Multi-Agentic Utility-Centered AI Digital Infrastructure for Storm Event Data Intelligence and Analytics Assistance
Lead PI: Dr. Suining (Henry) He, UConn
Co-PI: Dr. Xinxuan Zhang, UConn; Dr. Manos
Industrial Relevance & Need:
Utilities are expected to document storm impacts in increasing detail, provide transparent restoration metrics, and justify resiliency investments with data-driven evidence. Yet historical storm event data — wind speeds, precipitation, National Oceanic and Atmospheric Administration (NOAA) advisories, storm tracks, and local damage reports — often resides in separated systems or archives. Outage data, crew logs, and asset level failure histories are likewise siloed across Outage Management System (OMS), Geographic Information System (GIS), Supervisory Control and Data Acquisition (SCADA), and work management tools. Utility analysts as well as their domain researchers spend weeks assembling these datasets to reconstruct storm impacts or prepare regulatory filings. As storm-related reporting obligations expand, this approach is no longer scalable or sustainable. Such a gap between large-scale, extensive storm event data silos, and the actionable insights constrains a utility’s ability to respond proactively to grid stresses and meet regulatory and reliability requirements.
Relying on reactive and largely fragmented processes to extract information slows down decision-making, limits scalability, and makes it difficult to produce timely reports for compliance or long-term planning. With the advent of generative agentic AI, there emerges a pressing need to adapt such a state-of-the-art technology and develop a more intelligent, interactive, and utility-centered data platform for historical storm event impact analysis. This platform can be designed as a deployable, scalable service tool for utility analysts and other decision-makers. Such an agentic AI-powered tool will serve as a digital infrastructure to enable fast, intuitive search, and efficient analysis of vast, heterogeneous storm-event datasets.
Project Goals:
The overarching goal of this project is to design, develop, and validate MAISI, a novel Multi-agentic utility centered AI tool for Storm-event data Intelligence and analytics assistance. Our MAISI project will achieve following goals:
Design storm event data intelligence service that empowers utility companies to unlock the full operational and strategic value of their historical datasets.
Integrate multi-agentic, generative, and utility-centered AI capabilities to automate data retrieval, analytics, visualization, and reporting, via novel techniques of retrieval augmented generation, predictive intelligence, and natural language interaction.
Engage closely with utility stakeholders including operational decision makers and storm domain researchers toward co-designing and validating a usable, deployable service infrastructure with practical and tangible impacts.