Classification of power distribution system fault currents using wavelets associated to artificial neural networks

In a power distribution system with a resonant neutral grounding, traditional protection algorithms based on a steady state analysis are no longer adapted. Hence a good use of transients becomes essential. This paper deals with the possibility of using wavelet transform as a preprocess for artificial neural networks (ANN) in the algorithm of power system relays. The ANN decides, after training, if the measured signal is faulty or sound. The inputs of the ANN are the arguments of wavelet coefficients obtained after applying a recursive wavelet transform on faulty signals generated with EMTP (ElectroMagnetic Transient Program). A comparison between the wavelets and fast Fourier transform has been made.