Sei sulla pagina 1di 9
Available online at www.sciencedirect.com ScienceDirect Solar Energy 133 (2016) 35–43 www.elsevier.com/locate/solener

Available online at www.sciencedirect.com

ScienceDirect

ScienceDirect Solar Energy 133 (2016) 35–43 www.elsevier.com/locate/solener Assessment of PV modules

www.elsevier.com/locate/solener

Assessment of PV modules shunt resistance dependence on solar irradiance

Cristiano Saboia Ruschel a,, 1 , Fabiano Perin Gasparin b, 2 , Eurides Ramos Costa a, 1 , Arno Krenzinger a, 1

a Universidade Federal do Rio Grande do Sul, Brazil b Universidade Estadual do Rio Grande do Sul, Brazil

Received 13 August 2015; received in revised form 2 March 2016; accepted 21 March 2016 Available online 18 April 2016

Communicated by: Associate Editor Arturo Morales-Acevedo

Abstract

Modeling and simulation of photovoltaic systems, further than aiding on the project design phase, can be used to emulate the system performance in real time, therefore helping to identify any malfunction that may occur. Among the available performance models for photovoltaic systems, the single diode model is preferred by many authors, since it combines relative simplicity and accuracy. Previous works reported that this model has some limitations on describing the photovoltaic system I V curves under low irradiances, indicating that the variation of the shunt resistance parameter with the irradiance level can be adopted to minimize this drawback. This paper aims to study the shunt resistance dependence on the irradiance level in order to evaluate some of the usual expressions proposed on the literature. A large area pulsed solar simulator model PASAN SunSim 3C was used to acquire the IV characteristics of several photovoltaic modules of different brands and technologies under 20 distinct irradiance levels ranging from 75 W/m 2 to 1000 W/m 2 . The shunt resistance parameter was calculated as the inverse slope of the I V curve in the short circuit region, and fitting equations were derived for each photovoltaic technology. The results in general agreed with previous published works, showing the tendency of an increase of the shunt resistance on lower irradiance levels. Some empirical models tested did not present satisfactory accuracy to reproduce the experimental data. Although simpler, an inverse dependence of the shunt resistance on the irradiance using the measured value at STC as a reference was seen to describe adequately the experimental data. A preliminary study showed that the inclusion of this dependence on the single-diode model indeed increases the model accuracy, reducing the average error on the performed tests by more than half comparing to the original model. 2016 Elsevier Ltd. All rights reserved.

Keywords: Photovoltaic module; Single-diode model; Shunt resistance

Corresponding author. E-mail addresses: cristianosaboia@gmail.com (C.S. Ruschel), gasparin. fabiano@gmail.com (F.P. Gasparin), didircosta@gmail.com (E.R. Costa), arno.krenzinger@ufrgs.br (A. Krenzinger). 1 Address: LABSOL - Av. Bento Gonc¸alves, 9500 – Pre´dio 42712, Porto Alegre, RS CEP 91509-910, Brazil. 2 Address: Av. Bento Gonc¸alves, 8855, Bairro Agronomia, Porto Alegre, RS CEP 91540-000, Brazil.

0038-092X/ 2016 Elsevier Ltd. All rights reserved.

1. Introduction

The development of accurate simulation tools for photovoltaic (PV) systems is a crucial issue not only for project and performance prediction but also for plant oper- ation supervision. An unreliable prediction model may cause financial losses if the energy to be produced by the system is overestimated, discouraging future investments on PV projects. Using an overly cautious estimation to

36

C.S. Ruschel et al. / Solar Energy 133 (2016) 35–43

avoid this problem is not a reasonable alternative either,

since it may lead to a false conclusion that the system is not economically viable. The use of simulation as a tool to supervise photovoltaic systems can be of major assis- tance, since it may indicate when a device is not working correctly by comparing the actual operating conditions with the simulated ones. Evidently, an adequate accuracy

is required on these cases; otherwise it would not be possible

to properly identify a system malfunction. The characteristic I V curve, and therefore the power produced by a photovoltaic system, is highly dependent on the environmental conditions, namely the temperature and the solar irradiance. Several models are available to forecast the operation of photovoltaic systems. However, on the definition of the parameters for many models a problem arises: the data provided by the manufacturer datasheet usually contains only information at standard test conditions (STC), or at best also for the NOCT

