High speed offset free distance relaying algorithm using multilayer feedforward neural networks

Abstract

A new high-speed digital distance relaying algorithm using artificial neural networks is presented. The proposed artificial neural network models are trained with the input patterns of distorted voltage and current signals passed a lowpass filter and with the target patterns of real and imaginary components of DC-offset free current and voltage signals conditioned by a DC-offset removing algorithm. The artificial neural networks play the roles of a DC-offset removing filter and a Fourier filter. In accordance with a series of tests, the relay operating time is less than 9 ms after faults in a 80 km, 154 kV, 60 Hz overhead transmission line system.