EECS Faculty candidate seminar: Power electronics enabled resilience for renewable rich power systems by Yichen Zhang, Argonne National Laboratory (ANL)
About the event
OVERVIEW
Grid decarbonization and grid resilience are two national missions. However, they are contradicted with each other in a certain sense as integrating inverter-based resources degrades dynamic performance, increases cyber vulnerability, and induces uncertainty for grid operation and control. This presentation will mainly discuss two strategies to enhance grid resilience using inverter-based resources: performance-guaranteed control-based frequency regulation and learning-enabled automatic restoration. These strategies are both safety- and time-critical missions. Hence, theoretical framework, control architectures, and learning paradigms will be discussed to synergize real-time capability, online reactivity, computation complexity, and, more broadly, grid decarbonization and resilience.
In performance guaranteed control-based frequency regulation, set-theoretic frameworks and backward reachability will be employed to guide the mode switching of multi-mode inverter-based resources so that they can provide supportive power under contingencies to ensure adequate system-level performance. It is also found that the traditional approach of inertia emulation only provides time-varying inertia constant equivalently, making system assessment more difficult. A mode reference control-based inertia emulation strategy is proposed to provide a programmable and guaranteed inertial response. To address complex state and control constraints, numerical optimal control is embedded in a hierarchical structure for online deployment. Additionally, it is shown that the forward reachability can support the formal evaluation of linearization-induced uncertainty. This allows us to use linearized systems in the robust optimization framework to compute the control, which significantly reduces the computation complexity.
In learning-enabled automatic restoration, it is often challenging to compute the restoration actions for vast outage scenarios in the embedded environment. The reinforcement learning framework is proposed to embed the restoration policy into a deep neural net. To further improve the training efficiency, imitation learning algorithms are proposed, where the agent will learn the optimal actions from the expert, a mixed-integer program in this case. The successful results indicate that the imitation learning framework acts as a bridge between exploration-dominant reinforcement learning and mixed-integer programming and a way to leverage well-studied mixed-integer programming solvers for reinforcement learning-based automation.
Finally, the presentation will probe into preliminary studies that aim to bridge dynamic stability and optimization in power systems to address dynamics-constrained scheduling problems. Machine learning models will be built to map static operating conditions to dynamic performance and then converted to a set of optimization constraints. Active learning methods are employed to resolve the so-called labeling bottleneck in this constraint embedding process. The study illustrates the great potential of machine learning to address dynamics-constrained scheduling problems.
Bio
Yichen Zhang received the B.S. degree in electrical engineering from Northwestern Polytechnical University, Xi’an, China, in 2010, the M.S. degree in electrical engineering from Xi’an Jiaotong University, Xi’an, China, in 2012, and the Ph.D. degree in electrical engineering from the Department of Electrical Engineering and Computer Science, University of Tennessee, Knoxville, TN, USA, in 2018. He was also a research assistant with the Oak Ridge National Laboratory from 2016 to 2018. He is currently an energy systems scientist with the Energy Systems Division, Argonne National Laboratory, where he was a post-doctoral appointee from 2018-2021. His research interests include power system dynamics, grid-interactive converters, control, and decision-making for cyber-physical power systems.
Yichen Zhang received the Best Reviewer Award for IEEE Transactions on Power Systems, IEEE Transactions on Smart Grid, and Outstanding Research Assistant Award at University of Tennessee. He served as the student chair/co-chair for CURENT NSF/DOE Site Visit Committee from 2015-2017 and is co-chairing the IEEE task force on Flexible Grid-interactive Efficient Buildings to Enhance Electric Service Resilience from 2021.