Participation Factor-Based Model Reduction and Partitioning for Accelerating Power System Simulations

The increasing deployment of Inverter-Based Resources (IBRs) introduces fast dynamics such as sub-synchronous oscillations (SSOs), fault ride-through events, and stability challenges that are often inadequately captured by traditional phasor-domain simulators. Accurate representation of these dynamics necessitates computationally intensive electromagnetic transient (EMT) simulations. A practical solution is to focus EMT simulations on a localized critical zone—typically containing IBRs, generators, and associated network elements—while representing the rest of the system with simplified equivalents.

This research proposes a novel framework for EMT boundary determination and model reduction by leveraging both linear and nonlinear participation factors (PFs). Linear PFs are used to identify critical states for model simplification, while nonlinear PFs, computed using normal form theory, help capture higher-order dynamic interactions in near-resonant conditions. To enable scalable computation of high-order NPFs, a tensor contraction–based approach combined with a dynamic batching strategy is introduced, allowing memory-efficient computation up to arbitrary expansion orders.

For systems exhibiting time-periodic behavior—common in EMT simulations—a Floquet theory–based method is developed for PF calculation and EMT boundary determination. This method captures modal behavior across harmonics by analyzing the system over a single 60 Hz cycle, enabling efficient detection of SSOs and informing boundary placement. A data-driven version of this method is also proposed, requiring one cycle time-domain responses for PF computation, making it applicable to black-box models. Furthermore, a mathematical connection is established between participation factors in linear time-invariant (LTI) and linear time-periodic (LTP) systems, revealing that phasor-domain PFs are a special case corresponding to the zero-harmonic component in the broader Floquet-based framework.

Collectively, this dissertation aims to lay the foundation for more accelerated while accurate simulation of modern power systems with high IBR penetration, enabling new model reduction approaches, advanced boundary determination, and data-driven dynamic analysis.