(nominal operating cell temperature). So, the knowledge of the dependence of the I V curve parameters on the environmental conditions is necessary in order to establish

a model valid for every condition. Among the alternatives for representing the electrical characteristics of photovoltaic modules, the single diode five-parameter model is one of the most commonly used. Despite combining relative simplicity and accuracy, previous works have shown that as a drawback this model tends to fail on the description of I V curves on lower irradiance conditions (Ishaque et al., 2011). The variation of the shunt resistance parameter with the irradiance can be adopted to minimize this limitation. The aim of this paper is to study the behavior of the shunt resistance with the irradiance level for several photovoltaic modules, in order to check the validity of some usual expressions and propose viable alternatives based on experimental data.

2. Photovoltaic devices modeling

An ideal solar cell can be described, from the basic theory of semiconductors, as a current source in parallel with a diode. Representing the diode current with the expression proposed by Shockley (1950), the characteristic I V curve of the ideal solar cell is given by Eq. (1).

I

¼ I ph ; cell I 0; cell

exp

qV cell

mkT

1

ð1Þ

where I ph is the photogenerated current by the cell, I 0 the reverse saturation or leakage current, q is the electron charge, V the applied voltage on the cell terminals, m the diode ideality factor, k the Boltzmann constant and T the absolute cell temperature. Photovoltaic modules are assembled by connecting sev- eral cells in series. When representing a module, Eq. (1) is modified by adding one term representing the number of cells in series, N s , and by replacing the cell parameters by the module ones, leading to Eq. (2).

I

¼ I ph I 0

exp

qV N s mkT

1

ð2Þ

However, this ideal photovoltaic module equation has little practical use, since photovoltaic modules with such characteristics are impossible to build. For a better descrip- tion of the I V characteristic of PV modules, a series resistance ( R s ) is included. This component represents the resistance of the materials which compose the module and causes a reduction on the power converted by this device, and the resulting current I is then described by Eq. (3).

I

¼ I ph I 0

exp

q ð V þ IR s Þ

N

s mkT

1

ð3Þ

In addition to the parameters of the ideal cell, the series resistance R s must also be determined, increasing the number of unknown parameters to four. This model is known as the four-parameter model, and has been used for several authors such as Xiao et al. (2004) and Chenni et al. (2007), describing the behavior of PV devices adequately for irradiance level of 1000 W/m 2 under different tempera- tures. On other studies though, this method was shown to present a poor performance under certain conditions. Celik and Acikgoz (2007) exhibit a comparison between measured data for a sequence of five days, with two numerical simulations, one using the four-parameter model and another with a five-parameter model. The four-parameter model was shown inadequate to describe the PV system under some of the reported environmental conditions, with the five-parameter model presenting a better agreement. The main difference between the two aforementioned models is the inclusion of the so-called shunt resistance (R sh ), which is connected in parallel with the diode. The complete circuit for the five-parameter model is presented on Fig. 1. The current I is therefore given by Eq. (4).

I

¼ I ph I 0

exp

q ð V þ IR s Þ

N

s mkT

1

V

þ IR s

R

sh

ð4Þ

The shunt resistance accounts for alternative paths for the free carriers produced by the solar radiation. A high shunt resistance means that the vast majority of these carriers generate power, whereas a low resistance indicates large losses, affecting mainly the slope of the I V curve on the proximity of the short circuit region. Breitenstein et al. (2004) attempts to reach a better understanding of the

region. Breitenstein et al. (2004) attempts to reach a better understanding of the Fig. 1. Five-parameter

Fig. 1. Five-parameter model circuit.

C.S. Ruschel et al. / Solar Energy 133 (2016) 35–43

37

nature of this shunt current by applying lock-in thermogra- phy to investigate and classify different kinds of shunts. The author states that the majority of the shunts are process-induced, such as edge shunts, cracks, holes, scratches or aluminum particles, rather than material- induced shunts, these including crystal defects or inversion layers due to SiC inclusions. Although a more precise model for considering the shunt resistance, the five-parameter model still has some limitations, being reported to be less precise under low irradiance conditions, especially on the vicinity of the open circuit voltage (Ishaque et al., 2011). For that reason, double-diode models are frequently used, but these have the drawback of requiring the determination of additional parameters, raising the complexity of the problem. There- fore, an interesting approach to the problem is to improve the single-diode model by studying the behavior of the parameters under different conditions and then adding adjustment equations for some of these parameters appropriately.

