Planning Fully Renewable Powered Charging Stations on Highways: A Data-Driven Robust Optimization Approach

This paper proposes a comprehensive two-stage method for siting and sizing stand-alone electric-vehicle charging stations on highway networks. In the first stage, locations, where individual vehicles require charging services, are obtained from Monte Carlo simulation provided with the traffic demand and battery data; an integer programming model is proposed to determine the optimal sites of charging stations from potential candidates, ensuring that every vehicle is able to visit at least one charging station without depleting the battery; afterward, the spatial and temporal distribution of charging demand at individual selected sites can be simulated. In the second stage, a datadriven distributionally robust optimization model is developed to optimize the capacities of renewable generations and energy storage units in each charging station. The uncertain generation and demand are described by a family of inexact distributions around an empirical distribution, and their distance in the sense of Kullback-Leibler divergence is controlled by an adjustable scalar. Two reformulations of the robust model are suggested based on risk theory. The first one relies on Value-at-Risk (VaR) and gives rise to a mixed-integer linear program (MILP), which is more accurate; the second one offers a conservative approximation based on Conditional VaR and comes down to a linear program, which is more tractable. Numerical study on a test system demonstrates the effectiveness of the proposed methods.