You are on page 1of 4

ENERGY RATING OF PV MODULES: COMPARISON OF METHODS AND APPROACH

R.P. Kenny1, G. Friesen2, D. Chianese2, A. Bernasconi2, and E.D. Dunlop1 EC, JRC, IES, Renewable Energies Unit, via Fermi 1, 21020 Ispra (Va), Italy 2 SUPSI, DCT, LEEE-TISO, CH-6952 Canobbio-Lugano, Switzerland

ABSTRACT The Matrix and Performance Surface Methods of Energy Rating are related techniques under development for the determination of electrical yield and is intended to be a more useful predictor of performance for installers than Wp alone. Here a power matrix or performance surface as a function of irradiance and ambient temperature P(Gi, Ta) is linked to a climatic condition occurrence matrix N(Gi, Ta) for a particular location. The use of just two independent variables has the advantage of simplicity but it is important to evaluate the possible cost of reduced accuracy due to the exclusion of other variable parameters such as Air Mass for example. The power matrix may be determined by outdoor or indoor measurements, and the complete matrix may also be extrapolated from a reduced data set using models of cell behaviour thereby reducing measurement time. The present paper relates solely to crystalline Si and reports on the comparison of outdoor measurements at two sites, and their further comparison with indoor measurements.

1. ENERGY RATING MEASUREMENTS Global horizontal irradiances and average ambient temperatures are readily obtainable for many locations, which is not the often the case for other meteorological parameters which may be of relevance to the energy yield of photovoltaic modules. An energy rating method that is based on only these two parameters therefore has potentially great practical benefits. Techniques for Energy Rating employing the use of matrices and performance surfaces have previously been shown to be accurate for crystalline Si for the specific test conditions that were reported[1-3]. Here it is demonstrated that the method is robust enough to be able to be transferred between different sites, and hence differing meteorological conditions, and also when using different measurement and analysis techniques. A crystalline Si module has been extensively tested in two different locations, in Ispra (Italy) at the ESTI laboratory and in Lugano (Switzerland) at the LEEE-TISO laboratory. The first phase consisted in an extensive one-year monitoring campaign on the outdoor test facility of LEEETISO. The electrical behaviour of the module has been measured with respect to the meteorological parameters. The module was linked to a Maximum Power Point Tracker (MPPT) and the module and meteo parameters are measured at one-minute intervals, from 5:00 to 22:00 oclock over a one-year period. The data are then converted into a power matrix with a mesh size of 10 W/m2 x 1C. Two matrices are generated, one in dependence of ambient temperature and one in dependence of module temperature. The following phase of the study, executed at the ESTI

laboratory, consisted of two steps. Firstly, in an indoor measurement of the same module and secondly in a campaign of outdoor measurements similar to those performed at LEEE-TISO, but naturally under differing climatic conditions. The indoor measurements consist of a series of IV-measurements at different irradiances and module temperatures. After the indoor characterisation the module was mounted outdoors where its I-V characteristics were continuously monitored together with the relevant environmental data. A dedicated active load connected to the module enabled a complete I-V curve to be measured every five minutes whilst otherwise maintaining the module at its maximum power point. In both outdoor systems the solar irradiance Gi is monitored with a crystalline Si ESTI sensor mounted in the plane of the modules. The back-of-module temperature Tbom is measured with a Pt-100 temperature sensor attached to the rear of the module and the ambient temperature T a is recorded by a meteo station located close to the module mounting rack. At LEEE-TISO the module rack was mounted at a fixed 45 elevation all year round, whilst at ESTI the module rack is adjusted during the year to keep the elevation within 5 of normal incidence at solar noon. In both sites an open-rack mounting arrangement was employed. 2. OUTDOOR-OUTDOOR INTERCOMPARISON 2.1 LEEE-TISO Matrix Generation At LEEE-TISO the power matrix of the module is obtained by filtering and fitting the two raw data matrices Im(Gi, Tbom) and Vm(Gi,Tbom). Im and Vm are the average measured maximum current and voltage for each single climatic condition (Gi,Tbom) during the monitoring period of one year. Initial filtering of the raw data is performed to ensure that erroneous measurements, and data with low occurrence (N<10) and high standard deviation (StDev/Avg>20%) are excluded from further analysis. In the next step the current and voltage data are fitted with the semi-empirical equations (1) and (2) introduced by King et al [4], but with the simplification that angle of incidence (AOI) and air mass (AM) influences has been ignored in this study. These equations have been chosen since they lead to a physically more meaningful model of the module performance than previously used fitting procedures which employed arbitrary functions. The use of such a model allows extrapolation of the measured data set to any required matrix grid by completing and/or extending the matrix in the parts where data are missing. Im= (Gi/G0)(Im,STC+a Im(T-T0)) (1)

