You are on page 1of 12

Available online at www.sciencedirect.

com

Solar Energy 93 (2013) 322333 www.elsevier.com/locate/solener

A simple accurate model for the calculation of shading power losses in photovoltaic generators
ndez a, F. Almonacid b, P.J. Pe rez-Higueras b P. Rodrigo a,, Eduardo F. Ferna
a

n University, Las Lagunillas Campus, C6 Building, 23071 Jae n, Spain Center of Advanced Studies in Energy and Environment, Jae b n University, Las Lagunillas Campus, A3 Building, 23071 Jae n, Spain Electronic Engineering and Automatic Department, Jae Received 13 February 2013; received in revised form 10 April 2013; accepted 19 April 2013 Available online 20 May 2013 Communicated by: Associate Editor Jan Kleissl

Abstract This paper presents a simple accurate model that allows calculating the shading power losses in photovoltaic generators in a fast, easy to implement way. Calculated losses for dierent generators and shading scenarios are compared to those obtained with a detailed model based on solving the whole currentvoltage curve of the generators. Results show a good agreement between the proposed model and the detailed model. The proposed model is useful for the calculation of the instantaneous power losses and also gives good results in the calculation of the energy yield, requiring less information as input than other detailed models and requiring low computational eort. 2013 Elsevier Ltd. All rights reserved.

Keywords: Photovoltaic generators; Shading power losses

1. Introduction The presence of shading in photovoltaic (PV) generators causes disproportional power losses that are dicult to quantify (Drif et al., 2008, 2012; Quaschning and Hanitsch, 1995; Woyte et al., 2003). In order to calculate these losses, two approaches can be used. The rst one is to use a model based on solving the whole currentvoltage curve of the generator. These kinds of models are accurate, but are difcult to implement, require a great deal of information as input and require extensive computation. The second option is to use a simplied model, easy to implement and fast. This option is preferable in practical applications

Abbreviations: PV, photovoltaic; RMSE, Root Mean Square Error; MCS, Most Concentrated Shadow; MDS, Most Distributed Shadow; SEM, single exponential model; RMSD, Root Mean Square Dierence. Corresponding author. Tel.: +34 953213518; fax: +34 953212183. E-mail address: prodrigo@ujaen.es (P. Rodrigo). 0038-092X/$ - see front matter 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.solener.2013.04.009

and can be easily integrated in common yield estimation software. Nowadays, we can nd in the scientic literature models for shading losses calculation based on solving the whole a currentvoltage curve of the generator (Alonso-Garc et al., 2006; Brecl and Topi, 2011; Kajihara and Harakawa, 2005; Karatepe et al., 2007; Kawamura et al., 2003; Quaschning and Hanitsch, 1995, 1996; Silvestre and Chouder, 2008). These models require as inputs all the details concerning the basic components of the generator (solar cells and bypass diodes), as well as the details about the electrical interconnections between them. Knowledge of the amount of shading on each solar cell is also required. Once the behavior of the basic components is known, the behavior of the whole generator is calculated by applying the laws of the Circuit Theory. This implies dealing with non-linear equations systems that must be numerically solved. The main advantage of these models is their accuracy. However, their practical application presents diculties. First, a great deal of information is required as input in

P. Rodrigo et al. / Solar Energy 93 (2013) 322333

323

order to be able to reproduce the IV curves of the solar cells and bypass diodes. This information is not easy to get. On the other hand, dealing with the whole IV curve of the basic components is problematic from a computational point of view, taking into account that a minimum of 50 points is required to determine an IV curve (Blaesser and Munro, 1995) and that a PV generator is composed of thousands of solar cells. In addition, solving the non-linear equations that govern the behavior of the generator is time consuming. Finally, the implementation of these models and the numerical algorithms is not easy. So these models are oriented to research works and not to their common use by the PV professionals in yield estimation (Klise and Stein, 2009). When it is not possible to have a set of IV curves that characterize the components of the generator, the user is forced to apply a simplied model. Existing software packages rely on simplied expressions that allow a fast estimation of the shading losses (Institute of Environmental Sciences, 2012; Laplace systems, 2012; National Renewable Energy Laboratory, 2012). However, an experimental camnez-Moreno et al. (2010) showed paign carried out by Mart that these expressions lead to appreciable errors. These authors proposed a more accurate model based on an empirical equation that improves the previous methods. It shows good results in the calculation of the energy yield. However, it still gives important deviations in the calculation of the instantaneous power losses. Depending on the considered generator and its particular conguration, the authors reported Root Mean Square Errors (RMSEs) between the modeled and the experimental power losses of up to 26%. So, it would be desirable to have a model that combines simplicity, good results in the calculation of the energy yield and a better characterization of the instantaneous shading losses. Such a model is proposed in this paper. The model consists on several analytical equations that can be easily implemented, require less information than existing detailed models and require low computational eort. In this paper, the proposed model is compared to a detailed model based on solving the whole currentvoltage curve of the generator. The detailed model is very accurate, so that it is used as a baseline to compare the proposed model. There is wide experience that the detailed model is able to represent a real PV generator (Kawamura et al., 2003; Quaschning and Hanitsch, 1996; Silvestre and Chouder, 2008). Results of the present research show that the proposed model provides an adequate characterization of the instantaneous power losses, as well as a good estimation of the energy yield. The paper is organized as follows: in the second section, some denitions related to the research are stated; in the third section, a detailed model based on solving the whole currentvoltage curve of the generator is briey summarized; in the fourth section, the proposed model is explained; the fth section presents the results of applying the proposed model and the detailed model

