Optimal planning for electric vehicle fast charging stations placements in a city scale using an advantage actor-critic deep reinforcement learning and geospatial analysis

Abstract

The transition to Electric Vehicles (EVs) for reducing urban greenhouse gas emissions is hindered by the lack of public charging infrastructure, particularly fast-charging stations. Given that electric vehicle fast charging stations (EVFCS) can burden the electricity grid, it is crucial for EVFCS to adopt sustainable energy supply methods while accommodating the growing demands of EVs. Despite recent research efforts to optimize the placement of renewable-powered EV charging stations, current planning methods face challenges when applied to a complex city scale and integrating with renewable energy resources. This study thus introduces a robust decision-making model for optimal EVFCS placement planning integrated with solar power supply in a large and complex urban environment (e.g., Chicago), utilizing an advantage actor-critic (A2C) deep reinforcement learning (DRL) approach. The model balances traffic demand with energy supply, strategically placing charging stations in areas with high traffic density and solar potential. As a result, the model is used to optimally place 1,000 charging stations with a random starting search approach, achieving total reward values of 74.30 %, and estimated the capacities of potential EVFCS. This study can inform the identification of suitable locations to advance the microgrid-based charging infrastructure systems in large urban environments.