Unified Time-Scale Stability Analysis and Koopman-Based Modeling of Wide-Area Control and Grid-Forming Inverters
Date Issued
May 2026
Author(s)
Wang, Yichao
University of Tennessee at Knoxville
Advisor(s)
Seddik Djouadi
Additional Advisor(s)
Kevin Tomsovic
Fatima Taousser
Dan Wilson
Suzanne Lenhart
Abstract
Modern power systems are undergoing rapid transformation as communication networks and converter-interfaced resources alter their dynamic behavior. Wide-area measurement systems enable system-level monitoring and control, but the involvement of communication networks introduces time-varying delays and intermittent signal losses. These factors convolute the analysis and controller design, due to the fact that the closed-loop system ceases to exhibit purely continuous-time dynamics. Moreover, the increasing penetration of grid-forming inverters leads to hybrid behaviors originating from current-limiting and protective actions, which significantly affect inverter dynamics during large disturbances.
This dissertation addresses these challenges through two complementary frameworks. First, a unified modeling approach based on time-scale theory is developed to represent wide-area damping controllers operating over communication networks with intermittent transmission. The approach captures the switching between continuous-time plant dynamics and discrete-time behavior generated by zero-order holds during outages. Using this formulation, explicit stability conditions and upper bounds on allowable communication interruptions are derived, offering quantitative guidelines for delay-tolerant wide-area control.
Second, motivated by the hybrid behavior of grid-forming inverters under current saturation, a deep Koopman-based learning framework is proposed for the data-driven identification of inverter dynamics. By distinguishing voltage-source and current-limited operating modes and learning different Koopman operators for each, the method accurately captures nonlinear transient behavior, including mode switching and limit-cycle oscillations.
Collectively, these contributions demonstrate how hybrid systems theory and data-driven Koopman learning can provide useful tools to analyze the dynamics and stability of modern inverter-dominated power grids.
Degree
Doctor of Philosophy
Major
Electrical Engineering
File(s)![Thumbnail Image]()
Name
WangDiss_2026.pdf
Size
5.87 MB
Format
Adobe PDF
Checksum (MD5)
421092eae7b8e7bc82735d47b4e31632