PhD Oral Exam - Nassr Al-Dahabreh, Information Systems Security

When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.

Abstract

As the Electric Vehicle (EV) market continues to experience significant growth, there is a pressing need to expand the public EV charging infrastructure accordingly, so that rising EV demand for EV charging stations (CSs) is met while preserving a satisfactory quality of experience (QoE) for EV users during the charging service. Global policy commitments and financial incentives aimed at phasing out internal combustion engine vehicles are accelerating EV adoption. While this transition supports decarbonization goals, public EV charging network expansion has not kept pace uniformly: growth rates vary widely across countries, cities, and even neighboring regions, creating localized capacity shortfalls and undermining user confidence. Ensuring that the public EV charging infrastructure scales reliably, while preserving acceptable QoE requires systematic visibility into the EV session dynamics, robust EV demand forecasts, and principled deployment rules.

Although the growth of the EV market has increased the need for a reliable public EV charging infrastructure, the current ad-hoc approach to deploying additional charging stations cannot guarantee long-term control of critical QoE metrics or instill confidence in users. Proper tracking of events and occurrence times is necessary to give operators visibility into EV charging system dynamics, enabling accurate forecasting and optimal infrastructure expansion strategies that meet rising demand while maintaining acceptable QoE levels. In particular, uncertainty in per-CS waiting time estimation hinders public charging infrastructure (PCI) uptake: EV users avoid public CSs when waiting times are unpredictable, which both lowers QoE and leads to under-utilization or misplaced capacity expansion. Conversely, poorly designed expansion strategies can worsen congestion at EVCSs and displace demand rather than resolve it.

To address these challenges, this thesis contributes a set of analytical, theoretical, and data-driven methods that: (i) characterize per-public-charging-site session dynamics and waiting-time statistics; (ii) building on that empirical foundation, formulate reliable QoE performance metrics that give operators clear visibility into per-site service dynamics (waiting time, blocking probability, utilization, queue length, etc.); (iii) to turn the analysis into strategic decision support, develop a fully customized, robust client–server software tool for visualizing, analyzing, and interpreting EV charging station performance metrics at a granular, site-specific level, this tool functions as a decision-support system, empowering operators and stakeholders to assess the adequacy of existing infrastructure by tracking EV demand trends and key performance metrics with precision; (iv) propose demand-management incentives and queueing models to enhance QoE; and (v) quantify and forecast the QoE impact of new deployments. Concretely:

First, the thesis provides an end-to-end examination of public EV charging ecosystem components and their operational dynamics. Using large, real-world datasets, the thesis uncovers the statistical properties of EV charging sessions and waiting times, notably showing that charging times are better captured by an Erlang-k distribution for many practical settings. It further demonstrates that per-CS request processes and service behavior can be accurately modeled as single-server queueing systems under common scheduling policies. These empirical findings establish the probabilistic foundation required for accurate waiting-time estimation and capacity planning.

Building on that empirical foundation, the thesis formulates reliable QoE performance metrics that give operators visibility into per-site service dynamics (waiting time, blocking probability, utilization, queue length, etc.). These metrics serve as inputs to a tailored machine-learning forecasting model trained on recent real-world datasets to predict long-term EVCS loads. The forecasting pipeline is validated through extensive simulation so that operators and service providers can make evidence-based capacity-expansion decisions that maintain QoE for EV users during public charging service. To translate the analysis into effective decision support, this thesis introduces a robust, custom client–server platform that applies the study’s findings. The developed system-core provides granular, site-specific visualizations and diagnostics of EV charging metrics, providing operators with precise EV demand and performance insights for deployment and capacity planning.

To limit overload at individual sites and reduce congestion during service, this thesis proposes a Data-driven, Incentive-based Charging Truncation (DICT) scheme that encourages EV drivers to terminate charging once their battery reaches 80% state of charge (SoC) during the public EV charging service. An accurate and validated statistical and mathematical examination of the DICT policy is provided: simulation results are analyzed to derive a closed-form fit for the resulting service-time distribution; an analytical M/G/C/K queueing-theory model is implemented and applied to capture per-site dynamics under DICT in terms of QoE metrics. DICT is then benchmarked against common resizing and in-proximity deployment/expansion strategies to suggest recommended actions for improving QoE. Finally, moving beyond correlational studies, this thesis develops a with/without counterfactual machine-learning framework to quantify how new L3 fast-charging sites deployments affect congestion and QoE at nearby public EV charging sites. The framework controls for spatial proximity, charger capacity (port counts and power ratings), and amenity access; it uses large-scale ML models to estimate counterfactual demand trajectories; and it maps those projections into queueing inputs for M/G/C/K models to derive performance and QoE metrics.

Together, these contributions combine empirical discovery, principled queueing models, forecasting, decision-support software, and incentive strategy design to provide operators and policymakers with the tools and evidence needed to expand the public EV charging infrastructure in ways that are both capacity-aware and QoE preserving.