to the same shading scenarios, and the dierences between both models are discussed; nally, in the sixth section, the conclusions and future actions related to the research are commented. 2. Denitions The generic structure of a PV generator as considered in the present research is shown in Fig. 1. The generator is divided into blocks, where a block is composed of several series-connected solar cells protected by a bypass diode. The dimensions of the generator are dened by the parameters NCELL, NBLOCK and NSTR. A block is composed of NCELL solar cells in series, a string is composed of NBLOCK blocks in series and the generator is composed of NSTR strings in parallel. This way, the generator contains NBLOCKNSTR diode-protected blocks. The total number of cells is NCELLNBLOCKNSTR. The following denitions will be used: Geometric shading factor (sG): it is the ratio between the shaded area (ASH) and the total area (A). sG ASH =A 1

The geometric shading factor can be dened for a solar cell (sG,CELL), for a diode protected block (sG,BLOCK), for a string of series-connected blocks (sG,STRING) or for a generator (sG,GEN). Mean global irradiance on the plane of the shaded PV device (GSH): it can be determined from the components of the global irradiance (G) and the geometric shading factor (sG). The global irradiance G can be written as the sum of four components: a direct component (B), a diuse circumsolar component (DCIR), a diuse isotropic component (DISO) and an albedo component (R). The geometric shading factor only aects the two directional components: B and DCIR (Quaschning and Hanitsch, 1995, 1998). So, the following can be written: GSH 1 sG B DCIR DISO R
ISO

In practice, D + R can be measured with a calibrated module placed on the shaded area and G, with a calibrated module placed on the illuminated area. This way, B + DCIR is obtained by subtracting G (DISO + R) and GSH is obtained from Eq. (2). The mean global irradiance can be dened for a cell (GSH,CELL), a block (GSH,BLOCK), a string (GSH,STRING) or a generator (GSH,GEN). Shading factor (s): it is dened as the ratio between the global irradiance loss due to shading and the total global irradiance: s G GSH =G 3

324

P. Rodrigo et al. / Solar Energy 93 (2013) 322333

Fig. 1. Structure of a generic PV generator.

Again, it can be dened for a cell (sCELL), a block (sBLOCK), a string (sSTRING) or a generator (sGEN). From (3), an alternative way of expressing the mean global irradiance is: GSH 1 sB DCIR DISO R; 4

i.e. the factor s aects the four components of the global irradiance. Shading intensity factor (sI): it is the ratio between s and sG: s I s =s G 5

Number of shaded cells in a block (nS,ij): for a particular block (i, j), it is the number of cells in the block with a geometric shading factor sG,CELL > 0. Number of totally shaded cells in a block (nT,ij): for a particular block (i, j), it is the number of cells in the block with a geometric shading factor sG,CELL = 1. It is obvious that, for a block, the following is true: 0 6 nT ;ij 6 nS ;ij 6 N CELL 8

Number of shaded blocks in a string (nS,j): for a particular string j, it is the number of blocks in the string with a shading power losses factor LSH,BLOCK > 0. Dierent kinds of shadows will be considered in order to develop the proposed model for shading power losses calculation. For each block in the generator, four shadows are dened: the real shadow, the Most Concentrated Shadow (MCS), the Most Distributed Shadow (MDS) and the approximated real shadow. Fig. 2 shows the values of the cell shading factors for the four dened shadows in a generic block. The four shadows for a block are dened in the following way: Real shadow: the real shadow has a geometric shading factor of 1 for the rst nT,ij cells. Hence, the shading factor for these cells is sij,k = sI sG,ij,k = sI 1 = sI. The cells in the positions from nT,ij + 1 to nS,ij have shading factors s1, s2, . . . , sh. The remaining cells are unshaded. MCS shadow: the Most Concentrated Shadow of a block is obtained by concentrating the shaded area in the minimum possible number of solar cells. This way, the MCS

It also can be expressed as the ratio between the directional components of the global irradiance and the total global irradiance: sI B DCIR G 6

Shading power losses factor (LSH): it is the ratio between the power loss due to shading and the theoretical power that would be produced in absence of shadows: LSH P P SH =P 7

PSH being the power produced by the considered PV device under shading and P the theoretical power produced in absence of shadows. It can be dened for a cell (LSH,CELL), a block (LSH,BLOCK), a string (LSH,STRING) or a generator (LSH,GEN). In PV systems, the value of LSH is greater than the value of s, i.e. the relation between the irradiance losses and the power losses is not linear.

