Prognosis and health management (PHM) of Control moment gyroscope (CMG) plays a crucial role in ensuring the operational efficiency and safety of spacecraft. In order to improve the accuracy of PHM and supplement abundant monitoring data, this paper proposes a digital twin-driven degradation modeling method, which establishes a detailed simulation model based on degradation mechanisms at the CMG component level. The digital twin model not only provides a large amount of high-quality data for performance evaluation, but also serves as an important reference for on-orbit CMG state assessment. Finally, a case of estimating virtual sensor degradation information based on reinforcement learning is used to demonstrate the effectiveness of the proposed digital twin method.