Classification of partial discharge transient patterns using texture analysis algorithms

This paper, for the first time, evaluates the ability of four texture analysis algorithms for classifying partial discharge (PD) sources. The algorithms which have been examined are the spatial grey level dependence method (SGLDM), the grey level difference histogram method (GLDKM), the grey level run length method (GLRLM) and the power spectral method (PSM). A comparison of these algorithms with a conventional method for PD classification has been carried out. The optimum number of features of each algorithm have been selected according to the classification accuracy of each feature as well as by using the transformed divergence analysis. The maximum likelihood classifier has been used for classification. Results revealed that these algorithms offer good discrimination among different types of defects.