2013 IEEE Conference on Clean Energy and Technology (CEAT)

Solar cell parametters extraction using paarticle swarm

opttimization algorithm
Nurul Farhana Abdul Hamidd* Nasruddin Abdul Rahim, Jeyraj Selvaraj
School of Electrical System Engineeringg, Universiti Universiti Malaya Poweer Energy Dedicated Advance
Malaysia Perlis, Malaysia. Centre (UUMPEDAC),
Email:[email protected] m Universiti Malaya,
M Malaysia.

Abstract—This paper presents an applicationn of particle swarm with those of artificial bee swarm optimization (ABSO),
optimization (PSO) technique for extracting the parameters of simulated annealing (SA) and chaos particle swarm
single diode solar cell model. The proposed teechnique is used to optimization (CPSO).
estimate five different model parameters; namely,
n generated This paper is organized as follows. Section II provides an
photocurrent, saturation current, series resistance, shunt
resistance and ideality factor that govern the current-voltage
overview of the solar cell siingle diode model. Section III
relationship of a solar cell. A measuremen nt data of 57 mm presents the proposed optim mization method. Section IV
diameter commercial (R.T.C. France) silicon solar
s cell is used to presents the results and valiidates the performance of the
test and verify the consistency of accurately estimating various proposed method. Finally, Sectiion V summarizes the work.
parameters. The effectiveness of the prooposed method is
compared with the results found by the other parameter II.SOLAR CELL MODEL
estimation techniques.
The solar cell can be represented using equivalent electrical
IndexTerms—solar cell model, particle sw
warm optimization,
parameter extraction circuit as shown in Fig. 1. Itt consists of a photo-generated
current source (‫ܫ‬௣௛ ), anti-paralllel diode, a series resistance (ܴ௦ )
I.INTRODUCTION and a shunt resistance (ܴ௦௛ ). ThisT model is known as single
The single diode model of solar cells iss derived from the diode model and widely useed to represent the solar cell
equivalent circuit based on the internal phyysical mechanisms behaviour[1-8]. The output currrent,‫ܫ‬௔ is given as:
acting within solar cells[1-8]. The key param
meters that describe
solar cell models behaviour are the generaated photocurrent, ‫ܫ‬௔ ൌ ‫ܫ‬௣௛ െ ‫ܫ‬ௗ െ ‫ܫ‬௦௛ (1)
saturation current, series resistance, shunnt resistance, and
ideality factor. Since the model is a transceendental nonlinear Considering Shockley equationn for the diode current, ‫ܫ‬ௗ and
exponential equation, direct method to measuure the parameters substituting the current of thhe shunt resistor,‫ܫ‬௦௛ , Eq. 1 is
is lacked. Therefore, to provide the accurrate modeling and rewritten as the following form
m:
performance evaluation of a given solar system, a valid
estimation of this parameter is always requireed. ‫ܫ‬௔ ൌ ‫ܫ‬௣௛ െ ‫ܫ‬௢ ቂ‡š’ ቀ
௏ೌ ାூೌ ோೞ
ቁ െ ͳቃ െ
௏ೌ ାூೌ ோೞ
(2)
In recent years, the metaheuristic optimmization algorithms ௔௏೟ ோೞ೓
such as genetic algorithm (GA) [1-3], sim mulated annealing
(SA) [4], harmony search (HS) [5], artiificial bee swarm where ܸ௔ is the solar cell outpuut voltage, ܽ is the diode ideality
௞்
optimization algorithm (ABSO) [6], andd particle swarm factor and ܸ‫ ݐ‬ൌ represennts the thermal voltage. The
௤
optimization (PSO)[7], have attracted mucch attention in the
parameter ݇ is the Boltzmann constant (ͳǤ͵ͺͲ͸ͷͲ͵‫ିͲͳݔ‬ଶଷ
investigation of solar cell parameters identtification problem.
ͳǤ͸Ͳʹͳ͹͸Ͷ͸‫ିͲͳݔ‬ଵଽ °C) and ܶ is
J/K),‫ ݍ‬is the electron charge (ͳ
Metaheuristic algorithms are suitable choicees for solving this
the temperature of the solar celll in Kelvin.
problem due to their global search power as well as derivative-
free advantage.
The PSO algorithm is a swarm intelliggence optimization
algorithm based on observations of the sociall behaviour of bird
flocking[7, 9-19]. Numerous authors developped and improved
various versions of PSO algorithm[7, 14-177].However, every
version of PSO has different advantage for different complex
optimal problem. In this paper, a PSO alggorithm with time
varying inertia weight and acceleration coefficient (PSO-
TVIWAC) is proposed.The proposed PSO allgorithm is applied
Fig. 1: Single diode model equivalent circuit
to identify the optimal parameters of a 57 mm diameter
commercial (R.T.C. France) silicon solar cell[8]. In order to
evaluate the performance, the obtained resuults are compared

