This paper focuses on the gravity anomaly separation problem in dynamic reservoir monitoring and proposes a Multi- order and Multi- scale Surface Gravity Anomaly Collaborative Decomposition Method (MMSRGE). Aiming at the problems faced by traditional methods in processing reservoir microgravity anomalies, such as low signal- to- noise ratio, poor spatial resolution, and difficulty in anomaly extraction. The method achieves quantitative extraction of target layer anomalies by establishing a mapping relationship between basis function order and decomposition scale designing a hybrid basis function system of orthogonal polynomials and radial basis functions and integrating rock physics forward modeling constraints. In theoretical model tests, two types of spherical combination models were designed to simulate different reservoir structures, verifying the accuracy and reliability of the method in handling complex geological structures. In practical applications, empirical studies were conducted in two major reservoir development areas in western China: Area A (carbonate fracture—cavity type) and Area B (tight sandstone). In the gas injection monitoring of Area A, the separated residual anomalies accurately delineated the dominant fluid channels between wells; in the time- lapse microgravity inversion of Area B, the distribution range of high- density areas was consistent with the spatial distribution of the fracturing fluid diffusion zone. These results provide a new technical paradigm for dynamic monitoring of unconventional reservoirs, indicating the reliability and effectiveness of microgravity monitoring technology in predicting well connectivity in fractured—cavity reservoir development and monitoring tight sandstone fracturing, and offering strong technical support for well connectivity research and injection—production scheme adjustment during reservoir development.