A Decentralized MPC Strategy for Charging Large Populations of Plug-In Electric Vehicles

This paper presents a framework for decentralized coordination of plug-in electric vehicle (PEV) charging patterns in scenarios where the future cannot be perfectly predicted. We begin with the mathematical formulation of the decentralized problem, in which individual PEVs minimize their own charging costs, which are a function of total system demand. We summarize results from our prior work in this area, relating specifically to convergence and uniqueness of “valley filling” charging strategies in situations where future system states are known with perfect accuracy. We then present an approach to manage forecast uncertainty by allowing decentralized agents to continually update their optimal control trajectories subject to revised forecasts of system states. We show that the resulting trajectories are strongly influenced by the accuracy of the forecast over the charging period.