During storm events, a utility’s storm response team is tasked with appropriately allocating resources in advance of the event to allow for provision and positioning of crews. And other utilities may request mutual aid to adequately address the resulting outages. However, predicting the number of expected outages and the resources needed to restore power is extraordinary difficult because of the complexity of weather forecasts and the interaction with vegetation and utility infrastructure.

Rather than relying on weather forecasts alone, learn how data science can incorporate the relative risk from vegetation, various infrastructure types and conditions, terrain, geography, and inspection history before integrating high-resolution numerical weather prediction model outputs. We’ll explore how data-driven applications forecast nonweather causes as well as weather events for a more accurate and reliable total outage prediction. This approach predicts expected daily system-wide outages up to five days before storm impact and hourly location-specific outages up to 24 hours before storm impact. We’ll learn how a northeastern utility improved its outage prediction accuracy by 30% three days in advance of a storm. Armed with data science, the utility was able to put the right levels of staffing in place, appropriately secure or deploy mutual aid, and implement a data-driven response plan.

Join this session and we’ll cover the building blocks for how a utility can begin to implement a data science approach to storm outage prediction which will lead to: improved outage prediction modeling, more timely and accurate mutual calls, and improved response time for restoring outages after a storm.