Rural Charging Station Planning Method based on Multi-scale Graph Convolution and Bidirectional Long Short-term Memory Network

To develop a charging planning method tailored for rural areas and capable of addressing the charging needs of rural electric vehicles, this paper introduces a two-stage planning approach based on multi-scale graph convolution and a bidirectional long short-term memory network. Firstly, the graph-structured characterization modeling method is introduced to model the features of the coupled “vehicle-station-network” network in the first stage. The spatial topology and node association information are extracted using a multi-scale graph convolutional network, while the temporal characteristics of charging loads are analyzed with a bidirectional long short-term memory network, enabling precise predictions of rural charging demand. Then, a planning model of a rural EV charging station is constructed in the second stage. The Particle Swarm Optimization algorithm is applied to identify the optimal locations for charging stations and the configuration of charging piles. Finally, case studies are conducted based on the IEEE-123 node rural distribution network topology. The testing results show that this method can effectively reduce user travel cost and charging expenses.