta-heuristic algorithms results obtained by the modified ABC show its superiority compared to
have received much attention. Meta-heuristic techniques are applied to the traditional ABC. In [26,27], Particle Swarm Optimization (PSO) has
convert the photovoltaic module parameter estimation problem into an been applied to estimate the solar cell parameters of the three diode
optimization one. The resulting optimization problem is nonlinear, model (TDM). Salp Swarm Algorithm (SSA) [28] has been proposed to
bounded, and continuous, which is a complex problem. They are estimate parameters for single and double diode solar photovoltaic
effective in providing solutions for different weather conditions, but models. It uses three steps such, retrieving the parameters convention
may have convergence problems in the presence of local o ptimum ally, determining the uncertainties of each parameter and the instanta
points [21]. Many meta-heuristic algorithms and their variants have neous parameters using the results of the first two steps. In comparison
been used for extracting the parameters of the PV models. In [22], with other algorithms such as TLBA, ITLBO, ImCSA, ISCE, and HFAPS
Improved Gaining-Sharing Knowledge algorithm (IGSK) has been pro for SDM and HISA, ELPSO, CPSO, BSA and ABC for DDM, it shows that it
posed for parameter extraction of photovoltaic models, in which an is a valuable tool for estimating PV parameters. In [29] the
adaptive mechanism ensures the balance between the number of di Biogeography-based heterogeneous cuckoo search (BHCS) algorithm is
mensions updated by the junior gain sharing phase and the number of applied to solve four different PV parameters estimation problems of
dimensions updated by the senior gain sharing phase. A boundary different photovoltaic models, including SDM, DDM and two PV panel
constraint treatment method and a linear population size reduction modules. The cooperation of the Cuckoo Search (CS) strategy and the
technique are used to increase the convergence speed and maintain a Biogeographic Optimization (BBO) strategy allows BHCS to achieve an
2
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
effective balance between exploration and exploitation. However, the because the PV model parameter extraction problem is a multi-modal
performance of BHCS is very competitive in terms of accuracy and optimization problem. Therefore, developing a meta-heuristic algo
reliability compared to CS, BBO [30] and several other meta-heuristic rithm that effectively balances global and local search capabilities and
algorithms. that estimates PV model parameters accurately and reliably remains a
On the other hand, the Moth-Flame Optimization (MFO) [36] has great challenge.
been proposed for identifying the parameters of the three-diode model GA is an optimization technique that allows resolving restricted and
for the multi-crystalline solar cell/module. The performance of this al unrestricted optimization problem in science and engineering applica
gorithm is compared to recent and efficient algorithms such as Differ tions [36]. The GAs are based on natural genetics and natural selection
ential Evolution with Integrated Mutation per iteration (DEIM) and [37]. The three basic rules of natural genetics are: (1) selection rules to
Flower Pollination Algorithm (FPA). The performances of the three al select the individuals, considered the source for next generation; (2)
gorithms applied to the three models DDM, MDDM and TDM show that crossover rules that combine two parents in order to produce new
the MFO algorithm has the best performance when applied to TDM in offspring for the next generation [38], reproduction, and mutation are
terms of the most accurate representation of the physical behavior of the used in the genetic search procedure [11]. From the application point of
multi-crystalline solar cells/modules, the best (I-V) curve fit, the fastest view, the genetic algorithm (GA) is used as a search technique to find the
convergence speed and the shortest execution time for the DEIM [31] most efficient heliostat field for a given absorbed power, and to calculate
and FPA [32] algorithms. The MFO algorithm has proven to be superior the optimum slope and surface azimuth angles for solar collectors to
to the DEIM and FPA algorithms by the nature of its inspiration, which receive maximum solar radiation [39,40]. The non-dominated sorting
helps it to never lose its optimal solutions and to converge quickly to the genetic algorithm (NSGA-II) approach is used to optimize geometric
solution. Abd Elaziz et al. [34] have proposed a novel Enhanced Marine characteristics of photovoltaic/thermal (PV/T) system in order to ach
Predators Algorithm (EMPA) for identifying static PV models (SDM and ieve greater efficiency [41]. In the literature, many version of GA have
DDM) and dynamic PV models parameters. In EMPA, the differential been improved and developed in order to extract parameters of solar PV
evolution operator (DE) is incorporated into the original marine MPA in cell models [11,17,42–46]. To date, the estimation of the parameters of
order to achieve stable and reliable performance while handling that the non-linear PV models is a multimodal problem that contains mul
nonlinear optimization problem of PV modeling. The performance of tiple local optima. Several optimization algorithms reports in the liter
EMPA has been compared with other well-known algorithms. As a ature can offer unsatisfactory performances in the application on
result, EMPA is an accurate tool for identifying the unknown parameters multimodal objective functions. For example, the problems of conver
of PV models in terms of data fit, convergence rate, stability, and con gence to a sub-optimal solution remain a problem that GAs face. Due to
sistency. In [33], the Biogeography-based Heterogeneous Cuckoo Search low diversity in a population, it becomes difficult to locally exploit so
(BHCS) combines both explorations of CS and BBO uses the migration lutions [47].
