Adaptive Threshold for Electrical Disturbances Segmentation

Abstract

The detection of the electrical signal discontinuities in the oscillographies recorded in substations and/or points of common coupling allows their segmentation, which is crucial for implementation of automated methods for detection, classification, location and storage by classes of disturbances in electric power systems. In this context, this study provides a way of determining an adaptive threshold that allows the segmentation of voltage or current signals based on Wavelets decomposition. The disturbances considered in this work were the short-duration voltage variations, impulsive and oscillatory transients, and harmonic distortions. The signals were synthetically generated. Moreover, white noise was added on the signals. Thus, a Symlet Wavelet was applied to the signals in order to denoise them. In the sequence, a Daubechies Wavelet was used to decompose the filtered signals. So, to determine the initial and final points of each segment, an adaptive threshold was established based on the energy and entropy of energy for the second level of decomposition. Thus, the number and position of each segment were determined according to the intersections of the detail curves and the thresholds found.