P. Rodrigo et al. / Solar Energy 93 (2013) 322333

325

Fig. 2. Shading factors of the individual cells in a generic block for four cases of shadows: real shadow, MCS shadow, MDS shadow and approximated real shadow.

shadow has a shading factor of 1 for the rst Int(sBLOCK,ijNCELL) cells, where Int(sBLOCK,ijNCELL) designates the integer part of the product sBLOCK,ijNCELL. Afterwards, there is one partially shaded cell with a shading factor sr,ij. The remaining cells are unshaded. The value of sr,ij can be obtained taking into account that the block shading factor can be expressed as the mean value of the individual cells shading factors: sBLOCK;ij 1 N CELL
N CELL X k1

Distributed Shadow (MDS). In this case, each block in the string is replaced by an equivalent PV device whose shading factor is equal to the shading power losses factor of the block. Hence, if we name sBLOCK EQ,ij the shading factor of the equivalent PV device for the block (i, j), this parameter is dened as: sBLOCK EQ;ij LSH;BLOCK;ij 12

sij;k

IntsBLOCK;ij N CELL 1 sr;ij N CELL

and the equivalent shading factor of the string sSTRING EQ,j is dened as: sSTRING EQ;j 1 N BLOCK
NX BLOCK i1

sBLOCK EQ;ij

13

! sr;ij sBLOCK;ij N CELL IntsBLOCK;ij N CELL 9 MDS shadow: the Most Distributed Shadow of a block is obtained by distributing the shaded area between the nS,ij cells in a way that every cell is equally shaded. This way, the MDS shadow has a constant shading factor sd,ij for the rst nS,ij cells and the rest of the cells remain unshaded. The value of sd,ij can be obtained by: N CELL 1 X nS ;ij sd ;ij sBLOCK;ij sij;k ! sd ;ij N CELL k1 N CELL sBLOCK;ij N CELL 10 nS ;ij Approximated real shadow: in order to simplify the nal equations of the model, an approximated real shadow is dened in the following way: the rst nT,ij cells have a shading factor sI and the cells in the positions from nT,ij + 1 to nS,ij have a constant shading factor sa,ij. The rest of the cells remain unshaded. The value of sa,ij can be obtained by: N CELL 1 X nT ;ij sI nS ;ij nT ;ij sa;ij sBLOCK;ij sij;k N CELL k1 N CELL ! sa;ij sBLOCK;ij N CELL nT ;ij sI nS ;ij nT ;ij 11

Fig. 3 shows the values of the block equivalent shading factors for the three dened shadows in a generic string. The values of sr,j and sd,j for the case of a string can be obtained in a similar way than for the case of a block: sr;j sSTRING EQ;j N BLOCK IntsSTRING EQ;j N BLOCK s d ;j sSTRING EQ;j N BLOCK nS ;j 14 15

3. Model based on solving the whole IV curve of the generator Several authors have proposed models based on solving the whole currentvoltage curve of the generator for the calculation of shading power losses from similar points of view a et al., 2006; Brecl and Topi, 2011; Kajihara (Alonso-Garc and Harakawa, 2005; Karatepe et al., 2007; Kawamura et al., 2003; Quaschning and Hanitsch, 1995, 1996; Silvestre and Chouder, 2008). In the present research, a detailed model inspired in some of these works has been implemented. As it was commented in Section 1, there is experimental evidence that this model is very accurate and provides a good characterization of real PV generators (Kawamura et al., 2003; Quaschning and Hanitsch, 1996; Silvestre and Chouder, 2008). Most of the reviewed papers rely on the standard Single Exponential Model (SEM) in order to characterize the

On the other hand, for a given string j in the generator, three kinds of shadows are dened: the real shadow, the Most Concentrated Shadow (MCS) and the Most

326

P. Rodrigo et al. / Solar Energy 93 (2013) 322333

behavior of a solar cell. This model is composed of a current source, a diode, a series resistance and a shunt resistance. As, in shading scenarios, the operation of the cell with high reverse voltage is possible, the SEM model must be extended to take into account the avalanche eect. Several authors have proposed models for solar cells working in reverse voltage (Hartman et al., 1980; Lopez Pineda, 1986; Spirito and Abergamo, 1982). For the purposes of the present research, we have followed the formulation of Bishop (1988). It is based on including a voltage source in series with the shunt resistance. Fig. 4 shows the electrical circuit proposed for modeling the solar cell. The M term is dened as the exponential function n M a1 V j =V BR , Vj being the pn junction voltage. This way, the equation of the solar cell IV curve can be expressed as: I I PH I 0 fexpV IRS =mVT 1g V IRS =RP f1 a1 V IRS =V BR n g 16

Fig. 4. Solar cell model including the avalanche eect.

equation that depends on eight parameters: IPH (the photocurrent), I0 (the saturation current), m (the diode ideality factor), RS (the series resistance), RP (the shunt resistance), VBR (the avalanche breakdown voltage), a (the factor of the Bishops term) and n (the exponent of the Bishops term). VT is the semiconductor thermal voltage, VT = kT/ q, being k the Boltzmann constant, T the absolute temperature of the junction and q the electron charge. In general, these parameters depend on the operating conditions of the cell: G (global irradiance on the plane of the cell) and TC (cell temperature). In the present study, VBR is assumed to grow linearly with the cell temperature a and Ruiz, 2006; McKay, 1954), parameters (Alonso-Garc a, n and RP are kept constant and the rest of the parameters are allowed to vary with G and TC. There are dierent methods for determining the variable parameters of the model IPH, I0, m and RS (Chan et al., 1986; Charles et al., 1981; Chegaar et al., 2006; Easwarakhantan et al., 1986; Jain and Kapoor, 2004; Ortiz-Conde et al., 2006; rez et al., 2000). In this case, a Phang et al., 1984; Ranua numerical method has been chosen based on building a system of four non-linear equations: three equations are

