Distribution network fault section identification and fault location using wavelet entropy and neural networks

• This paper proposes a new 2-stage hybrid method for distribution network fault section identification and fault location.
• Level-5 detail coefficients obtained from wavelet decomposition are used in computing the proposed wavelet-based indices.
• Artificial neural network models are developed using the proposed wavelet transform-based indices as inputs.
• A performance analysis of the developed models trained using the proposed indices is presented.

Fault location in power system distribution networks is especially difficult because of the existence of several laterals/tap-offs in distribution networks. This implies that the calculated fault point can be wrongly estimated to be in any of the laterals. This paper proposes a new hybrid method combining Discrete Wavelet Transform (DWT) and artificial neural network (ANN) for fault section identification (FSI) and fault location (FL) in power system distribution networks. DWT was used in the analysis and extraction of the characteristic features from fault transient signals of the three phase line current measurements obtained at a single substation relaying point, rather than the double-ended approach used in the existing literature. Entropy Per Unit (EPU) indices are afterwards computed from the DWT decomposition, and are used as input to multi-layer ANN models serving as FSI classifiers and FL predictors respectively. The proposed hybrid method is tested using a benchmark IEEE 34-node test feeder. Comparisons, verification, and analysis made using the experimental results obtained from the application of the method showed very good performance for different fault types, fault locations, fault inception angles, and fault resistances. The proposed hybrid method is unique because of the pre-processing stage done with the DWT-EPU indices, the use of only line current measurements from a single relaying point, and the division of the FSI and FL tasks into sub-problems with respective ANN models.

Figure optionsDownload as PowerPoint slide