Hindawi Publishing Corporation

Journal of Applied Mathematics

Volume 2013, Article ID 362619, 8 pages
http://dx.doi.org/10.1155/2013/362619

Research Article
Parameter Estimation of Photovoltaic Models via Cuckoo Search

Jieming Ma,1,2 T. O. Ting,2 Ka Lok Man,2 Nan Zhang,2

Sheng-Uei Guan,2 and Prudence W. H. Wong1
1
Department of Computer Science, University of Liverpool, Ashton Building, Ashton Street, Liverpool, L69 3BX, UK
2
Xi’an Jiaotong-Liverpool University, 111 Ren’ai Road, Suzhou Dushu Lake, HET, Jiangsu Province 215123, China

Correspondence should be addressed to T. O. Ting; [email protected]

Received 27 June 2013; Accepted 11 July 2013

Academic Editor: Xin-She Yang

Copyright © 2013 Jieming Ma et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Since conventional methods are incapable of estimating the parameters of Photovoltaic (PV) models with high accuracy, bioinspired
algorithms have attracted significant attention in the last decade. Cuckoo Search (CS) is invented based on the inspiration of
brood parasitic behavior of some cuckoo species in combination with the Lévy flight behavior. In this paper, a CS-based parameter
estimation method is proposed to extract the parameters of single-diode models for commercial PV generators. Simulation results
and experimental data show that the CS algorithm is capable of obtaining all the parameters with extremely high accuracy, depicted
by a low Root-Mean-Squared-Error (RMSE) value. The proposed method outperforms other algorithms applied in this study.

1. Introduction and the environmental variables, such as the operating

voltage (𝑉), the ambient temperature (𝑇), and the irradiance
Photovoltaic (PV) cells, normally assembled into modules or (𝐺). Over the years, significant research efforts have been
arrays on mounting systems, are capable of producing elec- contributing to the development of the behavioral models [4–
trons when photons strike its surface. Taking the advantages 8]. Among numerous modeling approaches, the Single-Diode
of many promising features like renewability, less pollution, Model (SDM) is the most widely utilized PV model in the
and ease of installation, PV generators are envisaged to be an literature. A general SDM includes five parameters, namely,
important energy source for the future. photocurrent (𝐼pv ), saturation current (𝐼𝑜 ), diode ideality
Due to the high initial cost of a PV-supplied system, constant (𝑛), series resistance (𝑅𝑠 ), and shunt resistance (𝑅𝑝 ).
predictive performance tools are widely used by engineers to In order to adapt PV model behavior to different operating
optimize the system performance [1, 2]. PV manufacturers conditions, de Blas et al. [9] suggested to apply the procedure
normally provide limited tabular data measured under the described in the International Standard IEC 891 that relates
Standard Test Conditions (STCs), which correspond to a cell current and voltage of the PV characteristics at given values of
temperature of 25∘ C and an irradiance of 1000 W/m2 at 1.5 air 𝑇 and 𝐺 with the corresponding values at different operating
mass spectral distributions. As reported in [3], PV generators environments. The Improved Single Diode Model (ISDM)
always operate under environments far from the STCs. Due presented by De Soto et al. [5] includes the dependence
to this reason, the data available in the datasheet usually fail of the PV parameters on operating conditions. The normal
to fulfill the engineering requirements. parameters at the STCs are necessary to be determined in
PV model, with the ability to predict 𝐼-𝑉 characteristics this model. Both SDM and ISDM are adopted in this study
of PV generators under an operating environment other of parameter estimation.
than the STCs, is a predictive performance tool that allows Analytical methods [5, 10–12] are common approaches
consumers to maximize the cost effectiveness of the system in estimating the parameters by mathematical equations.
before installation [2]. They are generally analytical equations Although having the merit of simplicity, it is hard to fur-
based on a physical description that formulate PV generated ther reduce the errors of the estimated values. Further-
current (𝐼) with the most crucial technical characteristics more, analytical methods utilize the 𝐼-𝑉 curve features or
2 Journal of Applied Mathematics

