You are on page 1of 11

Available online at www.sciencedirect.

com

Solar Energy 90 (2013) 123–133


www.elsevier.com/locate/solener

Extraction of maximum power point in solar cells using bird


mating optimizer-based parameters identification approach
Alireza Askarzadeh a,⇑, Alireza Rezazadeh b
a
Department of Energy Management and Optimization, Institute of Science and High Technology and Environmental Sciences, Graduate University
of Advanced Technology, Kerman, Iran
b
Faculty of Electrical and Computer Engineering, Shahid Beheshti University, GC, Evin 1983963113, Tehran, Iran

Received 9 May 2012; received in revised form 10 January 2013; accepted 11 January 2013
Available online 9 February 2013

Communicated by: Associate Editor Nicola Romeo

Abstract

Maximum power point of solar cells can be extracted by knowing the values of the electrical parameters. The validity of the obtained
result depends on the accuracy of the model parameters. Hence, it is important to use a superior optimization technique to identify the
optimal values of the parameters. Recently, a metaheuristic optimization algorithm, bird mating optimizer (BMO), has been devised
which tries to metaphorically imitate the mating strategies of bird species. BMO employs several searching patterns to explore the region
under consideration. This ability helps the algorithm to maintain the diversity and avoid premature convergence, and therefore, get close
to the global solution. In this paper, the electrical parameters of a 57 mm diameter commercial (RTC France) silicon solar cell are iden-
tified using BMO. The optimal parameters are then used to extract the maximum power point of the system. The accuracy of the pro-
posed parameter identification approach is compared with the results found by the other optimization techniques. Simulation results
accentuate the superior potential of BMO algorithm.
Ó 2013 Elsevier Ltd. All rights reserved.

Keywords: Solar cell; Maximum power point; Electrical parameters identification; Bird mating optimizer

1. Introduction solved by several strategies in the literature such as fuzzy


control, perturbation, power–voltage differentiation and
The importance of using renewable energy sources orig- on-line method (Bahgat et al., 2004; Yu et al., 2004; Enri-
inates from the fossil fuels increasing cost, their feasible que et al., 2007), these strategies have deficiencies in stabil-
scarcity, and the resultant environmental pollutions. ity and response time as the illumination of solar varies
Among renewable energy sources, solar cells, with promis- abruptly (Zagroubaa et al., 2010). As an alternative, a
ing features, have significantly attracted the attention of direct MPP strategy using the electrical parameters has
many nations and are being used worldwide to contribute been introduced in the literature (Ikegami et al., 2001).
to meeting the rising demands of electric power. The precision of this strategy depends strongly on the accu-
For achieving the maximum available power of a solar racy of the identified parameters. Because the lack of infor-
cell system, it is essential to use that system at its maximum mation about the precise values of the parameters is the
power point (MPP). Though the problem of MPP has been main drawback, developing an accurate parameter identifi-
cation process is essential.
Two main modeling approaches are commonly used for
⇑ Corresponding author. Tel./fax: +98 342 6233176.
solar cell systems: the single and double diode models.
E-mail addresses: a.askarzadeh@kgut.ac.ir, askarzadeh_a@yahoo.com
However, the single diode model is much more common
(A. Askarzadeh), a-rezazade@sbu.ac.ir (A. Rezazadeh).

0038-092X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.solener.2013.01.010
124 A. Askarzadeh, A. Rezazadeh / Solar Energy 90 (2013) 123–133

