Collaborative Mobile Charging Vehicles Placement: A Reinforcement Learning Approach

Nowadays, there are increasing demand from electric vehicles that need to be charged during peak hours in a city. The number of traditional fixed charging stations is limited, which may not be sufficient to satisfy the charging demand of electric vehicle users. Mobile charging vehicles can be scheduled at fixed positions to serve as a fixed charging station, providing charging services for electric vehicle users, for relieving the pressure of mass charging demand. In this paper, the charging scheduling problem of mobile charging vehicle is modeled as a Markov Decision Process(MDP) problem, and a reinforcement learning method based on policy evaluation is adopted to solve this problem. The decision of the specific position to which the mobile charging vehicle is scheduled is determined by this method. In addition, the decision of how to dispatch electric vehicle charging orders is determined by a fixed algorithm. We propose two different order dispatching algorithms. According to the simulation, our proposed reinforcement learning method has superior results in terms of the efficiency of order processing and user satisfaction compared with the traditional greedy method. The two different electric vehicle charging order dispatching algorithms have their own advantages and disadvantages in terms of processing speed and effectiveness.