Tractor transmission shaft is one of the components that affect the reliability of the tractor. Compiling load spectrum which can reflect the actual working conditions of high-power tractor drive shaft is of great significance to improve the reliability of the tractor. It is time consuming and expensive to obtain the actual measurement for load spectrum of the whole life cycle, so extrapolated methods are always applied by using the load spectrum of the limited measurement duration. However, the traditional method of rain flow counting and rain basin extrapolation in the process of compiling load spectrum of transmission system has limitations, this paper proposed a POT (peak over threshold) model based time domain extrapolation method for driving shaft load of high-power tractor. Firstly, a set of wireless torque measurement system for tractor transmission shaft is built, which mainly includes strain gauges, batteries, wireless strain nodes and wireless acquisition terminals. The strain gauge is pasted on the proper part of the tractor drive shaft and connected with the wireless torque node. The wireless transmitter is encapsulated in the wireless torque node and transmits the measured torque signal to the wireless receiver synchronously. The test data of the drive shaft under field working conditions are stored in a portable computer connected to a wireless receiver. Then, POT model is established based on extremum theory to extract the peak and valley values of the test data and eliminate the small load cycles, and the intervals of the optimal thresholds are determined by using the mean excess function graph, which are [492, 501] N·m for the upper limit and [325, 334] N·m for the lower limit, respectively. Thirdly, the data in the threshold interval is divided according to a certain gradient, and the fitting effect corresponding to each threshold is preliminarily tested by calculating R-square. The grey relational degree analysis method is used to select the optimal threshold and calculate the grey relational degree of each threshold in the corresponding threshold range. the optimum threshold values 497 N·m for upper limit and 333 N·m for lower limit are achieved by comparing the values of grey correlation degree corresponding to different threshold. The excess threshold data are fitted by generalized Pareto distribution and the CDF and P-P figures are plotted to evaluate the effectiveness of fitting test. The results show that the fitting curve of generalized distribution function can accurately describe the distribution law of extreme samples. Finally, according to the fitted generalized distribution function, the load sequence with the same distribution is generated. The time domain extrapolation of the load is achieved by replacing the original extremum load with the generated sequence. The cumulative frequency curve of the load cycle is developed according to the original load data and the results obtained by the time domain extrapolation method and the rain flow extrapolation method. The results show that the frequency curves obtained by the traditional rain flow extrapolation method and the time domain extrapolation method are consistent. Compared to the original load data, both load extrapolation results have a small amount of large extreme load which did not occur during the test. Therefore, it can be indicated that the load extrapolation method based on the POT model not only increases the frequency of the load cycle, but also extrapolates the load extreme value to a certain extent based on the distribution law of the large extreme load. The accuracy of the time domain extrapolation method is verified. The load time domain extrapolation method developed in this paper has good applicability to the measured loads of high-power tractor transmission shaft in operation conditions. Compared with the rain basin extrapolation method, the load time domain extrapolation method based on POT model can not only obtain the load time domain sequence of arbitrary mileage, but also retain the order of the measured load cycle to a great extent, which can provide reliable data support for the indoor load spectrum loading test of high-power tractor transmission system in the future. [ABSTRACT FROM AUTHOR]