StormImpact’s Hurricane Outage Prediction Model

Hurricanes can cause significant damage to the electrical power system, leading to prolonged power interruptions to a large number of customers. The number of weather-related outages has increased significantly in recent years. This presentation summarizes the data-driven machine learning power outage models that StormImpact has developed and deployed for several major investor-owned electrical utilities in the United States. These models are used to support pre-storm preparation to determine the crew needs and the optimal staging of crews and materials. These models can also be used for emergency management training, and identifying the best approaches for system hardening and increasing long-term resilience. Here we provide an overview of model development and validation. We also highlight our approach for quantifying and communicating uncertainty. We will highlight the key features of our hurricane outage models using case studies from recent hurricanes that have affected the Gulf Coast and Eastern United States.