than its counterpart. To extract the MPP using the direct silicon solar cell (Easwarakhanthan et al., 1986) using both
method the parameters of the used model should be identi- the single and double diode models. The optimal parame-
fied with high degree of accuracy. Parameters identification ters are then used to extract MPP from the system. In order
is essential not only to extract the MPP but also it provides to evaluate the search power of BMO, the accuracy of the
a powerful tool for simulation, performance evaluation, obtained results are compared with those found by GA,
control, and optimization of solar cell systems. Further- chaos particle swarm optimization (CPSO), SA, pattern
more, solar cell performance parameters, namely, open cir- search (PS), HS-based algorithms, and ABSO.
cuit voltage, short circuit current, fill factor and conversion The rest of this paper is arranged as follows: Section 2
efficiency that are useful in analyzing the system perfor- provides a description of the solar cell models; In Section 3,
mance, are derived from the I–V curve and the information the parameters identification process will be represented in
provided by the model parameters. The parameters of the detail; MPP problem is given in Section 4, and finally, con-
electrical equivalent circuit for a solar cell which needs to clusion is presented is Section 5.
be identified are the photo-generated current, diode satura-
tion current, series resistance, shunt resistance, and diode
ideality factor. The ultimate aim of the parameter identifi- 2. Solar cell models
cation is to extract the optimal values of the parameters
with which the model can be able to describe the system Several models have been developed to represent the
performance as well as the actual one. behavior of the system under different operating conditions
Solving the parameter identification problem can be (Han et al., 2004; Villalva et al., 2009; Huld et al., 2010).
addressed with traditional (Easwarakhanthan et al., 1986; They vary from models with simple assumptions to
Chan et al., 1986; Jian and Kapoor, 2004; Saleem and Kar- advanced models accompanying with many physical vari-
malkar, 2009) and metaheuristic (Zagroubaa et al., 2010; ables. Nevertheless, two models are only used in practice.
AlRashidi et al., 2011; Wei et al., 2011; El-Naggar et al., These models will be briefly discussed in the following
2012; Askarzadeh and Rezazadeh, 2012, 2013) optimization subsections.
algorithms. Traditional optimization algorithms, needing
continuity, convexity and differentiability conditions for
being applicable, usually include heavy computations,
2.1. Double diode model
require derivative information, and converge to a local opti-
mal rather than a global one. Owing to the non-linearity of
Under illumination, an ideal solar cell is modeled as a
the solar cell models, these algorithms cannot successfully
light generated current source in parallel with a rectifying
solve the parameter identification problem. So, it is essential
diode. However, in practice the current source is also
to apply more efficient approaches to conquer the difficulty
shunted with another diode to consider the space charge
of the problem and obtain more promising results. In recent
recombination current and a shunt leakage resistor to take
years, because of their great potentials, metaheuristic opti-
into account the partial short circuit current path near the
mization algorithms such as genetic algorithm (GA)
cell’s edges related to the semiconductor impurities and
(AlRashidi et al., 2011; Zagroubaa et al., 2010), particle
non-idealities. In addition, a series resistor is connected
swarm optimization (PSO) (Wei et al., 2011), simulated
with the cell shunt elements due to the solar cell metal con-
annealing (SA) (El-Naggar et al., 2012), harmony search
tacts and the semiconductor material bulk resistance (Wolf
(HS) (Askarzadeh and Rezazadeh, 2012), and artificial bee
et al., 1977). The equivalent circuit of the double diode
swarm optimization algorithm (ABSO) (Askarzadeh and
model is shown in Fig. 1.
Rezazadeh, 2013) have been suggested for the parameters
The terminal current, It, can be expressed as follows:
identification problem. Metaheuristic algorithms are suit-
able choices for solving this problem owing to their global I t ¼ I ph  I d1  I d2  I sh ð1Þ
search power as well as derivative-free advantage.
Bird mating optimizer (BMO) is a recently devised meta-
heuristic algorithm (Askarzadeh and Rezazadeh, 2012,
2013) which tries to metaphorically imitate the mating
strategies of bird species. BMO has simple concept and
makes use of four distinct searching patterns of monog-
amy, polygyny, polyandry and promiscuity to increase
the probability of getting close to the global solution.
BMO has been successfully applied to different fields
(Askarzadeh and Rezazadeh, 2012, 2013) and has pro-
duced promising results in comparison with the other
metaheuristic algorithms.
In this paper, BMO is used to identify the optimal
parameters of a 57 mm diameter commercial (RTC France) Fig. 1. The double diode model of solar cell.
A. Askarzadeh, A. Rezazadeh / Solar Energy 90 (2013) 123–133 125

where Iph is the photo-generated current, Id1 denotes the


first diode current, Id2 is the second diode current, and Ish
denotes the shunt resistor current.
By using Shockley equation for the diodes currents and
substituting the current of the shunt resistor, Eq. (1) is
rewritten as represented in the following equation:
   
qðV t þ Rs I t Þ
I t ¼ I ph  I sd1 exp 1
n1 kT
   
qðV t þ Rs I t Þ V t þ Rs I t
 I sd2 exp 1  ð2Þ
n2 kT Rsh
Fig. 2. The single diode model of solar cell.

where Isd1 and Isd2 are the diffusion and saturation currents,
   
respectively, Vt is the terminal voltage, Rs and Rsh are the qðV t þ Rs I t Þ
series and shunt resistances, q is the electronic charge, k de- fðV t ; I t ; xÞ ¼ I t  I ph þ I sd1 exp 1
n1 kT
notes the Boltzmann constant, n1 and n2 are the diffusion    
qðV t þ Rs I t Þ
and recombination diode ideality factors, and T (K) is þ I sd2 exp 1
n2 kT
the cell temperature.
As Eq. (2) shows, the model consists of seven unknown V t þ Rs I t
þ ð4Þ
parameters, namely, Rs, Rsh, Iph, Isd1, Isd2, n1, and n2. To Rsh
model the system performance as well as the actual one,    
qðV t þ Rs I t Þ
the optimal values of the parameters must be identified. fðV t ; I t ; xÞ ¼ I t  I ph þ I sd exp 1
nkT
Accurate modeling can be performed by the help of a capa-
V t þ Rs I t
ble optimization technique and a set of experimental data þ ð5Þ
obtained from the system. Rsh
The value of f is calculated for each pair of the experi-
mental data. We use the root mean square error (RMSE)
2.2. Single diode model as a criterion to quantify the difference between the model
results and the experimental data. RMSE is defined by the
Single diode model is the most common representation following equation.
of the solar cell behavior. This model is built by combining rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
together both diodes currents, under the introduction of a 1 XN 2
RMSE ¼ ðfi ðV t ; I t ; xÞÞ ð6Þ
non-physical diode ideality factor, n. The equivalent circuit N i¼1

of this model is shown in Fig. 2.