semiconductor parameters that are unavailable in the I

datasheet. This problem often reduces its feasibility. Recently,
Rs +
PV parameter estimation is deemed as a multidimensional Id
optimization problem. Several computational intelligence Ipv
Rp V
methods, such as Genetic Algorithms (GA) [13], Chaos
Particle Swarm Optimization (CPSO) [14], Firefly [15], and
−
Pattern Search (PS) [16], were proposed in the literature.
These algorithms usually extract relevant parameters by
minimizing the Root Mean Square Error (RMSE) as the Figure 1: The equivalent circuit of the SDM.
objective function in the optimization process. Askarzadeh
and Rezazadeh [17] reported that the optimization methods
produce better results than analytical methods. where 𝑘 and 𝑞 represent the Boltzmann constant (1.380650
Cuckoo Search (CS) is a nature-inspired optimization × 10−23 J/K) and the electron charge (1.602176 × 10−19 C),
algorithm based on the fascinating breeding behavior such respectively.
as brood parasitism of certain species of cuckoos. In [18, 19], SDM assumes that the superposition principle holds;
Yang and Deb reported that the CS algorithm outperforms that is, the total characteristic is the sum of the dark and
Particle Swarm Optimization (PSO) and GA algorithms for illuminated characteristics [3–5]. As expressed in (3) below,
various standard test functions. In this paper, a CS-based the terminal current 𝐼 is therefore equal to 𝐼pv subtracting the
parameter estimation method for the SDM and ISDM is pre- current diverting through the diode and 𝑅𝑠 . The equivalent
sented. Simulation and experimental results show superior circuit of the SDM is shown in Figure 1.
accuracy and feasibility of the proposed parameter estimation
𝑉 + 𝐼𝑅𝑠
method. 𝐼 = 𝐼pv − 𝐼𝑜 (𝑒(𝑉+𝐼𝑅𝑠 )/𝑛𝑉𝑡 − 1) − . (3)
The rest of the paper is organized as follows. Section 2 𝑅𝑝
explains both PV models (SDM and ISDM) used in this work.
The objective function formulation is given in Section 3. This PV module is a particular case of the PV cells connected
is followed by results and discussions in Section 4. The results in series. If the number of the connected cells is up to 𝑁𝑠 , 𝑉𝑡
comparison is also available here, and finally the conclusions is scaled to 𝑁𝑠 times. The model equation is then rewritten as
are derived in Section 5. 𝑉 + 𝐼𝑅𝑠
𝐼 = 𝐼pv − 𝐼𝑜 (𝑒(𝑉+𝐼𝑅𝑠 )/𝑛𝑁𝑠 𝑉𝑡 − 1) − . (4)
𝑅𝑝
2. PV Modeling
In this sense, 𝐼pv , 𝐼𝑜 , 𝑅𝑠 , and 𝑅𝑝 are the corresponding
2.1. Single Diode PV Model (SDM). PV cells are made of parameters of a PV module.
a variety of semiconductor materials using different man-
ufacturing processes. The working principle of PV cells is
essentially on the basis of the PV effect, which refers to 2.2. Improved Single Diode Model (ISDM). The traditional
the generation of a potential difference at the 𝑃-𝑁 junction SDM ignores the operating conditions effect on these param-
in response to visible or other radiation. When a PV cell eters. However, some studies have shown that the parameters,
is exposed to light, the semiconductor materials absorb such as 𝐼pv and 𝐼𝑜 , vary under different environmental
photons, and accordingly charged carriers are generated. conditions. These are due to changes of temperature 𝑇
Potential difference and current in the external circuit lead to and irradiance 𝐺. Aiming to evaluate the PV behavior at
the separation of carriers in the internal electric field created environmental conditions other than the normal values 𝑇𝑛
by the 𝑃-𝑁 junction and collection at the electrodes. The and 𝐺𝑛 , the relations between the operating parameters and
photogenerated charge carriers can be subsequently captured the normal parameters are studied by numerous researchers.
in the form of an electric current, that is, electricity 𝐼pv . In [4], the value of light-generated 𝐼pv is reported to
Eliminated the PV effect, a PV cell behaves like a conventional be linearly dependent on the solar irradiation under the
diode that does not depend on any light parameters. The influence of temperature:
Shockley diode equation is generally used to describe the 𝐺
current flowing through the diode (𝐼𝑑 ): 𝐼pv = (𝐼 − 𝐾𝑖 Δ𝑇) , (5)
𝐺𝑛 pv𝑛
𝐼𝑑 = 𝐼𝑜 (𝑒𝑉𝑑 /𝑛𝑉𝑡 − 1) . (1) where 𝐼pv𝑛 is the light-generated current at the STCs. 𝐾𝑖 , the
In (1), 𝐼𝑜 is the normal diode current, and 𝑉𝑑 represents short-circuit current coefficient, is one of the ISDM parame-
the electrical potential difference between the two ends of the ters. The difference between the standard test temperature 𝑇𝑛
diode. The ideality factor 𝑛 is assumed to be independent of and 𝑇 is denoted by Δ𝑇.
the environment variables 𝑇 and 𝐺. 𝑉𝑡 denotes the thermal Based on the diode theory, Messenger and Ventre [20]
voltage of the PV, and its value can be written as a function of presented an approximate linear expression for the diode
𝑇: saturation current 𝐼𝑜 , which can be expressed as
𝑘𝑇 𝑇 [(𝑞𝐸𝑔 /𝑛𝑘)(1/𝑇𝑛 −1/𝑇)]
𝑉𝑡 = , (2) 𝐼𝑜 = 𝐼𝑜𝑛 ( )𝑒 , (6)
𝑞 𝑇𝑛
Journal of Applied Mathematics 3