derived by applying the solar cell equation, Eq. (16), to the short-circuit point, the open-circuit point and the maximum power point; the fourth equation is obtained by applying the condition that the derivative of the power with respect to the voltage must be zero at the maximum power point voltage (Kennerud, 1969). The four non-linear equations are then solved by means of a trust-region optimization method (Powell, 1970). Once the solar cell parameters have been determined, the generator model is based on building a system of equations that govern the behavior of the generator. These equations involve: (1) the equations of the partially shaded solar cells, (2) the equations of the bypass diodes and (3) the equations that relate the system currents and voltages obtained from the Kirchhos laws. The equations for the solar cells are:   ! V ij;k I ijC RS 1 sij;k I PH I 0 exp 1 mVT & ! n ' V ij;k I ijC RS V ij;k I ijC RS 1a 1 I ijC 0 RP V BR 17 sij,k being the shading factor of the cell in the block (i, j) position k. The equations for the bypass diodes are:   ! V ij I 0D exp 18 1 I ijD 0 mD V TD the bypass diode parameters being I0D (saturation current), mD (diode ideality factor) and VTD (thermal voltage).

Fig. 3. Equivalent shading factors of the individual blocks in a generic string for three cases of shadows: real shadow, MCS shadow and MDS shadow.

P. Rodrigo et al. / Solar Energy 93 (2013) 322333

327

These equations in combination with the equations derived from the Kirchhos laws are solved by means of the multivariable NewtonRaphson method. Finally, we have obtained two IV curves: the curve of the unshaded solar cell and the curve of the partially shaded generator. From the rst curve, we calculate the maximum power point power of the unshaded cell (PCELL) and, from the second curve, the maximum power point power of the shaded generator (PSH,GEN). The shading power losses factor of the generator is then obtained by: LSH;GEN 1 P SH;GEN N CELL N BLOCK N STR P CELL 19

LSH MDSij sBLOCK;ij N CELL =nS ;ij

21

4. Proposed model The explanation of the proposed model is organized in three subsections: Calculation of the shading power losses factors of the isolated blocks, LSH,BLOCK,ij (considering that each block is directly connected to an inverter, i.e. it is isolated from the rest of the blocks). Calculation of the shading power losses factors of the isolated strings, LSH,STRING,j (considering that each string is directly connected to an inverter, i.e. it is isolated from the rest of the strings). Calculation of the nal shading power losses factor of the generator, LSH,GEN.

This approximate expression considers that the shading power losses factor grows linearly with the shading factor and gets a value of 1 when sBLOCK,ij = nS,ij/NCELL, i.e. the power generated by the block cancels when the nS,ij cells are fully shaded. Fig. 5 shows an example of the LSH(MDS) and LSH(MCS) functions for a block containing 4 cells with 2 shaded cells. The double arrows indicate the possible values of the block shading power losses factor for two given shading factors sBLOCK1 and sBLOCK2. Once the LSH(MCS)ij and the LSH(MDS)ij shading power losses factors have been determined, the procedure for calculating the shading power losses factor of the block, LSH,BLOCK,ij, is as follows: (1) Calculation of the mean value of the shading factor for the Most Distributed Shadow case (lij): lij
nS ;ij 1 X

nS ;ij

sij;k

k 1

nS ;ij sd ;ij sBLOCK;ij N CELL sd ;ij nS ;ij nS ; ij

22

(2) Calculation of the standard deviation of the Most Concentrated Shadow shading factors with respect to lij (rMCS,ij):
v u nS ;ij u 1 X 2 rMCS;ij t sij;k lij nS ;ij k1 s 2 2 2 IntsBLOCK;ij N CELL 1 lij sr;ij lij nS ;ij IntsBLOCK;ij N CELL 1 0 lij nS ;ij 23

4.1. Shading power losses factors of the isolated blocks Two cases are considered for a partially shaded block; they are named Most Concentrated Shadow (MCS) and Most Distributed Shadow (MDS). The shading power losses factor of the block is calculated for each one of these cases, obtaining LSH(MCS)ij and LSH(MDS)ij. It will be shown that the real shading power losses factor of the block, LSH,BLOCK,ij, will have a value between LSH(MDS)ij and LSH(MCS)ij: LSH MDSij 6 LSH;BLOCK;ij 6 LSH MCSij Therefore, once the MCS and MDS cases are calculated, a procedure is proposed for calculating LSH,BLOCK,ij from these two values. The shading power losses factor of the block for the MCS case is modeled as: LSH MCSij minfsBLOCK;ij N CELL ; 1g 20

(3) Calculation of the standard deviation of the approximated real shadow shading factors with respect to lij (rij): v u nS ;ij u 1 X 2 rij t sij;k lij nS ;ij k1 s nT ;ij sI lij 2 nS ;ij nT ;ij sa;ij lij 2 24 nS ;ij (4) Calculation of the shading power losses factor of the block (LSH,BLOCK,ij): rij LSH;BLOCK;ij LSH MDSij LSH MCSij rMCS;ij LSH MDSij 25

This approximate expression considers that the shading power losses factor grows linearly with the shading factor and gets a value of 1 when sBLOCK,ij = 1/NCELL, i.e. the power generated by the block cancels when one cell is fully shaded. The shading power losses factor of the block for the MDS case is modeled as:

As can be seen, the standard deviation with respect to lij is an indicator of how far is the considered shadow with respect to the Most Distributed Shadow case. A standard deviation of zero corresponds to the MDS case, in which we have a minimum value of the block losses. A standard deviation of rMCS,ij corresponds to the MCS case, in which we have a maximum value of the block losses. The standard deviation of the approximated real shadow will have a value between 0 and rMCS,ij and the model supposes that

