Electric vehicles charging management using deep reinforcement learning considering vehicle-to-grid operation and battery degradation
Mostafa M. Shibl, Loay Ismail, Ahmed Massoud
Abstract
EVs are becoming more popular and widely used worldwide due to their environmentally friendliness as part of the world efforts to decrease the effects of climate change. Moreover, more users are buying EVs due to governmental incentives, development of charging technologies and cheaper maintenance costs. Thus, the increased electrical loads on the distribution grid caused by the charging of EVs can have negative impacts such as high voltage fluctuations, power losses and power overloads. Thus, a power system management solution is required to protect the distribution grid from the harmful effects of EVs charging through the regulation of the charging of EVs. In this paper, a deep RL-based EVs charging management solution is presented, while considering fast charging, conventional charging and V2G operation, in order to satisfy the requirements of the user and the utility. Deep RL is utilized to model the EV chargers and the EV users. The EV chargers are considered the RL environment and the EV users are considered the RL agent. Finally, the system was tested with a range of case studies using real-life EVs charging data, which proved the effectiveness and reliability of the system to protect the distribution grid and satisfy the EV user’s charging requirements.