Abstract: The unguided fast charging of the large-scale electric vehicles has brought great challenges to the coupled power-transportation networks. In this paper, a multi-objective optimization model including the electric vehicles, the charging stations, the transportation network, and the power grid is firstly established. Then, a request-driven recommendation model is proposed, and the irregular environment information extraction and fast charging guidance strategy learning are realized based on the graph reinforcement learning algorithm. Finally, some experiments are carried out on MATLAB-SUMO-Python joint simulation platform. The results show that the proposed algorithm can effectively reduce the time cost before charging of EVs and improve the service balance of charging stations, which ensures the optimization of transportation network and power grid and the long-term healthy and stable operation of the coupled networks. The proposed method has good optimization effect and real-time response capability.