Vm= Vm,STC+c1ln(Gi/G0)+c 2(ln(Gi/ G0))+ Vm(T- T0) (2)

The fit of the measured data resulted in parameters Im,STC= 2.74A, a Im= 3.3mA/C, Vm,STC= 34.6V, Vm= 159mV/C, c1= 1.15 and c2= -0.32. The maximum power point current and the maximum power point voltage are 1.56% and 0.45% lower then the respective values measured indoor on the LEEE-TISO solar simulator. These small differences are due in part to the outdoor non-constant air mass and to a difference between the indoor measured module temperature and the outdoor measured back of the module temperature, which is approximately 2C lower than the real cell temperature. 2.2 ESTI Matrix Generation At ESTI the data from the obtained outdoor measurement period of 10 weeks are analysed in the following manner. A binning process is performed to form a matrix of P(Gi, Tbom) on the same mesh as used for the indoor measurements described in Section 3. Measured powers corresponding to irradiances and module temperatures that are sufficiently close to the matrix points for all selected elements are then averaged to produce the final value for that operating point. A compromise must be made between the bin size and the number of data points selected. Too large a bin size results data far from the chosen operating point being included, while too small a bin size can result in not enough data being selected for statistically good results, particularly for shorter monitoring periods as is the case here. In practise it has been found that bins of 5 % of the relevant matrix mesh values produce satisfactory results. Unlike at LEEE-TISO no filtering or removal of outlying points is performed during this analysis. 2.3 Intercomparison For the relatively short 10-week measurement period undertaken at ESTI, the resultant power matrix is only halfcomplete since many conditions were not observed over this time, such as high temperatures/low irradiances. It would be expected that more of these would be observed if the measurements were to have continued over the summer months, for example. However, there are sufficient points to be able to perform a useful comparison with the LEEETISO data. Table I contains the results of this comparison over the observed conditions, and it is seen that generally the differences are quite small. The average of the differences over the matrix is 2.1 %. For the low irradiance level of 100 W/m2 the difference does however reach nearly 10 %. Increasing discrepancies at low irradiances could be expected since it is at these conditions that effects due to variations of angle of incidence, air mass or diffuse light become more significant. Other explanations could be that the maximum power point tracker accuracy may be reduced towards the lower power levels and the model used to fit the experimental data may also be less accurate here. However, as the net contribution to energy production is correspondingly lower at low irradiance levels the effect on total energy prediction is not so great.

500 600 700 800 900 1000

-3.72 -2.86 -3.85 3.55

-2.71 -3.54 -3.50 -2.32 0.62 -0.23

-1.87 -2.31 -2.60 -2.41 -1.74 -0.35

-2.22 -2.25 -2.07 -1.92 -1.51 0.40

3.79 -2.35 -1.12 -0.63 1.23

-0.67 0.80

-1.91

Table I. Difference in percent [%] between the LEEETISO and ESTI outdoor measured power over the P(G,T) matrix [W/m2, C]. A positive value indicates that LEEETISO measured value is larger. 3. INDOOR-OUTDOOR INTERCOMPARISON The module was characterised using standard indoor measurement methods employing pulsed solar simulators at ESTI. In particular, the module was characterised at each point on a 13 x 9 matrix of irradiances and temperatures. Table II shows the measured power for a subset of 72 of these measurement points. The irradiance is varied by measuring the I-V curve at different parts of the decaying lamp pulse and module temperature is varied by placing the module within a temperature controlled chamber fitted with a quartz window. This method may be expected to lead to a degree of spectral variation across the matrix, but this does not appear to have had a significant influence on the results.
G\T 100 200 300 400 500 600 700 800 900 1000 25
7.59 17.04 26.71 36.81 46.93 56.68 66.45 76.95 86.40 96.15

30
7.41 16.63 26.20 36.00 45.93 56.19 65.59 75.77 85.03 94.39

35
7.28 16.38 25.82 35.35 45.24 55.00 64.52 74.28 83.32 92.55

40
7.14 16.03 25.33 34.66 44.31 54.05 63.27 72.66 81.65 90.41

45
6.97 15.62 24.77 33.82 43.21 52.75 61.79 71.11 80.01 88.45

50
6.89 15.33 24.07 33.04 42.18 51.56 59.97 69.07 77.56 86.39

55
6.62 14.87 23.46 32.21 41.14 49.92 58.22 67.30 75.49 84.12

60
6.47 14.50 22.99 31.40 40.22 49.11 57.38 65.98 73.76 81.99

Table II. Matrix of indoor measured power [W] as a function of irradiance and temperature [W/m2 , C]. The complete indoor measured matrix is shown graphically in Fig. 1.