3. Dependence of the parameters on irradiance and temperature

According to the traditional approach for the five- parameter model, the photocurrent I ph depends on the irra- diance, I 0 is affected by the temperature of the cell and m , R s and R sh are constant. (Ciulla et al., 2014). The photo- current is known to vary linearly with the irradiance and the reverse saturation is frequently considered to change with temperature following Eq. (5) (Townsend, 1989):

e

q

g

I 0 ð T Þ ¼ DT 3 e AkT

ð5Þ

where D is the diode diffusion factor, approximately con- stant, T the absolute cell temperature, A = c Ns (for ideal cells c is equal to 1) and e g the material energy band gap, which is 1.12 eV for Silicon. Some studies, such as De Soto (2004) suggest that the series resistance has a dependence on the irradiance, decreasing its value for lower irradiance levels, and even assuming negative values for some of the operating condi- tions. As a matter of fact, earlier works also indicate nega- tive values for the series resistance on the five-parameter model on low irradiance conditions (Chan et al., 1986). Nevertheless, most authors do not consider these varia- tions relevant, treating R s as independent of the incident irradiance and temperature and obtaining sufficient preci- sion (de Blas et al., 2002; De Soto et al., 2006). Previous works have also observed a dependence of the shunt resistance on the irradiance. Ba¨tzner et al. (2001) measured the five characteristic parameters for three solar cells from different technologies with irradiance ranging from 0.1 W/m 2 to 1000 W/m 2 and verified an increase of the shunt resistance for lower irradiance levels in all cases. Eikelboom and Reinders (1997) also tested a cell under different irradiance levels and reached similar results,

although using the double diode model. This parameter was also studied for photovoltaic modules exposed to sunlight over the course of the day, and again the tendency was of an increase of shunt resistance with the decrease of irradiance levels. Some recent models propose to consider these variations on the shunt resistance with the irradiance. De Soto et al.

(2006) propose an inverse behavior of R sh with respect to the irradiance ( G), described by Eq. (6), following observa- tion based on experiments performed on the United States National Institute of Standards and Technologies and on an inference exposed by Schroeder (1998). However, the latter only guarantees this approach as valid for very low irradiance levels.

R sh ¼ R sh; ref

G

ref

G

ð6Þ

Other authors, such as Dongue et al. (2012) and Ma et al. (2014) followed on the consideration of a variable R sh with the irradiance, describing it also with Eq. (6). A study performed by Sandia National Laboratories (2015) presents a comparison between the shunt resistance pre- dicted with Eq. (6) and the one extracted from I V curves of several irradiance levels for a monocrystalline silicon module. Although following the same tendency of increas- ing R sh for lower irradiances, the experimental data does not match very closely the expected behavior for the tested PV module. Lo Brano et al. (2010) present a model with additional parameters, which requires the measurement of the IV curve in different irradiance and temperature levels to be obtained. Despite not explicitly defining an equation relat-

ing R sh and G , this model results on an inverse dependence of this resistance with the variation of G , similarly to the De Soto et al. (2006) model. Another procedure is suggested by the commercial soft- ware PVsyst. This software is a well-known tool for design and simulation of photovoltaic systems, which describes the PV panel using the five parameter model presented on Eq. (4). The shunt resistance is calculated by taking the virtual Maximum Power Point conductance (I sc I mp )/ V mp , corresponding to the absolute minimum value of R sh , and taking a given fraction of this quantity (PVsyst SA, 2012). The model also includes an option to vary the value of the shunt resistance with the irradiance, according to the exponential presented on Eq. (7).

R sh ¼ R sh; ref þ R sh ð0Þ R sh; ref

:e 5 : 5G=G ref

ð7Þ

The constant R sh (0) is suggested to assume the value of 12 times R sh for amorphous silicon modules, and 4 times R sh for crystalline silicon panels. However, it is stated that this equation is the result of few measurements, and there- fore the software does not use it as default, leaving as an option for the user. Following these studies, several photovoltaic modules of different brands and technologies were tested under

38

C.S. Ruschel et al. / Solar Energy 133 (2016) 35–43

different irradiance levels in order to evaluate the validity of Eqs. (6) and (7), and to help on the development of a more accurate single diode model. For that purpose, a discussion of the extraction methods for the shunt resistance is presented on the following section.

