Detection of magnetizing inrush current using artificial neural network

Describes the simulation of a neural network to detect magnetizing inrush currents on power transformers. The harmonic contents of the inrush current are captured using an FFT spectrum analyzer from a 1 kVA power transformer and are used for training of the network. Test results show that the number of network nodes can be reduced to only 6 for two layers to manipulate the harmonic contents. It is therefore suitable for real-time implementation.<>