Safe and Efficient Reinforcement Learning for Power System Control

Abstract

Smart grid is often viewed as the next generation electricity grid that enables the two-way communication between suppliers and consumers. How should operators coordinate the generation, the transmission and the consumption of electricity in this new infrastructure? In the first part of this talk, I will present several algorithmic approaches that are scalable to coordinate the operations of many distributed generators and storage units. Electric power grids are also increasingly stressed due to more variable renewable generation, increased demand and the requirement for real-time balancing. When multiple failures start to cascade in the grid, blackouts can occur. In the second part of this talk, I will present percolation-based models that can systematically study how and when cascading blackouts occur.

Bio

Baosen Zhang is an associate professor in the Electrical and Computer Engineering Department at the University of Washington. He received his B.A.Sc. degree from the University of Toronto in 2010 and his Ph.D. degree from University of California at Berkeley in 2015. His research interests are in machine learning, power systems and social networks. He is a recipient of the National Science Foundation CAREER Award and the Automatic Control Council Donald P. Eckman Award.