An Effective Model Based on Stmixing-LSTM for Short Term Wind Power Prediction

The precision of wind energy forecasting plays a vital role in maintaining grid security and optimizing reserve capacity distribution. The variability and intermittent characteristics of wind resources pose significant challenges to conventional power prediction methods that rely on historical pattern analysis. Therefore, this paper proposed a hybrid model named STmixing-LSTM for wind power prediction. Firstly, the initial input data undergoes a season-trend (ST) decomposition. The decomposed data is further mixed. Then the series are predicted by the decoder, which is consisted of a CNN layer and a LSTM layer. Case studies of forecasting using one wind farm dataset in China are conducted. The experiment result shows that our novel model proposed in this study can predict short-term wind power more accurately.