Neural Network based Transmission Line Fault Classifier and Locator using Sequence Values

Abstract

Faults can easily occur at transmission line because the line is exposed to the environment. The fast fault location after a fault occurrence will minimize the time to repair the faulty part thus reduce the stress of power system due to long outage time. This paper demonstrates the development of fault classifier and fault locator using neural network. The positive, negative and zero sequence values of three-phase voltage and current during fault time were used as the inputs to train the neural networks. Various fault conditions which have different fault types, fault resistance values and fault locations have been simulated to generate the training data for both neural network based fault classifier and fault locator. From the results, the fault classifier and locator successfully classified and located all the simulated fault conditions. One of the advantages of using sequence values as the inputs for the neural networks is only one fault type need to be simulated and used as the training data for single line-to-ground fault (RG, YG and BG), line-to-line fault (RY, YB, BR) and double line-to-ground fault (RYG, YBG, BRG).

Fault detection through discrete wavelet transforms and radial basis function neural network in shunt compensated distribution systems

“…A strategy has been demonstrated to diagnose faults in the contact line between different partitions of the large power grid based on RBFNN. The accuracy of the method has been verified independently or under complex faulty conditions [19]. A relay protection logic has been implemented for different values of fault incident angle and resistance through wavelet transforms in a hybrid compensated transmission line compensated through a series capacitor and a static synchronous series compensator with energy storage (SSSC-ES) and demonstrated wavelet transform accuracy for fault detection [20].…”