Comparison of Photovoltaic Generation Uncertainty Models for Power System Planning Using Regression Framework

Probabilistic photovoltaic (PV) generation forecasting is essential for the planning of a PV integrated power system. Compared to deterministic forecasting, probabilistic PV generation forecast results are more realistic in making critical decisions during the planning stage of power systems. Since its inspection, regression frameworks have gained enormous credit in the prediction literature owing to their decent prediction performance. The selection of a reduced but sensible theoretical relevant regressor set is an essential task for a befitting model. This paper elucidates various regression-based PV generation uncertainty models, mainly compares the forecasting result of two quantile regression models suitable for planning applications. Each model’s merits and demerits in the power system planning application are deeply analyzed. Historically collected PV generation time-series data from three different places in the USA are used for forecast performance comparison through suitable metrics. The suitability of a model out of the considered two with various choices of regressor(s) is critically analyzed through detailed result analysis.