Abstract: Aiming at the inherent volatility of wind power generation in power system and the problem of high frequency noise in power prediction. A short-term wind power prediction model based on VMD-WOA-ATLSTM is proposed. The Pearson correlation coefficient method is used in the model to obtain highly relevant data and determine the key impact characteristics of wind power. The variational mode decomposition (VMD) is used to decompose and reconstruct the data to highlight the main features related to power and reduce noise interference. Further the ATLSTM model combining long-term and short-term memory neural network and attention mechanism is built. The whale optimization algorithm (WOA) is used to globally optimize the parameters of ATLSTM network. Finally, the prediction results of each modal component are superimposed to improve the overall prediction performance of the model. The experimental results show that VMD-WOA-ATLSTM model has better prediction accuracy and robustness than LSTM, LSTM-Attention, VMD-ATLSTM and WOA-ATLSTM models in power prediction.
辛鹏, 郭玉强, 李键, 李超然, 刘培瑞, 杨建. 基于VMD-WOA-ATLSTM的短期风电功率预测[J]. 吉林化工学院学报, 2025, 42(5): 59-66.
XIN Peng, GUO Yu-qiang, LI Chao-ran, LI Pei-rui, YANG Jian. Prediction of Short-Term Wind Power Based on VMD-WOA-ATLSTM. Journal of Jilin Institute of Chemical Technology, 2025, 42(5): 59-66.
