Data Anomaly Resilience in Wide-area Oscillation Monitoring and Assessment of Damping Contribution from Power System Components
Open Access
- Author:
- Chatterjee, Kaustav
- Graduate Program:
- Electrical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 15, 2022
- Committee Members:
- Yan Li, Major Field Member
Greg Pavlak, Outside Unit & Field Member
Jing Yang, Major Field Member
Nilanjan Ray Chaudhuri, Chair & Dissertation Advisor
Kultegin Aydin, Program Head/Chair - Keywords:
- Power Systems
Oscillation Monitoring
Control Systems
Damping
Robust PCA
Data Recovery - Abstract:
- In bulk power systems, sustained low-frequency oscillations may appear due to the poorly-damped natural modes or as a result of external periodic disturbances. The poorly-damped modal oscillations may come from increased loading over tie-lines, improper tuning of controllers, and fast-acting generator excitation systems, among others. Such oscillations induce mechanical stress on the turbines, limit utilization of existing transmission assets causing congestion, and in the worst case, if not adequately damped, may lead to large-scale outages. An important first-step in preventing these is to detect the poorly-damped oscillations through mode meters and identify the sources contributing towards the deterioration of the system damping. This necessitates a system-wide monitoring system, which can quantify the mode-wise damping contribution from different power system components in real-time using the GPS time-synchronized voltage and current measurements obtained by phasor measurement units (PMUs). However, owing to the chances of corruption in PMU channels arising out of bad data outliers, missing data packets, and malicious injections from cyber attacks, the accuracy and reliability of the oscillation monitoring system may be compromised. This highlights the need for a data pre-processor to be a part of the monitoring system for online detection and correction of anomalies in the PMU data streams. Thus motivated, this dissertation presents the theoretical foundations and approaches for bad data resilience in wide-area oscillation monitoring with focus on quantifying mode-wise relative damping contributions from synchronous generators and power electronics-interfaced transmission devices. First, an online data pre-processing approach is proposed for ensuring data anomaly resilience. The proposed data pre-processor is capable of detecting outliers and other bad data in measurements; separating event-induced outliers, like fault outliers, from bad data outliers; and recovering true data from anomalous measurements using a fast online algorithm. The proposed approach leverages the spatio-temporal correlation and low-rankness of the true data and premises on the assumption that data anomalies are spatially sparse. The theoretical guarantees on the success of data recovery, in presence of any bad data with a given sparsity, are also derived. In this context, analytical connections are established between the denseness of a signal group and their modal observabilities. It is shown that grouping signals by minimizing variance in the modal observabilities for each poorly-damped mode, enhances denseness of the low-rank signal subspace and helps attaining the sufficiency condition that guarantees data recovery. Next, the theories and methods for assessing damping contribution from different power system components using the output of the pre-processor are presented. To that end, the synchronous generators, forming the bulk of today's generation systems, are analyzed first. Intuitively, the rate of dissipation of modal energy in a device gives a measure of the device's ability to damp the mode. However, without detailed knowledge of the generator model and network parameters, in a multimachine system, it is difficult to estimate the expressions of energy functions and their rates of dissipation. To circumvent this, for each generator it is proposed to estimate the damping power for the mode of interest, by estimating the relative modeshapes of damping torque and rotor speed from the measurements of real power output and bus voltage angle, respectively. To support this proposition, a mathematical proof showing the mode-wise equality of damping power and average rates of oscillation energy dissipation in a multimachine system is presented. For estimation, a recursive Kalman filter-based approach is proposed. In a power system, damping contribution may also come from the transmission systems with active power electronics-interfaced devices. With that in mind, the rates of dissipation of oscillation energy in the flexible ac transmission systems (FACTS) and voltage source converter-based high voltage dc (VSC-HVDC) transmission systems are also studied. For each of these devices, their damping contributions are assessed by comparing the average slopes of their transient energy expressions constructed from terminal measurements, for different control strategies and controller parameters. The operating scenarios under which these devices could worsen the overall damping by injecting oscillation energy into the rest of the system, are presented. Finally, the design conditions for enhancing dissipativity of the energy functions are also suggested. In the end, a real-time laboratory testbed is presented for running control-hardware-in-loop experiments relating to data anomaly resilience and oscillation monitoring. The testbed, built so far as a part of this doctoral research, comprises of two real-time digital simulators from RTDS and OPAL-RT, and a hardware PMU, a GPS-clock, and a real-time automation controller from SEL. A few of the algorithms presented in the thesis, have been simulated in the testbed for validating their real-time performance.