operator in order to create new solutions. A new hybrid algorithm based In order to solve these problems, the focus is now on achieving a
on grey wolf optimizer and cuckoo search (GWOCS) is applied to extract balance between the exploration and the exploitation characteristics of
the parameters of different PV cell models under different operating GAs. As a result, the objective of this study is to introduce a new
conditions [20]. The main advantage of GWOCS is its ability to balance improved genetic algorithm, namely genetic algorithm with non-
between exploration and exploitation. The experimental results show uniform mutation (GAMNU), which has shown high accuracy in opti
that GWOCS is a potential methodology for extracting parameters from mizing the parameters of different PV models. This algorithm has the
solar photovoltaic models. However, most of the algorithms published in merit of searching the space uniformly at the beginning stage and very
the literature have some limitations. For example, whale optimization locally at the further stage during the programming. The main search
algorithm (WOA) is good in the exploration phase, but it is limited in operators in this methodology are the mutation non-uniform and Blend
search local optimum due to a lack of population diversity [51]. The crossover (BLX-α). In the first phase, the crossover operator functions in
performance of memetic adaptive differential evolution (MADE) algo the survey of information that is accessible through the search space,
rithm deteriorates when dealing with a PV cell double-diode model [34]. which inadvertently improves the behavior of the GAMNU. Secondly,
In the same framework, particle swarm optimization (PSO) technique non-uniform mutation (NUM) [48] owns a fine adjustment capability
has some limitations such as being trapped in a local minimum espe whereby its action depends on the number of the population in order to
cially in the case of non-linear multi-modal problems as well as its slow reach equilibrium between exploration and exploitation. This balance,
convergence speed. Artificial bee colony (ABC) [35] is poor at exploi which critically controls the performance of the GAMNU, is controlled
tation. Cuckoo search (CS) algorithm suffers from slow convergence and by the good choices of the probability of crossover (pc ) and the proba
biogeography-based optimization (BBO) technique has a quick conver bility of mutation (pm ).
gence speed, but easily trapped in local optimal values due to their elitist The main contributions of this paper are organized as follows:
mechanism [33]. In the basic Salp Swarm Algorithm (SSA), the ability to
explore different areas of the feature space is limited. Furthermore, the
performance of these meta-heuristic algorithms needs to be improved
Fig. 1. Equivalent circuit of (a) single diode PV cell, (b) double diode PV cell.
3
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
where I01 , I02 , a1 , a2 are the saturation current and the ideality factor for
the first and second diode, respectively.
This model is characterized by five unknown parameters: Iph , I01 , I02 ,
a1 , a2 , Rs , and Rsh .
[ ( ) ] [ ( ) ]
(Vk + Rs Ik ) (Vk + Rs Ik ) Vk + Rs I k
h(Vk , Ik , Ψ) = Iph − I01 exp − 1 − I02 exp − 1 − − Ik (5)
a 1 Vt a 2 Vt Rsh
4
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
where:
7: Xi = rand(xi , xj
(t+1) (t+1) (t+1)
)
3.4. Selection
8: Xj = rand(xi , xj
(t+1) (t+1) (t+1)
)
9: end for
Parent selection involves assigning reproductive opportunities to
every individual within the population supported their own fitness.
3.3. Mutation non-uniform Literature shows that there are different selection schemes for genetic
algorithms, each one with different characteristics. In a tournament
The mutation operator changes one or more genes to increase the selection, chosen in the proposed algorithm, each individual in the
variability of the population. The main goal of mutation is to prevent population is accidentally combined with another. The comparison of
premature convergence of the solution, i.e., it prevents the algorithm the fitness of each pair of individuals allows the one with the lower
from parking at a local minimum. Each gene can mutate according to a fitness to move to the next round, while the other one is disqualified. The
genetic parameter called mutation probability pm . In this paper, non- number of parents selected depends on the selection probability ps . This
5
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
Fig. 4. Flowcharts of (a) The theoretical currents using the Newton Raphson method, (b) the root mean square error (RMSE).
last selected population is matched as parents of new individuals [62]. 4. Criteria for evaluating the proposed algorithm
Fig. 3 shows the flowchart of solar photovoltaic parameter estimation
process by the proposed GAMNU algorithm. In order to evaluate the performance of optimization methods, it is
Algorithm 2 presents the pseudo-code corresponding to this method. necessary to use performance indications, namely, individual absolute
Algorithm 2: Pseudo code: Method tournament selection current error (IAEC), individual absolute power error (IAEP), the sum of
1: Input: individuals in the population
individual absolute current error (SIAE), mean absolute error (MAE),
2: Output: selection of the fittest individuals relative error (RE), root mean square error (RMSE), percentage error p,
3: Start and correlation coefficient R2 [8,66-68]using Eqs. (8)–(15):
4: N← number of individuals in the population; ⃒ ⃒
5: k← parameter between 1 and Nof individuals to be IAEC = ⃒Icalculated − Iexperimental ⃒ (8)
6: selected;
7: if rand <= tournament probability ⃒ ⃒
IAEP = ⃒Pcalculated − Pexperimental ⃒ (9)
8: if fitness(parent_i1) < fitness(parent_i2)
9: selected (parent_i1)
10: else ∑
N
13: else
14: if fitness(parent_i1) < fitness(parent_i2) ∑
N
6
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
Table 1
Electrical characteristics of PV systems.
Parameter RTC solar cell Photowatt-PWP 210 STP6 120/36 ESP-160 PPW
Number of samples 26 25 22 43
Test Temperature (◦ C) 33 45 55 45
Test radiation (W/m2) 1000 1000 – –
Short circuit current (A) 0.7605 1.0300 7.48 5.51
Open circuit voltage (V) 0.5727 16.778 19.21 20.65
Impp (A) 0.6755 0.9120 6.83 16
Vmpp(V) 0.4590 12.649 14.93 4.95
Number of cells Ns 1 36 36 72
Technology Polycrystalline Polycrystalline Polycrystalline Polycrystalline
Table 2
Boundaries of the optimal parameters.
Parameters R.T.C France Solar Cell Photowatt PWP 201 STP6-120 ESP-160 PPW
Iph (A) 0 1 0 2 0 8 0 8
I0 , I01 , I02 (μA) 0 1 10− 6 10 0 50 0.01 10
a, a1 , a2 1 2 0 50 1 50 0 3
Rsh (Ω) 0 100 10− 3 2000 0 1500 0 0.55
Rs (Ω) 0 0.5 0 2 0 0.36 0 1000
Table 3
Comparison of different parameters extraction methods for single-diode model
Algorithms Iph (A) I0 (μA) a Rs (Ω) Rsh (Ω) RMSE( × 10− 4 )
lated and measured respectively; Iexperimental is the mean of the experi at 55 ◦ C and ESP-160 PPW PV module composed of 72 cells in series
mental current data. working at 45 ◦ C. For a plausible comparison, the range of the search for
The optimal parameters extracted by the optimization algorithms are each parameter is presented in Table 2. The performance of the proposed
used to calculate: method has been evaluated by comparing it to other methods. The
electrical characteristics of four PV systems are summarized in Table 1.
