A Large Neighborhood Search for Battery Swapping Station Location Planning for Electric Scooters

Abstract

We consider the Multi Objective Battery Swapping Station Location Problem (MOBSSLP) for planning the setup of new stations for exchanging depleted batteries of electric scooters with the aim of minimizing a three-part objective function while satisfying an expected amount of demand. Batteries returned at a station are charged and provided to customers again once they are full. We present a large neighborhood search (LNS) for solving MOBSSLP instances. The LNS makes use of a mixed integer linear program (MILP) to quickly find good solutions within a specified neighborhood. Multiple neighborhood structures given by pairs of destroy and repair operators are suggested. The proposed LNS is evaluated on instances generated by adapted approaches from the literature with up to 500 potential station locations and up to 1000 user trips. Solutions obtained from the LNS have on average ten to thirty percent better objective values on these instances than a state-of-the-art MILP solver.