Abstract: The development of hyperspectral microwave radiometers is of great significance for improving the accuracy of inversion of atmospheric parameters. In this paper, the radiant transmission model mpm93 and BP neural network are used to construct the models of forward modeling upward radiance brightness temperature and inversion of atmospheric temperature profiles, respectively, and the accuracy of inversion of atmospheric temperature profiles under the condition of clear sky by hyperspectral microwave radiometer is studied. Eighty channels are selected as the hyperspectral channels in the 54-58 GHz, 64-68 GHz oxygen absorption band, and simulation experiments are conducted based on the radiosonde data from May to December 2015 in Kunming. The nine temperature detection channels of the special sensor microwave imager/sounder (SSMIS) are selected for comparison experiments to evaluate the inversion effect. The experimental results show that the inversion accuracy of the hyperspectral channel is 0. 3-0. 6 K higher than that of the SSMIS in the altitude range of 3-10km, and the inversion accuracy is increased by 1 K in the height range of 0-3 km.