where 𝐸𝑔 is the material band gap. Usually, 𝐸𝑔 is set at a

reasonable level depending on the semiconductor materials Cuckoo Search via Lévy Flights
(1.12 eV for crystalline silicon and 1.75 eV for amorphous Initialization of 𝑛 host nests (population)
silicon) in simulation and design tools [21]. De Soto et al. [5] while within the stopping criterion,
present an estimation method for 𝐸𝑔 in a wide temperature Choose a cuckoo egg by Lévy flights and evaluate
range: its fitness (𝐹𝑖 );
Choose an egg in other’s nest randomly and
𝐸𝑔 = 𝐸𝑔𝑛 (1 − 0.0002677Δ𝑇) , (7) calculate its fitness (𝐹𝑗 );
If 𝐹𝑖 > 𝐹𝑗 , replace jth egg by ith egg;
where 𝐸𝑔𝑛 is a normal value at the STCs (𝐸𝑔𝑛 = 1.12 eV A fraction (𝑝𝑎 ) of worse nests are demolished and
for silicon cells and 𝐸𝑔𝑛 = 1.6 eV for the triple junction replaced by new ones;
amorphous cells). Preserve good nests (best solutions).
In [3], Lo Brano et al. study how the series and shunt end while
resistances are affected by the solar irradiance. On the basis of
the experimental data, the values of 𝑅𝑠 and 𝑅𝑝 are observed Pseudocode 1: Pseudocode of the Cuckoo Search (CS) [19].
varying in inverse linear modes with 𝐺:
𝐺𝑛
𝑅𝑠 = 𝑅 , (ii) The best nests with high quality eggs will be carried
𝐺 𝑠𝑛
(8) forward to the next generation.
𝐺𝑛
𝑅𝑝 = 𝑅 , (iii) There are a fixed number of available host nests. If
𝐺 𝑝𝑛 a host bird discovers that the eggs are not its own,
where the values of the resistances 𝑅𝑠𝑛 and 𝑅𝑝𝑛 are evaluated it will either throw these alien eggs away, or it may
under the STCs. abandon the nest and build a brand new nest at a
By using the aforementioned relations, the ISDM nearby location.
described in [5] is able to analytically describe the 𝐼-𝑉
Based on the three rules, the basic steps of CS can be
characteristic of a PV generator for each generic condition of
summarized by the pseudocode shown in Pseudocode 1. In
operative temperature and solar irradiance.
the CS algorithm, a pattern corresponds to a nest while each
individual attribute of the pattern corresponds to an egg laid
3. Parameter Estimation by the cuckoo. On the basis of random-walk algorithms, the
general system equation of the CS algorithm is given in:
3.1. Formulation of Parameter Estimation Problem. PV
parameter estimation is a process that minimizes the differ- 𝑋𝑔+1;𝑖 = 𝑋𝑔;𝑖 + 𝛼 ⊗ Lévy (𝛽) , (11)
ence between the measured data and the calculated current
by adjusting the normal PV parameters. When the number where 𝑔 and 𝑖 denote the generation number (𝑔 =
of experimental data is up to 𝑁, the objective function can be 1, 2, 3, . . . , MaxGen) and the pattern number (𝑖 = 1, 2, . . . , 𝑛),
formulated by RMSE as respectively. The product ⊗ means entry-wise multiplications.
Here 𝛼 > 0 is the step size scaling factor which should be
1 𝑁 2
related to the scales of the problem of interest [19]. The 𝑗th
RMSE = √ ∑ (𝑓 (𝑉, 𝐼, 𝑥)) , (9) attributes of the 𝑖th pattern is initiated by using (12),
𝑁 𝑖=1 𝑖
𝑋𝑔=0;𝑗,𝑖 = rand ⋅ (𝑈𝑏𝑖 − 𝐿𝑏𝑖 ) + 𝐿𝑏𝑖 , (12)
where 𝑥 = [𝐼pv , 𝐼𝑜 𝑛, 𝑅𝑠 , 𝑅𝑝 ] for SDM and 𝑥 =
[𝐼pv𝑛 , 𝐼𝑜𝑛 , 𝑛, 𝑅𝑠𝑛 , 𝑅𝑝𝑛 , 𝐾𝑖 , 𝐸𝑔 ] for ISDM. 𝑓(𝑉, 𝐼, 𝑥) is the where 𝑈𝑏𝑖 and 𝐿𝑏𝑖 are the upper and lower bounds of the
homogeneous form of (4) which expresses the 𝐼-𝑉 𝑗th attributes, respectively. In each computation step, the CS
characteristics of the SDM: algorithm checks whether the value of an attribute exceeds
𝑉 + 𝐼𝑅𝑠 the allowed search range. If this happens, the value of the
𝑓 (𝑉, 𝐼, 𝑥) = 𝐼pv − 𝐼𝑜 (𝑒(𝑉+𝐼𝑅𝑠 )/𝑛𝑁𝑠 𝑉𝑡 − 1) − − 𝐼. (10) related attribute will be updated with the corresponding
𝑅𝑝
boundary value.
For the case of ISDM, 𝐼pv , 𝐼𝑜 , 𝑛, 𝑅, and 𝑅𝑝 satisfy the Before the searching process, the CS algorithm detects the
relational expressions discussed in the previous subsection, most successful pattern as 𝑥best pattern. Among the existing
namely, (5)–(8). algorithms exist for generating Lévy flights in the literature,
Yang and Deb [18, 19] reported that Mantegna’s algorithm
[22] works well in most of the optimization problems.
3.2. Cuckoo Search. The CS algorithm [18, 19], proposed by Accordingly the evolution phase of the pattern initialized
Yang and Deb, is a nature-inspired stochastic global search with the detection step of 𝜙, which is given by (13) [23]:
algorithm that follows three idealized behavior rules.
1/𝛽
(i) A cuckoo lays an egg and dumps it randomly into Γ(1 + 𝛽) ⋅ sin (𝜋 ⋅ 𝛽/2)
𝜙=( ) , (13)
other bird species’ nests. Γ (((1 + 𝛽) /2) ⋅ 𝛽 ⋅ 2(𝛽−1)/2 )
4 Journal of Applied Mathematics

