eeh power systemslaboratory

The integration of a high number of plug-in electric (PEV) vehicles could lead to overloads in the systems assets and demand peaks if the charge of the fleet is left uncontrolled. However, with the use of smart-charging strategies these problems could be avoided. In this work the development of a smartcharging strategy is presented. The goal of each electric vehicle, modeled as an agent, is to minimize the cost of energy purchase while satisfying the energy requirements. To solve this problem, multi-agent system theory is used in combination with market-based control. The vehicles are considered as agents bidding on the market, optimizing their bidding to minimize their costs. An aggregator agentacts as communication middleman between the vehicles and the market. This way, a system with a high number of agents competing for the resources is established. The resources are allocated according to the demand-supply theory, and the equilibrium price of the day-ahead market is used as a control signal. Moreover, a Q-learning algorithm is used for the learning process of the vehicles, establishing their optimal bidding strategy. In our case studies, we analyze this approach both in a simple market clearing and an Optimal Flow setting. Moreover, we analyze the effect of uncertainties in driving patterns and non-PEV bids. The results show that the use of this strategy leads to a lower energy costs for the vehicles. The fleet charges mainly during the night hours, avoiding the charge during demand peaks.