Classification and localisation of multiple partial discharge sources within high voltage transformer windings

Partial discharge (PD) analysis is a common method for condition monitoring and diagnostics of power transformers, which can be used as a tool for assessing the lifespan of transformers and can detect insulation malfunctions before they lead to failure. This report describes the development of analytical tools for PD activities within HV transformer windings. In most cases, PD will occur in transformer windings due to ageing processes, operational over stressing or defects introduced during manufacture, and different PD sources have different effects on the condition and performance of power equipment insulation system. Therefore, for further analyses, the ability to accurately distinguish between the PD signals generated from different sources is seen as a critical function for future diagnostic systems. Under realistic field conditions, multiple PD sources may be activated simultaneously within the transformer winding. An experiment has been designed to assess different methodologies for the identification and localisation of multiple PD sources within a HV transformer winding. Previous work at Southampton developed a non-linear based technique that facilitates identification of the location of a single PD source within an interleaved winding. It is assumed that any discharge occurring at any point along a winding will produce an electrical signal that will propagate as a travelling wave towards both ends of the winding. This project is concerned with the feasibility of locating several sources simultaneously based only on measurement data from wideband radio frequency current transformers (RFCTs) placed at the neutral to earth point and the bushing tap-point to earth. The proposed processing technique relies on the assumption that the PD pulses generated from different sources exhibit unique waveform characteristics. Due to termination and path taken characteristics, the PD signals will suffer attenuation and distortion during the propagation of the PD signals along transformer windings. Therefore, it will cause changes in the energy characteristics of the PD pulses at both measurement points, which can be used to separate, identify and locate the multiple PDs within an HV transformer winding. Based on analysis of the captured data from experiment, various approaches for identifying multiple PD sources have been assessed. Obtained results indicate that the analysis of absolute energy distributions determined using Mathematical Morphology and the use of OPTICS for clustering will reliably separate PD data from two sources that are simultaneously active within a distributed winding