A decision-making system for managing electric vehicle fleets subject to multiple operational constraints

Abstract

Electric Vehicles (EVs) are attractive candidates to reduce transportation's environmental impact. Nonetheless, their low driving ranges, high recharging times, and the poor recharging infrastructure prevent their entire deployment. In this thesis work, we develop a strategy to manage EV fleets for delivery purposes efficiently. The main objective is to find and update the least-cost routes, charging plans, and departure times that allow an EV fleet to visit all destinations, subject to several operational constraints and real-world conditions, a problem known as the Electric Vehicle Routing Problem (E-VRP). The strategy consists of splitting the operation into two stages: pre-operation and online operation. We calculate initial routes in the pre-operation by solving an offline E-VRP (Off-EVRP). In the online stage, the dispatcher updates the routes based on traffic state realizations and EVs' state measurements by solving an online E-VRP (On-E-VRP). We solve both E-VRP variants with Genetic Algorithms (GA) using a novel encoding for each case. The overall strategy is tested with two experiments. Results show that solving the Off-E-VRP provides good initial route candidates, whereas solving the On-E-VRP can improve the operation and service quality. Finally, the developed decision-making system enables the fleet to fulfill the delivery purpose efficiently.