Date of Award
Doctor of Philosophy (PhD)
Electrical and Computer Engineering (Holcomb Dept. of)
Dr. Ramtin Hadidi
Dr. Richard Groff
Dr. Daniel Noneaker
Dr. Sukumar Brahma
Dr. John Wagner
This dissertation presents and evaluates model-free methodologies to process Phasor Measurement Unit (PMU) data. Model-based PMU applications require knowledge of the system topology, most frequently the system admittance matrix. For large systems, the admittance matrix, or other system parameters, can be time-consuming to integrate into supporting PMU applications. These data sources are often sensitive and can require permissions to access, delaying the implementation of model-based approaches. This dissertation focuses on evaluating individual model-free applications to efficiently perform functions of interest to system operators for real-time situational awareness. Real-time situational awareness is evaluated with respect to central digitization where the PMU data is archived, and delays from telecommunication and system architecture are not considered.
The PMU data available to utilities is often a subset of the overall system. Even without full observability, PMU data for observable portions of the system provides valuable, high-resolution information about the current system state. Methods are needed that can analyze and generate critical insight about the system in real-time to assist in detection and mitigation of major system events. All chapters address methodologies that can derive their output solely from the PMU signals. These methodologies are evaluated for their reliability and computational efficiency, considering a specific task of interest.
Inter-area oscillations and poorly damped electromechanical modes are dangerous when undetected for extended periods of time, eventually leading to blackouts when unstable parameters are present. Prony Analysis and Matrix Pencil Method were selected in Chapter 4 for their proven effectiveness of estimating the dominant modes of an input signal; for purposes of this dissertation, the signal of interest for oscillation analysis is real power. The speed of convergence, accuracy of the methods, and viability when applied to utility PMU data were assessed to determine suitability to online system operation. Matrix Pencil Method was determined to provide more robust and computationally efficient estimation of key system modes for both simulated and real utility PMU data.
The biorthogonal discrete wavelet transform, which can correlate frequency data to a time-domain solution, was utilized in Chapter 3 to create a methodology for event detection and classification for a subset of selected events. The derived methodology was shown to be effective for identification and classification of load and capacitor switch events, as well as breaker operation and faults.
Methods to mimic the power flow Jacobian from discrete measurements are derived to assess system stability and eigenvalues in Chapter 2. These methods were effective for fast detection of unstable system parameters. Chapter 5, the most significant contribution of this dissertation, details derivations of a mathematical reduced system model and power flow Jacobian variants for more robust instability detection, system weak point identification, mitigation techniques, and state estimation capabilities. Considering the functions of all evaluated and developed model-free methodologies, event detection, event classification, detection of poorly damped oscillatory modes, and instability detection and mitigation can be achieved for situational awareness.
Kantra, Sean, "Model-Free Methods to Analyze Pmu Data in Real-Time for Situational Awareness and Stability Monitoring" (2023). All Dissertations. 3460.