Through an advanced application of AI, EY and Eversource have created a proprietary algorithm to prevent sustained outages by accurately predicting the likelihood of one occurring. In a 2-month proof of concept, the model and subsequent fieldwork to fix issues the model has found has resulted in 40,000 avoided customer outages. This capability represents a significant leap in the utility industry, as utilities are under increasing pressures to improve grid reliability as a result of unprecedented weather patterns (i.e. climate change), evolving customer expectations, utility commissions requiring more details to approve capital investments, and dynamic changes to the grid with renewables and electric vehicles.

This cutting-edge machine learning model delves into historical data of power outages and assets, pinpointing the initial cause of momentary disruptions and assessing the risk of their escalation into sustained outages. The model integrates weather patterns, geographical nuances, and other critical geospatial data to establish correlations that bolster prediction accuracy. Furthermore, our collaboration with vegetation management teams infuses the model with insights from vegetation trimming cycles, enriching the predictive features. On-the-ground validation by dedicated patrol teams provides feedback serving as a vital component in the continuous refinement of the model. 

These predictions are based on “AI-ready data” through a cloud-based modern data platform that integrates data from disparate data sources/silos (SCADA, AMI, OMS, GIS, Circuit, EAM, Customer, inspection, weather, etc.) across the distribution grid. The data is integrated using a business-led flexible data model that responds to changing business needs and enables efficient root cause analysis of outages while enabling employee productivity, capital efficiency, and accelerated remediation of issues. The solution has three primary components:  outage, asset, and vegetation.  

Key Takeaways:

• Predicting the likelihood of sustained outages by leveraging existing data leads to improved SAIFI and customer satisfaction
• Momentary outages provide a wealth of information to improve customer service and reliability
• A holistic view of grid networks through analyzing actual grid investments leads to better investment modeling and capital efficiency
• Humans at the Center: Establishing feedback mechanisms will engage engineers while yielding additional data and circuit insights used to prevent future outages
• Reliability Data Hub brings “AI-ready data” to accelerate use case delivery

Note:  Technology/Solution is patent pending.