Neural network approach to fault detection in electric power systems

This paper describes a fault detector that uses artificial neural networks (ANN). It represents the first step to the development of a neural distance relay for protecting transmission lines. We envisage the fault detection problem as a pattern classification process. Our suggested approach is based on the fact that when a fault occurs, a change in the system impedance takes place and, as a consequence, the current phase and amplitude change. The ANN-based fault detector is trained to detect this changes as indicators of the instant of fault inception. Results showing the performance of the fault detector are presented in the paper, indicating that it is fast, robust and accurate