ANN and wavelet entropy based approach for fault location in series compensated lines

• Published in: 2016 International Conference on Microelectronics, Computing and Communications (MicroCom) pp. 1 - 6
• Main Authors: Singh, Sunil, Vishwakarma, D. N.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.01.2016
• Subjects: Artificial neural networks; Biological neural networks; Cascade-forward neural network; Circuit faults; Entropy; Feature extraction; Feed-forward neural network; Generalized regression neural network; Norm entropy; Wavelet coefficients; Wavelet transform
• DOI: 10.1109/MicroCom.2016.7522557

This paper presents a novel approach based on combined wavelet transform and artificial intelligence technique for estimating fault location in a series compensated transmission line. In proposed approach the samples of faulty current signals generated from simulink model are used for fault analysis. Wavelet transform is utilised for the purpose of feature extraction from the faulty current signals. The fault current signals are decomposed using Db5 mother wavelet. Features of faulty signals are extracted in terms of standard deviation and norm entropy value of the coefficients and are fed to designed artificial neural network (ANN) models for fault distance estimation. The paper also presents a comparison of error in estimating the distance of fault by different neural network i.e. Feed-forward, Cascade-forward and generalized regression neural network (GRNN). The preciseness and workability of the proposed algorithm has been evaluated on a 400 KV, 300 km series compensated transmission line for different fault cases using MATLAB simulation. The results acquired, indicate that the proposed approach can reliably located the faults points in series compensated transmission line with high accuracy.