A feedforward Artificial Neural Network approach to fault classification and location on a 132kV transmission line using current signals only

Transmission lines represent a major part of an electrical power system network but due to their long lengths and direct exposure to climate conditions, they are more prone to faults as compared to other power system components. The aim of this paper is to develop fast, reliable and accurate fault classification and location algorithms that can efficiently locate faults on transmission lines and thus reduce outage time. The algorithms have been implemented using feedforward Artificial Neural Networks (ANN) given their good generalization characteristics. Current signals measured at one end of the line only have been used as the inputs to the ANN algorithms since current transformers are always present at each end of the line for measurement and protection purposes while voltage transformers may sometimes be omitted for economic reasons. The test system is a 132 kV transmission line model based on the electrical power system network in Mauritius. The fast Fourier Transform (FFT) has been adopted for feature extraction since it is fast and easy to implement. Finally, the sensitivity of the algorithms to changes in fault inception angle, fault impedance and the length of the transmission line have been investigated.