Scheduling of Plug-in Electric Vehicle Battery Charging with Price Prediction

This paper proposes a reinforcement learning algorithm that solves the problem of scheduling the charging of a plug-in electric vehicle's (PEV) battery. The algorithm is employed in the demand side management of smart grids. The goal of the algorithm is to minimize the charging cost of the consumer over long term time horizon. The PEV battery charging problem is modeled as a Markov decision process (MDP) with unknown transition probabilities. A Sarsa reinforcement learning method with eligibility traces is proposed for learning the pricing patterns and solving the charging problem. The model uses true day-ahead prices for the current day and predicted prices for the next day. Simulation results using true pricing data demonstrate the cost savings to the consumer.