Multi-step short-term wind power prediction based on spatio-temporal graph convolutional networks

Wind power generation has been developed rapidly in recent years. However, the inherent volatility and intermittency of wind power bring challenges to the stability of the power system, so it is of great significance to improve the accuracy of wind power prediction. In this paper, a wind power prediction model based on spatio-temporal graph convolutional network is proposed. Here, turbines are considered as nodes of a graph with a learnable adjacency matrix that determines the strength of relationships between turbines based on historical wind speed and wind power data, and a self-loop connectivity matrix is added to the adjacency matrix to increase the self-connectivity of each node. In this way, we effectively integrate the spatial dependencies and temporal trends of neighboring turbines. Experiments conducted on a real dataset collected from onshore wind farms in the U.S. show that our proposed model outperforms a previously developed baseline model on the reference dataset.