Detection of arcing fault in underground distribution cable using artificial neural network

Abstract

Arcing faults can cause substantial damage if they are not detected and isolated promptly. Detection of arcing faults has always been a difficult issue. Those faults tend to be of high fault resistance and hence the fault current is well below maximum load limit and its detection is not possible through the use of overcurrent relays. In the case of overhead lines, the gas generated through arcing is dispersed rapidly. But in the case of underground cables, the generated gas could travel along cable duct and could result in explosion at manhole location, which is dangerous to personnel. The damage can be reduced if arcing faults are detected before they develop into major faults. The general aim of this study is to develop an arcing fault detection algorithm which can detect the presence of arcing fault in underground distribution cable. Arcing faults data are collected through simulations and experiments. The simulations involve the modelling of a simple underground distribution system and two TNB underground distribution systems using Power System Computer Aided Design 1 Electromagnetic Transient for Direct Current (PSCADIEMTDC) program. On the other hand, the experiments are conducted in research laboratory. The data collected from the simple underground distribution system are analysed in both time domain and frequency domain to identify the characteristics of arcing fault. A Multi-layer Perceptron (MLP) with Backpropagation (BP) learning is used to discriminate arcing faults from normal load condition. The detection results revealed satisfactory performance in all test cases.