4. Shunt resistance evaluation

A common method to assess the shunt resistance is by using the Eq. (8), where the R sh 0 is the inverse of the slope of the IV curve at the short circuit region, as given by Eq. (9), and R s the series resistance (de Blas et al., 2002, Lo Brano et al., 2010).

ð8Þ

R sh ¼ R sh0 R s

R sh0 ¼ dV

dI

I ¼ I sc

ð9Þ

On the method proposed by Bouzidi et al. (2006), which was also adopted on studies regarding the behavior of photovoltaic cells on different conditions (Cheegar et al., 2013; Cuce et al., 2013), although different coefficients are used, the shunt resistance is ultimately calculated by Eq. (8). However, the second term on Eq. (8) is often neglected as it is usually much smaller than the first. The inverse of the slope of the I V curve at the short circuit region is con- sidered to provide an accurate estimation for the shunt resistance for the single-diode model, and this method, R sh = R sh 0 , is preferred by several authors (Phang et al., 1984; Hadj Arab et al., 2004). Furthermore, Chan et al. (1986) compares different parameter extraction methods and states that for the assessment of the shunt resistance, Eq. (9) provides a better estimation than the use of curve fitting methods. Another approach for this problem is proposed by Villalva et al. (2009), who applies an iterative process to find the pair (R s , R sh ) in which the calculated maximum power coincides with the P mp provided by the manufac- turer. This strategy has the advantage of not requiring the measurement of the characteristic curve, relying only on data available on the PV module datasheet. Although useful for practical applications, this method can only be used for the extraction of parameters at the STC, since this is the usual condition for which the manufacturer provides data.

5. Measurement apparatus and procedure

The electrical characterization of the PV modules was performed using a LAPSS (Large Area Pulsed Solar Simu- lator), model PASAN SunSim 3C that is installed at the facilities of Solar Energy Laboratory (LABSOL) at Federal University of Rio Grande do Sul (UFRGS) – Brazil. The PASAN SunSim 3C LAPSS has a 10 ms flash and an illuminated area of 2 m 2 m for the device under test. The electrical accuracy is at least 0.2% and the light collimation is less than 15 . This LAPSS complies with

Table 1 LAPSS characteristics and IEC 60904-9 requirements for a class A solar simulator.

Characteristic

Class A solar simulator

LAPSS PASAN

SunSim 3C

Non-uniformity Temporal instability Spectral match ratio

2%

<1%

2%

<1%

0.75–1.25

0.875–1.125

the international standard IEC 60904-9 (2007) and is rated as AAA, even exceeding all standard requirements as shown in Table 1. Tests performed on site after installation of the solar simulator have shown a spatial non-uniformity of 0.27%, therefore better than the nominal features. The flash non-uniformity was tested by measuring the short circuit current of a multicrystalline solar cell in 64 positions over the device under test area. The I V curves were traced in 10 ms under a single volt- age sweep varying from I sc to V oc , acquiring 418 I V pairs. The raw I V curve is corrected to STC conditions according to IEC 60891 (2009) by the LAPSS software. The room temperature was maintained at 25 C±1 C, and for each measurement the module temperature was required to be stable between this range as well. A set of six masks with attenuation ratios of 0.7, 0.5, 0.4, 0.3, 0.2 and 0.1 was used to trace the I V curves over a wide range of irradiance levels. The I V curves were traced with irradiances varying from 1000 to 75 W/m 2 with 50 W/m 2 intervals. The measurements of 1000, 950, 900, 850, 800 and 750 W/m 2 were performed without the use of attenua- tion masks, only by regulating the flash intensity. The mea- surements of 700, 650, 600 and 550 W/m 2 were performed with the mask with ratio of 0.7 and so on, until the 75 W/ m 2 irradiance level, on which was employed the mask with the attenuation ratio of 0.1. The lowest irradiance level was 75 W/m 2 instead of 50 W/m 2 due to technical limitations. The LAPPS software gives several parameters from each I V curve, including short-circuit current, open circuit voltage, maximum power, maximum power current, maxi- mum power voltage and shunt resistance ( R sh ). The method for R sh extraction on this work is the one based on the direct use of Eq. (9), with R sh = R sh 0 , which gives suffi- ciently precise results with relative simplicity. The R sh value is obtained from linear regression of the measured I V pairs ranging from 0 V to V mp /2. This interval was seen to contain sufficient points to minimize the effects of mea- surement noises, at the same time that it is entirely inside the linear region of the I V curve for all measurements. The global uncertainty for the I V curve parameters would involve reference cell and temperature sensor cali- bration uncertainties, among several others typical for an accurate I V curve measurement, as presented by Mullejans et al. (2009). However, since the absolute values of the variables are not of major interest, the measurement uncertainty for the purpose of this work is only the one

