An improved MSD-based method for PD pattern recognition

In this paper a new method for multi source partial discharge (PD) pattern recognition by discrete wavelet transform, is presented. The proposed method is based on the application of multi-resolution signal decomposition (MSD) technique applied to a 3D PD pattern image. The multi-resolution technique has shown that some detail images (horizontal, vertical and diagonal) at different decomposition levels have frequencial characteristics typical of a single discharge phenomenon. By applying this method to a double-source PD pattern it is observed that one source prevails on the other. Taking advantage from the linearity of the biorthogonal wavelet functions by using a removal and reconstruction process, it is possible to eliminate the first characterized discharge phenomenon, so obtaining the pattern of the second source. Tests have been performed on specimens having single and multiple defects. The obtained results prove that the proposed improved recognition method is quite efficient and accurate.