A global detection relay for synchronous generator transient and loss of excitation stability using neural network

The paper presents an artificial neural network (ANN) based on-line relaying scheme for synchronous generator transient and loss of excitation in stability pattern detection. The protection scheme utilizes selected frequencies of an adjustable window FFT-frequency spectrum for a discriminant diagnostic vector comprising machine speed deviation, generator current, voltage, instantaneous power, and current voltage ratio as well as their cross spectra. The selected frequencies manifest the system condition, fault type, and allow the detection of allowable clearing time as well as identification of loss of excitation as either short circuit or open circuit conditions.