2016 International Conference on Industrial Informatics - Computing Technology, Intelligent Technology, Industrial Information

Integration

Parameter extraction for single-diode model of solar cell

R. Tai 1, 2, B. Y. Chen 1, 2, F. X. Chen 3
1. Hubei Collaborative Innovation Center for High-efficient Utilization of Solar Energy, Wuhan, China
2. Hubei University of Technology Wuhan, China
3. Northeastern University, Shenyang, China

Abstract—a method to extract five parameters in the single- Where Iph is cell generated photocurrent, I0 is reverse
diode equivalent circuit of a solar cell is described in detail. saturation current of diode, n is the diode ideality factor, Rs is
The parameter extraction is mainly from the knowledge of the the series resistance, Rsh is the shunt resistance. The
sole information contained in the datasheets given by the extraction of the five parameters is the key to determine the
manufacturer of the solar cell. Besides the information with I-V characteristic of solar cell under certain real conditions
regard to the most representative points of the I-V curve (short such as irradiance levels and temperature. Generally,
circuit, open circuit, and the maximum power point) provided
extensive testing results are necessary for experimental
by all manufacturers, it just needs one more operating point.
The extra information can be obtained either from the
characteristics of the I-V curve, based on the data of which
datasheet or by simple testing without setting up a complicated the parameters are identified. But in most cases,
measurement system. There is only one unknown parameter in manufacturers do not provide the testing results or I-V curve
the proposed objective function, reducing the dimension and and setting up a complicated measurement system is too
complexity of heuristic algorithm. The algorithm is applied to costly or needs a long period. The only information available
extract five parameters in the equivalent circuit and fit the I-V to adjust the circuit model is limited to only three points
curve of AzurSpace’s solar cell 3C40. Experimental results (short circuit, open circuit, and maximum power point) of the
show that this approach leads to a reduction of the I-V curve on the datasheet. To calculate the five parameters,
measurement effort and allows an accurate reconstruction of five boundary conditions are required, but with the three
the solar cell characteristics as well. points, only four boundary conditions can be derived and due
to the transcendental and implicit formulations, the
Keywords- solar cell; parameter extraction; single-diode model identification of five parameters is a nontrivial endeavor.
Current extraction approaches can be roughly classified
I. INTRODUCTION as follows. 1) Identification from experimental
characteristics [1-3]. Five parameters are identified
The solar cell is the elementary building block of the
simultaneously by iterative methods. Usually, the
photovoltaic system. Single and double diode equivalent
identification procedure is quite complex and requires an
circuit models are commonly used to simulate the solar cell’s
elaboration on algorithm design. 2) Identification from the
behavior. The use of diode models makes the performance
three representative points’ information on datasheet [4, 5]. By
evaluation, design and control of more complex power
several approximate solution, either shunt or series resistance
systems that include solar cell panels possible. However,
is neglected. Thus the number of parameters to be extracted
manufacturers of solar cells do not provide required
is reduced from five to four. But the simplifications will
knowledge of parameters for the modeling. Therefore, the
reduce the accuracy of the equivalent circuit model. In
parameter extraction based on the manufacturer data is a
addition, the simplified four parameter model may be
crucial stage in the accomplishment of a reliable simulation
unsuitable for some applications. 3) Identification from the
model for a solar cell, a solar panel or even a complex
three representative points’ information on datasheet together
photovoltaic system.
with some ancillary information [6-9]. The ancillary
The single-diode equivalent circuit of a solar cell is
information, such as the slopes of I-V curve at the open
shown in Fig.1.
circuit (Rs0) and short circuit (Rsh0) or the energy band gap
and material dependent constants (, ), can provide one
more boundary condition. However, not all manufacturers
provide the ancillary information and users cannot get it by
simple testing. Therefore, just based on representative
points’ information and a reduced amount of ancillary
information, to find a quick and easy way to extract five
parameters without neglecting any parameters is essential.
In the present work, heuristic algorithm is employed.
Figure1. the single-diode equivalent circuit of a solar cell Comparing to similar methods that generally take the I-V
The solar cell I-V relationship is given by equation as object function to iteratively calculate the five
parameters simultaneously, the customized object function
contains only one unknown parameter, thereby reducing the
ª § V + IR s
dimension of the problem. Besides the three remarkable
· º V + IRs . (1)
I = I ph − I 0 «exp¨¨ ¸¸ − 1» − points stated above, the algorithm in the current study only
¬ © nVth ¹ ¼ Rsh needs one more point, namely, the current and voltage value

978-1-5090-3575-5/16 $31.00 © 2016 IEEE 319

DOI 10.1109/ICIICII.2016.37
at an arbitrary operation point. Not all of the manufacturers Finallyˈconsidering equation (2), (4), (5), (6), (8) and
provide the extra point but it can be obtained from simple according to Cubas’ approximations [8], the following
measurement. relationships are obtained:
The photovoltaic system developed by the authors’
project group employs the AzurSpace’s 5mm*5mm solar cell Rsh 0 I sc (9)
3C40. So in the following content, the single-diode I ph =
Rsh0 - Rs
equivalent circuit model of 3C40 will be identified based on
its datasheet.
II. OBJECTIVE FUNCTION Rsh 0 I sc - Voc . (10)
I0 =
In this section, an appropriate object function is defined. § Voc ·
Firstly, let short-circuit current be Isc and equation (1) (Rsh 0 - Rs )exp¨¨ ¸¸
rewritten in the short-circuit point becomes: © nVth ¹

