Equity-Aware Power Restoration Framework Using Uncertainty-Aware Predict-Then-Optimize
A new research paper proposes an equity-aware power restoration strategy that balances efficiency and equity across communities. The study, available on arXiv (2508.04780), addresses the disparity in restoration request submission, where disadvantaged communities tend to submit fewer requests, leading to inequitable power restoration. The framework uses uncertainty-aware predict-then-optimize to handle heteroscedasticity in repair duration prediction and mitigate the tendency of reinforcement learning agents to favor low-uncertainty areas. The research highlights the urgent need for equitable restoration in the face of increasing extreme weather events like hurricanes.
Key facts
- The paper is titled 'Uncertainty-aware Predict-Then-Optimize Framework for Equitable Post-Disaster Power Restoration'.
- It is available on arXiv with ID 2508.04780.
- The research identifies disparities in restoration request submission, with disadvantaged communities submitting fewer requests.
- The proposed strategy balances restoration efficiency and equity across communities.
- The framework addresses challenges in predicting repair durations under dataset heteroscedasticity.
- It also addresses the tendency of reinforcement learning agents to favor low-uncertainty areas.
- The study is motivated by the increasing frequency of extreme weather events such as hurricanes.
- The paper is a replace-cross announcement type.
Entities
Institutions
- arXiv