Stochastic Model Predictive Control for Scheduling Charging of Electric Vehicle Fleets with Market Power

Abstract

With the penetration of electric vehicles in local markets, vehicle-induced electricity demand can cause power grid instability. Collaborative smart charging can help stabilize grid demand and mitigate those issues. This paper formulates charge scheduling when connected vehicles constitute a large portion of instantaneous demand. Allowing coordinated charging to sway electricity price, we formulate a multi-objective stochastic optimization problem to minimize cost while maximizing charge in each car. We model stochastic base electricity demand using a Gaussian Mixture Model (GMM) and solve the certainty-equivalent stochastic optimization problem. We then implement a stochastic model predictive control (SMPC) algorithm and compare performance between a naive policy, a certainty-equivalent optimized policy, and SMPC on a dataset derived from California ISO-serviced demand.