Short-term wind speed prediction based on deep learning and intelligent optimization algorithm

Abstract

In order to improve the accuracy of wind speed prediction, a wind speed prediction model combining complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), long short-term memory (LSTM) and gray wolf optimization (GWO) algorithm was proposed from the perspective of reducing wind speed nonstationarity and optimizing combination weight. First, CEEMDAN was used to decompose the observed wind speed into a series of sub-sequences reflecting the characteristics of the original wind speed. Then the subsequence is predicted by LSTM, and the predicted value of the subsequence is output. Finally, the combined weight of the sub-sequences was optimized by GWO, and the sub-sequences were combined to obtain the wind speed prediction results. The experimental results show that CEEMDAN-LSTM-GWO wind speed prediction model proposed in this study has better performance than the comparison model.