Research on Method of Nuclear Power Plant Operation Fault Diagnosis Based on a Combined Artificial Neural Network

To solve the online real-time diagnosis problem of the nuclear power plant in operating condition,a method based on a combined artificial neural network is put forward in the paper.Its main principle is: using the BP neural network for the fast group diagnosis,and then using the RBF neural network for distinguishing and verifying the diagnostic result.The accuracy of the method is verified using the simulation values of the key parameters in normal status and malfunction status of a nuclear power plant.The results show that the method combining the advantages of the two neural networks can not only diagnose the learned faults in similar power level of the nuclear power plant quickly and accurately,but also can identify the faults in different power status,as well as the unlearned faults.The outputs of the diagnosis system are in form of the reliability of the faults,and are changing with the lasting of the operation time of the plant.This makes the diagnosis results be more acceptable to operators.