where N is the number of the experimental data.
The representation of this model can be formulated as
During the optimization process, the objective function
follows:
is to be minimized with respect to the parameters range.
    The upper and lower bounds of the parameters, provided
qðV t þ Rs I t Þ V t þ Rs I t
I t ¼ I ph  I sd exp 1  ð3Þ by the literature survey, are shown in Table 1. The model
nkT Rsh
parameters (five or seven based on the selected model)
are successively adjusted by the optimization algorithm,
In the single diode model, the number of the unknown
until a termination criterion is met. It is clear that the smal-
parameters reduces to five, namely, Rs, Rsh, Iph, Isd, and n.
ler the objective function value, the better the solution is.

3. Parameter identification 3.2. Bird mating optimizer

3.1. Objective function The population of BMO algorithm is called a society


and each individual in the society is called a bird. The
The unknown parameters can be identified by the help
of an optimization technique using a set of experimental Table 1
I–V data obtained from the real system. In the optimiza- Upper and lower range of the solar cell parameters.
tion algorithm, each solution is defined by a vector, x, Parameter Lower Upper
where x = [Rs Rsh Iph Isd1 Isd2 n1 n2] in the double diode Rs (O) 0 0.5
model and x = [Rs Rsh Iph Isd n] in the single diode model. Rsh (O) 0 100
To define the objective function, Eqs. (2) and (3) are rewrit- Iph (A) 0 1
ten in their homogeneous forms as follows: Isd (lA) 0 1
n 1 2
126 A. Askarzadeh, A. Rezazadeh / Solar Energy 90 (2013) 123–133

society contains four types of birds: monogamous, polyg- for j¼1:d


ynous, polyandrous, and promiscuous, breeding in a d- if r1 < mcf
dimensional search space, S  Rd, to find the optimum xðbrood; jÞ ¼ xðl; jÞ þ w  r2  ðxðef ; jÞ  xðl; jÞÞ
solution. Assume that we have a set of birds in a society
else
indicated by v. The birds of the society are categorized
based on their fitness values so that v = l [ j [ w [ n, xðbrood; jÞ ¼ xðl; jÞ þ mw  ðr3  r4 Þ  ðuðjÞ  lðjÞÞ
where l, j, w, and n represent the set of monogamous, end
polygynous, polyandrous, and promiscuous birds, respec- end
tively. Each bird is associated with a predefined number ð7Þ
of genes and shown by a vector ~ x(v) = (x(v, 1), x(v, 2),
. . . , x(v, d)). In the society, any bird is a feasible solution where ~ xðlÞ, and ~
xðbroodÞ, ~ xðef Þ are, respectively, the resul-
of the optimization problem with a quality represented tant brood, monogamous bird, and interesting elite female,
by fit(~ x(v)). The birds attempt to pass on better genes d denotes the problem dimension, j is the variable index, w
to their broods. Consequently, as the algorithm pro- is a time-varying weight to adjust the importance of the
gresses, the quality of the bird’s society improves. For elite female, r1, r2, r3 and r4 are normally distributed ran-
convenience of computation, we assume that there is dom numbers between 0 and 1, mw denotes mutation
only one brood when a bird mates with other one(s). weight, and u(j) and l(j) are the upper and lower bounds
The society is then updated with the better birds. of variable jth, respectively.
The breeding among the society continues until a crite- In order to select the interesting elite, we resort to the
rion named maximum number of generations, genmax, roulette wheel approach. In this approach, as the quality
is met. of a bird increases, the probability of its selection increases,
In the proposed algorithm, it is also assumed that the too. In roulette wheel approach, the selection probability
birds of a society can switch their types during generations. of the bird kth from a group including m birds is defined
At each generation, society birds, which have the most by the following formula:
promising genes, are chosen as polyandrous birds
(females). They have the best fitness values among the soci- 1=fitð~

