Series Arc Fault Characteristics Based on Gray Level-Gradient Co-Occurrence Matrix

In order to obtain the characteristics of series arc fault in the nonlinear load circuit, arc fault experiments were carried out with self-developed experimental system under both converter and industrial computer load conditions. A kind of feature extraction method based on gray level-gradient co-occurrence matrix was proposed. Firstly, the current signal was preprocessed by using forward difference method. The obtained signal was decomposed and reconstructed by using wavelet packet. The reconstructed signal was put into a two-dimensional array according to signal frequency. Secondly, the energy in each frequency band at the same time was normalized. And it was converted to a gray image with gray value range from 0 to 255. Thirdly, the gray image was filtered with Wiener filter method and enhanced with Laplace operator. The gray level-gradient co-occurrence matrix was solved from the image. The signal frequency of the image is higher than 1 562.5Hz. Finally, fifteen kinds of features were calculated with the co-occurrence matrix and the typical characteristics of arc fault were selected. The arc fault identification tests were carried out by using support vector machine (SVM). The input vector of SVM was the selected characteristics. The tests verified the validity of the proposed feature extraction method. © 2018, Electrical Technology Press Co. Ltd. All right reserved.