ª §I R · º I sc Rs . (2) Rsh = Rsh 0 - Rs . (11)

I ph = I sc + I 0 «exp¨¨ sc s ¸¸ − 1» +
¬ © nVth ¹ ¼ Rsh

An additional equation is the slope of the I-V curve at the

n=
(V mp - I mp Rs )(I sc Rsh 0 - I mp Rsh 0 - Vmp ) (12)
Vth (I mp Rsh 0 - Vmp )
short circuit:

§ ∂V · (3)
Rsh0 = -¨ ¸ I =I
© ∂I ¹ sc ( A - B )Vmp B Voc . (13)
Rs = +
Differentiating Eq. (1) with respect to I and rewriting at ( A + B )I mp ( A + B ) I mp
the short-circuit point, yields:
Where:
1 I I R 1 . (4)
= 0 exp sc s +
Rsh 0 - Rs nVth nVth Rsh ªVmp + (I mp − I sc )Rsh 0 º .
[ ]
A = Vmp + (I mp - I sc )Rsh 0 ln « »
(14)
¬ Voc − I sc Rsh 0 ¼
Secondly, let open-circuit voltage be Voc and equation (1)
rewritten in the open-circuit point becomes:
B = Vmp − Rsh 0 I mp . (15)
ª §V · º Voc (5)
I ph = I 0 «exp¨¨ oc ¸¸ − 1» +
«¬ © nVth ¹ »¼ R sh From equation (9)-(15), all five parameters can be
explicitly represented as functions of Rsh0 directly or with the
Thirdly, let the maximum power point voltage and incorporation of other equations and the calculation order is
current be Vmp and Imp respectively, and substitute the values RsRsh/Iph/n/I0. However, Rsh0 hasn’t got yet and few
into equation (1): manufacturers give its reference value. Orioli experiment [7]
provides rules to estimate Rsh0. Suppose A is a point on I-V
ª § Vmp + I mp R s · º Vmp + I mp Rs (6)
curve of the solar cell and it has:
I mp = I ph − I 0 «exp¨¨ ¸¸ − 1» −
¬« © nVth ¹ ¼» Rsh
V A ≅ σVoc. (16)
It is well known that the voltage derivative of the power
is null at the maximum power point. Where, =0.2. Point A can be served to calculate Rsh0
appropriately:
∂P ∂I (7)
=V +I =0 VA .
∂V ∂V Rsh 0 = (17)
I sc - I A
Namely [8]:
Unfortunately, the parameters of I-V equation are
unknown and the formulation of IA versus VA is not explicit.
I mp I § I · ª § V + I R ·º 1 § I mp · (8)
= 0 ¨1 - mp Rs ¸ «exp¨¨ mp mp s ¸¸» + ¨1 − Rs ¸ Here, take one more operating point of solar cell and gain
Vmp nVth ¨© Vmp ¸¹ ¬« © nVth ¨
¹¼» Rsh © Vmp ¹
¸
the voltage and current value of the point (Vop, Iop), substitute
it into equation (1), yields:

320
 ª § Vop + I op R s · º Vop + I op Rs 5) Calculate the fitness
I op = I ph − I 0 «exp¨¨ ¸¸ − 1» − (18) Calculate the fitness of offspring population. The
¬ © nVth ¹ ¼ Rsh approach is same as 2).
6) Selection
Thus, the objective function is defined as follows: Combine the offspring population with the parent
population to form the cross generation population. Sort the
ª § Vop + I op R s · º Vop + I op Rs individuals by their fitness f(Rsh0) and select the individuals
f (Rsh0 ) = I ph − I op − I 0 «exp¨¨ ¸¸ − 1» − rank in the top 50% to constitute the new elite population.
¬ © nVth ¹ ¼ Rsh Take the elite population generated at 6) as the new
.(19) parent population and go back to step 2). So, 2-3-4-5-6-2-…,
Calculating the minimum value of f(Rsh0) based on follow by recycling, actualizing the evolution of population.
genetic algorithm, when the value of Rsh0 corresponding to The number of cycles (iterations) is 50.
the minimum f is achieved, the five parameters are solved
V. RESULTS
simultaneously.
The data of solar cell 3C40 from AzurSpace shown in
III. ALGORITHM DESIGN table 1 are input into the program. Results such as key
In this section, the key points in algorithm design are parameters and the I-V equation are shown in table 2.
Table 1 Datasheet clip of 3C40 by AzurSpace
illustrated. Version Glass, Measurement conditions: 1.5 AMd - 1000 W/m² (ASTM G
1) Population size 173-03), T= 25°C)
Population size of each generation is 80.
2) Encoding
Binary encoding is adopted. An example of data structure
for recording individual’s information is given here.
{{1, 1, 0, 0, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 Table 2 Conclusion of the program
}, 34210060/479349, 1.77096*10-8, 7, {53, 32}}
In this data structure, the first item is the chromosome
sequence of an individual, the length of which is determined
by input parameter. The second item is the value of the
chromosome, determined by the range of solution and the
length of the chromosome. The third item is the fitness. The
fourth item is the crossover position. The final one is the
numbers of two parent chromosomes that execute crossover.
3) Crossover The I-V curve of 3C40 plotted is shown in fig.2.
One-point crossover with random crossover position is
applied. The crossover probability is 95%.
4) Mutation
Each bit in a chromosome changes from 0 to 1 or 1 to 0
with the probability of 2%.
5) Selection
Cross generational elitist selection is used.
IV. PROGRAM DESIGN
In this section, the program flow of the algorithm is
described.
1) Generate the initial population randomly
2) Calculate the fitness
Calculate the fitness of initial (parent) population. Take
objective function as fitness function, namely, substituting
the individual into equation (19) to get f(Rsh0).
3) Crossover
Get individuals in parent population paired off. Then
judge the pairs whether execute crossover or not one by one.
Thus new individuals are generated and the offspring
population is formed.
4) Mutation
Mutate all individuals in offspring population. Judge Figure2. the I-V curve of 3C40
each bit of an individual whether mutate or not. Update the Evolution progress is plotted in fig.3. The horizontal
offspring population. coordinate is the number of generation while the vertical

321
coordinate is the best (plotted in blue) and mean (plotted in ACKNOWLEDGEMENTS
yellow) value of f. The two values of previous generations do The authors are grateful to the Doctoral Starting up
not display in this fig due to the relatively large values. Foundation of Hubei University of Technology for the
project under grant No.BSQD12066, the Open Foundation of
Hubei Collaborative Innovation Center for High-Efficiency
Utilization of Solar Energy under grant No.HBSKFMS20
14020 and the Student Innovation Foundation of Hubei
Collaborative Innovation Center for High-Efficiency
Utilization of Solar Energy under grant No.HBSDB201501.
REFERENCES
[1] J. B. Alexandre, A. Krenzinger. Method for photovoltaic parameter
extraction according to a modified double-diode model [J]. Progress
in photovoltaics: research and applications. 2013, 21(3): 884-893.
[2] S. Suckow, T. M. Pletzer, H. Kurz. Fast and reliable calculation of the
two-diode model without simplifications [J]. Progress in
photovoltaics: research and applications. 2014, 22(11): 494-501.
Figure3. evolution progress of the algorithm [3] A. Oritz-Conde, F. G. Sanchez, J. Muci. New method to extract the
model parameters of solar cells from the explicit analytic solutions of
Experiments show that the algorithm identifies the five their illuminated I-V characteristics [J]. Solar energy materials &
parameters successfully and calibrates a satisfactory single- solar cells. 2006, 90(4): 352-361.
diode equivalent circuit model for solar cell 3C40. During [4] J. B. Bai, Y. Z. Hao, Z. Zhang, et al. A complex method for
the evolution process, the optimal and mean fitness of the simulating multiple types of silicon photovoltaic modules [J]. Acta
population decrease steadily and finally converge. Energies Solaris Sinica. 2014, 35(9): 1586-1591.
[5] J. H. Gao, J. Y. Su, Y. Y. Li. Research on the solving algorithm of
VI. CONCLUSION solar cell model parameters [J]. Acta Energies Solaris Sinica. 2012,
33(9): 1458-1462.
In the present work a method to extract parameters of a
[6] A. Jain, A. Kapoor. Exact analytical solutions of the parameters of
single-diode equivalent circuit model is presented. The real solar cells using LambertW-function [J]. Solar energy materials
method limits the amount of measurements needed for the & solar cells. 2004, 81(11): 269-277.
parameter extraction. It is mainly based on the data typically [7] A. Orioli, A. Di Gang. A procedure to calculate the five-parameter
included in manufacturer’s datasheet. Only one more model of crystalline silicon photovoltaic modules on the basis of the
operating point is required, which is interesting for end users tabular performance data [J]. Applied energy. 2013, 102(6): 1160-
1177.
with little testing resources. The single target and single
parameter objective function designed for genetic algorithm [8] J. Cubas, S. Pindado, M. Victoria. On the analytical approach for
modeling photovoltaic systems behavior [J]. Journal of power
reduces the solving dimension of the algorithm. The sources. 2014, 247(9): 467-474.
algorithm is applied to fit the I-V curve of AzurSpace’s solar [9] G. Segev, G. Mittelman, A. Kribus. Equivalent circuit models for
cell 3C40 and performs well. triple-junciton concentrator solar cells [J]. Solar energy materials &
solar cells. 2012, 98(10): 57-65.

322