ety. A predefined percentage of the other birds which have p k ¼ Pm ð8Þ
xi Þ
i¼1 1=fitð~
the worst fitness are abandoned from the society and
replaced by new ones produced using a chaotic sequence. Based on its selection probability, each candidate bird is
The new birds are considered as promiscuous. The remain- devoted a range between 0 and 1. The birds with better
ing birds of the society are regarded as monogamous and quality have wider range than the others. Then, a random
polygynous birds. Monogamous birds have better fitness number is uniformly generated between 0 and 1. That
than polygynous ones. Monogamous, polygynous and pro- range which includes the generated number is specified
miscuous birds make the males in the society. In BMO, the and the corresponding bird is selected as the interesting
percentage of each type is determined manually, however, elite bird. It is obvious that the birds with better quality
monogamous and polygynous types have a great portion have more chance of being selected.
and polyandrous and promiscuous types have a low per- Polygynous birds are those males that have a tendency
centage of the society. to couple with multiple females. In nature, a polygynous
Monogamous birds are those males that tend to mate bird mates with several females resulting in a number of
with one female. During mating season, a monogamous broods, but in BMO this behavior is metaphorically
bird starts to sing and tries to attract female birds. Polyan- adopted in which by mating a polygynous bird with multi-
drous birds receive his song and gather at the vicinity of ple females only one brood is raised which its genes are a
him. They employ intelligent behaviors such as dancing combination of the females genes. After mating a polygy-
or tail drumming to catch the attention of the male bird. nous bird with his interesting elite females, the resultant
The ultimate aim of the male bird is to pass on better genes brood is given as follows:
to his brood by combining his genes with the genes of his
interesting elite female. Therefore, he evaluates the quality for j¼1:d
of the females, employs a probabilistic approach to select
if r1 < mcf
one of them as his interesting elite female, and mates with nef
X
her. Female birds with more promising genes have a more xðbrood; jÞ ¼ xðj; jÞ þ w  ri  ðxðefi ; jÞ  xðj; jÞÞ
chance of being selected. Besides, each bird might make a i¼1
mutation in each brood’s gene. The probability of muta- else
tion is controlled by a factor named mutation control fac- xðbrood; jÞ ¼ xðj; jÞ þ mw  ðr2  r3 Þ  ðuðjÞ  lðjÞÞ
tor, mcf, which varies between 0 and 1. This factor helps
end
the algorithm maintains the diversity and avoids premature
convergence. As a result, the resultant brood is produced end
by ð9Þ
A. Askarzadeh, A. Rezazadeh / Solar Energy 90 (2013) 123–133 127

where ~xðjÞ and ~xðefi Þ are, respectively, the polygynous bird interesting elite females. The behavior of each promiscuous
and ith elite female, nef denotes the number of elite females, bird is the same as that of monogamous bird. As a result,
and ri are normally distributed random numbers between 0 each gene of the resultant brood is given as follows:
and 1. A polygynous bird mixes the information of more
for j¼1:d
candidate solutions into a new one. This behavior may lead
to raising a brood with more promising genes. The collab- if r1 < mcf
oration of each interesting elite female in each gene of the xðbrood; jÞ ¼ xðn; jÞ þ w  r2  ðxðe; jÞ  xðn; jÞÞ
brood is random because the coefficients ri are indepen- else
dently generated.
xðbrood; jÞ ¼ xðn; jÞ þ mw  ðr3  r4 Þ  ðuðjÞ  lðjÞÞ
A polygynous bird mates with a female by use of an
annealing function with the following probability: end
end
Pr ¼ expðDf =T Þ ð10Þ
ð12Þ
where Pr is the probability of mating, Df denotes the abso-
lute difference between the objective functions of the polyg- where ~ xðnÞ denotes the promiscuous bird.
ynous bird and female one, and T is an adjustable Using a chaotic sequence to produce new feasible solu-
parameter to control the probability. The probability of tions in the search space increases the capability of the
mating is high when the quality of the polygynous bird is algorithm to discover potential solutions in as yet untested
as good as the favorite female’s quality, or when the value regions of the space. Chaos has some good properties such
of T is high. However, a random number between 0 and 1 as ergodicity, stochastic properties, and regularity. A cha-
is generated and compared with the calculated probability. otic sequence can go through every state in a certain area
If it is less than the calculated probability, that female bird according to its own regularity, and every state in experi-
is selected for mating. Otherwise, the selection of that fe- enced only once. Therefore, BMO can more easily escape
male is failed. from local optima by using chaotic movement. At the ini-
In BMO, each polyandrous bird seeks for superior males tial generation, each promiscuous bird is produced using
to breed a brood with high-quality genes. Consequently, Eq. (13), where z is chaos variable and its initial value is
she makes aware the best males of the society (monoga- a random number between 0 and 1 (not the points of
mous birds) of her conditions for mating. The males receive 0.25, 0.50 and 0.75). At the next generation, the parameter
her signs and participate in her ritual. In order to increase of z is firstly updated by the well-known Logistic map using
the probability of raising a good brood a predefined num- Eq. (14) and then, the new promiscuous bird is produced.
ber of monogamous birds which have a better quality than for j¼1:d
the others participate in this ritual. The female bird evalu-
xðn; jÞ ¼ lðjÞ þ zgen  ðuðjÞ  lðjÞÞ ð13Þ
ates the quality of the males, employs a probabilistic
approach to select her interesting elite males, and mates end
with them. Each gene of the resultant brood is obtained zgenþ1 ¼ 4zgen ð1  zgen Þ ð14Þ
as follows:
At the end of each generation, associated with each bird
for j¼1:d
one brood has been raised. Replacement is the next stage.
if r1 < mcf At this stage, any bird makes a decision to replace its brood
X
nem
xðbrood; jÞ ¼ xðw; jÞ þ w  ri  ðxðemi ; jÞ  xðw; jÞÞ instead of itself or not. Therefore, the bird evaluates the
i¼1 quality of its brood. If the brood has better genes than
else the bird, the bird abandons the society and the brood
xðbrood; jÞ ¼ xðw; jÞ þ mw  ðr2  r3 Þ  ðuðjÞ  lðjÞÞ attaches to it, otherwise, the brood is abandoned and the
end bird stays in the society. The flowchart and pseudocode
of BMO algorithm have been represented in Figs. 3 and
end
4, respectively.
ð11Þ
where ~ xðemi Þ is the ith elite
xðwÞ is the polyandrous bird, ~ 3.3. Simulation results
male, and nem denotes the number of interesting elite males.
Polyandrous birds use the annealing function to select In order to evaluate the usefulness of the BMO-based
their males, too. However, in order to increase the proba- parameter identification method, the I–V characteristic of
bility of producing good broods a predefined percentage a 57 mm diameter commercial (RTC France) silicon solar
of monogamous birds with better qualities participate in cell is considered. The experimental data has been adopted
this ritual. from the system under 1 sun (1000 W/m2) at 33 °C (Easwa-
As previously mentioned, promiscuous birds are pro- rakhanthan et al., 1986). The adjustable parameters of BMO
duced using a chaotic sequence. With different qualities, algorithm, tuned by trial, are as follows: The society size is
they attend during each generation and mate with their set to 30 of which 13, 7, 5, and 5 birds are, respectively,
128 A. Askarzadeh, A. Rezazadeh / Solar Energy 90 (2013) 123–133