978-1-4799-3238-2/13/$31.00 ©2013 IEEE 461

 2013 IEEE Conference on Clean Energy and Technology (CEAT)

III.PARTICLE SWARM OPTIMIZATION (PSO) ALGORITHM optimal solution was observed when inertia weight is varying
from 0.9 to 0.5 over entire search range [12].
PSO is a population-based stochastic swarm intelligence
optimization method introduced by Kennedy and Eberhart[9]. B. Time-Varying Acceleration Coefficients (TVAC)
It has the capability in finding global optimal points by using
the social interaction of unsophisticated agents. In PSO, each Acceleration coefficients ‫ܥ‬ଵ and ‫ܥ‬ଶ control the movement
particles flies or swim through a multidimensional space with of each particle towards its local and global best position,
velocity updated by movement inertia, self-cognition, and respectively. Small values of the acceleration coefficients limit
social interaction[18]. the movement of the particles, while large numbers may cause
The PSO algorithm is initialized with a group of random the particles to diverge[18].Instead of having fixed value, a
particles (solutions) and then searches for optima by updating time varying value of acceleration coefficients proved to
generations. Each particle (‫ݔ‬௜ ሻ flies through the D-dimensional provide better performance of PSO after certain number of
search space with a velocity ‫ݒ‬௜ , which is dynamically adjusted iteration [19]. With the aim to increase global search in the
according to its own previous best solution ܾܲ݁‫ݐݏ‬௜ and the beginning and encourage particles to converge to global
previous neighbourhood best solution ‫ ݐݏܾ݁ܩ‬of the entire optimum point at the end of search, the acceleration
swarm. The velocity updates are calculated as a linear coefficient, ‫ܥ‬ଵ and ‫ܥ‬ଶ is adjusted as follows:
combination of position and velocity vectors.
௠ି௡
‫ܥ‬ଵ௡ ൌ ‫ܥ‬ଵ௘ ൅  ሺ‫ܥ‬ଵ௘ െ ‫ܥ‬ଵ௦ ሻ (6)
௡ାଵ ௡ ௡ሻ ௠ିଵ
‫ݒ‬௜ ൌ ‫ݒ כ ݓ‬௜ ൅ ‫ܥ‬ଵ ‫ݎ כ‬ଵ ‫ כ‬ሺܾܲ݁‫ݐݏ‬௜ െ ‫ݔ‬௜ ൅ ‫ܥ‬ଶ ‫ݎ כ‬ଶ (3)
‫  כ‬ሺ‫ ݐݏܾ݁ܩ‬െ ‫ݔ‬௜ ௡ ሻ ௠ି௡
‫ܥ‬ଶ௡ ൌ ‫ܥ‬ଶ௘ ൅  ሺ‫ܥ‬ଶ௘ െ ‫ܥ‬ଶ௦ ሻ (7)
௠ିଵ
where each particle’s position ‫ݔ‬௜ represents a possible solution
point in the problem search space, ‫ ݓ‬is the inertia weight, ‫ܥ‬ଵ where ‫ܥ‬ଵ௦ ,‫ܥ‬ଵ௘ , ‫ܥ‬ଶ௦ and ‫ܥ‬ଶ௘ are the start and end value of
and ‫ܥ‬ଵ are the acceleration coefficient, ‫ݎ‬ଵ and ‫ݎ‬ଶ are random acceleration coefficients ‫ܥ‬ଵ , and ‫ܥ‬ଶ , respectively. Through
numbers between 0 and 1, ݊ is the iteration number, and ݅ is empirical studies, Ratnaweera et al.[14] have observed that the
the particle size, ݅ ൌ ͳǡ ʹǡ Ǥ Ǥ Ǥ ǡ ‫ܯ‬. Then, each particle is optimal solution can be improved by varying the value of
updating it position using the following equation: ‫ܥ‬ଵ from 2.5 to 0.5 and ‫ܥ‬ଶ from 0.5 to 2.5 over entire search
range.
‫ݔ‬௜ ௡ାଵ ൌ ‫ݔ‬௜ ௡ ൅ ‫ݒ‬௜ ௡ାଵ (4)
C. Parameter Extraction using PSO-TVIWAC Algorithm
IV.TIME-VARYING INERTIA WEIGHT AND ACCELERATION
COEFFICIENT PSO (PSO-TVIWAC) ALGORITHM Before continue with the optimization operations, a
performance criterion or an objective function should be
In this work, the velocity is calculated by varying the defined. In order to define the objective function, the current-
values of inertia weight and acceleration coefficient for the voltage (‫ ܫ‬െ ܸ) relationship given in of Eq. 2 is rewritten in
entire search range. The idea is to ensure that the particles the following homogeneous equation:
having better global search ability in the beginning and better
local search ability towards the end, avoiding premature ݂ሺ‫ܫ‬௔ ǡ ܸ௔ ǡ ‫ݔ‬ሻ ൌ Ͳ (8)
convergence and improved convergence speed.
where ‫ ݔ‬ൌ  ൣܴ௦ ܴ௦௛ ‫ܫ‬௣௛ ‫ܫ‬௢ ܽ൧. The value of ݂ is calculated for
A. Time-Varying Inertia Weight (TVIW) each pairs of measurement data. In this paper, the root mean
square error (RMSE) is chosen as a criterion to quantify the
The inertia weight ‫ ݓ‬in Eq. 3 is introduced by Shi and difference between the model results and the measurement
Eberhart[10], [12]. The idea is to provide balancing between data. RMSE is defined by the following equation:
the global search and local search. It can be a positive constant
or even a positive linear or nonlinear function of time [10]. In ଵ ଶ
ܴ‫ ܧܵܯ‬ൌ  ට σே
௜ୀଵ൫݂௜ ሺ‫ܫ‬௔௜ ǡ ܸ௔௜ ǡ ‫ݔ‬௜ ሻ൯ (9)
order to have better global search ability at the beginning, ே
whereas having better local search ability near the end, the
inertia weight is decreased linearly over the generations[12]. In where ܰ is the number of data points, ‫ܫ‬௔௜ and ܸ௔௜ are ݅ ௧௛
this work, the inertia weight is linearly varying using the measured current and voltage pair values, respectively. During
following equation: the optimization process, the objective function is to be
minimized with respect to the parameters range. The process
௠ି௡
‫ݓ‬௡ ൌ ‫ݓ‬௘ ൅  ሺ‫ݓ‬௘ െ ‫ݓ‬௦ ሻ (5) of parameter identification with PSO-TVIWAC method is
௠ିଵ
shown in Fig. 2.
where ‫ݓ‬௦ = start weight, ‫ݓ‬௘ = end weight, ݉ = maximum
iteration value and ݊ = current iteration index. Improved

