Research of Arc Fault Detection Based on Wavelet Entropy

There will be many high frequency signals next to current zero when the arc fault occurs; these brief high frequency mutation signals were extracted by wavelet transform and its wavelet energy entropy was put forward to reflect arc fault characteristic information. Then arc fault was effectively identified by using the least squares support vector machine (LS-SVM) where wavelet entropy was classified. Identification results show that arc fault was entirely recognized under the experiment conditions, it is conclude that this method can not only identify arc fault for the single loads, but also can avoid disturbances which generated by some inhibitory loads.