Candidate: Satish Kumar Polisetty
Title: Partial Discharge Classification Using Acoustic Signals and Artificial Neural Networks and its Application in Detection of Defects in Ceramic Insulators
Date: December 5, 2018
Time: 10:00 AM
Place: EIT 3141
Supervisor(s): Jayaram, Sheshakamal - El-Hag, Ayman A. (Adjunct)
Abstract:
On-line condition monitoring of critical assets is one of the ways the electrical insulation industry can contribute to safeguard the grids by avoiding system outages due to insulation failure. A novel approach for monitoring the condition of outdoor ceramic insulators based on the partial discharge (PD) measurement is demonstrated in this thesis. Presence of physical defects like punctures, broken porcelain, cracks will ultimately lead to initiation of PD activity in outdoor ceramic insulators. Apart from defects, surface discharges like corona and dry band arcing are very common, particularly in wet and polluted outdoor insulators. These discharge activities originated from such cases may lead to flash-over or cause insulator failure and cause power outages.
Hence, it is very important to measure the discharge activities in early stage to avoid catastrophic situations in power networks. In the presented thesis work, initial tests are conducted to distinguish different types of controlled discharges generated in the laboratory using commercial acoustic sensor. Artificial neural network (ANN) has been implemented to be able to classify the type of discharge based on selected features extracted from the measured acoustic signal. First the relatively high frequency acoustic signal has been transformed into a low frequency signal using imbedded envelope detection algorithm in the commercial sensor. Then, fast Fourier transform (FFT) has been applied on the low frequency signal where 60, 120 and 180 Hz have been used as an input feature vector for the developed ANN. The research has been extended to test the proposed diagnostic tool on practical insulation system. Outdoor ceramic insulators have been selected to achieve this goal. Three types of defects have been tested under laboratory conditions, i.e. cracked ceramic insulator, healthy insulator wetted with salt water and corona generated from a thin wire attached to ceramic insulator. Both single disc and three discs connected in an insulator string have been tested with the aforementioned defects. Recognition rate more than 85% has been achieved for both the controlled samples and full insulators.