Doctoral Dissertations

Orcid ID

https://orcid.org/0000-0003-3018-830X

Date of Award

5-2025

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Energy Science and Engineering

Major Professor

Budhendra L. Bhaduri, Liem T. Tran

Committee Members

Melissa R. Dumas, Katherine N. Luke

Abstract

Extreme winter storms have increasingly been experienced in the southern U.S., where unwinterized infrastructure and vulnerable populations have been subjected to widespread power outages, resulting in significant financial losses and health risks. In this dissertation, power outage vulnerability during extreme winter storms has been examined, with a focus on Winter Storm Uri and the Houston, Texas metropolitan area. Three interconnected dimensions were addressed: (1) the influence of infrastructure conditions on outage susceptibility, (2) the socioeconomic characteristics commonly associated with at-risk populations, and (3) the use of radar-derived storm characteristics for outage prediction.

A machine learning model, MaxEnt, was utilized to identify key predictors of outage vulnerability. Variables such as power line density, tree coverage, and proximity to schools were found to be most influential. Census blocks in Central and South-Central Houston were identified by the model as having the highest likelihood of experiencing outages.

To evaluate how socioeconomic conditions contribute to vulnerability, a synthetic population of Houston was developed. Self-Organizing Maps (SOMs) were employed to cluster socioeconomic characteristics alongside outage likelihood. It was found that individuals with overlapping vulnerabilities were concentrated in the areas most likely to experience outages. Planning and policy recommendations were informed by these SOM results to support more equitable outage response strategies.

A novel set of “stormscape metrics” was developed based on principles of landscape ecology to track spatial-temporal storm patterns using near real-time radar data from Winter Storm Uri. A negative binomial regression model was applied to examine the relationship between stormscape metrics and county-level outages. It was revealed that increased storm complexity and spatial dispersion were associated with higher outage counts. The approach was designed to enhance existing outage awareness platforms by enabling the generation of outage estimates directly from radar imagery before storms arrive.

Through this research, a holistic understanding of power outage vulnerability during winter storms was developed by integrating infrastructure characteristics, social vulnerability, and storm dynamics. It is expected that these findings will support more equitable preparedness, improved outage response, and enhanced early-warning systems in the context of increasingly severe winter weather events.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS