Fault Diagnosis Method of Nuclear Power Plant Based on PCA-TCN-KAN

In order to solve the problem that the fault types of nuclear power plants are complex and the occurrence of faults is uncertain, this paper proposes a nuclear power plant fault diagnosis method based on principal component analysis (PCA)-temporal convolutional network (TCN)-Kolmogorov-Arnold network (KAN). Based on the third-generation nuclear power plant simulation experiment platform to obtain the data of 16 different types of faults in the nuclear power plant, including reactor coolant pump power failure, reactor pressure vessel leakage and primary circuit pipe rupture and so on. Firstly, the fault data is preprocessed by PCA to reduce the dimension and highlight the main data features. Secondly, the TCN network is used to extract the time series features of the reduced dimension data. Finally, the KAN network is used to extract the deep vector features of the data and output the fault diagnosis results. The research results show that the fault diagnosis model based on PCA-TCN-KAN has an accuracy of 97.98%, which is higher than the traditional deep learning model.