Scalable Geospatial ML for Power-Line Asset Risk Management

Electric power networks are increasingly vulnerable to weather-related failures, necessitating precise, asset-level risk modeling. This research introduces a modular, robust, and explainable framework for predicting the probability of failure (PoF) for utility assets. The core innovation is an asset-level architecture designed for scalability, allowing integration of new environmental data sources and additional PoF types without requiring a complete pipeline overhaul. This adaptability is crucial for utilities facing evolving data availability, asset management priorities, and climate change impacts. The framework is demonstrated by modeling risks associated with vegetation and lightning, using a harmonized geospatial machine learning pipeline. It incorporates diverse predictors such as topography, vegetation condition (MODIS NDVI), lightning climatology, OpenStreetMap features, and operational utility records. The resulting system is computationally efficient, operationally extensible, and suitable for large-scale deployment. It provides actionable risk stratification, enabling better prioritization for inspections, vegetation management, asset hardening, and resilience planning, ultimately supporting proactive interventions and more climate-resilient network operations.