Graph Reinforcement Learning Method for Electric Vehicle Charging Recommendation Considering Voltage Deviation

In recent years, the quiet and environmentally friendly characteristics of electric vehicles (EVs) have made them increasingly popular, driving the construction of charging infrastructure. However, the large amount of charging access and the promotion of fast charging technology have led to an increasing total power demand for charging. If these demands arise simultaneously, it will cause excessive voltage deviation issues in the power distribution network. The charging recommendation method proposed in this paper distributes EVs to different charging stations, keeping the voltage deviation at the power distribution network nodes within the allowable range. We employ a graph reinforcement learning algorithm and demonstrate that it yields more stable training results compared to traditional reinforcement learning algorithms. The simulation results show that our algorithm can maintain the voltage deviation within the limits and improve the level of obtaining charging services.