Optimizing Charging Time Periods of Electric Vehicles with Reinforcement Learning

The increasing integration of electric vehicles into the grid leads the focus of research to Demand Side Management and thus the control of charging processes. Mainly complex and hardly realizable coordination strategies have been investigated so far. They often involve high efforts in data measurement, collection and communication. We therefore investigate the optimal design of fixed charging time periods as a counter approach and apply them in a grid simulation. In this paper we present a Reinforcement Learning Agent to derive these charging time periods. Our case study shows that today's curtailment strategies of distribution grid operators can increase asset utilization up to 145 % while our optimized strategy can decrease it to 40 % compared to the unregulated case.