Classification Of Partial Discharge Signals Through Acoustic Sensor On the Basis Of Propagation Paths

This study focuses on the vital task of detecting Partial Discharge (PD) signals inside a power equipment through acoustic PD detection technique. A major problem in the analysis of acoustic PD is that the signals are often received by the sensor after going through multiple reflections. In this paper, a method has been proposed to ascertain if the acoustic signals have reflection while their journey towards the sensor or not. Experimental studies were conducted in the laboratory by producing artificial PD inside a metal tank. Inside the tank transformer oil was filled and acoustic signals are generated due to partial discharge and this was recorded by a sensor kept on the outside part of the transformer tank wall in the presence and absence of a pressboard barrier in the path between source and tank wall. From the recorded data, statistical features were extracted and they are used to train an XG boost-based machine learning model. The testing results demonstrated an impressive accuracy of 90% in classifying PD, based on the propagation paths, showcasing the effectiveness of the proposed methodology.