Research About Application Of Online Monitoring Algorithm In Fault Diagnosis Of Nuclear Power Plant System

As a low-carbon energy source,nuclear power plays an important role in China's energy structure.Nowadays,China's nuclear power has entered a safe and efficient development stage.To ensure safe operation,the nuclear power plant recalibrates all safety-related sensors and crucial instruments during its shutdown and refueling,and then repairs related equipment.Traditional maintenance strategies are not only costly,but also increase the risk of system and equipment and reduce availability of sensors.With the rapid development of intelligent algorithms,state-based prediction strategies are catching more and more attention,such as online monitoring.The relevant technical standards for online monitoring are given by the U.S_Nuclear Regulatory Commission on the document named "NUREG-6895",which includes many intelligent algorithms.In this thesis,the sensor parameter calibration methods including online monitoring intelligent algorithms in nuclear power plants are classified by two different sensor systems.Three algorithms like Independent Component Analysis,Multivariate State Estimate Technique and Sequential Probability Ratio Test are introduced.The redundant and non-redundant sensor systems in the nuclear power plant are taken as research objects.Based on the MATLAB programming tool,the appropriate online monitoring algorithm is selected to perform abnormal monitoring on the corresponding system respectively.The main research contents of the thesis are as follows:1.For the anomaly monitoring of redundant sensor systems,a combination of independent component analysis and sequential probability ratio testing is proposed.In this paper,the independent component analysis model is used to predict the state of the redundant sensor system,and sequential probability ratio test is used to test the residual of the system state to determine the state of the system.The paper tests the algorithm model using the redundant pressure data in main steam system in the nuclear power plant,and verifies the availability of the ICA-SPRT method for abnormal monitoring of redundant sensor systems.2.For fault detection in non-redundant sensor systems,a combination of multivariate state estimate technique and sequential probability ratio testing is used.The nuclear power plant principle simulator data was used to test the method,and the availability of the method under the SGTR accident in the power-reducing process of the simulator was verified.The steam generator water level data was analyzed during the test.The results show that the MSET-SPRT method can be implemented in the situation.Early warning information is given before the alarm of simulator is reported.The research results of this paper show that the application of intelligent algorithms in nuclear power plants is feasibile and has vast potential for furture development,which provides theoretical and technical reference for the implementation of online monitoring technology in nuclear power plants.