Short-Term Wind Power Prediction Based on Wind Speed Interval Division and TimeGAN for Gale Weather

Abstract

Accurate short-term wind power prediction(WPP) is important for the grid dispatch of the power system. The accuracy of WPP is closely related to weather conditions, and gale weather has a great impact on WPP. To improve the accuracy of WPP in gale weather, a short-term WPP method based on wind speed interval division and TimeGAN for gale weather is proposed in this paper. First, the original data is divided into wind intervals according to Beaufort Wind Scale. Then, TimeGAN-based sample data augmentation is carried out for gale intervals with few data, and the expanded sample set is used as the training set. BPNN and LSTM neural network are proposed as WPP model for every interval. Finally, divide the input data according to the intervals to make prediction, and combine the outputs to obtain complete prediction result. Based on the proposed method, the 96h-short-term prediction RMSE and MAE can be decreased by 2.57% and 2.56%, which demonstrates that the proposed method is an effective WPP method that improves the prediction accuracy for gale weather.