Improving the reliability of the transmission gear module in the gearbox of a nuclear power plant is of considerable and far-reaching significance to ensure energy security. As a key transmission part of this specific type of power plant, this gear module often runs with heavy load in complex environments and without shutting down for a long time. In such severe working conditions, the failure of this transmission gear module is primarily caused by the deterioration of lubricating oils and the aggravated wear of gears. Once the gear module is damaged owing to the poor lubrication or serious wear of the gear, it will not only cause huge power loss, but also nuclear power accidents, seriously threatening the life of the staff. It is essential that the transmission gear module used in a nuclear power plant is not arbitrarily shut down, which makes testing and monitoring difficult. Moreover, it is very important to monitor the lubricating state of oils in the gearbox in real time by measurements with various suitable high-tech sensors. However, there were few studies concerning this issue. To solve this problem, we first designed and built an experimental test platform to simulate the real working conditions of the transmission gear module in the gearbox of a nuclear power plant. Second, an innovative online oil monitoring method was proposed on the basis of the concept of marginal values of the normal distribution theory in statistics. Finally, a real-time prediction model based on the gray system theory was established to predict the variation of the dielectric constant of lubricating oils.