Bushing Fault Detection and Diagnosis using Extension Neural Network

Author

Vilakazi, Christina B. ; Marwala, Tshilidzi

Author_Institution

Sch. of Electr. & Inf. Eng., Univ. of the Witwatersrand, Johannesburg

fYear

0

fDate

0-0 0

Firstpage

170

Lastpage

174

Abstract

This paper proposes an extension neural network (ENN) based bushing fault detection and diagnosis. Experimentation is done using dissolve gas-in-oil analysis (DGA) data from bushings based on IEEEc57.104, IEC599 and IEEE production rates methods for oil impregnated paper (OIP) bushings. The optimal learning rate for ENN is selected using genetic algorithm (GA). The classification process is a two stage phase. The first stage is the detection which identifies if the bushing is faulty or normal while the second stage determines the nature of fault. A classification rate of 100% and an average of 99.89% obtained for the detection and diagnosis stage, respectively. It takes 1.98s and 2.02s to train the ENN for the detection and diagnosis stage, respectively

Keywords

IEEE standards; bushings; condition monitoring; fault diagnosis; genetic algorithms; insulating oils; learning (artificial intelligence); neural nets; paper; pattern classification; power engineering computing; IEC599 rate method; IEEE production rate method; IEEEc57.104 rate method; bushing fault detection; classification process; dissolve gas-in-oil analysis data; extension neural network; fault diagnosis; genetic algorithm; oil impregnated paper bushing; optimal learning rate; Condition monitoring; Dissolved gas analysis; Fault detection; Fault diagnosis; Hydrogen; Insulators; Neural networks; Oil insulation; Petroleum; Power transformer insulation;

fLanguage

English

Publisher

ieee

Conference_Titel

Intelligent Engineering Systems, 2006. INES '06. Proceedings. International Conference on

Conference_Location

London

Print_ISBN

0-7803-9708-8

Type

conf

DOI

10.1109/INES.2006.1689363

Filename

1689363