Localization of partial discharge in power transformer winding using electromagnetic time reversal and reciprocity

Abstract

Power transformers are key elements of power system infrastructure. Their efficiency and reliability influence the overall performance of electric power systems. For this reason, proper maintenance of power transformers is essential for operators to ensure continuous and reliable power delivery to customers. Partial discharges (PD), which are localized electrical discharges in transformer insulation, can lead to insulation failure if they are not detected and accurately localized. To enable early detection and localization of PDs in transformer windings, this work employs the electromagnetic time reversal (EMTR) technique to determine the location of PD sources within transformer windings. The first step of this method is to model the transformer winding in the time domain. In this thesis, the transformer winding is modeled using the multiconductor transmission line (MTL) formulation in the time domain, where each turn of the winding is represented as a conductor in the MTL model. To simulate and solve the MTL equations, a SPICE-based model of the MTL is developed. The model is verified by comparing the input admittance of two different winding configurations (i.e., interleaved-disk and continuous-disk) in the time domain with corresponding frequency-domain results reported in the literature. To simulate PD signals, a damped sinusoidal waveform is used as a current source and injected into the transformer winding. The rise time of this waveform is adjusted to approximate an impulse function. For the localization of internal and inter-turn PDs, the impulse response of each turn is computed to form a set of reference signals, which are then normalized using the L^2-norm.

Each normalized reference signal is subsequently convolved with the time-reversed PD signal measured at the transformer neutral terminal. The PD location is identified based on the position of the maximum peak in the convolution results. For inter-turn PD localization, the same procedure is applied; however, due to the characteristics of inter-turn PDs, the location is determined from the minimum peak of the convolution results. The proposed method maintains its accuracy under noisy conditions, enabling reliable PD localization even at low signal-to-noise ratio (SNR) levels. This robustness is particularly important for practical applications, where measurements at the neutral terminal are often affected by external disturbances and system noise. Overall, the results demonstrate that the EMTR correlation-based approach provides an effective and practical solution for accurate PD source localization in transformer windings, supporting its application in transformer diagnostic systems.