328
1 0.9 Lsh(MCS) Lsh(MDS)

P. Rodrigo et al. / Solar Energy 93 (2013) 322333

Shading power losses factor of the block

Lsh,block2

0.8 0.7 0.6 0.5 0.4

Fig. 7 represents this function for a string containing 10 blocks and four values of nS,j (nS,j = 3, 6, 8, 10). It can be seen that the proposed model approximates the function for any value of sSTRING EQ,j and nS,j. Finally, the procedure for calculating the shading power losses factor of the string, LSH,STRING,j, consists of the following equations: lj sd ;j
rMCS;j

Lsh,block1

sSTRING EQ;j N BLOCK nS ;j

28

s IntsSTRING EQ;j N BLOCK 1 lj 2 sr;j lj 2 nS ;j IntsSTRING EQ;j N BLOCK 1 0 lj 2 nS ;j 29

0.3 0.2 0.1 sblock1 0 0 0.1 0.2 0.3 0.4 0.5 sblock2

v u nS ;j u 1 X rj t sBLOCK EQ;ij lj 2 nS ;j i1 LSH;STRING;j LSH MDSj rj rMCS;j LSH MCSj

30

Shading factor of the block


Fig. 5. Example of a block containing 4 cells (NCELL = 4) with 2 shaded cells (nS = 2). The LSH(MDS) function is a lower limit of LSH,BLOCK, while the LSH(MCS) function is an upper limit of LSH,BLOCK. When the block has a shading factor sBLOCK1, the double arrow LSH,BLOCK1 indicates the possible values of the block shading power losses factor. When the block has a shading factor sBLOCK2, the double arrow LSH,BLOCK2 indicates the possible values of the block shading power losses factor.

LSH MDSj 4.3. Shading power losses factor of the generator

31

the shading power losses factor of the block grows linearly with this value of the standard deviation. 4.2. Shading power losses factors of the isolated strings For a given string j in the generator, each block in the string is replaced by an equivalent PV device whose shading factor is equal to the shading power losses factor of the block as calculated in Section 4.1. Again, two cases are considered for the isolated string: the Most Concentrated Shadow (MCS) and the Most Distributed Shadow (MDS). The shading power losses factor of the string for the MCS case is modeled as: ns o STRING EQ;j LSH MCSj min ;1 26 0: 9 i.e. the shading power losses factor grows linearly with the equivalent shading factor of the string and gets a value of 1 when sSTRING EQ,j = 0.9. In order to justify this approximation Fig. 6 is presented. The graphs compare the function LSH(MCS) = f(sSTRING EQ) obtained with the detailed model to the function obtained with the proposed model. It can be seen that the more the number of blocks in the string, the better the proposed model approximates this function. The shading power losses factor of the string for the MDS case is modeled as: & ' 1:1 sSTRING EQ;j nS ;j 27 LSH MDSj min ; 1:1 ;1 0:1 nS ;j =N BLOCK N BLOCK

Once the shading power losses factors of the isolated strings LSH,STRING,j have been calculated, a rst approach to the calculation of the generator shading power losses factor would be: L SH;GEN 1 N STR
N STR X j 1

LSH;STRING;j

32

However, this is an optimistic estimation. Both the blocks and the strings operate in points dierent than the maximum power point due to mismatch eects. These mismatch losses were not considered in the previous calculations. So, in order to take into account these mismatch losses, the shading power losses factor calculated with Eq. (32) must be corrected. We propose the following expression for obtaining the nal value of the shading power losses factor of the generator: h i LSH;GEN 1 e1 L 33 SH;GEN LSH;GEN where a value of the mismatch parameter e = 0.3 has been established. This parameter was calibrated through simulations. While results presented in the next section have been obtained with e = 0.3, a number of simulations were previously done in order to get this value. An algorithm was implemented that varied the value of e and calculated at each iteration the whole set of generators and shading conditions. The algorithm computed the root mean squared dierence in the shading power losses factor between the proposed model and the detailed model for each value of e. Finally, the value of e = 0.3 was observed to be the value that minimized this dierence.

P. Rodrigo et al. / Solar Energy 93 (2013) 322333

329

Shading power losses factor of the string for the MCS case

Shading power losses factor of the string for the MCS case

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.1 0.2 0.3

Nblock=10

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.1 0.2 0.3

Nblock=20

Detailed model Proposed model

Detailed model Proposed model

0.4

0.5

0.6

0.7

0.8

0.9

0.4

0.5

0.6

0.7

0.8

0.9

Equivalent shading factor of the string

Equivalent shading factor of the string

Fig. 6. MCS case: shading power losses factor of a string containing 10 blocks (left) and 20 blocks (right) as a function of the equivalent shading factor. The function obtained with the detailed model is compared to the function obtained with the proposed model.

Shading power losses factor of the string for the MDS case

1 0.9 0.8 0.7 0.6 0.5


ns=3 ns=6

Nblock=10
ns=8

ns=10

0.4 0.3 0.2 0.1 0 0 0.1 0.2 0.3 0.4 0.5


Detailed model Proposed model

0.6

0.7

0.8

0.9

Equivalent shading factor of the string


Fig. 7. MDS case: shading power losses factor of a string containing 10 blocks as a function of the equivalent shading factor for four values of nS (nS = 3, 6, 8, 10). The function obtained with the detailed model is compared to the function obtained with the proposed model.