• The theoretical currents using the Newton Raphson method (NRM), When implementing GA to Eq. (4) for SDM and Eq. (5) for DDM, it is
whose flowchart is presented in Fig. 4(a) [65]; necessary to choose lower and upper bound the search of the individual
• Root mean square error (RMSE); the steps to follow are given by the parameter of single and double diode models. The lower and the upper
flowchart in Fig. 4(b) [69]. bounds of PV model parameters are revealed in Table 2.
The reliability, the accuracy, and the precision of the GAMNU pro 1.1.1. Single diode model
posed are at first checked by estimating the unknown parameters of the Table 3 illustrates the optimal solution of the five parameters and the
single model (SDM) and double diode model (DDM). In order to validate root mean square error (RMSE) values obtained by the proposed
the proposed method, four experimental dataset are chosen as follows: GAMNU method compared with other methods such as CPSO [70], HHO
one unit called commercial silicon solar R.T.C France with 57 mm [13], ABC [71], HS [72], Rcr- IJADE [47], BBO-M [73], GWO [20] MFO,
diameter working at 33 C and solar radiance of 1000W/m2 , and a solar CSA [74], JAYA, LBSA, ABSO, and IADE [54]TLBO, DE/BBO, and CLPSO
◦
module Photo-watt-PWP-201 composed of 36 polycrystalline silicon [75]. It can be seen that GAMNU also obtains the best RMSE (9.8618 ×
cells in series operating under 1000W/m2 at 45 C, a commercial module 10− 4 ) among all algorithms, followed by BBO-M, TLBO, MFO, JAYA,
◦
type STP6-120/36 composed of 36 polycrystalline silicon cells in series IADE, CSA, ABSO, LBSA and HS. The values found are compared with
7
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
Table 4
Absolute error of proposed GAMNU on Single Diode Model (SDM) RTC France commercial silicon PV cell.
Item Experimental measures Simulated current data Absolute individual Error Relative Error
Fig. 5. Comparison between experimental data and simulated data obtained by GAMNU for single diode model of silicon RTC France: (a) I-V curve (b) P-V curve.
other estimates by different optimization algorithms for the 26 instances 1.1.2. Double diode model
of the I-V curve. The five unknown parameters extracted by the pro The double diode model accounts for the effect of recombination; it
posed algorithm are substituted in Eq. (1), in order to reconstruct the has two additional parameters that increase the number of unknowns to
simulated current values at the 26 experimental voltage points and to seven, and the objective function becomes more complex and implicit.
calculate the individual absolute error IAE, relative error RE and to Table 5 shows the seven parameters estimated and the RMSE calculated
represent the I-V and P-V characteristics. The values obtained are by the proposed algorithm and by other optimization algorithms.
summarized in Table 4 and represented in Fig. 5. The values found show In agreement with the results obtained for the single diode model,
that IAE, RE and their sum are low, also the I-V and P-V characteristics the value of RMSE found by the proposed algorithm (9.8683 × 10− 4 ) is
indicate that the simulated data obtained by the method agree closely better in comparison with other optimization algorithms quoted in the
with the experimentally measured data over the entire voltage range. As literature such as the following ones: ITBLO (9.9021 × 10− 4 )[69], PSO-
illustrated in Fig. 6, GAMNU and its closest competitor HS and Rcr- WOA (16.70 × 10− 4 ), [68] TLBO (10.069 × 10− 4 ), [34] LBSA (1.65 ×
IJADE give small values of IAEi in many points. This shows concretely 10− 4 ) [54], and CLPSO (9.9894 × 10− 4 ) [54]. The experimental and the
that the parameter values extracted by the GAMNU algorithm are more simulated I-V characteristic are presented in Fig. 7. The results show that
accurate. The SIAE is 1.77 × 10− 4 , which is a small amount of current. the data calculated by GAMNU coincide well with the experimental
To compare the minimization percentage of the proposed algorithm data, which means that the parameters extracted by the proposed
with other algorithms, we determined the error percentage by Eq. (14). method are very accurate.
The error percentage obtained by the proposed algorithm is 0.12342% In addition, Table 6 shows the values of the Individual Absolute Error
and 0.12344% by Rcr-IJAYA, 0.12375% by ABSO, 0.15% by IJAYA, (IAE), the relative error (RE) and the sum of Individual Absolute Error. It
0.96018% by GACCC and 4.8939% by BBO-M. The results show that the is easy to see that the measured and simulated data coincide well, and
GAMNU is more efficient in terms of error minimization. that the sum of errors for the two-diode model is slightly smaller
8
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
[78]. It can be seen that the GAMNU algorithm provides the optimal
value of RMSE (2.3742256 × 10− 3 ) among all the comparative
algorithms.
For further clarification on the quality of the parameter values
identified by the proposed GAMNU, the I − V and P − V characteristics
have been plotted in Fig. 8(a) and (b) and in Table 8 the IAE and RE as
well as the sum of the absolute errors.
It is clear from the I − V and P − V characteristics in Fig. 8 that the
simulated data obtained by GAMNU are in good agreement with the
experimental data over the entire voltage range. The estimated current,
the individual absolute error and the sum of errors, for the proposed
method and the comparative optimization methods, such as GACCC, SA
and EHHO have been calculated by the substitution of the parameters
Iph , Is , Rs , Rsh and a in Eq.1 and using Eqs. (9) and (11). The obtained
values are shown in Table 8, the sum IAE value suggests that GAMNU
has the smallest sum (0.0413280) compared with GACCC (0.0468364)
EHHO (0.0452045) and SA (0.0642). From Fig. 9, it is clear that the
proposed method has a better performance compared with the optimi
zation methods listed above.
