Spatio-Temporal Model for Power Outage Prediction and Resilience Enhancement Planning
Speaker
Dr. Feng Qiu
Argonne National Laboratory.
Title
"Spatio-Temporal Model for Power Outage Prediction and Resilience Enhancement Planning"
Abstract
As extreme weather events grow more frequent and grid topologies more complex, electric utilities require reliable tools to forecast where and when outages will occur and to simulate plausible outage scenarios for resilience assessment. Existing approaches for power outage analysis are typically tailored to a specific machine learning task or local service territory, which limits their scalability and transferability. Inspired by the success of large-scale generative pretraining in language and vision, we propose the first spatio-temporal model for power outages that learns from massive archives of customer outage records across the U.S. alongside multi-modal covariates (e.g., weather, census data, and network topology).
Preliminary experiments demonstrate that our model delivers strong predictive accuracy for outage forecasting and generates realistic scenarios for ``what-if'' resilience analyses. The parameters extracted from the model could give statistic indications for performing resilience enhancement planning. This equips utilities and planners with a flexible, data-efficient tool to anticipate emerging risks and stress-test grid resilience under diverse conditions.
About Speaker
Dr. Feng Qiu is a principal computational scientist and the group manager for the Advanced Grid Modeling, Optimization, and Analytics (AGMOA) and serves as the Lab Relationship Manager for the OE Advanced Grid Modeling program. AGMOA provides fundamental modeling and computing tools for power system research and solutions for operations, planning, cybersecurity, etc. Dr. Qiu received his PhD from the School of Industrial and Systems Engineering at the Georgia Institute of Technology in 2013. He holds a BS in Automatic Control and an MS in Systems Engineering from Huazhong University of Science and Technology. He joined ANL in 2013.
His current research interests include power system modeling and optimization, electricity markets, power grid resilience, and machine learning and data analytics. He is a PI for a number of Department of Energy (DOE) projects from multiple DOE offices, including the Solar Energy Technology Office, Water Power Technology Office, and Office of Electricity, as well as projects from the National Science Foundation (NSF), among others. He also serves as the Lab Relationship Manager for the Office of Electricity Advanced Grid Modeling program. Additionally, he holds joint positions at Iowa State University as an Affiliate Associate Professor and at Northwestern University as a NAISE Fellow.