Please join us for the following presentation at 11am in ITE 336:
Talk title: Learning-Enabled Modeling, Estimation, and Decision-Making for Resilient Renewable-Rich Power Systems
Abstract: The rapid proliferation of distributed energy resources (DERs), smart inverters, sensors, and AI data centers is reshaping modern electric grids into dynamic, data-rich ecosystems. This talk will introduce learning-enabled power system modeling, estimation, and decision-making, which link physics and machine learning (ML) to significantly enhance energy efficiency, reliability, and resilience. Physics-informed ML-enabled methods will be presented for grid modeling, situational awareness, and dynamic control, fully unlocking the potential of data while respecting physical constraints. The proposed learning-enabled grid modeling focuses on high-fidelity data-driven surrogates of unbalanced distribution networks and dynamic power systems with inverter-based resources (IBRs). Building on these foundations, this talk will showcase two real-time solutions: uncertainty-adaptive Bayesian voltage estimation and post-fault emergency control via graph reinforcement learning. The proposed approaches exhibit high cross-task generalization, efficiency, robustness, and physical interpretability. One direction for future work is to develop lightweight, energy-efficient TinyML that can run on low-power devices at the grid edge, a key next step toward sustainable electrification in the AI era.
For more information, contact: Brandy Ciraldo at brandy.ciraldo@uconn.edu