Electric Vehicle Optimal Charging Algorithm using Reinforcement Learning

The number of Electric vehicles is increasing significantly with the increase in environmental concerns and technological advancements, among many reasons, which will affect the power grid load management system. To manage the electric grid successfully, it is required to charge the electric vehicle optimally, especially during peak load hours. It will help enhance energy efficiency, lower costs, and promote grid integration. In this paper, the optimal charging algorithm is developed using a finite Markov Decision Process based reinforcement learning approach. Here, the aim is to formalize a model based on the Markov property for sequential decision-making problems. We have considered a practical scenario of a single charging station with three charging points to implement the algorithm. We aim to find the best charging rate for each time step to satisfy the feeder transformer constraints while fulfilling the energy requirement of consumers' EVs. The finite MDP approach is developed based on policy evaluation, policy iteration, and value function iteration.