Modeling and optimization of electric vehicle networks in future sustainable cities

Abstract

Motivated by the increasing concern of environmental pollution and fossil fuel shortage, transportation electrification, namely the process of integrating a large fleet of public and private electric vehicles (EVs) into the transportation system, is conceived to be one of the promising solutions in future sustainable cities. Enormous efforts have been made on modeling and optimizing EV networks and their extensions. Here, an EV network is an integration system of EVs and EV refueling stations, where EVs can get refueled in stations and drivers/passengers with mobility demand can take EVs to travel. Therefore, two basic problems are considered in EV networks, namely, i) energy allocation and ii) mobility management problems. First, considering EV networks as energy consumers, a properly designed refueling strategy is highly necessary in order to relieve the impact of EVs' refueling load on power grids and meanwhile satisfy EVs' quality-of-service. Second, considering EV networks as mobility service providers, efficient mobility management is needed to accommodate temporally and spatially different mobility demand from passengers. This thesis develops three distinct real-time decision-making models and algorithms to handle the two basic problems in practical scenarios of EV networks. In the first technical chapter, we investigate a network of battery swapping stations (BSSs), where battery swapping is considered as a more time-efficient method for EV refueling compared to plug-in charging. Specifically, in this chapter, we focus on a joint long-term battery inventory planning and real-time vehicle-to-station (V2S) routing problem, where EVs' refueling demand arrives randomly and sequentially. An online decision-making framework is proposed to model and optimize the operation of BSS networks, and a closed-form performance bound is theoretically guaranteed. In the second technical chapter, we consider the scenario of electric mobility-on-demand (EMoD) system (e.g., vehicle-sharing and ride-sharing), where EVs are directly managed by a system operator to provide mobility services. EVs can be dispatched to serve passengers' individual mobility demand or reallocated to accommodate unbalanced demand in different locations. In addition to that, we also make recharging decisions to refuel EVs and keep their energy levels. An efficient decision policy is derived to accommodate the real-time demand and maximize the long-term system revenue. In the third technical chapter, we propose a dynamic pricing mechanism in the EMoD system to incentivize passengers with spatially and temporally unbalanced demand to make different mobility choices. In this way, passengers' traveling demand can be reshaped and the vehicle reallocation cost is reduced. We formulate a bi-level optimization problem, with the system revenue maximization and customers' utility maximization as the upper-level and lower-level problems, respectively. A near-optimal decision policy is derived to make real-time pricing decisions and maximize the expected long-term system revenue.

Date of Award	2021
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Danny Hin Kwok Tsang (Supervisor)