To compare the minimization percentage of the proposed algorithm
with three other algorithms such as GACCC, EHHO, and SA, we deter
Fig. 6. Graphical representation of absolute current error for the single
mined the error percentage for the different algorithms. The error per
diode model.
centage obtained by the GAMNU is 0.2881 % and 0.3299 % by GACCC,
0.3152% by EHHO, and 0.4476% by SA. The value obtained by the
(2.145146 × 10− 2 ) than for the single-diode model. To compare the proposed algorithm is less than that obtained by the three other algo
minimization percentage of the proposed algorithm with two other al rithms by 0.0418%, 0.0271%, and 0.1595% respectively.
gorithms, we determined the error percentage by Eq. (14). The error
percentage obtained by the GAMNU is 0.1248 % and 0.1207 % by
ITLBO, and 0.14300% by PSO-WOA. The value obtained by ITLBO al 1.3. Case 3: STP6-120/36 model
gorithm is less than that obtained by the GAMNU by 0.0041% .
The characteristics I − V and P − V obtained by GAMNU and the
experimental data are presented in Fig. 10. The optimal model param
1.2. Case 2: Photo watt-PWP 201 PV module eters and the average root mean square error (RMSE), the individual
absolute error of current (IAEC )) and the individual absolute power
For the single diode model of the Photowatt-PWP201 module, the (IAEP ), obtained by proposed algorithm and others published in the
results of the parameter extraction of the proposed algorithm are pre literature for single diode model are presented in Tables 9 and 10. The
sented in Table 7 and compared with other algorithms, such as Rcr- best value of RMSE is achieved by MADE [34], SDO (RMSE = 1.66 ×
IJADE [47], GACCC [43], CPSO [76], EHHO [20], SGDE [77] and SA 10− 2 ) [79], and the proposed method GAMNU (RMSE = 1.6735 ×
Table 5
Comparison of different parameters extraction methods for double-diode model of RTC France.
Algorithms Iph (A) I01 (μA) I02 (μA) a1 a2 Rs (Ω) Rsh (Ω) RMSE ( × 10− 4 )
Fig. 7. Comparison between experimental data and simulated data obtained by GAMNU for double diode model of silicon RTC France (a) I-V curve (b) P-V curve.
9
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
Table 6
Absolute Error of proposed GAMNU on double Diode Model (DDM)
RTC France commercial silicon PV cell
Item Experimental measures Simulated current data Absolute individual Error Relative Error
Table 7
Comparison with other methods in the literature.
Algorithms Iph (A) I0 (μA) A = Ns × a Rs (Ω) Rsh (Ω) RMSE( × 10− 3 )
Fig. 8. Comparison between experimental data and simulated data obtained by GAMNU for single diode model of Photowatt-PWP201 PV module:(a) I-V curve (b) P-
V curve.
10− 2 A), in comparison with BMO ((RMSE = 1.6985 × 10− 2 A)[80], measured current shown in Table 8 indicates that the estimated currents
CLPSO (RMSE = 1.7495 × 10− 2 A)[75], CWOA (RMSE = 1.76 × are in good agreement with the experimental currents. From Fig. 10, it is
10− 2 A)[80], SA (RMSE = 1.7879 × 10− 2 A)[64], ACS (RMSE = clear that the simulated data obtained by GAMNU agrees closely with
1.98653 × 10− 2 A), DE/BBO (RMSE = 2.20596 × 10− 2 A)) [33]and the experimentally measured data over the entire voltage range.
LAPO (RMSE = 4.1306 × 10− 2 A)[81]. The error of the simulated and
10
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
Table 8
Optimal values of IAE extracted by GAMNU, GACCC, EHHO and SA.
GAMNU GACCC EHHO SA
Fig. 9. Optimal IAE values obtained by proposed method and others optimization methods.
Fig. 10. Comparison between experimental data and simulated data obtained by GAMNU for single diode model of STP6-120/36 module:(a) I-V curve (b) P-V curve.
11
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
Table 9
Comparison GAMNU with other algorithms for the single diode model of the STP6-120/36.
Algorithms Iph (A) Is (μA) A = Ns × a Rs Rsh RMSE( × 10− 2 )
Table 10
Simulated results of GAMNU for STP6-120/36 module.
Vexp (V) Iexp (A) Isim (A) IAEC (A) Pexp (W) Psim (W) IAEP (W)
model, shown in Fig. 11, reveal that the simulated curve obtained by
Table 11
method proposed is in good agreement with the measured data. The
Parameters for the proposed model and SFS.
quality and the performance of the proposed method are tested by the
Parameters GAMNU SFS IAE. In Table 12, the comparison between GAMNU and SFS algorithm is
Iph (A) 5.50609664 5.534 illustrated. The sum absolute errors obtained by GAMNU and SFS are
I0 (μA) 4.9378 0.5859 1.68944108 and 1.69271237 respectively.
a 0.7511308 0.6518 Additionally, Fig. 12 shows the IAE curves between the experimen
Rs (Ω) 0.161934 0.2175 tally measured and the estimated current. All the IAE values are smaller
Rsh (Ω) 112.7891 77.962 than 0.1742 for proposed method.
RMSE(A) 0.082942 0.095006
MAE(A) 0.03928932 0.0393654 6. Conclusion
R2 0.99927781 0.999322097
In this study, a novel optimization genetic algorithm is proposed in
order to identify accurately the unknown parameters of the single and
1.4. Case 4: ESP-160 PPW PV module. the double diode for different PV cell and module technologies. The
proposed algorithm, namely genetic algorithm based on non-uniform
The measurements data of polycrystalline solar panel ESP-160 PPW mutation (GAMNU), is an improved of the classical genetic algorithm.
