4. 1 PV Module Design, Manufacture, Performance and Reliability
Energy Rating, PV Module, Energy Performance
Summary / Abstract:
The aim of this work was to re-evaluate the Climate Specific Energy Rating (CSER) methodology with regard to modelling the temperature of PV modules as defined in the IEC 61853 series of standards "Photovoltaic (PV) module performance testing and energy rating". The temperature behaviour of PV modules is described by a simple model in which the effects of light absorption and wind are considered by two modelling parameters u0 and u1. These are derived from a series of measurements under natural sunlight, from which only data points that meet certain stability criteria for solar irradiance and wind speed are used for linear regression (Method A). Our work addresses the shortcomings of the IEC data filter reported by test laboratories in the past: a) small number of filtered data points, b) filtered data points are not well distributed in the wind speed range. Both can lead to considerable uncertainty in the modelling parameters. We investigated an alternative data filter that provides a higher number of usable data points and improves the robustness of the regression by wind speed binning (Method 2). Furthermore, we introduce two indicators for the quality of the regression: a) the mean bias error with respect to the differences between measured and modelled temperatures and b) the irradiance-weighted module temperatures. Based on these indicators, a third method for determining u0 and u1 was investigated that does not require data filtering (Method 3) and focuses only on the best fit of modelled and measured module temperatures. For a measurement series from New Mexico with mostly clear skies, the application of the three methods resulted in a large scatter of the u0 and u1 parameters, but comparable NMOT values. Furthermore, the evaluation of a measurement series from Germany showed a scattering of the modelling parameters due to the measurement season mainly for Methods A and B. Our results confirm clear advantages for Method C. It is easy to apply, the data collection period can be kept short and the parameters u0 and u1 are less sensitive to seasonal variations.