Understanding how extreme weather causes widespread power system failures requires utilities to group thousands of individual outages into coherent events. Researchers at arXiv have developed an automated algorithm that clusters utility outage records based on when and where they occur, enabling more accurate analysis of power system resilience during storms and natural disasters.
The core innovation uses a geometric model where each outage is represented as a cylinder in three-dimensional space. The cylinder's circular base represents the geographic location of an outage, while its height corresponds to the outage duration. Two outages belong to the same event if their cylinders intersect in both time and space—meaning they overlap geographically and temporally.
This approach improves upon simple time-based grouping methods that can mistakenly merge unrelated outages across large service territories. By requiring spatial overlap, the algorithm prevents distant outages that happen to occur within the same hour from being incorrectly categorized as components of a single storm event.
The grouping process is implemented as a graph algorithm where outages represent nodes and edges connect overlapping outages. Events then emerge as connected components—clusters of mutually related outages. The researchers propose a tuning metric to optimize the algorithm's temporal and spatial thresholds, minimizing false groupings while capturing genuine weather-driven events.
Validation testing compared automatically extracted events against NOAA storm records and DOE Form 417 reports (the federal standard for outage documentation). Results showed strong agreement between the algorithm's output and independently documented major weather events, demonstrating practical reliability.
The method works with both granular utility data and web-scraped information from EAGLE-I, a publicly available outage tracking system. This flexibility makes the algorithm valuable for utilities, grid operators, and researchers seeking to correlate outage patterns with meteorological conditions and improve grid hardening strategies during severe weather seasons.