5. Results Fig. 8 shows the generators and congurations for two analyzed cases: A and B. The case A is based on the module A, a module composed of 32 solar cells in series with 2 bypass diodes, one diode each 16 cells. A generator is dened composed of six modules. The seriesparallel connection of the modules in the generator is varied in order to obtain four dierent congurations: A1, A2, A3 and A4. The case B is based on the module B, a module composed of 30 solar cells in series with 3 bypass diodes, one diode each 10 cells. A generator is dened composed of eight modules. The seriesparallel connection of the modules in the generator is varied in order to obtain four dierent congurations: B1, B2, B3 and B4.

Shadows on a PV generator can be caused by a lot of dierent external obstacles: trees, buildings, clouds, adjacent arrays, etc. The possible shading cases are innite. So, in the present analysis, we have selected a set of representative shading conditions in order to make the problem aordable. This way, dierent rectangular shadows have been reproduced for the generators. The shape of these shadows is as shown in Fig. 9. This is the shape of the shadows projected on a two-axis solar tracker by an adjacent tracker. Also, these shadows could be caused by a row of xed modules on an adjacent xed array, or by a wall placed near the analyzed generator. Each shading prole is dened by two parameters: the horizontal geometric shading factor (sG,HOR = XSH/XGEN) and the vertical geometric shading factor (SG,VER = YSH/YGEN). This way, 20 dierent shading proles have been simulated for each generator by means of the two implemented models. The values of sG,HOR and sG,VER for the 20 reproduced shadows are indicated in Table 1. Table 2 shows in detail the results of ve example shadows applied to the generators A1, A2, A3 and A4. The table indicates the shading power losses factor obtained by the detailed model and by the proposed model for the ve selected shadows. Also, the relative dierence between models in percentage is shown in brackets: relative difference% LSH;proposed model LSH;detailed model 100 34

For these generators and example shadows, the relative dierences vary between 9.87% and 15.91%. These large values can surprise the reader. However, it must be taken into account that these values correspond to specic examples. The advantages of the proposed model are better shown if we examine the averaged values of the whole set of simulations. For this purpose, the Root Mean Square Dierence (RMSD) between the proposed model and the detailed model has been calculated for each generator by means of:

330

P. Rodrigo et al. / Solar Energy 93 (2013) 322333

Fig. 8. Generators A1, A2, A3, A4 and B1, B2, B3, B4 under study.

v u N u1 X RMSD% t relative difference%2 N i1

35

N being the number of analyzed shadows (N = 20). Results are shown in Table 3.

The global RMSD is 8.0%, what is a good value. Another existing simplied model (that developed by nez-Moreno et al. (2010)) provides Root Mean Mart Square Errors between modeled and experimental power losses between 12% and 26% depending on the analyzed generator. This means that the proposed model seems to

P. Rodrigo et al. / Solar Energy 93 (2013) 322333

331

Fig. 9. Type of shadows reproduced for the generators.

reduce the error in the shading power losses calculation. However, it must be taken into account that the calculated RMSDs correspond to the specic shading proles dened in Table 1. If other proles were simulated, the global RMSD could be dierent. The presented results only indicate that, for the considered generators and shading proles, the model is accurate. In order to generalize the results, many other generators and shadings should be simulated. In this paper, we have limited the number of shading scenarios and therefore, the global RMSD of 8.0% cannot be considered denitive. The proposed model also seems to be adequate for the calculation of the energy yield. Next, we analyze an example of annual energy calculation. Obviously, it is only a particular example. The number of possible cases is innite and therefore the results of this example cannot be considered denitive. A PV plant composed of 25 two-axis solar trackers has been simulated. The 25 trackers are equally spaced on a grid of L L m (Fig. 10) and the shadows projected on the central tracker by the 24 adjacent trackers have been reproduced. The reproduced shadows are fully realistic, i.e. the exact geometric shape has been computed at each

timepoint from the plant conguration and the solar position. Each tracker supports a PV generator composed of 3 strings of 12 modules type A. Fig. 11 shows the evolution of the central tracker power production for the earliest hours of a typical day in March (from sunrise at 6:05 AM to 8:10 AM) considering the typ n, South of Spain, 37.766N ical meteorological data of Jae 3.790W and a distance between trackers L = 10 m. Three curves are represented: the curve of the detailed model, the curve of the proposed model and the curve of the simnez-Moreno et al. (2010). plied model developed by Mart It can be seen that the proposed model approximates better the curve of the detailed model, so that it better characterizes the shading power losses for this particular example. Finally, Table 4 shows the annual shading energy losses of the central tracker for two cases of distance between trackers: L = 10 m and L = 11 m. It can be seen that the proposed model provides a very good approximation of the energy yield in these cases. Experimental annual energy a losses in real tracking plants are in the range 56% (Garc et al., 2008). In the simulations, we have obtained losses slightly higher. This can be due to the fact that, in real plants, the trackers located at the boundaries are less aected by shade and contribute to reduce the plant energy losses. The examples analyzed in this section show that the proposed model seems to provide a good characterization of the shading power losses as well as an accurate estimation of the energy yield. However, it will be necessary to validate the model for many other scenarios in order to ensure the ability of the model to adapt to every case.

