Combined Approach for Short-Term Wind Power Forecasting Based on Wave Division and Seq2Seq Model Using Deep Learning

Abstract

The accuracy of short-term wind power forecasting (WPF) can be improved by effective mining of numerical weather prediction data. In this article, a novel short-term WPF approach is proposed by combining wave division (WD), improved grey wolf optimizer based on fuzzy C-means clusters (IGFCM), and Seq2Seq model with attention mechanism based on long short-term memory model (LSTMS), named the WD-IGFCM-LSTMS model. Based on the fluctuation trend, the wind speed sequences of NWP are divided into a series of waves. Six fluctuation features that reflect the shape characteristics are extracted to quantify the partitioned waves. A new strategy is proposed to improve the global searching ability of the GWO to select the initial clustering center of FCM more effectively. The Seq2Seq deep learning model based on LSTM, named LSTMS, is applied for wave-oriented forecasting. The proposed approach outperforms the traditional point-to-point forecasting and realizes continuous sequence forecasting. The simulation results demonstrate that the WD-IGFCM-LSTMS model can perform better than other benchmark forecasting models.