Document Type

Dissertation - Open Access

Award Date

2025

Degree Name

Doctor of Philosophy (PhD)

Department / School

Electrical Engineering and Computer Science

First Advisor

Timothy Hansen

Abstract

The rapid integration of distributed energy resources (DERs) is transforming the operational dynamics of electric power systems. As inverter-based resources (IBRs) with fast, low-inertia characteristics increasingly replace conventional synchronous generators (SGs), ensuring system stability has become significantly more challenging. Accurate power system modeling, dynamic state estimation (DSE) , and advanced control are therefore essential to capture fast-changing system behavior, maintain frequency stability, and enable reliable protection and coordination. Recent large-scale disturbances, such as the April 28, 2025 blackout across Spain, Portugal, and France, have underscored the consequences of inadequate modeling and control under high-renewable, low-inertia conditions. Leveraging high-resolution, time-synchronized measurements together with computational advancements, data-driven modeling and estimation techniques have emerged as vital tools for enhancing situational awareness, improving control performance, and ensuring the stability and resilience of future power grids. In this dissertation, a Neural Ordinary Differential Equations (Neural ODEs) framework is developed to accurately model and estimate the frequency dynamics of modern power systems with high penetration of DERs. The proposed approach enables data-driven learning of system dynamics directly from measured states and inputs, eliminating the need for detailed system models and parameters. Neural ODEs are first validated on simplified systems and then extended to DERs integrated microgrids (IMGs) with diverse and dynamic configurations. The method demonstrates strong robustness against measurement noise and variations in initial conditions, while maintaining low computational complexity suitable for real-time applications. Through transfer learning (TL) , the model effectively adapts to changing system topologies and generator dispatch conditions, making it particularly suitable for evolving IBR-dominated grids. Furthermore, by integrating Neural ODEs-Neural Network (Neural ODEs-NN) based estimator with a Soft Actor-Critic Reinforcement Learning (SAC-RL) based fast frequency controller, a complete model-free estimation and control framework is developed. This unified framework enables rapid frequency support and enhanced stability under high renewable and low-inertia operating conditions. This contribution is especially valuable for grid operators, offering an adaptive, data-driven, and computationally efficient toolset for dynamic system modeling, state estimation, and frequency control in future DER-rich power networks. The secondary objective of this dissertation is to evaluate and compare various state and parameter estimation and control techniques for modeling the frequency dynamics of power systems. Estimation methods including the Kalman Filter (KF), extended Kalman filter (EKF), unscented Kalman filter (UKF), and moving horizon estimation (MHE), were assessed for accuracy and computational efficiency. While all filters performed comparably under Gaussian noise, EKF and UKF exhibited convergence issues with poor initial estimates, and MHE, though stable, required higher computation time. The study further compares model-based estimation using KF with data-driven approaches such as Neural ODEs and System Identification (SysId) , along with control strategies involving Model Predictive Control (MPC) and SAC-RL for Fast Frequency Response (FFR) . Results demonstrate that Neural ODEs offer improved scalability for complex systems, while SAC-RL provides faster and more effective frequency support than MPC, highlighting the potential of model-free, data-driven methods for future DER-integrated power grids. The comparative study guides the reader towards computationally tractable algorithms for modeling, estimation and control for the future DER-integrated power grid.

Publisher

South Dakota State University

Download

Available for download on Wednesday, December 15, 2027

Share

COinS
 

Rights Statement

In Copyright