Table 1 Dimensions of the 20 reproduced shadows for the generators A1, A2, A3, A4, B1, B2, B3 and B4. Shadow number 1 2 3 4 5 sG,HOR 1.000 1.000 1.000 0.875 0.771 sG,VER 0.094 0.250 0.438 0.094 0.094 Shadow number 6 7 8 9 10 sG,HOR 0.771 0.771 0.490 0.490 0.313 sG,VER 0.250 0.438 0.094 0.750 0.063 Shadow number 11 12 13 14 15 sG,HOR 0.313 0.135 0.135 0.031 0.031 sG,VER 0.688 0.094 0.906 0.375 1.000 Shadow number 16 17 18 19 20 sG,HOR 0.125 0.271 0.333 0.458 0.625 sG,VER 1.000 1.000 1.000 0.875 0.875

Table 2 Results of applying the two implemented models to the generators A1, A2, A3 and A4 for ve example shadows. Shadow number 1 7 12 16 20 sG,HOR sG,VER sG LSH detailed model Gen A1 1.000 0.771 0.135 0.125 0.625 0.094 0.438 0.094 1.000 0.875 0.094 0.337 0.013 0.125 0.547 0.5477 0.8236 0.1320 0.1854 0.7407 Gen A2 0.5477 0.6950 0.1002 0.1432 0.5764 Gen A3 0.5477 0.7383 0.1954 0.2788 0.5760 Gen A4 0.5477 0.7977 0.1647 0.3269 0.6334 LSH proposed model (relative dierence, %) Gen A1 0.7068 (15.91) 0.8266 (0.30) 0.2023 (7.03) 0.2283 (4.29) 0.7920 (5.13) Gen A2 0.7068 (15.91) 0.7436 (4.87) 0.1130 (1.28) 0.1782 (3.50) 0.6403 (6.39) Gen A3 0.7068 (15.91) 0.7708 (3.25) 0.1795 (1.58) 0.2283 (5.06) 0.6403 (6.44) Gen A4 0.7068 (15.91) 0.7964 (0.13) 0.2023 (3.76) 0.2283 (9.87) 0.6798 (4.64)

332

P. Rodrigo et al. / Solar Energy 93 (2013) 322333 Table 3 Root Mean Square Dierence in the shading power losses factor between the proposed model and the detailed model. Generator A1 A2 A3 A4 B1 B2 B3 B4 Global RMSD (%) 8.5 7.2 7.3 7.5 11.0 7.0 7.2 8.4 8.0 Table 4 n for two Annual shading energy losses of the central tracker located at Jae values of the distance between trackers, L. Distance between trackers, L (m) Ground Cover Ratio, GCR Annual energy losses, % Detailed model 7.17 6.01 Annual energy losses, % (relative dierence, %) Proposed model 7.13 (0.04) 5.94 (0.07)

10 11

0.177 0.147

Fig. 10. Conguration of a PV plant composed of 25 two-axis solar trackers.

6. Conclusions and future actions A simple model that allows calculating the shading power losses in photovoltaic generators has been developed. It

require as inputs the geometric shading factor, the number of shaded cells and the number of totally shaded cells for each block in the generator and it also takes into account the seriesparallel interconnection of the blocks. The model does not require solving the full IV curve of the generator so that it is fast and easy to implement. Like other existing simplied models, it requires a reduced amount of information in order to be applied, so that it is useful when it is not possible to get a set of IV curves for the components of the generator. It has shown good performance in the calculation of both the energy yield and the instantaneous shading power losses for some analyzed examples. So, it is a good candidate to be integrated in existing photovoltaic software packages and to be used by the photovoltaic professionals. However, as the number of possible cases of generators and shading scenarios is innite, the model needs to be applied to more cases in order to be completely validated. Also, the presented results are based on comparing the proposed model to a detailed model based on solving the whole IV curve of the generators, so that, as a future action, an experimental campaign will be required in order to prove the validity of the model when applied to real shading scenarios.

1200

Central tracker power production (W)

1000

Detailed model Martnez-Moreno model Proposed model

800

600

400

200

0 6:00

6:10

6:20

6:30

6:40

6:50

7:00

7:10

7:20

7:30

7:40

7:50

8:00

8:10

Time (h:min)
n from sunrise to 8:10 AM. Fig. 11. Central tracker power production as calculated by three models for a typical day in March in Jae