978-1-4799-3238-2/13/$31.00 ©2013 IEEE 462

 2013 IEEE Conference on Clean Energy and Technology (CEAT)

minimum RMSE value. Therefore, the optimal parameters

START found by PSO-TVIWAC are closer to the real parameters of
the system.

Initialize randomly position and velocity of each TABLE I.ADJUSTABLE PARAMETERS

FOR PSO-TVIWACALGORITHM
Items PSO-TVIWAC
Evaluate the fitness of each particle No of Iteration 5000
No of particles 100
ws 0.9
Compare the personal best of each particle to its we 0.4
current fitness, and set Pbest to the better performance C1s 2.5
C1e 0.5
C2s 0.5
C2e 2.5
Set the global best Gbest to the position of the particle
with the best fitness within the swarm
The ‫ܫ‬Ȃ ܸ characteristic is reconstructed to investigate the
quality of the identified parameters. This is simply achieved
Calculate inertia weight acceleration coefficients by calculating Eq. 2 using Newton method when ‫ܫ‬௔ is
considered unknown while ܸ௔ is known. Figure 3 shows the
comparison data between the results produced by the optimal
parameters and the measurement data. It can be observed that
Update velocity and position of each particle the ‫ ܫ‬െ ܸ characteristic obtained by the identified model using
PSO-TVIWAC algorithm is in good agreement with the
measurement data. The power vs. voltage ሺܲȂ ܸሻ
characteristic of the system is also shown in Fig.4.
N Stop
Condition?
TABLE II. THE COMPARISON OF PARAMETERS EXTRACTION
PSO- CPSO
Items ABSO SA
Y TVIWAC
Rs (Ÿ) 0.03638 0.03659 0.0345 0.0354
Return best value Rsh (Ÿ) 53.6894 52.2903 43.1034 59.012
Iph (A) 0.76078 0.76080 0.7620 0.7607
Io (μA) 0.32267 0.30623 0.4798 0.4000
END a 1.48107 1.47583 1.5172 1.5033
RMSE 9.86024e-4 9.9124e-4 0.0017 0.00139