C.S. Ruschel et al. / Solar Energy 133 (2016) 35–43

39

related to the R sh determination. In order to calculate this expanded uncertainty, repeatability tests were conducted by measuring the I V curve of a multicrystalline PV module 26 times consecutively. Assuming that the measurement noise is constant, it is reasonable to conjecture that their influence on a smaller current measurement is more pronounced, and therefore the uncertainty would be larger at lower irradiances. The repeatability test was performed for two levels of

irradiance: 1000 W/m 2 and 100 W/m 2 . The

expanded

uncertainty for the R sh value was determined by multiply- ing the standard deviation of the R sh measurements by a coverage factor k = 2.105, the coefficient t of Student for a confidence interval of 95%. The relative expanded uncer- tainty was determined dividing the expanded uncertainty by the mean value of R sh , which resulted in ±4.8% for the 1000 W/m 2 irradiance level and ±7.7% at 100 W/m 2 , confirming an increase in the uncertainty of R sh for lower irradiances.

6. Results and discussion

In order to better analyze the behavior of the photo-

voltaic modules, the measured data was sorted into smaller groups according to their cell technology. Modules built with the same photovoltaic technology are expected to have more similar characteristics with one another. Hence, six groups were established: one including 17 multicrystalline modules, another with 7 monocrystalline modules, a third with 2 CIGS panels, and three groups with one individual PV module on each: amorphous silicon, tandem with amorphous/microcrystalline silicon, and Cadmium Telluride. The measurements from two of the modules, one multicrystalline and one monocrystalline,

were discarded due to negative values for the shunt resis- tance on low irradiance levels. These unexpected results can be credited to measurement noises, since on modules which present a very small slope it is possible that these

noises may cause the linear regression to have an upward direction. Characteristic I V curves with upward-trending current, and therefore negative shunt resistances, were also observed by Sandia National Laboratories (2015), and the approach of discarding the corresponding measurements was taken as well. Moreover, this occurrence was observed for only 2 of the 580 performed I V curve measurements on this work, representing 0.34% of the tests. Fig. 2 presents the obtained R sh for every measurement, as well as a fitting power law equation for each group of PV modules tested. For the sake of comparing the shunt resistance dependence of each module on the irradiance, instead of its absolute value, the results were normalized taking the value at STC conditions as the reference.

A tendency of increase on the shunt resistance for

lower irradiances was observed on all groups. Further- more, differences were observed between the distinct photovoltaic technologies. CIGS modules presented less variation on their shunt resistance with the irradiance, with

its higher value being nearly 7 times larger than the reference value at STC. By contrast, the Tandem module displayed a variation of 40 times between the reference value and the lowest measured irradiance. The Cadmium

Telluride module increased about 9 times its shunt resis- tance decreasing the radiation from 1000 W/m 2 to

100 W/m 2 , and at 75 W/m 2 the obtained value for this

parameter was 76 times larger than the reference one. The correlation coefficient, R 2 , for the fitting equations was calculated, with its value being in general larger for the groups with less measured modules. The amorphous silicon technology, from which only one module was measured, presented the largest correlation coefficient. The Cadmium Telluride technology, although also a single-module group, presented a relatively low correlation coefficient, meaning that the fitting equation does not represent precisely the measured points. As for the groups with several measured modules, namely the monocrys- talline and multicrystalline technologies, it was noted a large dispersion on the shunt resistance behavior with the irradiance. Moreover, it was perceived that even panels from the same manufacturer and belonging to the same model exhibited fairly different behaviors in some cases. In order to compare the different groups, each obtained fitting equation was plotted on Fig. 3. This figure includes also a representation of Eq. (6), corresponding to the De Soto et al. model and Eq. (7), proposed on the PVsyst model, to help on the evaluation of their validity for the tested modules. The two suggested values for R sh (0) on

