A high impedance fault detector using a neural network and subband decomposition

High impedance faults (HIFs) are not easily detectable using conventional overcurrent protection relays. The fault current for HIF is usually less than the normal load current, thus the overcurrent relays cannot easily distinguish HIFs from normal currents. A new method based on a subband decomposition of the current is presented. The energies from the different subbands are used as input to train an artificial neural network (ANN) for the detection of HIFs. The technique, not only detects HIF faults, but also classifies the signals into one of several classes. The main advantage of this method is that it is less sensitive to noise and HIF can be distinguished from similar events, even in the presence of high levels of noise.