Fig. 2. Flowchart of the PSO-TVIWAC algorithm

I.RESULT AND DISCUSSION

The performance of PSO-TVIWAC algorithm in

estimating solar cell model parameters is tested in this section.
Measurement data of 57 mm diameter commercial (R.T.C.
France) silicon solar cell are used for validation. The data has
been taken from the system under 1 sun (1000 W/m2) at 33 °C
[8].The adjustable parameters for PSO-TVIWAC algorithm
are given in Table 1. The algorithm is coded using Matlab
environment.
Table 2 shows optimal parameters and RMSE value for the
single diode model using PSO-TVIWAC. The results are
compared with results obtained by artificial bee swarm
optimization algorithm(ABSO)[6], simulated annealing
algorithm(SA)[4], and chaos particle swarm optimization
algorithm(CPSO)[7]. It is observed that PSO-TVIWAC Fig. 3.Comparison between the I–V characteristics resulted from the
outperforms ABSO, SA, and CPSO, because it has found the experimental data and the single diode model

978-1-4799-3238-2/13/$31.00 ©2013 IEEE 463

 2013 IEEE Conference on Clean Energy and Technology (CEAT)

ூ೘೐ೌೞೠೝ೐೏ ିூ೎ೌ೗೎ೠ೗ೌ೟೐೏
݁ൌ (11)
ூ೘೐ೌೞೠೝ೐೏

The computed value of e for each measurements and

MAE is tabulated in Table 3. The value of MAE indicates the
high accuracy of the proposed method.

II. CONCLUSION

This paper presents the application of PSO-TVIWAC

algorithm for extracting solar cell parameters using single
diode model. The proposed algorithm is implemented and
tested using measurement data of a 57mm diameter
commercial (R.T.C France) silicon solar cell. Outcomes
obtained using proposedmethod are quite promising and
surpasses the ABSO, SA, and CPSO method. Therefore, the
PSO-TVIWAC algorithm can be an effective candidate for
Fig. 4.Comparison between the P–V characteristics resulted from solving solar cells optimization problems.
the experimental data and the single diode model

Figure 5 illustrate the best value of the objective function TABLE III.RELATIVE ERROR FOR EACH MEASUREMENT
during the iterations. As can be seen, the convergence rate of
the proposed algorithm is fast. This is due to enhancement
make by varying value of inertia weight and acceleration
coefficient during iteration allowing the algorithm to have
better global search in the beginning and converges to the
optimal solution quickly.

Fig. 5.Convergence process of PSO-TVIWACalgorithm during the

identification process
ACKNOWLEDGMENT

In order to confirm the accuracy of the extraction process, The authors would like to thank Universiti Malaya Power
the mean absolute error (MAE) is calculated using the Energy Dedicated Advance Centre (UMPEDAC) for providing
following formulae: the facilities to conduct this research and Universiti Malaysia
Perlis for providing the scholarship for this study.
ଵ
‫ ܧܣܯ‬ൌ  σே
௜ୀଵȁ݁௜ ȁ (10)
ே
REFERENCES
where ܰ is number of measurement and ݁ is relative error
obtained as follow:
[1] J. K. Maherchandani, C. Agarwal, and M. Sahi, “Estimation of
Solar Cell Model Parameter by Hybrid Genetic Algorithm Using

