A Technique Based on an Archimedean Copula for the Localization of Partial Discharge in a Transformer Winding

Precise localization of a Partial Discharge (PD) source inside a winding is one of the important tasks for early detection and prevention of catastrophic failure of a transformer. However, due to the highly complex structures of the windings and statistical nature of PD, the exact location of a PD source is quite difficult to discover. In this paper, a technique based on an Archimedean copula is proposed to find the location of a PD source in a transformer winding. The winding responses to the PD signals of known pulse-widths applied at all the sections are taken as reference data. Further, the winding responses, obtained from the PD pulses of random shapes and magnitudes applied across various sections of the winding, are taken as test data. To predict the location of the test signals, the best-fit copula and its parameters are established using the reference and test data. The maximum value of log-likelihood between the reference data and the test signal indicates the location of the PD source. Firstly, simulation case-studies are conducted to demonstrate how the proposed approach can be implemented to locate the PD source. The results of the proposed method are then compared with the time-domain and frequency-domain based correlation techniques. Finally, the proposed method is tested with experimental data acquired from a laboratory scale model of the transformer winding. The simulation and experimental results demonstrate that the proposed approach is capable of achieving an accurate prediction of the location of the PD source.