Partial Discharge Classification Using Acoustic Signals and Artificial Neural Networks and its Application in detection of Defects in Ceramic Insulators
MetadataShow full item record
Online condition monitoring of critical assets constitutes one method whereby the electrical insulation industry can help safeguard grids through the avoidance of system outages due to insulation failure. This thesis introduces a novel approach for monitoring the condition of outdoor ceramic insulators based on partial discharge (PD) measurements. The presence of physical defects such as punctures, broken porcelain, and cracks will ultimately lead to the initiation of PD activity in outdoor ceramic insulators. In addition to defects, surface discharges such as that caused by corona and dry band arcing are also very common, particularly in wet and polluted outdoor insulators. Such a discharge activity that originates in these kinds of conditions can cause flashover or insulator failure, resulting in power outages. Measuring early-stage discharge activity is thus very important as a means of avoiding catastrophic situations in power networks. The work presented in this thesis involved initial tests conducted to distinguish between different types of controlled discharges generated in the laboratory. The next step was the implementation of an artificial neural network (ANN) for classifying the type of discharge based on selected features extracted from the measured acoustic signals. First, relatively high-frequency acoustic signals are transformed into low-frequency signals using an envelope detection algorithm imbedded in the commercial acoustic sensor. A fast Fourier transform (FFT) is then applied to each low-frequency signal, and finally, 60 Hz, 120 Hz, and 180 Hz are used as input feature vectors for the developed ANN. This initial research was then extended to include testing of the proposed diagnostic tool on a practical insulation system, and outdoor ceramic insulators were selected for this purpose. Three types of defects were tested under laboratory conditions: a cracked ceramic insulator, a healthy insulator contaminated by wetting with salt water, and a corona generated from a thin wire wound to the ceramic insulator. Both a single disc, and three discs connected in an insulator string were tested with respect to these defects. For both controlled samples and full insulators, a recognition rate of more than 85 % was achieved.
Cite this version of the work
Satish Kumar Polisetty (2019). Partial Discharge Classification Using Acoustic Signals and Artificial Neural Networks and its Application in detection of Defects in Ceramic Insulators. UWSpace. http://hdl.handle.net/10012/14415