With the rapid development of the global nuclear power industry, the safety of nuclear power equipment has received increasing attention. In the event of an accident in its equipment, it is likely to cause significant losses and harm to personnel and the environment. Therefore, real-time monitoring of nuclear power plants has very important practical significance. Based on Kernel Principal Component Analysis (KPCA) and Jensen-Shannon (JS) divergence, equipment-level state monitoring of electric gate valves of nuclear power plants will be conducted. The traditional Principal Component Analysis (PCA) method, which can only reduce the dimensionality of nonlinear data, is improved by KPCA. The traditional statistics T2 and Q, which have the problem of low recognition rate of small faults and poor timeliness, can be greatly improved by the JS divergence. The main model process is to first perform KPCA dimensionality reduction on the data, calculate the JS divergence value of the data after dimensionality reduction, and compare the change trend of the JS divergence value to reflect the operating status of the electric gate valve, and perform real-time status monitoring. The experimental results show that the state detection method combining KPCA data dimensionality reduction and JS divergence has improved the accuracy of monitoring compared with other methods, and has better timeliness.