Research Of Loss-of-Excitation Protection Of Generator Based On Artificial Neural Network

Various states of generator are simulated by Power System Blockset of Matlab software.The proper algorithm is chosen to recognize the operating domain of the active power-rotor voltage curve; real compactly supported biothogonal wavelet is used for analysis of stator current and the fault is locked with wavelet-threshhold mothod;author abstracts characteristic vectors of oscillation to construct the training data of neural network which is used to lock oscillation.Combining the 3 parts above constructs the complete loss-of-excitation protection.This paper also researches the mothod of using loosening wavelet neural network to lock the protection on the condition of oscilation and fault and verifies the probability of using complex wavelet to diagnose the loss-of-excitation fault based on wavelet analysis.