Event
Ph.D. Dissertation Defense: Byungchul Kim
Friday, March 27, 2026
2:00 p.m.
AV Williams 1146
Emily Irwin
301 405 0680
eirwin@umd.edu
ANNOUNCEMENT: Ph.D. Dissertation Defense
Name: Byungchul Kim
Committee:
Professor Eyad H. Abed (Chair)
Professor Nuno Martins
Professor Alireza Khaligh
Assistant Professor Kaiqing Zhang
Professor David Akin (Dean's Representative)
Date/time: Friday, Marth 27, 2026 at 2:00 PM
Location: AVW 1146
Title: Frequency and voltage stability of power systems with high penetration of renewable sources
Abstract:
The growing integration of converter-interfaced renewable energy sources has fundamentally transformed modern power systems, leading to a significant decrease in system inertia and presenting urgent challenges for frequency control. Distributed energy resources, which represent a substantial portion of the emerging power generation landscape, are anticipated to play a vital role in facilitating frequency regulation in future low-inertia systems. Nevertheless, their effectiveness is still limited due to constraints in active and reactive power generation from droop control, varying response capabilities, and communication difficulties. This dissertation explores the application of reinforcement learning control as a pivotal technology to overcome these challenges by providing real-time active and reactive power generation from distributed energy resources and predictive capabilities for distributed energy generation. This approach enables the development of new frequency service dispatch and control schemes for future power systems.
Three control strategies for power systems and microgrids are proposed, namely modified finite control set model predictive control, Heaviside Pontryagin’s optimal control, and reinforcement learning, each offering unique benefits for the dynamic control of distributed energy resource behavior.
The suggested methodologies consider the fidelity and efficiency of control for distributed energy resources applications, and the control strategies developed are the first of their kind to facilitate real-time management of frequency and active power responses in distributed energy resources. Furthermore, a new strategy for managing rapid frequency control is introduced, which is based on the optimization of the controller or goal function, leading to a significant enhancement in the robustness of frequency and voltage control within microgrids and power systems. Simulation experiments were performed using MATLAB and Simulink. The results of the tests validate that the proposed control strategies can effectively manage distributed energy resources in both microgrids and power systems.
The microgrid control methods based on neural networks within leader-follower configurations allow the system to evaluate the leader-follower responses of distributed energy resources, while also pinpointing the discrepancies between anticipated and actual time differences error in real-time. Simulation outcomes indicate that the suggested reinforcement learning control framework minimizes control errors and improves system resilience during frequency disturbances.
A coordinated control scheme based on an energy internet is proposed to further improve the frequency and voltage control capabilities of distributed energy resources. Both centralized (cloud-hosted) and distributed (edge-hosted) architectures are examined to overcome local area voltage and frequency control. The energy internet-based coordinated control scheme facilitates traditional coordinated control of distributed energy resources by eliminating communication delays in the optimal power flow transmission of distributed energy resources. Simulation results confirm the effectiveness of this approach, demonstrating enhancements over conventional control regarding speed, overshoot, and accuracy of active power responses. The power system and microgrid frequency are supported more effectively by coordinated distributed energy resources, leading to increased frequency stability during potential events.