where 𝛽 is 1.5 in the standard software implementation of the Table 1: A comparison between the parameter results obtained by
CS algorithm [30]. Γ denotes the gamma function. the CS algorithm and that of other algorithms from the SDM.
After initialization, the evolution phase of the 𝑥𝑖 pattern
CS CPSO [14] GA [13] PS [16]
starts by defining the donor vector V, where V = 𝑥𝑖 . The
required step size of the 𝑗th attributes can be calculated by 𝐼pv 0.7608 0.7607 0.7619 0.7617
the following equation: 𝐼𝑜 3.23𝐸–07 4.00𝐸–07 8.09𝐸–07 9.98𝐸–07
𝑛 1.4812 1.5033 1.5751 1.6
1/𝛽 𝑅𝑠
𝑢𝑗 0.0364 0.0354 0.0299 0.0313
𝑠𝑗 = 0.01 ⋅ ( ) ⋅ (V − 𝑥best ) , (14) 𝑅𝑝 53.7185 59.012 42.3729 61.1026
V𝑗
RMSE 0.0010 0.0014 0.0191 0.0149
where 𝑢 = 𝜙 ⋅ rand𝑛[𝐷] and V = rand𝑛[𝐷]. The rand𝑛[𝐷]
function generates a uniform integer between [1, 𝐷] [25]. The
applied to evaluate in this paper. Equations (17) and (18)
donor pattern V is then randomly adjusted by
preset the IAE and MAE, respectively:
V = V + 𝑠𝑗 ⋅ rand𝑛 [𝐷] . 󵄨 󵄨
(15) IAE = 󵄨󵄨󵄨𝐼calculated − 𝐼measured 󵄨󵄨󵄨 , (17)
The CS algorithm will evaluate the fitness of the random 1 𝑁
pattern. If a better solution is caught, the 𝑥best pattern will be MAE = ∑ IAE𝑖 . (18)
𝑁 𝑖=1
updated. The unfeasible patterns are revised by the crossover
operator given in (16) as follows: The optimization algorithms applied in this paper are
programmed in MATLAB. Similar simulation conditions,
𝑥 + rand ⋅ (𝑥𝑟1 − 𝑥𝑟2 ) , rand𝑖 > 𝑝0 , including population size, maximum generation number, and
V𝑖 = { 𝑖 (16)
𝑥𝑖 , others, search ranges, are set to ensure a fair evaluation (population
size = 25; maximum generation number = 5000).
where 𝑝0 is the mutation probability value (𝑝0 = 0.25 in the
standard software implementation [30]). The final step of a 4.1. Case Study 1: Parameter Estimation for a PV Cell at the
generation is to check if the revised infeasible patterns deliver Certain Irradiance Level. Table 1 lists the model parameters
a better solution. of the R.T.C France PV cell at 33∘ C, which are extracted
from the experimental data in [26]. The parameters obtained
4. Results and Discussions from the CS algorithm are compared with three different
parameter estimation approaches: CPSO [14], GA [13], and
With the aim of providing a thorough evaluation of the CS PS [16]. From the RMSEs of these methods, which are listed
algorithm in estimating the PV parameters, both SDM and in the last row of Table 1, the CS algorithm [30] outperforms
ISDM are considered in this paper. Two case studies are the other three optimization methods. CS obtained slightly
designed to estimate the CS algorithm in model parameters lower RMSE, recording 0.0010 in numerical value.
estimation: During the parameter estimation process for the SDM,
the values of the objective function in different optimization
(i) a commercial 57 mm diameter solar cell (R.T.C. algorithms are shown in Figure 2. The function “ga” in
France [26]) operating at the standard irradiance MATLAB [31], whose crossover rate 𝑃𝑐 = 0.8 and mutation
level; rate 𝑃𝑚 = 0.2, is utilized for the convergence process test. As
(ii) a PV module (KC200GT Multicrystal Photovoltaic for PSO implementation [24], the algorithm parameters are
Module) operating under varied environment condi- set as learning factors 𝑐1 = 𝑐2 = 2, inertia factors 𝑤max = 0.9,
tions. 𝑤min = 0.4, and velocity clamping factor 𝑉max = 0.5. In
Figure 2, no further improvement by GA is observed after
During the parameter extraction process, the objective 500 iterations. On the contrary, the CS algorithm showed
function 𝑓(𝑉, 𝐼, 𝑥) is minimized with respect to the param- continuous improvement until the maximum generation. The
eters range. In theory, the value of 𝐼pv𝑛 is slightly larger than CS algorithm, whose convergence speed is slightly faster than
that of 𝐼sc𝑛 . 𝐸𝑔𝑛 is in a loose range from 1 eV to 2 eV. 𝐾𝑖 is PSO, shows the best accuracy result in the minimization task
around the value provided by the datasheet (normally less after 5000 iterations.
than 0.02%/∘ C). The 𝐼𝑜𝑛 is usually less than 50 𝜇A. As stated in Table 2 lists the parameters of the ISDM obtained by the
[27], the ideality factor ranges between 1 and 2. PV modules CS algorithm. In order to evaluate the accuracy of the CS-
produced by most manufacturers have 𝑅𝑠 less than 0.5 Ω and based estimation, these parameters are substituted into the
𝑅𝑝 between 5 and 170 Ω [8, 28]. As for PV cell, the ranges of ISDM. Since the 𝐼-𝑉 demonstrates nonlinear characteristics,
𝑅𝑠 and 𝑅𝑝 can be scaled by simply dividing 𝑁𝑠 [29]. the PV terminal current 𝐼 is solved by the Newton-Raphson
Statistical analysis is performed to evaluate the quality of method [32] in this paper. In Table 3, the calculated results
the fitted models to the experimental data. Besides RMSE, 𝐼ISDM are compared with the experimental data 𝐼measured to
other two fundamental measures, namely, Individual Abso- observe the agreement between them. The notations IAESDM
lute Error (IAE) and the Mean Absolute Error (MAE), are and IAEISDM denote the IAE for SDM and ISDM, respectively.
Journal of Applied Mathematics 5

