Power cable fault feature extraction based on wavelet and segmentation

To extract the features of the online power cable fault information for the fault recognition, a new method which combines the wavelet transform and segmentation is presented in this paper. In the method, the power cable fault information would be broken into segments, and then those segments are processed by wavelet packet decomposition. After that, the weighted sum of decomposed wavelet coefficients is calculated by using the logarithm values of the summations to build the new fault features. Finally, power cable fault model is built and Fisher criterion is used to measure the performance of the new features compared with the original features. Simulation experiments show that the new features have the better performance which is evident from the Fisher values. The within-class distances of the single-phase faults have the smaller Fisher values which are in favor of the latter classifier design and the fault recognition accuracy.