Classification of partial discharge under different voltages using acoustic emission techniques

The partial discharge is one of the major failure mechanisms for insulation of power equipments subjected to high electrical stress. In order to understand the service conditions of insulating materials, it is very important to investigate on partial discharge characteristic in electrical apparatus. In the present paper, a set of acquisition system for acoustic signal generated by partial discharge is established in laboratory. The acoustic emission signals are generated by pinplane electrodes in oil insulation system in the voltage range of 4-6kV, the distance between electrodes and sensors in the range of 10-15cm. The test results show that voltage applied and distance between sensor and pin-plane regularly influence energy distribution of acoustic signal in different frequency component. The discrete wavelet transform is used to analyze acoustic signal acquired. A feature extraction method to classify partial discharge and a new method to identify different types of acoustic signal based on combined wavelet network and wavelet transform are respectively suggested.

I. Introduction

Partial discharge (PD) is electrical discharges that do not completely bridge the distance between two electrodes under high electrical stress. During last several years, a lot of technologies have been developed to diagnose insulation condition of high voltage apparatus. The representative methods include insulation resistance, gas density, dielectric loss and partial discharge measurements [1],[2]. All these methods mentioned above should be used in offline state. So they can only be applied to periodic diagnosis of apparatus. At present, gas chromatography and acoustic emission (AE) methods have found their technical application in estimating electrical discharges.