Fig. 1 3-Dimensional visualisation of the indoor measured matrix and fitted performance surface The indoor performance surface as fitted according to equations (1) and (2) is also shown in Fig. 1 and it may be observed that a good fit is obtained. The differences in percent between the indoor measurements and both sets of outdoor measurements are

G\T 100 200 300 400

25
9.42 1.83 -1.86 -2.86

30
0.94 -0.40 -1.89

35

40

45

50

55

60

-2.43

reported in Table III for a reduced set of matrix points where it is seen that the differences are small. The LEEETISO data used are the extrapolated outdoor matrix which slightly increases the error in the inter-comparison. The ESTI data are the real outdoor measured power points resulting in some missing points in Table III. The average differences for the comparisons between the indoor data and the LEEE-TISO and ESTI outdoor data are 2.1 % and 1.3 % respectively. G\T
200 W/m 400W/m 600 W/m 800 W/m 1000 W/m 30C 3.1/2.14 -0.56/1.36 -2.51/1.07 -2.43/-0.12 -1.31/-1.08 40C 2.77/ -0.58/ -2.34/-0.09 -1.90/0.01 -0.61/-1.00 50C 2.91/ 0.13/ -1.61/ -0.77/ -0.07/-0.73 60C 3.81/ 0.82/ -1.03/ -0.39/ 1.15/

Table III. Difference in percent [%] between indoor and outdoor measured power matrices (TISO/ESTI) 4. ENERGY RATING CALCULATIONS 4.1 Power Matrix Temperature Transformation The main utility of the power matrices is the possibility to predict the annual energy production of a PV module under specific climatic conditions. For this the power matrix has to be linked to a climatic condition occurrence matrix N(Gi, Ta) of a particular location. The energy production of the module is obtained by multiplying the elements of the power matrix P(Gi,Ta) by the corresponding climatic condition matrix elements. The link in between the two matrices is only possible if the power matrix is known in dependence of ambient temperature. These can be either measured directly outdoor or indirectly by transforming the measured power matrix in dependence of module temperature into one in dependence of ambient temperature. The transformation is based on the approach of Reference [5] employing the Nominal Operating Cell Temperature (NOCT) which is determined from outdoor measurements according to a standard procedure[6]. Here we have not measured NOCT according to the precise definition which places severe restrictions on the specific mounting and measurement conditions. Instead we have estimated what we term the Nominal Operating Specific Temperature NOST defined as the site and mounting specific module temperature of a module operating @ maximum power point, 800 W/m and 20C ambient temperature. This is established from the complete measured data set by plotting the module and ambient temperature differences against irradiance analogously to the standard NOCT determination procedure. Different values for NOST have been found at LEEE-TISO and ESTI due to differing mounting and climatic conditions, and for the purposes of the present study the average value of 42C has been used in the calculations reported below. The transformation on the measured data is performed according equation (3), Tbom or cell = (NOST - 20) Gi/800+Ta (3)

previously it has been found [9] that in order to produce accurate energy predictions a relatively fine matrix mesh size is required for the meteo matrix. The extrapolation of the ESTI indoor data using equations (1) and (2) results in Im,STC= 2.73A and Vm,STC= 35.2V, comparable to the values reported for the LEEETISO data in Section 2.1. The STC measured values measured at ESTI are Im,STC= 2.84A and Vm,STC= 33.9V, which deviate more from the extrapolated values, but the resultant P m,STC values of 96.3 W and 96.1 W respectively are very close. The extrapolated indoor data matrix is then transformed into an equivalent matrix referenced to Ta using equation (3), but in this case the NOST value was increased by 3C to 45C since the indoor data refers to Tcell not Tbom as for the outdoor case. This increase is justified by an analysis of the Equivalent Cell Temperature ECT[8], calculated using the Voc values of the outdoor test data at ESTI, which finds this mean discrepancy between Tcell and Tbom. 4.3 Energy Prediction Comparisons A comparison of the real module energy production measured at LEEE-TISO with the matrix method calculated energy is made for different data sets (LEEE-TISO outdoor, ESTI outdoor and ESTI indoor). Table IV summarises the energy predictions obtained by using the different LEEETISO and ESTI indoor and outdoor measured power matrices and the percent differences from the energy actually measured for the various cases are also reported. ENERGY PREDICTION USING Energy difference METHOD EMPLOYED kWh % LEEE-TISO energy counter 125.71 --outdoor measured power LEEE-TISO integrated outdoor 125.62 -0.1 measured power LEEE-TISO outdoor measured 123.14 -2.0 power matrix P(Gi, Ta) LEEE-TISO outdoor measured 122.25 -2.8 power matrix P(Gi, Tbom) ESTI outdoor measured power 121.68 -3.2 matrix P(Gi, Tbom) ESTI indoor measured power 121.13 -3.6 matrix P(Gi, Tcell) Table IV. Summary of energy predictions for the different power matrix cases using the LEEE-TISO measured meteo matrix. The percent differences in the last column relate to the energy counter measured value. The real energy production measured by an energy counter and the energy obtained by the integration of the outdoor measured power at LEEE-TISO are seen to be almost identical. The LEEE-TISO predicted energies obtained from the two matrices P(Gi,Tbom) and P(Gi, Ta) are quite close, within 0.8 %, suggesting that the Tbom to Ta transformation does not introduce significant errors. Both however are at least 2 % from the actual measured energy, suggesting that the matrix extrapolation is introducing a net error of this order of magnitude, although it is not clear if the numerical fitting or the form of the model itself is the more significant factor accounting for this difference. The ESTI estimates are close to each other, within

