Optimal Placement of Public Electric Vehicle Charging Stations Using Deep Reinforcement Learning

As climate change becomes a pressing issue, encouraging the use of electric vehicles (EVs) has become a popular solution to address the pollution associated with fossil fuel automobiles. Placing charging stations in areas with developing charging infrastructure is a critical component of the accessibility and future success of electric vehicles. In this work, we implement a supervised learning-based demand prediction model along with Deep Q-Network (DQN) reinforcement learning (RL) model to select optimal locations for charging stations. We use a Gradient Boosting model to predict EV charging demand at existing charging stations based on traffic data and points of interest (POIs) for select counties in the state of New York. The model outperforms other supervised learning methods with an R score of 0.87, with the main predictors of demand being the traffic density near a station and the number of EV registrations