A new approach for interval forecasting of photovoltaic power based on generalized weather classification

Summary Photovoltaic (PV) power forecasting is of great significance to the grid connection and safe operation of PV plants. Problems such as complex weather conditions, numerous weather types, and limited weather classification methods make such forecasting a highly challenging endeavor. The point forecasting model is limited to apply due to the lack of error information. To solve above problems, a novel interval forecasting method based on generalized weather conditions is proposed. The uncertainty of PV power under different weather conditions is first analyzed, then a generalized weather classification method based on solar irradiance reduction index K is performed. Next, a PV power forecasting multi‐model is established based on the extreme learning machine under different generalized weather types. The confidence interval of forecasted PV power is determined by kernel density estimation. Comparative experiments demonstrate the effectiveness of the proposed method in terms of training time, model performance, and interval accuracy.