Eq. (7) were included. For the sake of clarity, Table 2 compiles the represented equations. On this analysis, the curves were extrapolated to irradiance levels lower than the ones used on the experiment. Except for the tandem technology, the six fitting equa- tions provided very similar results for irradiances over

500 W/m 2 . For lower irradiance values, larger deviations

occurred, with the multicrystalline and the CIGS modules showing a lower increase of the shunt resistance for low radiation levels, while the CdTe and the Tandem techno- logy panels presented a much more significant growth. The values of R sh for these groups under the lower levels

of irradiance are not visible on the chart as they surpass the chosen axis limitation. The fitting for the amorphous silicon module was shown to have a very close agreement with Eq. (6), which seems therefore to be suitable for this photovoltaic technology. As for the exponential pro- posed by Eq. (7), the suggested value for crystalline silicon modules did not agree with any of the verified fitting equa- tions, while the recommended coefficient for amorphous silicon presented a behavior which was quite similar to that of the crystalline silicon modules. A further analysis is introduced by Fig. 4, which shows a histogram representation of the absolute values of R sh , ref . Both crystalline silicon groups presented a high variability on the reference shunt resistance value, ranging from

173 X to 729 X for the multicrystalline and from 148 X to

1320 X for the monocrystalline technology. Once again,

40

C.S. Ruschel et al. / Solar Energy 133 (2016) 35–43

40 C.S. Ruschel et al. / Solar Energy 133 (2016) 35–43 Fig. 2. Dimensionless shunt resistance

Fig. 2. Dimensionless shunt resistance versus irradiance for each measured group.

C.S. Ruschel et al. / Solar Energy 133 (2016) 35–43

41

C.S. Ruschel et al. / Solar Energy 133 (2016) 35–43 41 Fig. 3. Comparison between the

Fig. 3. Comparison between the obtained fit equations and the methods proposed by De Soto and PVsyst.

Table 2 Summary of the tested equations.

Model

Equation

De Soto et al. (2006)

PVSYST

PVSYST

Multicrystalline fit

Monocrystalline fit

CIGS fit

Tandem fit

Amorphous fit

CdTe fit

amorphous

crystalline

R sh ¼ 1000G 1

R sh ¼ R sh ¼

R sh ¼ 219: 7G 0: 75 R sh ¼ 540: 7G 0: 9 R sh ¼ 144: 7G 0: 7 R sh ¼ 12132: 6G 1:31 R sh ¼ 832: 8G 0: 96 R sh ¼ 3918: 1G 1: 19

1 þ 11 exp

1 þ 3 exp

5: 5G

1000

5:5G

1000

even modules with the same nominal characteristics presented large variability on this parameter. For the other groups the results are shown on Table 3. Nothing can be stated about these technologies variability since only one or two modules of each were tested. It is interesting to notice, though, that the absolute value of shunt resistance for these four groups is, in general, similar or superior to that of the best tested crystalline silicon modules.

Table 3 R sh , ref for the other photovoltaic groups measured at STC.

Module

CIGS A

CIGS B

DA142

DA100A5

FS280

R sh , ref (X )

