Abstract
Detecting, analyzing, and predicting power outages is crucial for grid riskassessment and disaster mitigation. Numerous outages occur each year,exacerbated by extreme weather events such as hurricanes. Existing outage dataare typically reported at the county level, limiting their spatial resolutionand making it difficult to capture localized patterns. However, it offersexcellent temporal granularity. In contrast, nighttime light satellite imagedata provides significantly higher spatial resolution and enables a morecomprehensive spatial depiction of outages, enhancing the accuracy of assessingthe geographic extent and severity of power loss after disaster events.However, these satellite data are only available on a daily basis. Integratingspatiotemporal visual and time-series data sources into a unified knowledgerepresentation can substantially improve power outage detection, analysis, andpredictive reasoning. In this paper, we propose GeoOutageKG, a multimodalknowledge graph that integrates diverse data sources, including nighttime lightsatellite image data, high-resolution spatiotemporal power outage maps, andcounty-level timeseries outage reports in the U.S. We describe our method forconstructing GeoOutageKG by aligning source data with a developed ontology,GeoOutageOnto. Currently, GeoOutageKG includes over 10.6 million individualoutage records spanning from 2014 to 2024, 300,000 NTL images spanning from2012 to 2024, and 15,000 outage maps. GeoOutageKG is a novel, modular andreusable semantic resource that enables robust multimodal data integration. Wedemonstrate its use through multiresolution analysis of geospatiotemporal poweroutages.