Artificial Neural Network-Based Fault Identification for Grid-connected Electric Traction Network

Identifying the fault type and faulted phase prior to protection coordination and restoration of the remaining healthy part of the utility power grid in the presence of railway traction load is an important process to ensure power supply system reliability of the grid-connected traction network. An...

Full description

Saved in:
Bibliographic Details
Published in:IEEE access Vol. 12; p. 1
Main Authors: Myint, Shwe, Dey, Prasenjit, Kirawanich, P., Sumpavakup, Chaiyut
Format: Journal Article
Language:English
Published: IEEE 01.01.2024
Subjects:
Accuracy
ANN classifier
Backpropagation
Bayes methods
Bayesian Regulation backpropagation
Daubechies-6 mother wavelet
Discrete wavelet transforms
Fault diagnosis
Fault identification
Karrenbauer transform
Power system reliability
Protection
Reliability
Traction power supplies
Traveling wave
Wires
ISSN:2169-3536, 2169-3536
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Identifying the fault type and faulted phase prior to protection coordination and restoration of the remaining healthy part of the utility power grid in the presence of railway traction load is an important process to ensure power supply system reliability of the grid-connected traction network. An artificial neural network (ANN) based fault classifier has been proposed. The input features to the classifier are derived from multiple detail coefficients of modal current traveling wave signals using the three-level discrete wavelet transform (DWT) with the Daubechies-6 mother wavelet (db6). The Bayesian regularization backpropagation as a supervised machine learning algorithm performs through more than a thousand fault scenarios. The robustness of the proposed DWT-ANN algorithm is verified by testing with the IEEE 9-bus network connected with the large railway traction system through MATLAB Simulink simulations. The superiority in fault identification performance of the proposed algorithm is evident with the highest accuracy of 100% when compared with similar methods.
ISSN:2169-3536
2169-3536
DOI:10.1109/ACCESS.2024.3489802