Fig. 3. Flowchart of BMO algorithm.

monogamous, polygynous, polyandrous, and promiscuous; parameter is selected 0.9; and maximum number of genera-
T, w, and mw are defined as decreasing linear functions, tions is set to 5000.
where Tmax = 300, Tmin = 50, wmax = 2.5, wmin = 0.5, BMO-based method for the parameters identification of
mw,max = 0.01, and mw,min = 0.0001; mutation control the double and single diode models is coded and executed
A. Askarzadeh, A. Rezazadeh / Solar Energy 90 (2013) 123–133 129

Fig. 4. Pseudocode of BMO algorithm.

in the Matlab environment. The optimal parameters of the it has found the smallest RMSE value. Table 3 shows the
I calculated
double diode model along with the RMSE value are listed relative error, e ¼ I measured
I measured
, for each measurement
in Table 2. This table also shows the results found by the along withPN the mean absolute error (MAE),
other state-of-the-art metaheuristic methods: pattern MAE ¼ N1 i¼1 jei j. The calculated value of the MAE con-
search (PS) (AlHajri et al., 2012), simulated annealing firms the high accuracy of the identification process.
(SA) (El-Naggar et al., 2012), harmony search (HS) and In order to illustrate the fitting precision, the optimal
two of its variants, namely, grouping-based global har- parameters found by BMO are returned to the double
mony search (GGHS) as well as innovative global harmony diode model and the I–V characteristic is reconstructed.
search (IGHS) (Askarzadeh and Rezazadeh, 2012), and This is simply performed by applying Newton method
artificial bee swarm optimization algorithm (ABSO) when It is unknown while Vt is known. Fig. 5 shows the
(Askarzadeh and Rezazadeh, 2013). The best value found comparison between the experimental data and the results
by the algorithms is shown in bold. A comparison among produced by the double diode model using the optimal
the performance of the algorithms indicates that BMO pro- parameters. It can be seen that the I–V characteristic
duces better results than the other solution methods, since obtained by the identified model using BMO algorithm is
130 A. Askarzadeh, A. Rezazadeh / Solar Energy 90 (2013) 123–133

