Subcycle Voltage Dip Classification Using Matrix Pencil Method With Ellipse Fitting Algorithm

Abstract

We have developed a technique using the matrix pencil method (MPM), which was augmented with an ellipse fitting algorithm, that can classify three-phase voltage dips only using subcycle voltage data. This technique uses the MPM to extract the fundamental frequency components and reconstructs the fundamental voltage space vector that often changes into an ellipse during a fault. This ellipse's parameters are then estimated by an ellipse fitting algorithm to classify the dip. We have demonstrated that this technique can differentiate between highly distorted and similar voltage dips only using a quarter-cycle of data, which is equivalent to 5 ms in a 50-Hz alternating-current system. This method is further improved by prefiltering and downsampling the data to reduce the computation time and is implemented on a National Instruments CompactRIO platform. This system classified the dip in real time at about 10-12 ms after fault detection when tested with reproduced fault voltage waveforms amplified to a voltage level of 415 V. Our method thus executes faster than the conventional method of the fast Fourier transform that requires at least a full fundamental cycle time of 20 ms. This technique serves as a method of analyzing subcycle voltage dip phenomena and can be potentially extended to other fast subcycle electrical events.