FDCA-DSTGCN：一种基于频域信息增益与动态趋势感知的风电场群功率日前预测模型

Accurate wind farm cluster power prediction (WFCPP) is of vital significance for new power systems with large-scale wind power integration. The current WFCPP modeling method ignores the important role of wind direction and frequency domain information, resulting in insufficient use of spatial information and difficult to improve the prediction accuracy. In order to solve the above problems, a day-ahead power prediction model for wind farm cluster power is proposed based on information gain in frequency domain and dynamic trend sensing. Firstly, a cluster division method based on graph theory and progressive fusion of multiple information is proposed, and a virtual information node is set up. Secondly, a method to identify prevailing wind direction under time window is proposed for subsets, and a dynamic weighted directed graph structure is designed based on prevailing wind direction and wind speed. Thirdly, a dynamic spatio-temporal graph convolutional network model with frequency domain gain channel attention mechanism (FDCA-DSTGCN) is proposed to integrate the above information and complete the prediction. Finally, the proposed method is applied to a wind farm cluster in Inner Mongolia of China. Compared with the existing and traditional prediction methods, the normalized root mean square error of the proposed method is reduced by 1.13%–2.15%, which verifies the effectiveness of the proposed method.