PV module with 72 cells in series used in this section are presented in In this algorithm, Non-uniform mutation is used to maintain diversity in
[65] at 45 ◦ C. Table 11 shows the extracted values of PV module by the the explored solutions and the crossover operator follows an adaptive
proposed method and the SFS method, with the RMSE, MAE, and the search strategy, which consists of searching the entire space at the
correlation coefficient R2 . The values of RMSE and MAE obtained by beginning while maintaining a focused search when the population
GAMNU are respectively 0.082942 and 0.03928932. The value of R2 tends to converge in a certain region of the search space. The perfor
obtained is 0.99927781, close to 1, indicating a perfect agreement be mance of the proposed GAMNU algorithm has been validated using
tween the simulated and the measured value. experimental data on different solar cell models, including single diode,
For more detail, the I − V and P − V characteristics for the single diode double diode, and single diode PV modules, of a commercial R.T.C
12
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
Fig. 11. Comparison between experimental data and simulated data obtained by GAMNU for single diode model of ESP-160 PPW model. (a) I-V curve (b) P-V curve.
Table 12
Simulated results of proposed method and SFS for the single diode model
GAMNU SFS
France silicon solar cell at 33 ◦ C, ESP-160 PPW PV module with 72 cells method has been compared with other reported algorithm from litera
in series at 45 ◦ C, STP6-120/36 and Photowatt-PWP201 module with 36 ture. The obtained RMSE values by GAMNU for RTC cell, Photowatt-
cells in series at 33◦ and 55 ◦ C respectively, and compared with various PWP201, STP6-120/36 and ESP-160 PPW polycrystalline PV modules
meta-heuristic methods. The performance results of the proposed are 9.8618 × 10− 4 , 2.3824 × 10− 3 2.382420230900 × 10− 3 , 1.6735 ×
GAMNU algorithm show a very high similarity between the I-V and P-V 10− 2 1.6735786505085 × 10− 2 and 8.2942 × 10− 2 respectively when
curves and the experimental data. However, the efficacy of the proposed single diode model SDM is used, and 9.8683 × 10− 4 for RTC solar cell
13
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
Fig.12. Optimal IAE values obtained by proposed method and SFS method.
using double diode model DDM. The obtained correlation coefficient R2 [11] Ismail MS, Moghavvemi M, Mahlia TMI. Characterization of PV panel and global
optimization of its model parameters using genetic algorithm. Energy Convers
for STP6-120/36 is 0.9992. The obtained results of simulation are Manag 2013;73:10–25. https://doi.org/10.1016/j.enconman.2013.03.033.
smaller than all the RMSE obtained by other optimization methods. The [12] Ram JP, Babu TS, Dragicevic T, Rajasekar N. A new hybrid bee pollinator flower
statistical obtained results demonstrate that the proposed algorithm pollination algorithm for solar PV parameter estimation. Energy Convers Manag
2017;135:463–76. https://doi.org/10.1016/j.enconman.2016.12.082.
outperforms in terms of accuracy compared to other recently optimi [13] Abbassi A, Abbassi R, Heidari AA, Oliva D, Chen H, Habib A, et al. Parameters
zation algorithms reports in the literature. Based on this study, the identification of photovoltaic cell models using enhanced exploratory salp chains-
proposed GAMNU method is designed to be very useful in solving real based approach. Energy 2020;198:117333. https://doi.org/10.1016/j.
energy.2020.117333.
optimization problems in the energy sector.
[14] Wolf P, Benda V. Identification of PV solar cells and modules parameters by
combining statistical and analytical methods. Sol Energy 2013;93:151–7. https://
Declaration of Competing Interest doi.org/10.1016/j.solener.2013.03.018.
[15] Makrides G, Zinsser B, Georghiou GE, Schubert M, Werner JH. Temperature
behaviour of different photovoltaic systems installed in Cyprus and Germany. Sol
The authors declare that they have no known competing financial Energy Mater Sol Cells 2009;93(6-7):1095–9. https://doi.org/10.1016/j.
interests or personal relationships that could have appeared to influence solmat.2008.12.024.
[16] Bai J, Liu S, Hao Y, Zhang Z, Jiang M, Zhang Y. Development of a new compound
the work reported in this paper.
method to extract the five parameters of PV modules. Energy Convers Manage
2014;79:294–303. https://doi.org/10.1016/j.enconman.2013.12.041.
References [17] Bastidas-Rodriguez JD, Petrone G, Ramos-Paja CA, Spagnuolo G. A genetic
algorithm for identifying the single diode model parameters of a photovoltaic
panel. Math Comput Simul 2017;131:38–54. https://doi.org/10.1016/j.
[1] De Broe AM, Drouilhet S, Gevorgian V. A peak power tracker for small wind
matcom.2015.10.008.
turbines in battery charging applications. IEEE Trans Energy Convers 1999;14:
[18] Khan F, Baek SH, Park Y, Kim JH. Extraction of diode parameters of silicon solar
1630–5. https://doi.org/10.1109/60.815116.
cells under high illumination conditions. Energy Convers Manage 2013;76:421–9.
[2] AlHajri MF, El-Naggar KM, AlRashidi MR, Al-Othman AK. Optimal extraction of
https://doi.org/10.1016/j.enconman.2013.07.054.
solar cell parameters using pattern search. Renew Energy 2012;44:238–45. https://
[19] Saha C, Agbu N, Jinks R. 2 - Review article of the Solar PV Parameters Estimation
doi.org/10.1016/j.renene.2012.01.082.
using Evolutionary Algorithms. MOJ Sol Photoen Sys 2018;2:66–78. https://doi.
[3] Mojallizadeh MR, Badamchizadeh MA. Adaptive passivity-based control of a
org/10.15406/mojsp.2018.02.00026.
photovoltaic/battery hybrid power source via algebraic parameter identification.
