Short-Term Power Forecasting of Wind Farm Cluster Based on Global Information Adaptive Perceptual Graph Convolution Network

At present, the installed capacity of wind power has increased significantly. Improving the power forecasting accuracy of Wind Farm Cluster (WFC) is of great significance to promote the consumption of renewable energy. How to effectively extract the complex spatio-temporal correlation between wind farms is the key to improve the power forecasting accuracy of WFC. Aiming at the above problems, this paper proposes a short-term WFC power forecasting method based on global information adaptive perception graph convolution. Compared with the traditional static graph structure, this paper designs the construction method of multiple Characteristic Graph Structures (CGSs) according to the correlation between farms. Firstly, a set of calculation methods suitable for Dynamic Correlation Coefficient (DCC) between wind farms is proposed, and according to this method, the dynamic correlation of WFC is described from multiple perspectives to obtain the graph structure at each moment, then, the graph embedding and clustering algorithm are used to obtain multiple CGSs containing key features of WFC. Finally, by establishing an Adaptive Graph Convolution Network (AGCN), the prediction accuracy is improved by graph switching. Compared with the traditional WFC forecasting method, the root mean square error of the proposed forecasting method is reduced by 1.14%–3.42%.