1978

2086

2806

1218

3472

The effect of increasing the shunt resistance is larger when its value is low, since the leakage current decays asymptotically with it, as can be derived from Eq. (4). Therefore, the measured data indicate that considering a variation of the shunt resistance with the irradiance is especially relevant for the crystalline silicon modules. On an additional study, the IV curve of the multi- crystalline and monocrystalline modules measured at every irradiance level was simulated with the single diode model using two different approaches. In both cases, the parameters were extracted from the I V curve at the 1000 W/m 2 irradiance. While on the first one all the parameters were maintained constant, with the exception of I ph , on the second one Eq. (6) was used to determine the R sh value. Fig. 5 displays a comparison between I V curves at 150 W/m 2 to illustrate the effectiveness of implementing the variation of the shunt resistance with the irradiance on the single-diode model for one monocrys- talline module of this study. The measured IV curve is better represented by the simulated model which includes the dependence of the shunt resistance on the irradiance level. In order to calculate the overall improvement verified for the single-diode model, the mean absolute error between the measured I V curves and the simulated ones for each module on all irradiance levels was determined. Taking the average value of all errors, the error decreased 57% on the inverse R sh model comparing with the constant one. Despite the occurrence of some variation of this decrease on the error for each individual module, it was seen that the implementation of shunt resistance variation with irradiance provided superior results for every tested module.

provided superior results for every tested module. Fig. 4. Distribution of R s h , r

Fig. 4. Distribution of R sh , ref for (a) multicrystalline and (b) monocrystalline groups.

42

C.S. Ruschel et al. / Solar Energy 133 (2016) 35–43

1.2 0.8 Measured Simulated: R sh =R sh,ref (G ref /G) Simulated: R sh =
1.2
0.8
Measured
Simulated: R sh =R sh,ref (G ref /G)
Simulated: R sh = R sh,ref
0.4
0
0
10
20
30
40
Voltage [V]
Current [A]

Fig. 5. Measured and simulated IV curves for a monocrystalline module at 150 W/m 2 .

silicon modules. This equation lies in fact between the

obtained fits for the experimental data, overstating the

shunt resistance value on low irradiance for CIGS and

crystalline cells and understating it for CdTe and Tandem

technologies. This method has the advantage of its simplicity,

since no empirical coefficients are used, and seems to be an

appropriate option if a single equation is desired to describe

the behavior of every photovoltaic technology.

A brief discussion on the inclusion of an inverse

dependence of the shunt resistance with the irradiance

showed that the results are indeed superior when this effect is considered. The mean absolute error for this model was

less than half the error of the original single diode model,

considering the average of all tested crystalline modules

at all irradiance levels. The analysis of the I V curve of one of the modules at a low irradiance level with and

without the R sh dependence on the irradiance corroborated this result, since unlike the original single-diode model, the modified one represented closely the measured data.

Acknowledgments

7. Conclusions

The results obtained from the experiments agreed with previous studies published on the literature, confirming that the shunt resistance on the five-parameter model increases at lower irradiance levels when directly extracted from measured IV curves. Six fitting equations were obtained, one for each tested photovoltaic technology, demonstrating that there are considerable differences between the distinct groups. It must be stressed that the use of the equations here proposed for the behavior of each technology requires caution, since few modules were tested, and there is no guarantee that other modules from the same technology will follow it. Additionally, a high variability among different PV modules of the same technology was verified on the shunt resistance indicating a difficulty to generalize the results, as even PV modules with the same nominal features presented considerable variability on this parameter. The comparison of the obtained fits with the PVsyst method (Eq. (7)) indicated that it may be an interesting alternative to represent the variation of the shunt resis- tance. However, the proposed experimental coefficients in PVsyst SA (2012) may not to be adequate to be used as stated, since the recommended coefficient for the amor- phous silicon modules presented results which were in fact in better agreement with the crystalline modules, whereas the recommended one for the crystalline modules failed to describe any of the tested photovoltaic groups. Further tests should be carried on to establish an improved set of empirical coefficients if this method is chosen to be used. Considering the shunt resistance inversely proportional to the irradiance (Eq. (6)) as proposed by De Soto et al. (2006) was seen to agree very closely with the amorphous

This work was financially supported by CNPq – Brazil (Conselho Nacional de Desenvolvimento Cientı´fico e Tecnolo´ gico) and CAPES – Brazil (Coordenac¸a˜o de Aperfeic¸oamento de Pessoal de Nı´vel Superior).