[20] Long W, Cai S, Jiao J, Xu M, Wu T. A new hybrid algorithm based on grey wolf
IEEE J Photovoltaics 2016;6(2):532–9. https://doi.org/10.1109/
optimizer and cuckoo search for parameter extraction of solar photovoltaic models.
JPHOTOV.2016.2514715.
Energy Convers Manage 2020;203:112243. https://doi.org/10.1016/j.
[4] Darmawi, Sipahutar R, Bernas SM, Imanuddin MS. Renewable energy and
enconman.2019.112243.
hydropower utilization tendency worldwide. Renew Sustain Energy Rev 2013;17:
[21] Niu Q, Zhang H, Li K. An improved TLBO with elite strategy for parameters
213–5. https://doi.org/10.1016/j.rser.2012.09.010.
identification of PEM fuel cell and solar cell models. Int J Hydrogen Energy 2014;
[5] Barbier E. Geothermal energy technology and current status: an overview. Renew
39(8):3837–54. https://doi.org/10.1016/j.ijhydene.2013.12.110.
Sustain Energy Rev 2002;6(1-2):3–65. https://doi.org/10.1016/S1364-0321(02)
[22] Sallam KM, Hossain MA, Chakrabortty RK, Ryan MJ. An improved gaining-sharing
00002-3.
knowledge algorithm for parameter extraction of photovoltaic models. Energy
[6] Wilberforce T, Baroutaji A, El Hassan Z, Thompson J, Soudan B, Olabi AG.
Convers Manage 2021;237:114030. https://doi.org/10.1016/j.
Prospects and challenges of concentrated solar photovoltaics and enhanced
enconman.2021.114030.
geothermal energy technologies. Sci Total Environ 2019;659:851–61. https://doi.
[23] Jamadi M, Merrikh-Bayat F, Bigdeli M. Very accurate parameter estimation of
org/10.1016/j.scitotenv.2018.12.257.
single- and double-diode solar cell models using a modified artificial bee colony
[7] Sun B, Yu Y, Qin C. Should China focus on the distributed development of wind and
algorithm. Int J Energy Environ Eng 2016;7(1):13–25. https://doi.org/10.1007/
solar photovoltaic power generation? A comparative study. Appl Energy 2017;185:
s40095-015-0198-5.
421–39. https://doi.org/10.1016/j.apenergy.2016.11.004.
[24] Chen X, Xu B, Mei C, Ding Y, Li K. Teaching–learning–based artificial bee colony
[8] Chin VJ, Salam Z, Ishaque K. Cell modelling and model parameters estimation
for solar photovoltaic parameter estimation. Appl Energy 2018;212:1578–88.
techniques for photovoltaic simulator application: a review. Appl Energy 2015;
https://doi.org/10.1016/j.apenergy.2017.12.115.
154:500–19. https://doi.org/10.1016/j.apenergy.2015.05.035.
[25] Ketkar M, M. Chopde A. Efficient Parameter Extraction of Solar Cell using Modified
[9] Fernández EF, Montes-Romero J, de la Casa J, Rodrigo P, Almonacid F.
ABC. Int J Comput Appl 2014;102:1–6. https://doi.org/10.5120/17776-8535.
Comparative study of methods for the extraction of concentrator photovoltaic
[26] Khanna Vandana, Das BK, Bisht Dinesh, Vandana, Singh PK. A three diode model
module parameters. Sol Energy 2016;137:413–23. https://doi.org/10.1016/j.
for industrial solar cells and estimation of solar cell parameters using PSO
solener.2016.08.046.
algorithm. Renew Energy 2015;78:105–13. https://doi.org/10.1016/j.
[10] Ayodele TR, Ogunjuyigbe ASO, Ekoh EE. Evaluation of numerical algorithms used
renene.2014.12.072.
in extracting the parameters of a single-diode photovoltaic model. Sustain Energy
Technol Assessm 2016;13:51–9. https://doi.org/10.1016/j.seta.2015.11.003.
14
D. Saadaoui et al. Energy Conversion and Management: X 12 (2021) 100129
[27] Yousri D, Allam D, Eteiba MB, Suganthan PN. Static and dynamic photovoltaic [54] Yu K, Liang JJ, Qu BY, Chen X, Wang H. Parameters identification of photovoltaic
models’ parameters identification using chaotic heterogeneous comprehensive models using an improved JAYA optimization algorithm. Energy Convers Manage
learning particle swarm optimizer variants. Energy Convers Manage 2019;182: 2017;150:742–53. https://doi.org/10.1016/j.enconman.2017.08.063.
546–63. https://doi.org/10.1016/j.enconman.2018.12.022. [55] Wang Feng, Zhang Heng, Zhou Aimin. A particle swarm optimization algorithm for
[28] Messaoud Ramzi Ben. Extraction of uncertain parameters of single and double mixed-variable optimization problems. Swarm Evol Comput 2021;60:100808.
diode model of a photovoltaic panel using Salp Swarm algorithm. Meas J Int Meas https://doi.org/10.1016/j.swevo.2020.100808.
Confed 2020;154:107446. https://doi.org/10.1016/j.measurement.2019.107446. [56] Holland JH. Adaptation in natural and artificial systems: an introductory analysis
[29] Chen X, Yu K, Du W, Zhao W, Liu G. Parameters identification of solar cell models with applications to biology, control, and artificial intelligence. MIT press; 1992.
using generalized oppositional teaching learning based optimization. Energy 2016; [57] Bäck T, Fogel DB, Michalewicz Z. Handbook of evolutionary computation. Release
99:170–80. https://doi.org/10.1016/j.energy.2016.01.052. 1997;97:B1.
