Statistical analyses for capacity testing of photovoltaic systems

Data filtering and regression methods are compared in the context of PV plant capacity testing analysis, specifically for cloudy sites, with the goal of decreasing the amount of data required to produce a reliable AC output power capacity rating. Robust regression methods such as least absolute residual and bisquare are applied in contrast to the ubiquitous least squares regression method. Additionally, datasets are filtered using three different irradiance thresholds. Analysis is performed on data from individual days; cumulative datasets are built by adding daily data with the best agreement to the model.