The artificial neural networks based relay algorithm for distribution system high impedance fault detection

The paper presents a practical artificial neural network (ANN) based relay algorithm for electric distribution high impedance fault detection. The scheme utilizes the characteristics of high impedance faults (HIFs) in the resulting waveforms of the three-phase residual current, voltage, admittance and power. By using Fourier analysis, their low order harmonic vectors were evaluated and then fed to a neural network. The network was based on either perceptron or feedforward algorithms. The trained network was verified using other distribution systems.