A Two-Stage Genetic Algorithm for Battery Sizing and Route Optimization of Medium-Duty Electric Delivery Fleets

Abstract

For delivery fleets with medium-duty electric trucks, battery sizing and energy efficient routing play an important role in cost reduction. In scenarios where periodic trends such as weekly or monthly delivery demands exist, the optimization problem needs to consider the requests for each day in the time horizon simultaneously, which is a challenging task to address. The existing heuristic and metaheuristic approaches designed for a single instance of vehicle routing problem (VRP) may not guarantee an optimal solution over the longer period of interest. In this paper, a two-stage grouping genetic algorithm (GGA) that builds on the existing GA approach for delivery cost minimization is proposed. Independent VRPs are solved using a GGA with battery degradation in consideration. The resulting population is reformulated in a way such that a second GGA can utilize the information to converge the solution to a single battery combination and the respective daily optimal routes that minimize the operational cost over a fixed time horizon. Compared to the approaches using single-day data, the two-stage GGA was able to obtain solutions that give up to 7.31% lower daily cost of operation.