978-1-4799-3238-2/13/$31.00 ©2013 IEEE 464

 2013 IEEE Conference on Clean Energy and Technology (CEAT)

MATLAB,” International Journal of Advanced Research in 7929, Y. Tan, Y. Shi, and H. Mo, Eds. Springer Berlin
Computer Engineering & Technology, vol. 1, no. 6, pp. 78–81, Heidelberg, 2013, pp. 35–43.
2012. [18] Y. Valle, S. Member, G. K. Venayagamoorthy, S. Member, and
[2] M. Zagrouba, A. Sellami, M. BouaÕ, and M. Ksouri, R. G. Harley, “Particle Swarm Optimization : Basic Concepts ,
“Identification of PV solar cells and modules parameters using Variants and Applications in Power Systems,” IEEE
the genetic algorithms : Application to maximum power Transactions On Evolutionary Computation, vol. 12, no. 2, pp.
extraction,” Solar Energy, vol. 84, pp. 860–866, 2010. 171–195, 2008.
[3] S. J. Patel, A. K. Panchal, and V. Kheraj, “Solar Cell Parameters [19] A. Khare and S. Rangnekar, “A review of particle swarm
Extraction from a Current-Voltage Characteristic Using Genetic optimization and its applications in Solar Photovoltaic system,”
Algorithm,” Journal of Nano and Electronic Physic, vol. 5, no. Applied Soft Computing Journal, vol. 13, no. 5, pp. 2997–3006,
2, pp. 5–7, 2013. 2013.
[4] K. M. El-Naggar, M. R. Alrashidi, M. F. Alhajri, and A. K. Al-
Othman, “Simulated Annealing algorithm for photovoltaic
parameters identification,” Solar Energy, vol. 86, no. 1, pp.
266–274, 2012.
[5] A. Askarzadeh and A. Rezazadeh, “Parameter identification for
solar cell models using harmony search-based algorithms,”
Solar Energy, vol. 86, no. 11, pp. 3241–3249, Nov. 2012.
[6] A. Askarzadeh and A. Rezazadeh, “Artificial bee swarm
optimization algorithm for parameters identification of solar cell
models,” Applied Energy, vol. 102, pp. 943–949, 2013.
[7] H. Wei, J. Cong, X. Lingyun, and S. Deyun, “Extracting solar
cell model parameters based on chaos particle swarm
algorithm,” in Electric Information and Control Engineering
(ICEICE), 2011 International Conference on, 2011, pp. 398–
402.
[8] T. Easwarakhantan, J. Bottin, I. Bouhouch, and C. Boutrit,
“Nonlinear minimization algorithm for determining the solar
cell parameters with microcomputers,” International journal of
solar energy, 1986, vol. 4, no. 1, pp. 1–12.
[9] J. Kennedy and R. Eberhart, “Particle swarm optimization,” in
Proceedings of ICNN’95 - International Conference on Neural
Networks, 1995, vol. 4, pp. 1942–1948.
[10] Y. Shi and R. Eberhart, “A modified particle swarm optimizer,”
in Proceedings. IEEE World Congress on Computational
Intelligence, 1998, pp. 69–73.
[11] Y. Shi and R. Eberhart, “Parameter selection in particle swarm
optimization,” in in Evolutionary Programming VII SE - 57,
vol. 1447, V. W. Porto, N. Saravanan, D. Waagen, and A. E.
Eiben, Eds. Springer Berlin Heidelberg, 1998, pp. 591–600.
[12] Y. Shi and R. C. Eberhart, “Empirical study of particle swarm
optimization,” Proceedings of the 1999 Congress on
Evolutionary Computation-CEC99, pp. 1945–1950, 1999.
[13] X. Hu, Y. Shi, and R. Eberhart, “Recent advances in particle
swarm,” in Evolutionary Computation, 2004. CEC2004.
Congress on, 2004, vol. 1, pp. 90–97 Vol.1.
[14] A. Ratnaweera, S. K. Halgamuge, and H. C. Watson, “Self-
Organizing Hierarchical Particle Swarm Optimizer With Time-
Varying Acceleration Coefficients,” IEEE Transactions On
Evolutionary Computation, vol. 8, no. 3, pp. 240–255, 2004.
[15] A. T. Al-Awami, A. Zerguine, L. Cheded, A. Zidouri, and W.
Saif, “A new modified particle swarm optimization algorithm
for adaptive equalization,” Digital Signal Processing, vol. 21,
no. 2, pp. 195–207, Mar. 2011.
[16] Y. A. N. Chun-man, G. U. O. Bao-long, and W. U. Xian-xiang,
“Empirical Study of the Inertia Weight Particle Swarm
Optimization with Constraint Factor,” International Journal of
Soft Computing And Software Engineering, vol. 2, no. 2, pp. 1–
8, 2012.
[17] X. Liang, Z. Yin, Y. Wang, and Q. Sun, “Impulse Engine
Ignition Algorithm Based on Genetic Particle Swarm
Optimization,” in in Advances in Swarm Intelligence SE - 5, vol.

978-1-4799-3238-2/13/$31.00 ©2013 IEEE 465