Table 2
A comparison between the results obtained by BMO algorithm and the other ones for the double diode model parameter identification.
Item BMO PS SA HS GGHS IGHS ABSO
Rs (O) 0.03682 0.0320 0.0345 0.03545 0.03562 0.03690 0.03657
Rsh (O) 55.8081 81.3008 43.1034 46.82696 62.7899 56.8368 54.6219
Iph (A) 0.76078 0.7602 0.7623 0.76176 0.76056 0.76079 0.76078
Isd1 (lA) 0.21110 0.9889 0.4767 0.12545 0.37014 0.97310 0.26713
Isd2 (lA) 0.87688 0.0001 0.0100 0.25470 0.13504 0.16791 0.38191
n1 1.44533 1.6000 1.5172 1.49439 1.49638 1.92126 1.46512
n2 1.99997 1.1920 2.0000 1.49989 1.92998 1.42814 1.98152
RMSE 9.8262e4 0.01518 0.01664 0.00126 0.00107 9.8635e4 9.8344e4

Table 3
Relative error for each measurement and MAE value based on the
extracted parameters (double diode model).
Measurement Vt (V) It measured It calculated Relative
(A) (A) error
1 0.2057 0.764 0.763965 4.64e05
2 0.1291 0.762 0.762593 0.00078
3 0.0588 0.7605 0.761334 0.0011
4 0.0057 0.7605 0.760177 0.000425
5 0.0646 0.76 0.759117 0.001162
6 0.1185 0.759 0.758135 0.001139
7 0.1678 0.757 0.757205 0.00027
8 0.2132 0.757 0.756262 0.000975
9 0.2545 0.7555 0.755193 0.000406
10 0.2924 0.754 0.753732 0.000355
11 0.3269 0.7505 0.751397 0.0012
12 0.3585 0.7465 0.747287 0.00105
Fig. 5. Comparison between the I–V characteristics resulted from the
13 0.3873 0.7385 0.739973 0.00199
experimental data and the double diode model.
14 0.4137 0.728 0.727243 0.001039
15 0.4373 0.7065 0.706819 0.00045
16 0.459 0.6755 0.675224 0.000408
17 0.4784 0.632 0.630895 0.001749
individuals is more flexible than that algorithm which uses
18 0.496 0.573 0.572157 0.001472 a same pattern to update the position of its seekers. Seeking
19 0.5119 0.499 0.499589 0.00118 a search space with different patterns increases the proba-
20 0.5265 0.413 0.413569 0.00138 bility of maintaining the diversity and avoiding premature
21 0.5398 0.3165 0.317245 0.00235 convergence.
22 0.5521 0.212 0.212075 0.00035
23 0.5633 0.1035 0.102659 0.008123
The results of BMO and those found by chaos particle
24 0.5736 0.01 0.00931 0.069173 swarm optimization (CPSO) (Wei et al., 2011), genetic algo-
25 0.5833 0.123 0.12439 0.01131 rithm (GA) (AlRashidi et al., 2011), pattern search (PS)
26 0.59 0.21 0.20914 0.004117 (AlHajri et al., 2012), simulated annealing (SA) (El-Naggar
MAE 0.00439 et al., 2012), harmony search-based algorithms (Askarzadeh
and Rezazadeh, 2012), and artificial bee swarm optimization
algorithm (ABSO) (Askarzadeh and Rezazadeh, 2013) for
in good accordance with the experimental one meaning the parameters identification of the solar cell single diode model
high quality of the identification process. The power vs. are summarized in Table 4. It can be seen that BMO yields
voltage (P–V) characteristic of the system is also indicated better results than CPSO, GA, PS, SA, HS, GGHS, IGHS,
in Fig. 6. and ABSO. Table 5 lists the relative error of each measure-
The convergence rate of the proposed algorithm during ment along with the MAE. The optimal parameters are
the identification process is illustrated in Fig. 7, represent- put into the single diode model and Newton method is used
ing the best value of the objective function during the iter- to reconstruct the I–V characteristic. Fig. 8 indicates that the
ations. It is clear that the convergence rate of BMO is I–V characteristic found by BMO is in good agreement with
superior. As Fig. 7 shows, at the first iterations the algo- the experimental data. P–V characteristic and convergence
rithm discovers good regions of the search space and con- process of BMO are also represented in Figs. 9 and 10,
verges to the optimal solution quickly. The success of BMO respectively. The superior performance of BMO is clear.
for solving the identification problem can be explained by A comparison between the obtained results indicates
this fact that it makes use of distinct moving patterns to that the accuracy of the double diode model is slightly
explore the search space. An optimization algorithm which more than that of the single diode model, because the
employs distinct patterns to update the position of its RMSE value for the double diode model is 9.8262e4
A. Askarzadeh, A. Rezazadeh / Solar Energy 90 (2013) 123–133 131

Table 4
A comparison between the results obtained by BMO and the other ones
for the single diode model parameter identification.
Item BMO CPSO GA PS SA
Rs (O) 0.03636 0.0354 0.0299 0.0313 0.0345
Rsh (O) 53.8716 59.012 42.3729 64.1026 43.1034
Iph (A) 0.76077 0.7607 0.7619 0.7617 0.7620
Isd (lA) 0.32479 0.4000 0.8087 0.9980 0.4798
n 1.48173 1.5033 1.5751 1.6000 1.5172
RMSE 9.8608e4 0.00139 0.01908 0.01494 0.01900
HS GGHS IGHS ABSO
Rs (O) 0.03663 0.03631 0.03613 0.03659
Rsh (O) 53.5946 53.0647 53.2845 52.2903
Iph (A) 0.76070 0.76092 0.76077 0.76080
Isd (lA) 0.30495 0.32620 0.34351 0.30623
n 1.47538 1.48217 1.48740 1.47583
RMSE 9.9510e4 9.9097e4 9.9306e4 9.9124e4
Fig. 6. Comparison between the P–V characteristics resulted from the
experimental data and the double diode model.

