PERFORMANCE STUDY OF VARIABLE SHORT-TIME FOURIER FOR TRANSIENT DETECTION IN POWER SYSTEMS

Abstract

This paper investigates the efficiency of time-frequency analysis in locatingtransients in power systems signals. Two experimental cases were simulated by the Matlabsoftware. The first case is a linear system and the second is a three-phase synchronousgenerator. The transient analysis was performed by a Fourier technique called variableshort-time Fourier transform, which uses a varying window size over time that is adjusted bylocal kurtosis given by the fourth normalized cumulant. The results obtained have evidencedthe changes in frequency regime in both cases. The kurtosis defined by the fourth normalizedcumulant effectively captures the nonlinearities present in both scenarios. The technique usedproved to be able to identify and classify transients over time.