[30] Abd Elaziz Mohamed, Thanikanti Sudhakar Babu, Ibrahim Ibrahim Anwar, [58] Z. Michalewicz Genetic Algorithms+ Data Structure= Evolution Programs Springer
Lu Songfeng, Nastasi Benedetto, Alotaibi Majed A, et al. Enhanced marine 1992 New York.
predators algorithm for identifying static and dynamic photovoltaic models [59] Zhao Xinchao, Gao Xiao-Shan, Hu Ze-Chun. Evolutionary programming based on
parameters. Energy Convers Manage 2021;236:113971. https://doi.org/10.1016/ non-uniform mutation. Appl Math Comput 2007;192(1):1–11. https://doi.org/
j.enconman.2021.113971. 10.1016/j.amc.2006.06.107.
[31] Muhsen DH, Ghazali AB, Khatib T, Abed IA. Parameters extraction of double diode [60] Eshelman LJ, Schaffer JD. Real-Coded Genetic Algorithms and Interval-Schemata
photovoltaic module’s model based on hybrid evolutionary algorithm. Energy 1993;vol. 2. https://doi.org/10.1016/b978-0-08-094832-4.50018-0.
Convers Manage 2015;105:552–61. https://doi.org/10.1016/j. [61] Satoh H. Minimal generation gap model for GAs considering both exploration and
enconman.2015.08.023. exploitation. proc. 4th Int. Conf. Soft Comput. 1997.
[32] Alam DF, Yousri DA, Eteiba MB. Flower Pollination Algorithm based solar PV [62] Michalewicz Zbigniew, Schoenauer Marc. Evolutionary algorithms for constrained
parameter estimation. Energy Convers Manage 2015;101:410–22. https://doi.org/ parameter optimization problems. Evol Comput 1996;4(1):1–32. https://doi.org/
10.1016/j.enconman.2015.05.074. 10.1162/evco.1996.4.1.1.
[33] Chen X, Yu K. Hybridizing cuckoo search algorithm with biogeography-based [63] Easwarakhanthan T, BOTTIN J, bouhouch I, Boutrit C. Nonlinear minimization
optimization for estimating photovoltaic model parameters. Sol Energy 2019;180: algorithm for determining the solar cell parameters with microcomputers. Int J Sol
192–206. https://doi.org/10.1016/j.solener.2019.01.025. Energy 1986;4(1):1–12. https://doi.org/10.1080/01425918608909835.
[34] Li S, Gong W, Yan X, Hu C, Bai D, Wang L. Parameter estimation of photovoltaic [64] Tong NT, Pora W. A parameter extraction technique exploiting intrinsic properties
models with memetic adaptive differential evolution. Sol Energy 2019;190: of solar cells. Appl Energy 2016;176:104–15. https://doi.org/10.1016/j.
465–74. https://doi.org/10.1016/j.solener.2019.08.022. apenergy.2016.05.064.
[35] Oliva Diego, Cuevas Erik, Pajares Gonzalo. Parameter identification of solar cells [65] Rezk H, Babu TS, Al-Dhaifallah M, Ziedan HA. A robust parameter estimation
using artificial bee colony optimization. Energy 2014;72:93–102. approach based on stochastic fractal search optimization algorithm applied to solar
[36] Man K-F, Tang K-S, Kwong S. Genetic algorithms: concepts and applications in PV parameters. Energy Rep 2021;7:620–40. https://doi.org/10.1016/j.
engineering design. IEEE Trans Ind Electron 1996;43:519–34. egyr.2021.01.024.
[37] Jervase Joseph A, Bourdoucen Hadj, Al-Lawati Ali. Solar cell parameter extraction [66] Ziegel Eric R, Mathews John H. Numerical methods for computer science,
using genetic algorithms. Meas Sci Technol 2001;12(11):1922–5. https://doi.org/ engineering, and mathematics. Technometrics 1988;30(2):245. https://doi.org/
10.1088/0957-0233/12/11/322. 10.2307/1270193.
[38] El Shahat A. Maximum Power Point genetic identification function for photovoltaic [67] De Castro F, Laudani A, Riganti Fulginei F, Salvini A. An in-depth analysis of the
system. Int J Res Rev Appl Sci 2010;3:264–73. modelling of organic solar cells using multiple-diode circuits. Sol Energy 2016;135:
[39] Talebizadeh P, Mehrabian MA, Abdolzadeh M. Prediction of the optimum slope 590–7. https://doi.org/10.1016/j.solener.2016.06.033.
and surface azimuth angles using the Genetic Algorithm. Energy Build 2011;43 [68] Xiong G, Zhang J, Yuan X, Shi D, He Y, Yao G. Parameter extraction of solar
(11):2998–3005. photovoltaic models by means of a hybrid differential evolution with whale
[40] Talebizadeh Pouyan, Mehrabian Mozaffar Ali, Rahimzadeh Hasan. Optimization of optimization algorithm. Sol Energy 2018;176:742–61. https://doi.org/10.1016/j.
heliostat layout in central receiver solar power plants. J Energy Eng 2014;140(4): solener.2018.10.050.
04014005. https://doi.org/10.1061/(ASCE)EY.1943-7897.0000162. [69] Gnetchejo Patrick Juvet, Ndjakomo Essiane Salomé, Ele Pierre, Wamkeue René,
[41] Shahsavar A, Talebizadeh P, Tabaei H. Optimization with genetic algorithm of a Mbadjoun Wapet Daniel, Perabi Ngoffe Steve. Important notes on parameter
PV/T air collector with natural air flow and a case study. J Renew Sustain Energy estimation of solar photovoltaic cell. Energy Convers Manage 2019;197:111870.
2013;5(2):023118. https://doi.org/10.1063/1.4798312. https://doi.org/10.1016/j.enconman.2019.111870.
