Study of Transformer´s PD Sources Number Estimation and Falsity PD Sources Elimination

The PD phenomena in transformer oil/paper insulator were investigated experimentally with (UHF) partial discharge detection technique. PD in the transformer could be detected on-line by UHF technique. It is hopeful that problem of anti-interfere in transformer on-line PD detection could be settled with UHF partial discharge detection technique. Using electromagnetic vector sensor (EMVS) to detect PD´s UHF signal can realize the PD differentiation and location. However, owing to the complexity of transformer insulation, UHF signal radiated from PD may be refracted and reflected many times in its transmission, may be also diffracted by windings, namely, there will be two or more path UHF signals enter into the EMVS. Hence, falsity PD sources may be appearing in PD sources number estimation by adopting MDL rule of information theory. To eliminate the falsity PD sources, in this paper, a new algorithm is presented, which is using smoothed rank sequence (SRS) method to estimate the number of signal sources and identify their correlation configuration. According to the correlation between different path signals radiated from the same PD source, we can judge these signal sources correlated each other be come from the same PD source and eliminate the falsity PD sources. Furthermore, the effect of windings on the UHF signal is also analysed. The simulation experiment results show that the SRS method can not only estimate the number of signal sources but also identify their correlation configuration.