Preemptive Scheduling of EV Charging for Providing Demand Response Services

TL;DR

This paper introduces a novel preemptive scheduling algorithm for EV charging that optimizes demand response services by modeling the problem as a robust, monotone dynamic program with a simulation-based solution approach.

Contribution

It presents a new dynamic programming formulation for EV charging scheduling with preemption, including a fitted value iteration algorithm and analysis of its sample complexity.

Findings

The proposed algorithm effectively schedules EV charging under uncertain arrival distributions.

The dynamic program formulation is monotone with Lipschitz continuous value functions.

The simulation-based method provides an approximately optimal solution with quantifiable sample complexity.

Abstract

We develop a new algorithm for scheduling the charging process of a large number of electric vehicles (EVs) over a finite horizon. We assume that EVs arrive at the charging stations with different charge levels and different flexibility windows. The arrival process is assumed to have a known distribution and that the charging process of EVs can be preemptive. We pose the scheduling problem as a dynamic program with constraints. We show that the resulting formulation leads to a monotone dynamic program with Lipschitz continuous value functions that are robust against perturbation of system parameters. We propose a simulation based fitted value iteration algorithm to determine the value function approximately, and derive the sample complexity for computing the approximately optimal solution.