100 Table 3: A comparison between the errors of ISDM and SDM. The
parameters are extracted by the CS algorithm.

No 𝑉measured 𝐼measured 𝐼ISDM IAEISDM IAESDM

Fitness (RMSE)

10−1
1 −0.2057 0.7640 0.7639 0.0001 0.0001
2 −0.1291 0.7620 0.7626 0.0006 0.0007
10−2 3 −0.0588 0.7605 0.7614 0.0009 0.0009
4 0.0057 0.7605 0.7602 0.0003 0.0003
5 0.0646 0.7600 0.7592 0.0008 0.0009
10−3
0 1000 2000 3000 4000 5000 6 0.1185 0.7590 0.7583 0.0007 0.0010
Iteration 7 0.1678 0.7570 0.7574 0.0004 0.0001
8 0.2132 0.7570 0.7565 0.0005 0.0009
GA [31]
PSO [24] 9 0.2545 0.7555 0.7555 0.0000 0.0004
CS [30] 10 0.2924 0.7540 0.7540 0.0000 0.0003
Figure 2: Convergence process of different optimization algorithms 11 0.3269 0.7505 0.7517 0.0012 0.0009
during the parameter estimation process of the SDM. 12 0.3585 0.7465 0.7476 0.0011 0.0009
13 0.3873 0.7385 0.7402 0.0017 0.0016
14 0.4137 0.7280 0.7273 0.0007 0.0006
Table 2: Parameters of the ISDM obtained by the CS algorithm. 15 0.4373 0.7065 0.7066 0.0001 0.0005
𝐼pv𝑛 𝐼𝑜𝑛 𝑛 𝑅𝑠𝑛 𝑅𝑝𝑛 𝐾𝑖 𝐸𝑔𝑛 16 0.4590 0.6755 0.6748 0.0007 0.0002
0.7361 1.84𝐸–07 1.5009 0.0355 57.8394 0.0031 1.0020 17 0.4784 0.6320 0.6304 0.0016 0.0011
18 0.4960 0.5730 0.5717 0.0013 0.0009
19 0.5119 0.4990 0.4994 0.0004 0.0005
20 0.5265 0.4130 0.4137 0.0007 0.0005
Although the RMSE of the ISDM is less than that of CPSO,
GA, and PS, it is similar to the RMSE of the conventional 21 0.5398 0.3165 0.3176 0.0011 0.0007
SDM under a certain environmental condition. 22 0.5521 0.2120 0.2127 0.0007 0.0001
23 0.5633 0.1035 0.1033 0.0002 0.0008
4.2. Case Study 2: Parameter Estimation for a PV Module 24 0.5736 −0.0100 −0.0089 0.0011 0.0008
under Different Environment Conditions. In this section, the 25 0.5833 −0.1230 −0.1244 0.0014 0.0014
validity of the CS algorithm is evaluated using KC200GT PV 26 0.5900 −0.2100 −0.2095 0.0005 0.0009
module operating under different environment conditions. MAE 0.0007 0.0007
The estimated parameters, both in the SDM and ISDM, RMSE 0.0010 0.0010
are shown in Table 4. As illustrated in Section 1, the main
application of the parameter extraction is to predict the
𝐼-𝑉 characteristics for design purpose. It is worth pointing Table 4: Parameters of the KC200GT PV module obtained by the
out that the SDM parameters can only be extracted by the CS algorithm.
experimental data measured under a certain test condition.
Significant errors may occur as the experimental data are (a) SDM parameters (extracted by the CS algorithm)
measured under varying operating conditions. In the com- 𝐼pv 𝐼o 𝑛 𝑅s 𝑅𝑝
mercial simulation tool like PSIM [21], the PV parameters of
8.1729 4.23E–10 1.0090 0.2665 140.4875
the SDM are firstly estimated at the STCs, then the equations
(given in the appendix) are applied to calculate the electrical (b) ISDM parameters (extracted by the CS algorithm)
characteristics of different operating conditions. The ISDM- 𝐼pv𝑛 𝐼𝑜𝑛 𝑛 𝑅𝑠𝑛 𝑅𝑝𝑛 𝐾𝑖 𝐸𝑔𝑛
based parameter estimation, however, can be performed by
the data measured under any conditions. 8.1847 5.12E–10 1.0170 0.2574 117.9224 0.0028 1.2474
Figure 3 shows the 𝐼-𝑉curves generated using the param-
eters obtained by the CS algorithm. The simulated results are
compared with the experimental data, which are collected SDM is calculated as 0.2837, while the RMSE of 𝐼 in ISDM
