Location-allocation of capacitated electric vehicles charging stations on directed highway networks

Abstract

According to the electric vehicle (EV) market research report, the global EV market is expected to increase at a compound annual growth rate of 26.8% from 4 million units in 2021 to 35 million units by 2030. One of the reasons is that reducing greenhouse gas (GHG) emissions in the transportation sector is one important step in combating global warming in the United States. Furthermore, recently $5 billion have been allocated to states to fund EV chargers over five years along interstate highways as part of the bipartisan infrastructure package. On September 2022, the U.S. Department of Transportation announced that it approved EV charging station (EVCS) plans for highway systems in all 50 states, Washington, D.C., and Puerto Rico. Accordingly, with the rapid growth of the EV market in recent years, the location-allocation models for EVCSs have received increasing attention to effectively provide charging services to EV users. In this study, we aim to solve the location-allocation problem of capacitated EVCSs on directed highway networks considering path-based demands and limited EV driving ranges. On the one hand, since the proposed model is path-based, EV traffic flow rates for a round trip between origin-destination pairs in a given highway network system are regarded as demands. Furthermore, it is assumed that EVs can be driven within the maximum driving range with the full-charged battery, and the driving range is assumed to decrease linearly over time with the remaining battery energy. On the other hand, many capacitated flow refueling location models have assumed that the number of chargers installed at each station is deterministic and predetermined. In this study, we relax the impractical assumption, i.e., the flexible number of chargers per EVCS is allowed. Accordingly, the selection of possible EVCS combinations and the charging capacity of EVCSs are substantially addressed in our model. A modified Bayesian Optimization (BO) algorithm is used as the solution method to resolve high computational effort to find the optimal EVCS combination. The expected improvement acquisition function is exploited to maximize the total traffic flow covered by the selected EVCSs and simultaneously minimize the epistemic uncertainty in the surrogate model. To validate the applicability of the proposed model, various experiments are modeled considering the Pennsylvania Turnpike and South Dakota interstate highway networks. Sensitivity analysis with respect to the installation budget, the driving ranges of EVs, and the battery energy remaining level are conducted. Overall, the results demonstrate that, in both cases of Pennsylvania Turnpike and South Dakota interstate highway networks, our solution approach adopting the modified BO to allow flexible number of chargers per EVCS always finds better solutions than the base solution approach with fixed number of chargers per EVCS in terms of both installation costs and effective traffic coverage covered by the opened set of stations. Especially in the South Dakota case, our solution approach provided charging services with significantly more traffic coverage while spending less installation cost than the base solution approach because our solution approach can distribute flexible number of chargers through the highway network.