Models and algorithms for the management of traffic flows with wireless charging lanes

Abstract

The rapid growth of Electric Vehicles (EVs) presents promising solutions to environmental and energy crises but introduces challenges in traffic management and charging infrastructure. This thesis explores advanced real-time traffic flow management models and algorithms specifically tailored for roads equipped with Wireless Charging Lanes (WCLs), a technology that allows EVs to charge while in motion. The primary objective is to optimize the overall operational efficiency (traffic and charging efficiencies) of the traffic systems with WCLs.

The research is structured around three key studies, each addressing different aspects of traffic management with WCLs. The first study explores a ramp metering control problem on WCLs, considering optimal traffic and charging efficiencies. First, we incorporate the state of charge (SOC) of electric vehicles (EVs) into the cell transmission model (CTM) in a mathematically convenient way, reformulating the model as a piecewise-affine (PWA) system. Using a hybrid model predictive control (MPC) approach, the control problem at each time stage is formulated as a mixed integer linear programming (MILP) problem, which is solved by well-established solvers. We conduct numerical experiments on an 8-km WCL for two sample scenarios and another with real traffic demand. We demonstrate both the efficacy and the limitation of ramp metering control in WCLs in terms of maximizing charging efficiency. We also reveal the inherent conflict between traffic efficiency and charging efficiency on a fully covered WCL. The proposed method and experiment results provide a novel tool and valuable insights for traffic authorities and policymakers regarding the management and operations of WCLs.

The second study addresses the variable speed limit (VSL) control problem in wireless charging lanes (WCLs), considering optimal traffic and charging efficiencies. Firstly, we introduce a system predictive model designed to anticipate the evolution of both traffic flow characteristics and the SOC of EVs with consideration of variable speed limits. The model is formulated as a PWA system through various linearization techniques. Subsequently, we propose a series of control models that account for the delicate balance between traffic and charging efficiencies, enabling the exploration of effective control strategies under varying priorities. The optimal control problem at each stage is cast as a MILP by a hybrid MPC approach. Our simulation results offer valuable insights for traffic operators engaged in the operation and management of WCLs.

The third study considers a dynamic pricing problem in a dual-lane system consisting of a general purpose lane (GPL) and a WCL. The electricity price is dynamically adjusted to affect the lane-choice behaviors of incoming EVs, thereby regulating the traffic assignment between the two lanes. The aim of dynamic pricing is to maximize operational efficiency (traffic efficiency and charging efficiency). First, we establish the dynamic traffic model tailored to the context of the dual-lane system by an Agent-Based Model (ABM) method, in which each EV acts as an independent agent with distinct characteristics. Next, we propose a model-free reinforcement learning (RL) algorithm, i.e., deep q-learning, to derive the optimal dynamic pricing strategy. A traditional machine learning (ML) method, that is, a classification and regression tree (CART) algorithm, and a static pricing strategy are also proposed for comparison. The simulation results reveal that both the dynamic pricing strategies (CART and deep q-learning) outperform the static pricing strategy in maximizing operational efficiency. In particular, the deep q-learning algorithm demonstrates a superior capability in optimizing dynamic pricing strategies by leveraging system dynamics more effectively and future traffic demand information. These insights also contribute to the real-time management of WCLs. This study serves as pioneering work to explore dynamic pricing issues in a multi-lane system with WCLs. The methodology adopted in this paper serves as a template for other researchers interested in similar issues.

In summary, these studies contribute to the exploration of real-time traffic flow management problems in the context of WCLs, providing models and algorithms tailored to this traffic context. They also offer insights that aid traffic authorities and policymakers in managing systems equipped with WCLs. This thesis not only addresses significant gaps in real-time traffic management strategies for DWC scenarios but also lays a foundation for future research.

Date of Award	15 Oct 2024
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Zhen Tan (Supervisor), Hing Kai Chan (Supervisor) & Liang Zheng (Supervisor)