A Short-term Wind Power Forecasting Method Based on NWP Wind Speed Fluctuation Division and Clustering

Abstract

High-precision wind power prediction is an indispensable tool in the process of wind power integration operation. In order to improve the accuracy of wind power forecasting, this paper proposes a combined forecasting method based on NWP wind speed fluctuation division, Fuzzy C-means clustering (FCM) and Deep Confidence Network (DBN) for forecasting short-term wind power generation. Firstly, the Savitzky-Golay (SG) filter is used to filter the NWP wind speed sequence to obtain the wind speed fluctuation trend sequence. Then, according to the extreme value points of the wind speed fluctuation trend series, the NWP wind speed series is divided into multiple wind speed waves, and the characteristic parameters of the waves are extracted. In addition, wave-based feature parameters utilize FCM to divide waves into multiple classes. Finally, different DBN models are constructed for wind power forecasting according to different wave classes. The results show that the proposed combined method has better performance than the benchmark forecasting method.