Partial discharge location using a hybrid transformer winding model with morphology-based noise removal

Abstract

This paper employs a hybrid model of a power transformer winding for accurate partial discharge (PD) location. The hybrid model greatly reduces the computation complexity of the multi-conductor transmission line (MTL) model, yet retains the ability to accurately represent a winding. Having obtained the transfer functions of the hybrid model, PD signals can be calculated from both line-end and neutral-end at a set of detecting points. Subsequently, the pairs of the calculated signals are compared and a PD is located at the point where the minimum normalized Euclidean distance or F-ratio of the two signals is detected. In order to enhance computation efficiency a two-round technique, which includes a coarse round and a precise round, is developed and a guideline for section division is given. The influence of noise introduced by measurement devices has also been considered. To reduce the noise meanwhile preserve the feature waveforms, an adaptive morphological filter is employed. Simulation studies have shown that it is accurate and feasible to perform PD location using the hybrid model even with the presence of Gaussian noise.