Research

Electric power systems are the foundation of modern society, sustaining every aspect of daily life from industry to homes. With the rapid integration of multi-type energy resources such as photovoltaic systems, electric vehicles, and data centers, the grid’s dynamics have become increasingly complex, introducing challenges in stability and overall reliability. At the same time, advances in sensing and communication technologies have flooded the grid with multimodal data from phasor measurement units (PMUs), smart meters, and imaging devices. These data streams capture the grid’s physical and dynamic states in unprecedented detail yet remain fragmented and underutilized. Left unaddressed, they pose growing risks to modern power grids and society.

My research aims to enhance the intelligence, resilience, and operational reliability of modern power systems by jointly leveraging multimodal data analytics and optimization based control of multi type energy resources, through data-informed power system modeling and AI-driven operational strategies.

Data-driven fault analysis and optimal PMU placement in smart grids

With the increasing penetration of inverter based resources, conventional fault detection methods become less effective due to weak or masked fault signatures. To address this challenge, I developed data driven frameworks that exploit spatiotemporal correlations in PMU measurements to identify weak and hidden faults in distribution networks.

By integrating random matrix theory with high resolution PMU data, my work statistically distinguishes abnormal system behavior from normal dynamics and enables accurate fault localization. I further proposed a data driven PMU placement strategy that ranks node importance based on vulnerability and fault sensitivity, achieving comprehensive observability with minimal instrumentation cost.

Multimodal data fusion for intelligent grid monitoring

Modern power systems generate heterogeneous data streams, including electrical measurements, environmental information, and visual data. Effectively integrating these sources is essential for robust system awareness under complex operating conditions.

I developed multimodal fusion frameworks that integrate PMU measurements, weather forecasting information, images, and physical parameters from icing monitoring systems. Through correlation driven feature alignment and adaptive weighting of different modalities, these methods achieve accurate and fault tolerant monitoring of transmission line icing, even under adverse environmental conditions.

Unbalanced distribution optimization and closed loop control

High penetration of distributed energy resources introduces voltage fluctuations and phase imbalance that require coordinated control across multiple devices. I developed unbalanced optimal power flow formulations that incorporate IEEE 1547 compliant inverter functions together with regulators, capacitor banks, and other legacy grid devices.

A key contribution of my work is the closed loop implementation of optimization, where control decisions are automatically exchanged with CYME for time series validation. This framework is further extended through a hybrid reinforcement learning and optimization approach, enabling scalable and efficient multi period decision making under mixed integer constraints.