Learning-based Control for Large-scale Sustainable Energy Systems

Abstract: Sustainable energy systems are crucial for climate mitigation. However, as these energy systems continue to expand in scale, complexity, and interconnectivity, the associated technical challenges also intensify. For example, desirable system behaviors must be maintained despite system uncertainties and frequent changes induced by the fluctuations of renewable energy sources. To address these key challenges, I will first present a non-asymptotic diameter bound for a popular uncertainty set estimation method called set membership (SM) for linear dynamical systems. Motivated by the new theoretical insights for SM, I will introduce a novel SM uncertainty set-based online adaptive control framework that leverages online learning techniques (e.g., convex online optimization) and control methods (e.g., distributed control and MPC) that enable novel safety guarantees for unknown systems under non-stochastic and potentially adversarial disturbances. I will present the first distributed learning-based controller that provably achieves adversarial stabilization (bounding the worst-case transient behavior) for networked systems under communication delays. I will also describe the application of the framework to the voltage control problem in the power distribution grids, as well as the online comfort-constrained HVAC control problem in novel buildings. The experiments empirically demonstrate that our approach remains effective in realistic nonlinear simulations with real-world data.

Bio: Jing Yu is a postdoctoral researcher at University of Michigan. She received her Ph.D. in Control and Dynamical Systems at Caltech, advised by John Doyle and Adam Wierman. She is broadly interested in the interplay between control theory and machine learning, with a focus on online decision making and distributed algorithms for large-scale sustainable energy systems. She is the recipient of several awards, including the Amazon AI4Science Fellowship, the best paper finalist award for ACM e-Energy, and the Caltech Amori Doctoral Prize.