Document Type

Thesis - Open Access

Award Date

2026

Degree Name

Master of Science (MS)

Department / School

Electrical Engineering and Computer Science

First Advisor

Robert Fourney

Abstract

The continued growth of electric vehicles (EVs) is increasing the demand for charging infrastructure and creating new operational challenges for charging network operators (CNOs). Uncertainty in charging demand, electricity market prices, and user charging behavior can significantly impact both the economic performance and reliability of charging network operations. To address these challenges, this thesis develops an integrated data-driven framework that combines charging demand modeling, risk-aware energy procurement, machine learning, and reinforcement learning for intelligent EV charging network operation. The research begins by developing queueing-based models to characterize EV charging demand under time-varying and stochastic arrival conditions. Time-varying Mt/M/1 and Mt/M/c queueing models are used to estimate charging demand, station utilization, waiting times, and charging network performance using both synthetic and real-world charging datasets. These demand estimates are incorporated into a dynamic pricing framework that links charging prices with wholesale electricity procurement costs. Building upon the demand modeling framework, a two-stage stochastic optimization model based on Conditional Value-at-Risk (CVaR) is developed to determine risk-aware day-ahead energy procurement decisions under uncertain demand and electricity prices. The results demonstrate that different market conditions require different levels of risk aversion and that procurement performance is highly sensitive to the selection of the risk preference parameter. To address this challenge, machine learning (ML) methods are introduced to adaptively select risk preferences using historical market information and scenario-based uncertainty features. Several learning approaches, including cost-surface learning and contextual bandit models, are evaluated and shown to reduce realized operating costs and regret relative to fixed-risk procurement strategies. Finally, a reinforcement learning (RL) framework based on the Soft Actor-Critic (SAC) algorithm is developed for joint day-ahead energy procurement and dynamic charging price optimization. Unlike traditional optimization approaches, the proposed reinforcement learning agent directly learns operational policies from market observations and simulated interactions. The results demonstrate that the learned policy improves profitability, reduces procurement costs, and outperforms fixed and rule-based benchmark strategies across multiple operating conditions. Overall, this thesis presents a unified framework for intelligent EV charging network operation that integrates charging demand estimation, risk-aware energy procurement, machine learning, and reinforcement learning within a single decision-making architecture. The proposed methodologies provide practical tools for managing uncertainty, improving economic performance, and supporting the continued expansion of electric vehicle charging infrastructure. Future work will extend the framework by incorporating spatial charging network dynamics and congestion-aware reinforcement learning to support geographically distributed charging systems.

Publisher

South Dakota State University

Download

Share

COinS
 

Rights Statement

In Copyright