How StormGeo’s and Elvia’s AI Model Can Predict Power Outages Three Days in Advance

The Norwegian power grid operator Elvia has teamed up with StormGeo to develop an AI model that predicts power outages up to 72 hours in advance. First launched in 2023, the partnership project aims to improve grid reliability by enabling faster response times and reducing outage duration.

In a recent article in Europower, StormGeo’s Director of Data Science, Nina Winther-Kaland, elaborates on how StormGeo supports Elvia in developing this AI model for enhanced power outage prediction capabilities and reduced unplanned downtime. 

Winther-Kaland explains that the AI model learns from historical data and environmental factors to identify high-risk areas for power outages across the 69,500 kilometers of power lines that Elvia operates. While wind speed and direction have traditionally been key predictors of outages, this AI model incorporates additional factors, such as topography, changes in land use, soil moisture levels, and power line density.

The model has been trained on multiple years of data, validated on a year of historical data, and, since its operational launch in 2024, been tested during major storms. Recent results show that the model can predict the number of power outages in 6-hour intervals up to 72 hours in advance, with over 90% prediction accuracy for moderate wind-related outages. 

Looking ahead, Winther-Kaland emphasizes that AI-driven outage prediction is a significant step toward a more proactive grid management approach. By integrating AI with advanced weather data and solutions like StormGeo’s vegetation management solution GridEyes, StormGeo aims to provide energy companies with better decision-making tools to enhance supply security and reduce costs.

Read the full article in Europower (in Norwegian).