Socioeconomic disparities in hurricane-induced power outages: Insights from multi-hurricane data in Florida using XGBoost
Year: 2025
Type: article
Abstract: This study explores the importance of socioeconomic factors in hurricane-induced power outages in Florida. An XGBoost regression framework that incorporates a comprehensive feature set, including diverse socioeconomic factors, hurricane hazards, and physical exposure, is introduced. To reduce random deviations in importance observed in prior single hurricane studies, data for 11 Florida hurricanes is processed and analyzed, sourced from various state and federal agencies. To further enhance the robustness of model findings, analysis was conducted on 66 independent repetition runs filtered from 250 model iterations to control for overfitting. An extended formulation of SHAP values across iterations is introduced to enable a nuanced assessment of feature importance. Results show that socioeconomic variables account for 19% of the model prediction. This finding underscores the presence and significance of social inequities in hurricane outages. The unemployment rate, percentage of disabled, and racial/ethnic minorities are found as the most important predictors. Two new variables – flooding and substations per county – are assessed in this study, but they are found to have no notable contribution to power outages. The findings of this study provide new insights into the interplay between socioeconomic conditions and power system performance, aiding outage prevention efforts by identifying socioeconomic inequalities in pre-existing conditions and system operations. The findings of this study highlight systemic socioeconomic vulnerabilities in power grid resilience, offering critical insights for policymakers to allocate resources and improve disaster response strategies. While the model is tailored for Florida, its structure could be adapted to assess power outage disparities in other hurricane-prone regions.
Source: Sustainable Cities and Society
Authors: Alexys H. Rodríguez-Avellaneda, Ryan Rodriguez, Abdollah Shafieezadeh, Alper Yilmaz
Institutions: The Ohio State University, Georgia Institute of Technology
Language: English
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $4,870
OpenAlex ID: W4409464202
