Satellite microwave remote sensing data is one of the most important data required by global/regional assimilation system. The brightness temperatures of the upper atmosphere (middle and upper stratosphere) from microwave remote sensing are affected by many factors, and have certain errors. The accuracy evaluation of microwave remote sensing of the upper atmosphere has become an important research content due to the relatively lack of high-accuracy sounding data of the upper atmosphere. In this paper, the accuracy of the observation data of the on-orbit microwave temperature radiometers like AMSU-A、ATMS and FY-3D MWTS-2 is examined and evaluated in 2020. Three evaluation methods, i.e., comparison with radio occultation observation, with radiation transfer simulation and with cross calibration, are comprehensively used for the analysis. From different aspects, the three approaches reveal the error characteristics of the data at the upper atmosphere from various instruments, and the possible sources of the errors are analyzed. The specific deviation values of the brightness temperatures of the high-level channels are different from the three evaluation schemes, but the basic trend of the variation are consistent, that is, the high-level channels have greater noise than the middle and low level channels. In addition to the cross calibration method, the other two methods show the seasonal variation of the deviation. On the whole, AMSU-A performs better than MWTS-2 and ATMS. The temporal and spatial variation characteristics of brightness temperature accuracy at the upper atmosphere can provide reference for microwave data assimilation and climate application.