4.2 ESTI Data Extrapolation and Transformation For the energy rating calculations using ESTI data the existing indoor and outdoor data had to be extrapolated to match the LEEE-TISO meteo matrix mesh size since

0.4 % and are at most 1.6 % from the LEEE-TISO estimates. The largest difference observed is between the energy measurement and the ESTI indoor estimate, and at 3.6 % is at an acceptable level. In order to estimate the effect of the choice of the NOST parameter in equation (3), which in practise can vary between systems, a simple sensitivity analysis was performed. This results in the finding that the energy prediction will have a variation of the order of -1.5 % per 5C increase in its value. The overall error in energy prediction is therefore expected to be small if approximate NOST values are used, or even if in fact the module NOCT value is used, at least for open-rack mounting conditions. In conclusion therefore, good agreement is seen throughout the predictions made from the range of data taken in the different conditions reported. Furthermore, it is important to note that standard measurement techniques have been employed to obtain the data used in these predictions[7,8]. 5. CONCLUSIONS The performance of a crystalline Si module has been measured at two different locations enabling comparisons of power output over a range of conditions and predictions of total energy production to be made. The output power comparisons result in very close matches between measurements made at the two sites, and also between these and the reference indoor measurements. The energy produced by the module has also been measured and compared with that predicted using both indoor and outdoor generated matrices. Again the comparison is very good, with the indoor and outdoor estimations being within approximately 1.6 % of each other. The test conditions and measurement and analysis techniques have been significantly different at both sites, demonstrating the fundamental robustness of the methods employed. We can conclude that for a typical crystalline Si module, even when factors such as angle of incidence, Air Mass and diffuse fraction are not taken into consideration, the matrix method employing two independent parameters (ambient temperature, global irradiance) can be a sufficiently accurate means of determining module energy production. Furthermore, it has been shown that with a combination of existing standard test methods an energy rating prediction is possible without the need for the creation of additional tests.

[5]

[6]

[7] [8]

[9]

Characterization Procedure for Photovoltaic Arrays, 2nd World Conference and Exhibition on Photovoltaic Solar Energy Conversion, Vienna, (1998). D. Anderson, T. Sample, E. Dunlop, Obtaining Module Energy Rating from Standard Laboratory Measurements, 17th European PV Solar Energy Conference, Munich, (2001). Crystalline silicon terrestrial photovoltaic (PV) modules Design qualification and type approval CEI/IEC 61215 Photovoltaic Devices - Parts 1-10 CEI/IEC 60904 Procedures for Temperature and Irradiance Corrections to Measured I-V Characteristics of Crystalline Silicon Photovoltaic Devices CEI/IEC 60891 D. Chianese, N. Cereghetti, A. Realini, S. Rezzonico, G. Travaglini, Energy Rating of PV Modules, 17th European PV Solar Energy Conference, Munich, (2001).

REFERENCES [1] D. Anderson, J. Bishop, E. Dunlop, Energy Rating of Photovoltaic Modules, 16th European PV Solar Energy Conference, Glasgow, (2000). G. Friesen, D. Chianese, S. Rezzonico, A. Realini, N. Cereghetti, and E. Bur Matrix method for energy rating calculations of PV modules, PV in Europe, Rome, (2002). R. P. Kenny, E. D. Dunlop, T. Sample, K. Reitz, D. Anderson, Energy Rating of Diverse PV Module Technologies Through Indoor and Outdoor Characterisation, PV in Europe, Rome, (2002). D.L. King, J.A. Kratochvil, W.E.Boyson, and W.I. Bower, Field Experience with a new Performance

[2]

[3]

[4]

You might also like