The DC Arc Fault Detection Method Taken Advantage of WT and MFE

Abstract

Compared with AC arc faults, there isn’t zero-crossing points in the current waveform when the DC arc faults occur. Dc arc fault brings great harm to the safe operation of power supply system. Wavelet transform (WT) is suitable for analyzing nonstationary signal, and multi-scale fuzzy entropy (MFE) is of excellent performance in detecting the uncertainty and complexity of the signal. The random fluctuation and uncertainty of current will be greatly enhanced when arc faults occur. This paper aims to elevate the property of detection of dc arc faults, WT and MFE are utilized to construct the fault features. Least squares support vector machine (LSSVM) is employed to be as the classifier to make the detection of dc arc faults. The result of the experiment shows the availability of the method this paper proposed.