Coordinated and Generalizable Planning and Operation of PV-Storage-Charging Facilities in Coupled Power and Transportation Nexus

Transportation electrification, driven by large-scale electric vehicle adoption, is pivotal to global decarbonization. However, rapid charging infrastructure deployment entails high urban investment costs, risks of local grid overloads, and challenges to renewable integration due to temporal mismatches between charging demand and generation. Photovoltaic– storage–charging (PSC) stations provide a promising solution, yet coordinated planning of coupled transportation and power distribution networks (CPTN) remains underdeveloped. Existing models rarely integrate charging price signals, and analytical or heuristic approaches for hierarchical optimization under multi-scale uncertainties and nonlinear dynamics remain computationally prohibitive. This paper proposes a hierarchical multi-agent reinforcement learning framework for coordinated PSC planning and operation in CPTN. A multi-time-scale nested optimization model is formulated as a Markov Game with options, enabling collaborative learning of long-term investment and short-term pricing and dispatching policies. Spatial–temporal graph convolutional networks capture dynamic nodal and edge features, while a proximal policy optimization scheme with advantage decomposition and sequential updates enhances perception and convergence. Case studies based on real-world settings demonstrate superior performance against state-of-the-art benchmarks in terms of policy quality, computational and convergence efficiency, while revealing significant cost-efficiency improvements and operational benefits for CPTN.