Application of Matrix pencil method in sub-cycle voltage dip classification

Abstract

Matrix Pencil Method (MPM) has been applied to sub-cycle information of space vector and zero-sequence voltage to classify voltage dip phenomena. The feature extraction performance of MPM using different sampling window width has been statistically analyzed and found to perform relatively well between 0.1 and 1 cycle of a damped complex sinusoidal signal. The results show that MPM is able to estimate the fundamental frequency space vector even in highly distorted signals. The space vector ellipse's estimation is further enhanced by augmenting an ellipse fitting algorithm to MPM. This enhanced method differentiated between two highly similar two-phase voltage dips using only a quarter-cycle of data, demonstrating the feasibility of this scheme. This is demonstrated in a fault simulation on IEEE 34-bus system using a 5 ms sampling window.