Optimal EV Charging Scheduling Considering the Lack of Charging Facilities Based on Deep Reinforcement Learning

• Published in: 2023 8th Asia Conference on Power and Electrical Engineering (ACPEE) pp. 1825 - 1829
• Main Authors: Li, Hang, Li, Guojie, Li, Shidan, Han, Bei, Wang, Keyou, Xu, Jin
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.04.2023
• Subjects: Costs; Deep Deterministic Policy Gradient (DDPG); Deep learning; dynamic energy boundary (DEB); Dynamic scheduling; Electric vehicle charging; electric vehicles (EVs) charging scheduling; Markov Decision Process (MDP); Numerical simulation; Reinforcement learning; Simulation
• DOI: 10.1109/ACPEE56931.2023.10135734

In order to tackle the problem of intelligent charging and discharging management of electric vehicles (EVs) in the situation of lack of charging facilities, a method based on deep reinforcement learning is proposed. First, according charging demands of EVs both plugging in a certain charging pile and in the subsequent queuing, a dynamic energy boundary (DEB) is proposed to adjust the charging and discharging power boundary for the charging EV. Then, the charging scheduling problem is converted to a Markov Decision Process (MDP), and the reward function is designed to minimize the charging cost, charging time and maximize the satisfaction for users' charging demands. Finally, deep deterministic policy gradient (DDPG) algorithm is used to solve the MDP of continuous charging states and actions. The numerical simulation results show that the proposed method can effectively deal with the charging scheduling problem for EVs considering the lack of charging facilities. According to the real-time situation of EVs queuing, the proposed method can help charging piles adjust the charging process locally, so as to satisfy the charging demand of users as soon as possible and effectively reduce the charging cost.