A wavelet packet transform-based technique for three phase power transformer protection

Saleh, Saleh A. M. (2003) A wavelet packet transform-based technique for three phase power transformer protection. Masters thesis, Memorial University of Newfoundland.

[English] PDF (Migrated (PDF/A Conversion) from original format: (application/pdf)) - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (15MB) [English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. (Original Version)

Abstract

The protection of power transformers has been always a challenge due to the problem of distinguishing between the magnetizing inrush and different internal fault currents. There have been many proposed techniques to carry out the protection of power transformers. These techniques are based on the harmonic content of the differential current, transformer equivalent circuit, voltage and flux restraint, pattern recognition using trained artificial neural network, etc. These techniques have either limitation or some assumptions to validate their performance that may affect their reliability, speed of operation, complexity of the required computations, validity under transformer parameters changes and/or change in type of core material. In this work, a wavelet packet transform based algorithm for differential protection of three phase power transformers is developed, implemented and tested on-line for two different three phase power transformers using DS1102 digital signal processing board. The wavelet packet transform based algorithm is used to detect and classify various disturbances, in particular, the magnetizing inrush and internal fault currents with neither a harmonic analysis nor any dependency on the power transformer parameters, ratings, loading conditions and/or winding configurations. Wavelet transforms can create a multi resolution analysis, where the frequency contents of the analyzed signal are divided into octal bands beginning from the highest frequency component, ending with the lowest one and preserving the time location of these frequency components. This analysis is carried out using a cascaded filter banks associated with the selected mother wavelet and the number of levels of resolution. Both optimal mother wavelet and number of levels of resolution can be selected using the minimum description length criteria, that is based on the minimum entropy calculations. Applying any form of wavelet transforms produces details, which are related to high frequency components, as well as approximations, which are related to low frequency components at each level of resolution. In the case of applying the wavelet packet transform, the details are downsampled by 2 and filtered with the same filter bank at each level of resolution. An experimental setup is developed for collecting various differential currents waveforms for purposes of selecting the optimal mother wavelet and the number of levels of resolutions as well as off-line testings. The Daubechies (db4) mother wavelet with two levels of resolution are found to be optimal in providing adequate information to discriminate the magnetizing inrush current from internal fault currents. The proposed algorithm is tested off-line using Matlab on a data collected from a three-phase core type power transformer. An experimental setup is developed to accommodate the on-line testings, which are carried out on two different power transformers. The results of these tests show no difference in its ability to distinguish between the different magnetizing inrush and various internal fault currents. Also, the on-line testing results are consistent with those obtained from off-line testings in terms of reliability, accuracy, speed of response and applicability.

Actions (login required)

View Item