Partial Discharge Source Classification Based on Machine Learning and PRPD

A method for analyzing partial discharges is proposed to detect faults and evaluate the operational status of electrical system assets. Initially, the effectiveness of the method was hindered by electromagnetic interferences affecting the sensor, including corona, surface discharges, and radio signals. To address this challenge, multiple layers of signal processing were introduced. This paper proposes a new methodology for noise removal and classification of partial discharge sources based on clustering pulses by attributes derived from their shape. Consequently, it became feasible to distinguish partial discharges from the various sources of interference present in the measurements. The methodology was tested using a dataset of partial discharge measurements obtained from Current Transformers at an Eletrobras group substation in Brazil. Through this testing, the methodology proved capable of assessing the status of the analysed equipment.