Abstract: Compared with traditional radiometers, hyperspectral microwave radiometers have significantly increased the number of channels and can obtain continuous and high-frequency resolution brightness temperature detection results. However, due to issues such as channel correlation and reduced sensitivity after channel subdivision, further theoretical research is needed to support the improvement of hyperspectral microwave remote sensing capabilities. In view of issues above and the fact that there is no on-orbit hyperspectral microwave radiometer for Earth observation at present, a platform for brightness temperature simulation and atmospheric temperature profile retrieval analysis of spaceborne hyperspectral microwave radiometer is built. This platform simulates the brightness temperature detected by spaceborne hyperspectral microwave radiometer based on radiation transfer theory and atmospheric microwave gas absorption coefficient model, and constructs and trains an inversion model based on neural network algorithm using the simulated data. The brightness temperature observed in the oxygen absorption band of 50 GHz to 70 GHz and the retrieval results of atmospheric temperature profile are analyzed based on this platform, and the observation brightness temperature, weighting function and statistical errors of retrieval results under different channel numbers are compared. The experimental results show that the hyperspectral microwave radiometer can improve the remote sensing detection accuracy of atmospheric temperature profiles in various pressure layers, especially for the high-altitude atmospheric temperature profile.