Fault Diagnosis of Hydro-Generator Unit via GA-Nonlinear Principal Component Analysis Neural Network

Abstract

Based on the complicated relationships between the symptoms and the defects of hydro-generator units, An approach to diagnosing the faults in hydro-generator units via a neural networks combined with genetic algorithm (GA) and nonlinear principal analysis neural network (NLPCA NN) is presented in this paper. At first, both the structure and the connection of the NLPCA NN are optimized by GA. The so called GA-NLPCANN is employed to extract main features from high dimension samples. And then the Bayesian neural network (BNN) is also added to test the final diagnosis performance. Finally, the proposed scheme is applied to diagnose the faults samples of hydro-generator unit and the simulation results have proved the effectiveness of this method.