Data Mining on Partial Discharge Signals of Power Transformer’s Defect Models

Please enable javascript in your browser.

Parvin Darabad, Vahid | 2013

870 Viewed

Type of Document: Ph.D. Dissertation
Language: Farsi
Document No: 45685 (05)
University: Sharif University of Technology
Department: Electrical Engineering
Advisor(s): Vakilian, Mehdi; Phung, Bao Toan; Blackburn, Trevor
Abstract:
In this thesis the goal is to use data Mining techniques for finding features of different Partial discharges happen in power transformer insulation defect models in order to monitor the insulation condition of in-service power transformers continuously and on-line and to identify any type of insulation defect in power transformers at the early stage of formation.For this purpose, power transformer Insulation defects are modeled physically and partial discharge current pulses are recorded as Partial discharge data.Tow format of data are investigated, in first one, recorded data in one power frequency cycle are explored and in second one, individual PD pulses are considered for study.In this thesis, features like statistical, texture, FFT and cepstral are employed on Data. In addition, classification algorithms such as decision tree, neural network, k-nearest neighbor and support vector machines are used. Self-organizing Map as a clustering algorithm is also employed.
Results showed that the highest accuracy of classification would be obtained by using ANN algorithm on cepstral feature space which are extracted from partial discharge pulses
Keywords:
Data Mining ; Power Transformer ; Partial Discharge ; Insulation Defect Models

No TOC