at five different irradiance levels (1000 W/m2 , 800 W/m2 , is only 0.0776.
600 W/m2 , 400 W/m2 , and 200 W/m2 ) and three different Figure 4 shows the absolute current errors of different
temperature levels (25∘ C, 50∘ C, and 75∘ C). It can be seen performance predicting methods under different operating
that the 𝐼-𝑉 curves of the ISDM fit the whole range of the conditions. The curves denoted by the label “analytical SDM”
experimental dataset. On the other hand, the errors of SDM are obtained from the analytical SDM model [4]. Ignoring
seem larger at lower irradiance and higher temperature levels. the effect of incidence angle and air mass, the curves labeled
With the experimental data, the RMSE of the current 𝐼 in by “analytical ISDM” denote the 𝐼-𝑉 curves from De Soto’s
6 Journal of Applied Mathematics

9 9
8 8
7 7
6 6
5 5
I (A)

I (A)
4 4
3 3
2 2
1 1
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
V (V) V (V)
Experimental data Experimental data
DSM DSM
ISDM ISDM
(a) (b)

Figure 3: The simulated 𝐼-𝑉 characteristic curves of the KC200GT PV module: (a) under different irradiance levels; (b) under different
temperature levels.
Individual absolute error (IAE)

Individual absolute error (IAE)

1 2
0.8 1.5
0.6
1
0.4
0.2 0.5
0 0
1000
800 30 75 30
600 20 25 50 20 25
G (W 400 10 15 T(∘ 25 10 15
/m 2 200 0 5 C) 0 0 5
) V (V) V (V)

Analytical SDM[4] CS estimation-based SDM Analytical SDM[4] CS estimation-based SDM

Analytical ISDM CS estimation-based ISDM Analytical ISDM CS estimation-based ISDM
(a) (b)

Figure 4: A comparison of the individual absolute errors among different PV modeling methods: (a) under different irradiance levels; (b)
under different temperature levels.

analytical ISDM model [5]. It is evident the ISDM with the The maximum absolute error of the GA-based ISDM is up to
parameters extracted by the CS algorithm is more accurate about 0.8 A, while the absolute error of the CS is kept below
than the analytical model. As for the SDM, the CS algorithm 0.2 A.
is capable of extracting a set of PV parameters with a good
fit for the experimental data at the STCs. However, the SDM
with the equations in the appendix does not exhibit a good 5. Conclusion
prediction performance under other operating conditions.
To further validate the accuracy of the CS algorithm, In this work, the Cuckoo Search (CS) algorithm is applied
the extracted parameters are compared to the ones obtained to estimate the parameters of two PV models, namely, Single
using GA in Figure 5. In general, the CS algorithm gives the Diode Model (SDM) and its improved version (ISDM). The
better performance than GA for all cases. The Maximum feasibility of the proposed method has been validated by
Power Point (MPP), usually locating around 74% of the estimating the parameters of two commercial PV generators.
open circuit voltage, is an important technical data in PV The simulation and experimental results showed that the CS
modeling. However, a negative point of the GA-based ISDM algorithm is capable of not only extracting all the parameters
is that the errors in the high voltage range are relatively high. of the SDM under a certain condition but also successfully
Journal of Applied Mathematics 7

