Direct Interval Forecasting of Wind Power

Abstract

This letter proposes a novel approach to directly formulate the prediction intervals of wind power generation based on extreme learning machine and particle swarm optimization, where prediction intervals are generated through direct optimization of both the coverage probability and sharpness, without the prior knowledge of forecasting errors. The proposed approach has been proved to be highly efficient and reliable through preliminary case studies using real-world wind farm data, indicating a high potential of practical application.