New Framework Tackles Equity Gap in Power Restoration After Disasters

Researchers have developed a computational framework designed to make emergency power restoration more equitable after major weather disasters. The work addresses a structural problem in how utilities currently operate: prioritizing repairs based on the number of restoration requests, which inadvertently disadvantages less organized or resourced communities.

The Equity Problem in Power Restoration#

When utilities respond to hurricanes and other extreme weather events, they typically allocate repair crews according to the volume of service requests they receive from each area. A data-driven analysis revealed a troubling pattern—disadvantaged communities submit significantly fewer requests, even when facing the same outages as more affluent neighborhoods. This mismatch means that vulnerable populations often endure longer power outages, compounding the hardship that disasters inflict.

Two-Part Technical Solution#

The researchers designed a framework called EPOPR (Equity-aware Post-disaster Power Restoration) with two complementary components:

• Repair-time forecasting: A method called Equity-Conformalized Quantile Regression that predicts how long repairs will take while properly accounting for uncertainty—a particular challenge when training data varies unpredictably across different regions and conditions.

• Smart scheduling: A machine-learning agent using spatial-temporal attention that makes restoration decisions while considering uncertainty levels unique to each area, rather than defaulting to low-risk actions that would favor already-served communities.

Results#

Testing showed that EPOPR reduced average outage duration by 3.6% and decreased inequity between communities by 14.19% when compared to existing approaches. The gains suggest that explicitly modeling both efficiency and fairness can yield tangible improvements in disaster response.