Abstract: The geostationary microwave (GEO-MW) observation system has the all-weather observation capability with high frequency of real-time observation, large field-of-view and cloud-penetrating ability, which makes it a necessity in China's second-generation geostationary meteorology satellite "Fengyun-4" series development. Observation System Simulation Experiments (OSSEs) are conducted to understand how the characteristics of GEO-MW such as the time resolution, the frequency band selection and observation errors affect the performance of typhoon numerical prediction based on the candidate instrumental concept for China GEO-MW satellite mission with five bands of 50, 118, 183, 380 and 425 GHz. The OSSEs assimilate the simulated GEO-MW observation data using the GRAPES global 4D-Var data assimilation system and demonstrate the impact of GEO-MW observations on forecasts of two typhoon cases, "Maria" and "Mangkhut", in 2018. The experiment results indicate that improving the time resolution, increasing the number of frequency band and decreasing the noise level of GEO-MW observation data can effectively improve the performance of typhoon numerical prediction by the application of data assimilation.