Preprocessing of acoustic emission signals from partial discharge in oil-pressboard insulation system

Partial discharge (PD) is one of the major contributors of the problem in high voltage electrical system. It can cause the insulation of a high voltage equipment to fail and lead to catastrophic incident. In a power transformer, detection of PD using acoustic emission technique has been gaining popularity due to its nonintrusive application and capability of locating PD sources. In this research work, acoustic emission (AE) detection system is used to detect PD in an experimental tank filled with transformer oil. Three different types of PD sources to generate PD in the experimental tank are created from pressboards which are the plain pressboard, the floating metal in the pressboard and the bubble in pressboard. Several samples of AE signals due to the occurrence of PD from the same discharge source are captured and recorded. In order to characterize the different types of PD sources, five features or descriptors were extracted from the Short-Time Fourier Transform spectrogram of the AE signals. Then, the preprocessing of the AE signals are carried out from the extracted features using Self-Organizing Map (SOM) Neural Network. Finally, the characteristics of the AE signals from the acquired samples can be obtained and the outlier samples can be determined.