A Reinforcement Learning-based Assignment Scheme for EVs to Charging Stations

Due to recent developments in electric mobility, public charging infrastructure will be essential for modern transportation systems. As the number of electric vehicles (EVs) increases, the public charging infrastructure needs to adopt efficient charging practices. A key challenge is the assignment of EVs to charging stations (CSs) in an energy efficient manner. In this paper, a Reinforcement Learning (RL)-based EV Assignment Scheme (RL-EVAS) is proposed to solve the problem of assigning EV to the optimal CS in urban environments, aiming at minimizing the total cost of charging EVs and reducing the overload on Electrical Grids (EGs). Travelling cost that is resulted from the movement of EV to CS, and the charging cost at CS are considered. Moreover, the EV's Battery State of Charge (SoC) is taken into account in the proposed scheme. The proposed RL-EVAS approach will approximate the solution by finding an optimal policy function in the sense of maximizing the expected value of the total reward over all successive steps using Q-learning algorithm, based on the Temporal Difference (TD) learning and Bellman expectation equation. Finally, the numerous simulation results illustrate that the proposed scheme can significantly reduce the total energy cost of EVs compared to various case studies and greedy algorithm, and also demonstrate its behavioural adaptation to any environmental conditions.