Charging Load Forecasting of Highway Infrastructure Considering Spatiotemporal Coupling

The deep electrification of the transportation system has fostered a tight coupling between the transportation and energy systems, leading to a trend of profound integration between the transportation network and the energy network. This study concentrates on the ‘energy-highway’ integrated system. Firstly, we conduct a thorough study and analysis of the load characteristics of service areas and toll stations. Through the load data of the expressway service area, the different load characteristics on weekdays and holidays are obtained. By examining the charging behavior of electric vehicles, this paper derives the in-transit charging demand for electric vehicles. Finally, the distribution characteristics of load are obtained. Based on the data from each charging station, we propose a load forecasting method for highway infrastructure utilizing an adaptive graph neural network, and the highway infrastructure is regarded as the node of the graph network. Analyze the correlation degree of each dimension data in the service area, and form the feature vector of the node with the data with large correlation degree. A single-step prediction model of charging load based on graph neural network is established. The graph neural network is used to capture the spatial topological relationships between service areas, while the gated recurrent unit tracks the variations in charging load over time. The prediction performance of the proposed model is ultimately validated using real data from an expressway service area.