References

Ba¨tzner, D.L., Romeo, A., Zogg, H., Tiwari, A.N., 2001. CdTe/CdS solar cell performance under low irradiance. In: Proceedings of the 17th European Photovoltaic Solar Energy Conference and Exhibition, paper VB1.40. Munich, Germany. de Blas, M.A., Torres, J.L., Prieto, E., Garcı´a, A., 2002. Selecting a suitable model for characterizing photovoltaic devices. Renew. Energy 25, 371–380. Bouzidi, K., Chegaar, M., Bouhemadou, A., 2006. Solar cells parameters evaluation considering the series and shunt resistance. Sol. Energy Mater. Sol. Cells 91, 1647–1651. Breitenstein, O., Rakotoniaina, J.P., Al Rifai, M.H., Werner, M., 2004. Shunt type crystalline silicon solar cells. Prog. Photovoltaics: Res. Appl. 12, 529–538. Celik, A.N., Acikgoz, N., 2007. Modelling and experimental verification of the operating current of mono-crystalline photovoltaic modules using four- and five-parameter models. Appl. Energy 84, 1–15. Chan, D.S.H., Phillips, J.R., Phang, J.C.H., 1986. A comparative study of extraction methods for solar cell model parameters. Solid-State Electron. 29, 329–337. Cheegar, M., Hamzaoui, A., Namoda, A., Petit, P., Aillerie, M., Herguth, A., 2013. Effect of illumination intensity on solar cells parameters. Energy Procedia 36, 722–729. Chenni, R., Makhlouf, M., Kerbache, T., Bouzid, A., 2007. A detailed modeling method for photovoltaic cells. Energy 32, 1724–1730. Ciulla, G., Lo Brano, V., Di Dio, V., Cipriani, G., 2014. A comparison of different one-diode models for the representation of I-V characteristic of a PV cell. Renew. Sustain. Energy Rev. 32, 684–696. Cuce, E., Cuce, P.M., Bali, T., 2013. An experimental analysis of illumination intensity and temperature dependency of photovoltaic cell parameters. Appl. Energy 111, 374–382. De Soto, W., 2004. Improvement and validation of a model for photovoltaic array performance. M.S. Thesis, Mechanical Engineer- ing, University of Wisconsin-Madison.

C.S. Ruschel et al. / Solar Energy 133 (2016) 35–43

43

IEC 60891 (2009) (International Electrotechnical Commission) Photo- voltaic devices – Procedures for temperature and irradiance corrections to measured I-V characteristics. IEC 60904-9 (2007) (International Electrotechnical Commission) Photo- voltaic devices – Part 9: Solar simulator performance requirements. Ishaque, K., Salam, Z., Taheri, H., Syafaruddin, 2011. Modeling and Simulation of photovoltaic (PV) system during partial shading based on a two-diode model. Simul. Modell. Practice Theory 19,

Ma, T., Yang, H., Lu, L., 2014. Development of a model to simulate the performance characteristics of crystalline silicon photovoltaic mod- ules/strings/arrays. Sol. Energy 100, 31–41. Mullejans, H., Zaaiman, W., Galleano, R., 2009. Analysis and mitigation of measurement uncertainties in the traceability chain for the calibra- tion of photovoltaic devices. Meas. Sci. Technol. 20, 075101. Phang, J.C.H., Chan, D.S.H., Phillips, J.R., 1984. Accurate analytical method for the extraction of solar cell model parameters. Electron. Lett. 20 (10), 406–408. PVsyst SA, 2012. User’s Guide – Pvsyst Contextual Help. Satigny, Switzerland. Sandia National Laboratories, 2015. Parameters Estimation for Single Diode Models of Photovoltaic Modules. Sandia Report, Albuquerque, United States of America. Shockley, W., 1950. Electrons and Holes in Semiconductors. Van Nostrand, New York. Schroeder, D.K., 1998. Semiconductor Material and Device Characteri- zation, second ed. John Wiley & Sons Inc., New York. Townsend, T.U., 1989. A method of estimating the long-term perfor- mance of direct-coupled photovoltaic systems. M.S. Thesis, Mechan- ical Engineering, University of Wisonsin-Madison. Villalva, M.G., Gazoli, J.R., Filho, E.R., 2009. Comprehensive approach to modeling and simulation of photovoltaic arrays. IEEE Trans. Power Electron. 24 (5), 1198–1208. Xiao, W., Dunford, W.G., Capel, A., 2004. A novel modeling method for photovoltaic cells. In: Proc. IEEE 35th Annual Power Electronics Specialists Conference (PESC), vol. 3, pp. 1950–1956.