[42] Zagrouba M, Sellami A, Bouaïcha M, Ksouri M. Identification of PV solar cells and [70] Wang W, Wu JM, Liu JH. A particle swarm optimization based on chaotic
modules parameters using the genetic algorithms: application to maximum power neighborhood search to avoid premature convergence. 3rd Int Conf Genet Evol
extraction. Sol Energy 2010;84(5):860–6. https://doi.org/10.1016/j. Comput WGEC 2009 2009:633–6. https://doi.org/10.1109/WGEC.2009.168.
solener.2010.02.012. [71] Wang R, Zhan Y, Zhou H. Application of artificial bee colony in model parameter
[43] Hamid Naima, Abounacer Rachida, Idali Oumhand Mohammed, identification of solar cells. Energies 2015;8:7563–81. https://doi.org/10.3390/
Feddaoui M’barek, Agliz Driss. Parameters identification of photovoltaic solar cells en8087563.
and module using the genetic algorithm with convex combination crossover. Int J [72] Askarzadeh Alireza, Rezazadeh Alireza. Parameter identification for solar cell
Ambient Energy 2019;40(5):517–24. https://doi.org/10.1080/ models using harmony search-based algorithms. Sol Energy 2012;86(11):3241–9.
01430750.2017.1421577. https://doi.org/10.1016/j.solener.2012.08.018.
[44] Balasubramanian K, Jacob B, Priya K, Sangeetha K, Rajasekar N, Babu TS. Critical [73] Niu Q, Zhang L, Li K. A biogeography-based optimization algorithm with mutation
evaluation of genetic algorithm based fuel cell parameter extraction. Energy strategies for model parameter estimation of solar and fuel cells. Energy Convers
Procedia 2015;75:1975–82. https://doi.org/10.1016/j.egypro.2015.07.244. Manage 2014;86:1173–85. https://doi.org/10.1016/j.enconman.2014.06.026.
[45] Kumari PA, Geethanjali P. Adaptive genetic algorithm based multi-objective [74] Ismaeel Alaa AK, Houssein Essam H, Oliva Diego, Said Mokhtar. Gradient-based
optimization for photovoltaic cell design parameter extraction. Energy Procedia optimizer for parameter extraction in photovoltaic models. IEEE Access 2021;9:
2017;117:432–41. https://doi.org/10.1016/j.egypro.2017.05.165. 13403–16. https://doi.org/10.1109/Access.628763910.1109/
[46] Dizqah AM, Maheri A, Busawon K. An accurate method for the PV model ACCESS.2021.3052153.
identification based on a genetic algorithm and the interior-point method. Renew [75] Hu Z, Gong W, Li S. Reinforcement learning-based differential evolution for
Energy 2014;72:212–22. https://doi.org/10.1016/j.renene.2014.07.014. parameters extraction of photovoltaic models. Energy Rep 2021;7:916–28. https://
[47] Gong W, Cai Z. Parameter extraction of solar cell models using repaired adaptive doi.org/10.1016/j.egyr.2021.01.096.
differential evolution. Sol Energy 2013;94:209–20. https://doi.org/10.1016/j. [76] Xu S, Wang Y. Parameter estimation of photovoltaic modules using a hybrid flower
solener.2013.05.007. pollination algorithm. Energy Convers Manag 2017;144:53–68. https://doi.org/
[48] Xinchao Zhao. Simulated annealing algorithm with adaptive neighborhood. Appl 10.1016/j.enconman.2017.04.042.
Soft Comput J 2011;11(2):1827–36. https://doi.org/10.1016/j.asoc.2010.05.029. [77] Liang Jing, Qiao Kangjia, Yuan Minghua, Yu Kunjie, Qu Boyang, Ge Shilei, et al.
[49] Liang Jing, Qiao Kangjia, Yu Kunjie, Ge Shilei, Qu Boyang, Xu Ruohao, et al. Evolutionary multi-task optimization for parameters extraction of photovoltaic
Parameters estimation of solar photovoltaic models via a self-adaptive ensemble- models. Energy Convers Manage 2020;207:112509. https://doi.org/10.1016/j.
based differential evolution. Sol Energy 2020;207:336–46. https://doi.org/ enconman.2020.112509.
10.1016/j.solener.2020.06.100. [78] El-Naggar KM, AlRashidi MR, AlHajri MF, Al-Othman AK. Simulated Annealing
[50] Ram JP, Manghani H, Pillai DS, Babu TS, Miyatake M, Rajasekar N. Analysis on algorithm for photovoltaic parameters identification. Sol Energy 2012;86(1):
solar PV emulators: a review. Renew Sustain Energy Rev 2018;81:149–60. https:// 266–74. https://doi.org/10.1016/j.solener.2011.09.032.
doi.org/10.1016/j.rser.2017.07.039. [79] Xiong Guojiang, Zhang Jing, Shi Dongyuan, Yuan Xufeng. Application of supply-
[51] Md Sabudin SN, Jamil NM. Parameter Estimation in Mathematical Modelling for demand-based optimization for parameter extraction of solar photovoltaic models.
Photovoltaic Panel. IOP Conf Ser Mater Sci Eng 2019;536. https://doi.org/ Complexity 2019;2019:1–22. https://doi.org/10.1155/2019/3923691.
10.1088/1757-899X/536/1/012001. [80] Oliva D, Abd El Aziz M, Ella Hassanien A. Parameter estimation of photovoltaic
[52] Yahya-Khotbehsara A, Shahhoseini A. A fast modeling of the double-diode model cells using an improved chaotic whale optimization algorithm. Appl Energy 2017;
for PV modules using combined analytical and numerical approach. Sol Energy 200:141–54. https://doi.org/10.1016/j.apenergy.2017.05.029.
2018;162:403–9. https://doi.org/10.1016/j.solener.2018.01.047. [81] El-Fergany AA. Parameters identification of PV model using improved slime mould
[53] Rezaee Jordehi A. Enhanced leader particle swarm optimisation (ELPSO): an optimizer and Lambert W-function. Energy Rep 2021;7:875–87. https://doi.org/
efficient algorithm for parameter estimation of photovoltaic (PV) cells and 10.1016/j.egyr.2021.01.093.
modules. Sol Energy 2018;159:78–87. https://doi.org/10.1016/j.
solener.2017.10.063.
15