Day-ahead electricity price prediction in multi-price zones based on multi-view fusion spatio-temporal graph neural network

Factors such as high penetration of renewable energy, load, geographic location, and interactions between price zones make accurate electricity price forecasting (EPF) very challenging, especially day-ahead electricity price forecasting (DAEPF). To address the issue, A spatio-temporal graph neural network prediction model based on multi-view fusion is proposed in this paper, which learns and analyzes distance relationships, price correlations, and similarities in price distributions across multiple regions, four kinds of graph matrix are constructed to represent the complex spatio-temporal interaction in electricity market. To realize information aggregation between multiple perspectives, a novel multi-view fusion module (MVF) is proposed, which actively mines and utilizes the correlation between nodes within the graph and nodes across the graph through spatial attention and graph attention mechanism, and a temporal embedding module is proposed. The temporal information between nodes is represented by multi-head temporal attention mechanism and the time dependence of multiple receptive fields is obtained by multi-scale gated convolution. Massive experiments are conducted on multiple price zones in the European power market with a high proportion of new energy sources. The results show that MVF can effectively integrate multiple scenario information and improve the prediction accuracy of the network, and the proposed combined network has significant advantages over other models involved in this study.