Research on fault diagnosis of turbine generator unit based on improved CPN neural network

Steam turbine generator unit is the core of thermal power plant, whose structure is complex and the operating environment is special. The fault diagnosis research of turbine generator unit has a practical significance in many aspects because of the inevitable failure of the turbine, which can improve the operational safety, reliability and the economic efficiency for the unit. In this paper, it takes the advantages of the combination of supervised and unsupervised types learning process of the Counter-Propagation Network, uses the fault spectrum feature vectors of turbine generator unit as the learning samples to train the CPN, and then improves the algorithm of CPN training process, intervenes the neurons artificially so that the information of the failure modes can be recorded within different neurons. In this way, the network can reflect the mapping relationship between the fault spectrum feature vectors and the fault types directly. Compared with the BP neural network and the improved CPN neural network, the simulation results show that the improved CPN neural network can overcome the shortcomings and deficiencies of BP neural network, can be better applied to the fault diagnosis of turbine generator unit.