Short-term wind power prediction based on ICEEMDAN-Correlation reconstruction and BWO-BiLSTM

To solve the problems of high volatility and low prediction accuracy of wind farm output power, this paper proposes a short-term wind power prediction model with improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), dispersive entropy combined with zero crossing rate (DE-ZCR) correlation reconstruction and beluga whale optimization (BWO) to optimize bidirectional long short-term memory (BiLSTM) neural network. Firstly, the original wind power is decomposed into multiple modal components by ICEEMDAN; secondly, the DE-ZCR method is used to evaluate the complexity and correlation of each component, and each modal component is reconstructed into a high frequency oscillation component, a medium frequency regular component, and a low frequency stable component; then the BWO-BiLSTM is used to predict each reconstructed power component separately, and finally the predicted values are superimposed to obtain the final results. The prediction model constructed in this paper is compared with four other models under different wind seasons, the results show that the model of this paper is superior to other models, validating the effectiveness of the combined prediction model.