Traditionally, process measuring instruments in various process systems of nuclear power plants are regularly repaired, maintained, calibrated or calibrated during each refueling or overhaul, especially the safety-critical instruments used for protective action shutdown. The workload of each regular test and inspection of measuring instruments is large and prone to problems such as “over-maintenance” or “under-maintenance”, which affects the economy and safety of nuclear power plants. Therefore, many nuclear power plants have begun to shift from time-based calibration to state-based calibration, that is, using online state monitoring technology to monitor the state of sensors during nuclear power plant operation and identify sensor drift or failure. This paper summarizes the research and algorithm models of online monitoring systems at home and abroad, summarizes the application of kernel regression, neural network, transformation method and simple average algorithm in establishing online monitoring models, reviews the online monitoring methods of nuclear power plant measuring instruments published in recent years, and discusses the opportunities and challenges of implementation to guide the reasonable selection of modeling algorithms for online monitoring of nuclear power plant sensor states.