A Progressive Planning Scheme for Dynamic Expansion of Electric Vehicle Charging Resources

The mass adoption of electric vehicles (EVs) continuously requires the expansion of charging resources in the urban area. Such a continuing and dynamic problem associated with multi-level of uncertainties can hardly be tackled by deterministic models. To overcome this challenge, this paper newly proposes a progressive planning scheme (PPS) for dynamic expansion of charging resources deployment. Specifically, a simulation environment is constructed by considering EV user behavior model and evaluation model. Based on this, a tailored deep reinforcement learning (DRL) model is proposed to direct the concerned expansion problem, of which the objective is to enhance social welfare while adhering to specific economic and technical constraints. The proposed model leverages an attention mechanism and an inflation coefficient to enable the agent to selectively focus on dominant features and effectively adapt to variable planning schemes. Moreover, the DRL model is incorporated with Monte Carlo Tree Search (MCTS) to obtain the exploration-exploitation trade-off, through which more valid and generalizable decisions can be generated. Comparative case studies are carried out to verify the efficiency and robustness of the proposed methods, which demonstrate averaging 4.79% and 11.38% improvement over the advanced reinforcement learning algorithm and the genetic algorithm in an approximating real-world environment, respectively.