DATA-DRIVEN DIAGNOSTICS OF ISSUES RELATED TO POWER SYSTEM DYNAMICS USING PMU MEASUREMENT
1 online resource (156 pages) : PDF
University of North Carolina at Charlotte
This work investigates the use of data driven techniques to diagnose issues related to power system dynamics. The source behind the data are simulated Phasor Measurement Unit (PMU) measurements. First, this study examines the application of Weighted Support Vector Machine (WSVM), ensemble WSVM and Adaptive Neuro Fuzzy Inference System (ANFIS) for prediction of post-fault transient stability condition. The performance of the ensemble classifier is compared with other methods for accuracy of prediction. The method is tested on the IEEE 39-bus test system. Second two methods are introduced for predicting the voltage and rotor angle stability status of a power system instantly after a large disturbance. Generator voltages and angles gathered instantly after clearing a fault are used as inputs of Feedforward Neural Networks to classify stability status of a system. Grey Wolf Optimizer (GWO) and Particle Swarm Optimization (PSO) are applied as training methods for Feedforward Neural Networks. The performance of two methods are determined by applying the two methods on the IEEE 39 bus test system. Results show that the ensemble method achieves the highest accuracy using rotor angles of generators and voltage magnitudes after fault clearance. The resulting accuracy of GWO and PSO are compared. Examination showed that the applied methods can predict the stability status 30 cycles after fault clearance. This study also presents a model-free method for detecting coherency of generators in power systems by means of the Wavelet Packet Decomposition and the Recurrence Quantification Analysis (RQA). The time-series rotor angles of generators are used in this approach to detect coherent group of generators. Noise is a critical issue that needs to be considered when evaluating the performance of the coherency detection methods based on monitored data. This work also focuses on a method to detect low-frequency oscillations and identify frequency modes by applying measured data from PMUs. An algorithm based on reduced dimensionality of the data for detection of oscillations is proposed. The Slow Feature Analysis is applied to extract low-dimensional features from the PMU data. Based on the RQA, two thresholds are applied to identify low-frequency oscillations by using the slow features. The recurrence-derived Fourier Transform is applied to determine frequency modes. A 29-machine 179-bus system is considered for the study. Studies on the system shows the effectiveness of the proposed methods. Finally, a method is proposed for disturbance event detection and classification by applying the RQA. To evaluate the characteristics of location and type of disturbance events, nonlinear measures of the power system, such as voltage and frequency, are examined based on recurrence plots. For dimensionality reduction, the Principle Components Analysis is used. An unsupervised clustering method is applied to determine two types of disturbance events, which are fault and generator tripping. Simulations conducted on the 29-machine 179-bus system model and results reveal that the RQA might be an effective tool to identify the location and type of disturbance events.
Kakad, YogendraCecchi, ValentinaHong, Tao
Thesis (Ph.D.)--University of North Carolina at Charlotte, 2018.
This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s). For additional information, see http://rightsstatements.org/page/InC/1.0/.
Copyright is held by the author unless otherwise indicated.