Individual absolute error (IAE)

Individual absolute error (IAE)

1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
1000
800 30 75 30
600 20 25 50 20 25
G (W 400 10 15 T(∘ 25 10 15
/m 2 200 0 5 C) 0 0 5
) V (V) V (V)

CS estimation-based ISDM CS estimation-based ISDM

GA estimation-based ISDM GA estimation-based ISDM
(a) (b)

Figure 5: A comparison of the individual absolute errors between CS- and GA- based ISDM: (a) under different irradiance levels; (b) under
different temperature levels.

estimating all the parameters of ISDM under different operat- References

ing conditions. In statistical analysis, CS algorithm recorded
the lowest RMSE value compared to other algorithms such as [1] B. Amrouche, A. Guessoum, and M. Belhamel, “A simple
behavioural model for solar module electric characteristics
GA, PSO and PS.
based on the first order system step response for MPPT study
and comparison,” Applied Energy, vol. 91, no. 1, pp. 395–404,
2012.
Appendix [2] A. Orioli and A. di Gangi, “A procedure to calculate the five-
parameter model of crystalline silicon photovoltaic modules on
PV Physical Model Adopted in PSIM the basis of the tabular performance data,” Applied Energy, vol.
102, pp. 1160–1177, 2013.
By using the parameters extracted at the STCs, the 𝐼-𝑉 char-
[3] V. Lo Brano, A. Orioli, G. Ciulla, and A. di Gangi, “An improved
acteristics of a PV generator under nonstandard operating
five-parameter model for photovoltaic modules,” Solar Energy
conditions can be calculated via the following equations: Materials and Solar Cells, vol. 94, no. 8, pp. 1358–1370, 2010.
[4] M. G. Villalva, J. R. Gazoli, and E. R. Filho, “Comprehensive
approach to modeling and simulation of photovoltaic arrays,”
𝐼 = 𝐼pv − 𝐼𝑑 − 𝐼𝑅 , IEEE Transactions on Power Electronics, vol. 24, no. 5, pp. 1198–
𝐺 1208, 2009.
𝐼pv = 𝐼sc𝑛 ⋅ − 𝐾𝑖 ⋅ (𝑇 − 𝑇𝑛 ) , [5] W. De Soto, S. A. Klein, and W. A. Beckman, “Improvement and
𝐺𝑛 validation of a model for photovoltaic array performance,” Solar
Energy, vol. 80, no. 1, pp. 78–88, 2006.
𝐼𝑑 = 𝐼𝑜 ⋅ (𝑒𝑞𝑉𝑑 /𝑛𝑘𝑇 − 1) ,
[6] K. Ishaque, Z. Salam, and H. Taheri, “Simple, fast and accurate
two-diode model for photovoltaic modules,” Solar Energy Mate-
𝐺 3 (𝑞𝐸𝑔 /𝑛𝑘)⋅(1/𝑇−1/𝑇𝑛 ) rials and Solar Cells, vol. 95, no. 2, pp. 586–594, 2011.
𝐼𝑜 = 𝐼𝑐𝑛 ⋅ ( ) ⋅𝑒 ,
𝐺𝑛 [7] V. Lo Brano, A. Orioli, and G. Ciulla, “On the experimental
𝑉 validation of an improved five-parameter model for silicon
𝐼𝑅 = 𝑑 , photovoltaic modules,” Solar Energy Materials and Solar Cells,
𝑅𝑝
vol. 105, pp. 27–39, 2012.
𝑉 [8] A. N. Celik and N. Acikgoz, “Modelling and experimental
𝑉𝑑 = + 𝐼 ⋅ 𝑅𝑠 . verification of the operating current of mono-crystalline pho-
𝑁𝑠
tovoltaic modules using four- and five-parameter models,”
(A.1) Applied Energy, vol. 84, no. 1, pp. 1–15, 2007.
[9] M. A. de Blas, J. L. Torres, E. Prieto, and A. Garcı́a, “Selecting a
suitable model for characterizing photovoltaic devices,” Renew-
Acknowledgments able Energy, vol. 25, no. 3, pp. 371–380, 2002.
[10] J. P. Charles, G. Bordure, A. Khoury, and P. Mialhe, “Consis-
The authors are grateful to Professor Xin-She Yang for the tency of the double exponential model with physical mecha-
sharing of Cuckoo Search source code online. Without his nisms of conduction for a solar cell under illumination,” Journal
generosity, this work would not be possible. This research of Physics D, vol. 18, no. 11, pp. 2261–2268, 1985.
is supported by the National Natural Science Foundation of [11] D. S. H. Chan and J. C. H. Phang, “Analytical methods for
China under Grant 61070085. the extraction of solar-cell single- and double-diode model
8 Journal of Applied Mathematics