Table 5
Relative error for each measurement and MAE value based on the
0.14 extracted parameters (single diode model).
Measurement Vt (V) It Measured It Calculated Relative
0.12
(A) (A) error
0.1 1 0.2057 0.764 0.764076 9.89e05
Objective function

2 0.1291 0.762 0.762655 0.00086


0.08 3 0.0588 0.7605 0.76135 0.00112
4 0.0057 0.7605 0.760153 0.000456
5 0.0646 0.76 0.759058 0.001239
0.06
6 0.1185 0.759 0.758048 0.001254
7 0.1678 0.757 0.757099 0.00013
0.04
8 0.2132 0.757 0.756152 0.001121
9 0.2545 0.7555 0.755099 0.000531
0.02 10 0.2924 0.754 0.753676 0.000429
11 0.3269 0.7505 0.751399 0.0012
0 12 0.3585 0.7465 0.747357 0.00115
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
13 0.3873 0.7385 0.740102 0.00217
Iteration 14 0.4137 0.728 0.727395 0.000831
Fig. 7. Convergence process of BMO algorithm during the identification 15 0.4373 0.7065 0.706945 0.00063
process of the double diode model. 16 0.459 0.6755 0.675281 0.000324
17 0.4784 0.632 0.630868 0.001791
18 0.496 0.573 0.572067 0.001628
while this value is 9.8608e4 for the single diode model. 19 0.5119 0.499 0.499482 0.00097
20 0.5265 0.413 0.41349 0.00119
However, the error of 0.35% implies the suitability and pre- 21 0.5398 0.3165 0.317223 0.00228
cision of the single diode model. 22 0.5521 0.212 0.212111 0.00053
23 0.5633 0.1035 0.102731 0.007433
4. Maximum power point (MPP) 24 0.5736 0.01 0.00924 0.075761
25 0.5833 0.123 0.12438 0.01124
26 0.59 0.21 0.2092 0.003803
A derivative of the output power, Pt, with respect to the MAE 0.004621
terminal voltage, Vt, is equal to zero at MPP. Using the sin-
gle diode model, at the MPP Eq. (15) will be satisfied
2   3 After solving Eq. (15), the current and voltage values
dP t
q
nkT
I ph þ I sd  I t  V t þR
Rsh
sIt
þ R1sh corresponding to the MPP are 0.692 (A) and 0.449 (V),
¼ It  V t4   5 respectively. The corresponding power value is 0.311 (W).
dV t 1þ qRs
I ph þ I sd  I t  V t þRs I t
þ Rs
nkT Rsh Rsh In general, the well used approach to determine the MPP
in solar cells is based on maximum power selection from
¼0 ð15Þ
a set of I–V multiplication. The precision of this approach
MPP can be extracted by putting the identified electrical depends on the accuracy of the experimental data, which
parameters given in Table 4 found by BMO algorithm into should surround the real value of the MPP. However,
Eq. (15) and using standard numerical non-linear method. our proposed approach does not include these limitations
This can be acquired simply with the optimization Toolbox and makes use of the electrical parameters identified by a
of Matlab software. global search algorithm.
132 A. Askarzadeh, A. Rezazadeh / Solar Energy 90 (2013) 123–133

5. Conclusion

Maximum power point (MPP) of solar cells can be


extracted by using the electrical parameters. However, the
accuracy of the parameters will determine the precision of
the final result. In this paper, we propose bird mating opti-
mizer (BMO) algorithm to provide an efficient parameter
identification process. Using several seeking patterns is the
main ability of BMO to look for the global solution. The
accuracy of BMO for parameter identification is more than
those of the other studied algorithms. The better perfor-
mance of parameter identification leads to obtaining the
coordinates of the MPP more accurately. It can be drawn
that BMO algorithm can be a helpful and efficient candidate
for solving optimization problems related to solar cells.
Fig. 8. Comparison between the I–V characteristics resulted from the
experimental data and the single diode model.
Acknowledgement

The financial support of the Graduate University of Ad-


vanced Technology is greatly acknowledged.

