Graph RL-Based Route Planning Method for EVs with Synchronous Charge Station Selection

Electric vehicles (EV s) represent a crucial solution for energy conservation and emission reduction in response to global warming. However, widespread usage of EVs presents challenges to the optimization of both route planning and charging planning, due to the considerations of road network, distribution of charging stations, and vehicle interactions. This study proposes a synchronous method of route and charging plan-ning which employs graph neural networks and reinforcement learning. City maps are represented by graph structure data, and topology information of the road network is extracted by the Structure2Vector method, which is embedded into the state value function of the Deep Q Network (DQN) algorithm, so as to plan the route and optimize the selection of charging stations. The proposed method can ensure the timeliness of charging while reducing the traveling distance, and demonstrate good adaptability in different scales of map instances.