Extracting Value from Time-synchronous Phasor Measurement Data

The power industry faces unprecedented challenges driven by the rapid advancements of revolutionary yet power-hungry large language models (LLMs) alongside with the growing penetration of renewable energy sources (RES). Wide-area monitoring systems (WAMS) offer a unique way to collect and analyze ultra-high-resolution time-synchronous phasor measurements, also known as synchrophasors, thereby providing a comprehensive view of the entire power grid. Synchrophasors offer a multitude of benefits making them an indispensable tool for power systems. This dissertation focuses on extracting the values of synchrophasors through comprehensive analysis of historical data and developing various applications. The contributions of this dissertation are below: First, a comprehensive data analysis of worldwide power grid frequency is conducted, accompanied by detailed frequency visualizations of international grids. Two novel metrics are introduce: one for intra-grid frequency correlation, facilitating the identification of local noise, inter-area oscillation, potential weak links, and islanding; and the other for unveiling underlying inter-day correlation of the frequency dispersion. Second, a short-term frequency prediction model is formulated. This dissertation evaluates state-of-the-art time-series forecasting deep learning (DL) model, simple linear regression model and polynomial regression model, concluding that polynomial model is preferred for its robustness and interoperability. Third, to enhance real-time awareness of power grids, this dissertation proposes a fusion model that integrates existing disturbance magnitude estimation methods and an AI-assisted physics-based approach for disturbance source location estimation. The former reduces the error by nearly half across all evaluated metrics, and the latter significantly reduces the maximum error and the error deviation. Given the importance of the real-time disturbance analysis for the power industry, a high availability strategy with redundant execution is proposed and implemented.