References

AlHajri, M.F., El-Naggar, K.M., AlRashidi, M.R., Al-Othman, A.K.,


2012. Optimal extraction of solar cell parameters using pattern search.
Renewable Energy 44, 238–245.
AlRashidi, M.R., AlHajri, M.F., El-Naggar, K.M., Al-Othman, A.K.,
2011. A new estimation approach for determining the I–V character-
istics of solar cells. Solar Energy 85 (7), 1543–1550.
Askarzadeh, A., Rezazadeh, A., 2012a. A new heuristic optimization
algorithm for modeling of proton exchange membrane fuel cell: bird
mating optimizer. International Journal of Energy Research doi.
http://dx.doi.org/10.1002/er.2915.
Askarzadeh, A., Rezazadeh, A., 2012b. Parameter identification for solar
cell models using harmony search-based algorithms. Solar Energy 86,
3241–3249.
Fig. 9. Comparison between the P–V characteristics resulted from the Askarzadeh, A., Rezazadeh, A., 2013a. Artificial bee swarm optimization
experimental data and the single diode model. algorithm for parameters identification of solar cell models. Applied
Energy 102, 943–949.
Askarzadeh, A., Rezazadeh, A., 2013b. Artificial neural network training
using a new efficient optimization algorithm. Applied Soft Computing
13, 1206–1213.
0.35 Bahgat, A.B.G., Helwa, N.H., Ahamd, G.E., El Shenawy, E.T., 2004.
Estimation of the maximum power and normal operating power of a
0.3 photovoltaic module by neural networks. Renewable Energy 29 (3),
443–457.
0.25
Chan, D.S.H., Phillips, J.R., Phang, J.C.H., 1986. A comparative study of
Objective function

extraction methods for solar cell model parameters. Solid-State


Electronics 29 (3), 329–337.
0.2
Easwarakhanthan, T., Bottin, J., Bouhouch, I., Boutrit, C., 1986.
Nonlinear minimization algorithm for determinng the solar cell
0.15 parameters with microcomputers. International Journal of Solar
Energy 4, 1–12.
0.1 El-Naggar, K.M., AlRashidi, M.R., AlHajri, M.F., Al-Othman, A.K.,
2012. Simulated annealing algorithm for photovoltaic parameters
0.05 identification. Solar Energy 86, 266–274.
Enrique, J.M., Duran, E., Sidrach-de-Cardona, M., Andujar, J.M., 2007.
Theoretical assessment of the maximum power point tracking
0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 efficiency of photovoltaic facilities with different converter topologies.
Iteration Solar Energy 81 (1), 31–38.
Han, L., Koide, N., Chiba, Y., Mitate, T., 2004. Modeling of an
Fig. 10. Convergence process of BMO algorithm during the identification equivalent circuit for dye-sensitized solar cells. Applied Physics Letters
process of the single diode model. 13, 2433–2435.
A. Askarzadeh, A. Rezazadeh / Solar Energy 90 (2013) 123–133 133

Huld, T., Gottschalg, R., Beyer, H.G., Topic, M., 2010. Mapping the Wei, H., Cong, J., Lingyun, X., Deyun, S., 2011. Extracting solar cell
performance of a PV modules, effects of module type and data model parameters based on chaos particle swarm algorithm. In:
averaging. Solar Energy 84, 324–328. International Conference on Electric Information and Control Engi-
Ikegami, T., Maezono, T., Nakanishi, F., Yamagata, Y., Ebihara, K., neering (ICEICE), pp. 398–402.
2001. Estimation of equivalent circuit parameters of PV module and its Wolf, M., Noel, G., Stirm, R., 1977. Investigation of the double
application to optimal operation of PV system. Solar Energy Materials exponential in the current–voltage characteristics of silicon solar cells.
& Solar Cells 67, 389–395. IEEE Transactions on Industrial Electronics 53 (4), 1017–1026.
Jian, A., Kapoor, A., 2004. Exact analytical solutions of the parameters of Yu, G.J., Jung, Y.S., Choi, J.Y., Kim, G.S., 2004. A novel two-mode
real solar cells using Lambert W-function. Solar Energy Materials and MPPT control algorithm based on comparative study of existing
Solar cells 81 (2), 269–277. algorithms. Solar Energy 76, 455–463.
Saleem, H., Karmalkar, S., 2009. An analytical method to extract the Zagroubaa, M., Sellamia, A., Bouaichaa, M., Ksourib, M., 2010.
physical parameters of a solar cell from four points on the illuminated Identification of PV solarcells and modules parameters using the
J–V curve. IEEE Electron Device Letters 30 (4), 349–352. genetic algorithms: application to maximum power extraction. Solar
Villalva, M.G., Gazoli, J.R., Filho, E.R., 2009. Comprehensive approach Energy 84 (5), 860–866.
to modeling and simulation of photovoltaic arrays. IEEE Transactions
on Power Electronics 24 (5), 1198–1208.

You might also like