Data-Filtering-Dependent Variability of Long-Term Degradation Rates of MW-Scale Photovoltaic Power Plants from 'Non-Ideal' Monitoring and Weather Data

35th European Photovoltaic Solar Energy Conference and Exhibition; 2069-2074
In contrast to research photovoltaic (PV) systems, commercial PV systems are usually not equipped with sophisticated weather metrology and quite frequently even lack an irradiance sensor. Still, the determination of their degradation rate is of paramount importance for the plant owners. This paper compares various approaches to determine the degradation rates of commercial MW-scale PV systems from such “non-ideal” data sets as they are frequently obtained from monitoring data. In particular, the use of model-based irradiance and temperature data is evaluated and several novel methods for data filtering are introduced. It is investigated, for which weather conditions the use of externally weather station data is superior to on-site data. By reducing the data set to mostly sunny conditions, reliable DRs can be determined. Furthermore, the data scattering can be reduced significantly so that sudden performance drops or changes in the DRs become visible, which were previously masked by scattering and data drifts. These results demonstrate that significantly different DRs and uncertainties are obtained depending on the method of data filtering. Hence, care must be taken, when deriving DRs from “non-ideal” monitoring and weather data and when comparing DRs determined by different methods.