Electric Vehicle Charging and Discharging Based on Multi-Agents Reinforcement Learning

With the increase in the number of electric vehicles, the connection of EVs greatly increases the load on the power grid. By controlling the charging and discharging of EVs in a reasonable way, EVs can play a key role in the regulation of the power grid. Taking the charging and discharging control algorithm of EVs as the research object, we propose a deep deterministic policy gradient algorithm based on multiple intelligences to control the continuous charging and discharging behaviors of EVs. Aiming at the dynamic change of load timing in the distribution network, the charging and discharging power of different vehicles under different tariffs is controlled so that the users can get the maximum benefit. Considering the impact of EV charging on the distribution network, the optimal current is introduced to quantify the load caused by charging behavior on the network. Multiple EV systems in different environments are constructed as Markov decisions, and a multi-intelligence reinforcement learning algorithm is used to enable the vehicles to learn personalized charging and discharging strategies. The simulation results show that the algorithm can effectively regulate the charging and discharging behaviors of different vehicles and play the role of regulating the power grid, and the load caused by the algorithm on the power grid is more stable when the vehicles are charging.