Robust Three-Stage Dynamic Mode Decomposition for Analysis of Power System Oscillations

A spatio-temporal method based on Robust Three-stage Dynamic Mode Decomposition (RTDMD) is proposed to improve the modal parameter estimation of synchrophasor data in the presence of noise, missing data and outliers. The proposed method is based on the PMU correlated spatio-temporal data to estimate missing information using an iterative singular value decomposition (SVD) combined with the adjustment of the reduced Koopman operator. Moreover, it is proposed to identify and correct outliers using an inherent statistical extension of the SVD-based Mahalanobis distance. As a result of the proposed enhancements, modal parameters such as damping and power system oscillation frequency are extracted as a byproduct of the RTDMD algorithm process. Data resulting from real-time digital simulators (RTDS) with self-clearing three-phase fault and Phasor Measurement Units (PMUs) of Low-Frequency Oscillation (LFO) incidents at hydroelectric power plants are used to illustrate the effectiveness of the proposed method over other well-established methods.