P. Rodrigo et al. / Solar Energy 93 (2013) 322333

333

References
a, M.C., Ruiz, J.M., 2006. Analysis and modelling the Alonso-Garc reverse characteristic of photovoltaic cells. Solar Energy Materials and Solar Cells 90, 11051120. a, M.C., Ruiz, J.M., Herrmann, W., 2006. Computer Alonso-Garc simulation of shading eects in photovoltaic arrays. Renewable Energy 31, 19861993. Bishop, J.W., 1988. Computer simulation of the eects of electrical mismatches in photovoltaic cell interconnection circuits. Solar Cells 25, 7389. Blaesser, G., Munro, D., 1995. Guidelines for the assessment of photovoltaic plants, document C: initial and periodic tests on photovoltaic plants. Joint Research Centre of the European Communities, Ispra Establishment, Report EUR 16340 EN. Brecl, K., Topi, M., 2011. Self-shading losses of xed free-standing PV arrays. Renewable Energy 36, 32113216. 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 Electronics 29, 329337. Charles, J.P., Abdelkrim, M., Muoy, Y.H., Mialhe, P., 1981. A practical method of analysis of the currentvoltage characteristics of solar cells. Solar Cells 4, 169178. Chegaar, M., Azzouzi, G., Mialhe, P., 2006. Simple parameter extraction method for illuminated solar cells. Solid-State Electronics 50, 1234 1237. rez-Higueras, P.J., Aguilera, J., Aguilar, J.D., 2008. A new Drif, M., Pe estimation method of irradiance on a partially shaded PV generator in grid-connected photovoltaic systems. Renewable Energy 33, 2048 2056. rez-Higueras, P.J., 2012. A compreDrif, M., Mellit, A., Aguilera, J., Pe hensive method for estimating energy losses due to shading of GCBIPV systems using monitoring data. Solar Energy 86, 23972404. Easwarakhantan, T., Bottin, J., Bouhouch, I., Boutrit, C., 1986. Nonlinear minimization algorithm for determining the solar cell parameters with microcomputers. International Journal of Solar Engineering 4, 1 12. a, M., Maruri, J.M., Marroyo, L., Lorenzo, E., Pe rez, M., 2008. Garc Partial shadowing, MPPT performance and inverter congurations: observations at tracking PV plants. Progress in Photovoltaics: Research and Applications 16, 529536. Hartman, R.A., Prince, J.L., Lathrop, J.W., 1980. Second quadrant eect in silicon solar cells. In: IEEE Photovoltaic Specialists Conference, pp. 119122. Institute of Environmental Sciences (ISE), 2012. PVSYST V5.62, software for photovoltaic system. University of Geneva. <http://www.pvsyst.com/en/> (accessed 19.12.12). Jain, A., Kapoor, A., 2004. Exact analytical solutions of the parameters of real solar cells using Lambert W-function. Solar Energy Materials and Solar Cells 81, 269277. Kajihara, A., Harakawa, T., 2005. Model of photovoltaic cell circuits under partial shading. In: IEEE International Conference on Industrial Technology, pp. 866870. Karatepe, E., Boztepe, M., C olak, M., 2007. Development of a suitable model for characterizing photovoltaic arrays with shaded solar cells. Solar Energy 81, 977992.

Kawamura, H., Naka, K., Yonekura, N., Yamanaka, S., Kawamura, H., Ohno, H., Naito, K., 2003. Simulation of IV characteristics of a PV module with shaded PV cells. Solar Energy Materials and Solar Cells 75, 613621. Kennerud, K.L., 1969. Analysis of performance degradation in CdS solar cells. IEEE Transactions on Aerospace and Electronic Systems 5, 912 917. Klise, G.T., Stein, J.S., 2009, Models used to assess the performance of photovoltaic systems. Sandia National Laboratories, Report SAND2009-8258. Laplace systems, 2012. SolarPro V4.1. Photovoltaic Systems Simulation Software. <http://www.lapsys.co.jp/english/products/pro.html> (accessed 19.12.12). Lopez Pineda, C.F., 1986. Experimental evaluation of reverse bias stress induced on photovoltaic modules for dierent congurations. Solid Wind Technology 3, 8588. nez-Moreno, F., Mun Mart oz, J., Lorenzo, E., 2010. Experimental model to estimate shading losses on PV arrays. Solar Energy Materials and Solar Cells 94, 22982303. McKay, K.G., 1954. Avalanche breakdown in silicon. Physical Review 94, 877884. National Renewable Energy Laboratory (NREL), 2012. PVWatts V2. A Performance Calculator for Grid-Connected PV Systems. <http:// www.nrel.gov/rredc/pvwatts/> (accessed 19.12.12). a Sa nchez, F.J., Muci, J., 2006. New method to Ortiz-Conde, A., Garc extract the model parameters of solar cells from the explicit analytic solutions of their illuminated IV characteristics. Solar Energy Materials and Solar Cells 90, 352361. Phang, J.C.H., Chan, D.S.H., Phillips, J.R., 1984. Accurate analytical method for the extraction of solar cell model parameters. Electronics Letters 20, 406408. Powell, M.J.D., 1970. A Fortran subroutine for solving systems of nonlinear algebraic equations. In: Rabinowitz, P. (Ed.), Numerical Methods for Nonlinear Algebraic Equations. Quaschning, V., Hanitsch, R., 1995. Numerical simulation of photovoltaic generators with shaded cells. In: Universities Power Engineering Conference, pp. 583586. Quaschning, V., Hanitsch, R., 1995. Shade calculations in photovoltaic systems. In: ISES Solar World Conference. Quaschning, V., Hanitsch, R., 1996. Numerical simulation of current voltage characteristics of photovoltaic systems with shaded solar cells. Solar Energy 56, 513520. Quaschning, V., Hanitsch, R., 1998. Irradiance calculation on shaded surfaces. Solar Energy 62, 369375. rez, J.C., Ortiz-Conde, A., Garc a Sa nchez, F.J., 2000. A new Ranua method to extract diode parameters under the presence of parasitic series and shunt resistance. Microelectronics Reliability 40, 355358. Silvestre, S., Chouder, A., 2008. Eects of shadowing on photovoltaic module performance. Progress in Photovoltaics: Research and Applications 16, 141149. Spirito, P., Abergamo, V., 1982. Reverse bias power dissipation of shadowed or faulty cells in dierent array congurations. In: European Photovoltaic Solar Energy Conference, pp. 296300. Woyte, A., Nijs, J., Belmans, R., 2003. Partial shadowing of photovoltaic arrays with dierent system congurations: literature survey and eld results. Solar Energy 74, 217233.

You might also like