An Ultra-short-Term Wind Speed Prediction Method Based on Spatio-Temporal Feature Decomposition and Multi Feature Fusion Network.

Wind energy plays an important role in alleviating global warming. To improve the efficiency of wind energy utilization, an ultra-short-term wind speed forecasting method based on spatio-temporal feature decomposition and multi feature fusion network is proposed. The method is divided into two stages: data preprocessing and prediction. The data preprocessing stage includes the construction and decomposition of the spatio-temporal feature, which can reduce the impact of wind speed fluctuations on the model while ensuring the integrity of spatio-temporal features. In the prediction stage, a parallel prediction network consisting of multiple multi-feature fusion networks (MFFNets) is proposed. Considering the high semantic and high information density characteristics of spatio-temporal features, MFFNet integrates shallow, intermediate, and deep features to combine local detail information with global feature information, which reduces the impact of local wind speed fluctuations on the accuracy of predictions. The proposed method was validated on a wind farm located in the Midwestern United States. Compared with current advanced methods, MFFNet achieved improvements of more than 6.8% in all indicators. The results demonstrate that the proposed method has promising applications in large-scale wind farm forecasting.