At present, in the nuclear power plants in China, the periodic calibration is often adopted to manage the drift of safety instruments, however it is too conservative and uneconomical. Thus, this paper studies the technology of online monitoring system for safety instruments. First of all, the actual drift data of the safety instruments is analyzed, which defines the main types of the drift of safety instruments in nuclear power plants, and proves the feasibility of applying online monitoring to safety instruments. Secondly, through the analysis and research of related regulations and standards, the basic requirements of online monitoring technology for safety instruments in nuclear power plants were clarified. Finally, the data analysis research of the online monitoring system technology was carried out, an equivalent average algorithm was proposed for redundant instruments, the non-redundant instruments algorithm was analyzed, and the multivariate state estimation mode (MSET) method was verified by modeling based on the actual data of nuclear power plants, which proved the feasibility of this method in nuclear power plants.