parameters from I-V characteristics,” IEEE Transactions on [30] X. S. Yang, Cuckoo Search Algorithm (Source Code), http://
Electron Devices, vol. 34, no. 2, pp. 286–293, 1987. www.mathworks.com/matlabcentral/fileexchange/29809-
[12] J. C. H. Phang, D. S. H. Chan, and J. R. Phillips, “Accurate analyt- cuckoo-search-cs-algorithm.
ical method for the extraction of solar cell model parameters,” [31] Optimization Toolbox, The MathWorks Inc., http://www
Electronics Letters, vol. 20, no. 10, pp. 406–408, 1984. .mathworks.com/products/optimization/index.html.
[13] A. J. Joseph, B. Hadj, and A. L. Ali, “Solar cell parameter [32] R. L. Burden and J. D. Faires, Numerical Analysis, Cengage
extraction using genetic algorithms,” Measurement Science and Learning, Singapore, 2010.
Technology, vol. 12, no. 11, pp. 1922–1925, 2001.
[14] W. Huang, C. Jiang, L. Xue, and D. Song, “Extracting solar cell
model parameters based on chaos particle swarm algorithm,” in
Proceedings of the International Conference on Electric Informa-
tion and Control Engineering (ICEICE ’11), pp. 398–402, April
2011.
[15] I. Fister, I. Fister Jr, X. S. Yang, and J. Brest, “A comprehensive
review of firefly algorithms,” Swarm and Evolutionary Compu-
tation, 2013.
[16] M. F. AlHajri, K. M. El-Naggar, M. R. AlRashidi, and A. K.
Al-Othman, “Optimal extraction of solar cell parameters using
pattern search,” Renewable Energy, vol. 44, pp. 238–245, 2012.
[17] A. Askarzadeh and A. Rezazadeh, “Parameter identification
for solar cell models using harmony search-based algorithms,”
Solar Energy, vol. 86, pp. 3241–3249, 2012.
[18] X. Yang and S. Deb, “Engineering optimisation by cuckoo
search,” International Journal of Mathematical Modelling and
Numerical Optimisation, vol. 1, no. 4, pp. 330–343, 2010.
[19] X. Yang and S. Deb, “Cuckoo search via Lévy flights,” in
Proceedings of the World Congress on Nature and Biologically
Inspired Computing (NABIC ’09), pp. 210–214, December 2009.
[20] R. A. Messenger and J. Ventre, Photovoltaic Systems Engineering,
CRC Press, New York, NY, USA, 2nd edition, 2004.
[21] PSIM User Manual, Powersim, Woburn, Mass, USA, 2001.
[22] R. N. Mantegna, “Fast, accurate algorithm for numerical sim-
ulation of Lévy stable stochastic processes,” Physical Review E,
vol. 49, no. 5, pp. 4677–4683, 1994.
[23] A. H. Gandomi, X. Yang, and A. H. Alavi, “Cuckoo search
algorithm: a metaheuristic approach to solve structural opti-
mization problems,” Engineering with Computers, vol. 29, pp. 17–
35, 2013.
[24] B. Birge, Particle Swarm Optimization Toolbox, http://www
.mathworks.com/matlabcentral/fileexchange/7506-particle-
swarm-optimization-toolbox.
[25] P. Civicioglu and E. Besdok, “A conceptual comparison of the
Cuckoo-search, particle swarm optimization, differential evolu-
tion and artificial bee colony algorithms,” Artificial Intelligence
Review, vol. 39, pp. 315–346, 2013.
[26] T. Easwarakhanthan, J. Bottin, I. Bouhouch, and C. Boutrit,
“Nonlinear minimization algorithm for determining the solar
cell parameters with microcomputers,” International Journal of
Solar Energy, vol. 4, pp. 1–12, 1986.
[27] M. Bashahu and P. Nkundabakura, “Review and tests of meth-
ods for the determination of the solar cell junction ideality
factors,” Solar Energy, vol. 81, no. 7, pp. 856–863, 2007.
[28] S. J. Jun and L. Kay-Soon, “Photovoltaic model identification
using particle swarm optimization with inverse barrier con-
straint,” IEEE Transactions on Power Electronics, vol. 27, pp.
3975–3983, 2012.
[29] K. Ishaque, Z. Salam, S. Mekhilef, and A. Shamsudin, “Param-
eter extraction of solar photovoltaic modules using penalty-
based differential evolution,” Applied Energy, vol. 99, pp. 297–
308, 2012.
Copyright of Journal of Applied Mathematics is the property of Hindawi Publishing
Corporation and its content may not be copied or emailed to multiple sites or posted to a
listserv without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.