Short-term wind speed prediction based on CEEMD decomposition and optimization BP model

Abstract

As one of the important factors of wind power generation, wind speed directly affects the stability and safety of wind power system. Wind speed has the characteristics of non-linear and nonstationary series, so how to predict wind speed accurately is an outstanding problem. In this paper, a method based on CEEMD and GA-BP is used to forecast the wind speed of wind farm in short term. According to the characteristics that wind speed is a nonlinear and non-stationary sequence, firstly, the wind speed sequence is decomposed by CEEMD to obtain a subsequence (IMF) which is more stable than the original sequence. Then, each subsequence of the original sequence is modeled and predicted one by one by using BP model. Finally, the prediction results of all subsequences are added to obtain the final wind speed prediction value. Because the number of hidden nodes and learning rate of BP model will affect the modeling efficiency and prediction accuracy of BP model. GA algorithm is used to optimize these two parameters, and the best parameters are iterated according to the minimum MSE to improve the prediction accuracy. The simulation results show that, compared with the short-term wind speed prediction results directly using BP model and CEEMD-BP model, the method adopted in this paper has higher prediction accuracy, which is conducive to the safe and stable operation of power system.