Abstract: Objectives： Gravity satellites are an important tool for monitoring global mass redistribution. However, current GRACE-type missions are constrained by polar-orbit configurations, and their inversion results are easily affected by north-south striping noise, making it difficult to meet the needs of fine-scale monitoring in small regions. Methods： Taking the North China Plain as the experimental region, this study constructs a “real-world” mass variation field by integrating the PCR-GLOBWB hydrological model, groundwater well observations, and high-frequency atmospheric-oceanic signals. Based on the orbital parameters of China’s gravity satellite mission, multiple polar-orbit formations and Bender-type formations are designed to evaluate the effects of different constellation configurations and filtering methods on striping noise suppression and water storage signal recovery. Results： The results show that striping noise dominates the unfiltered inversion results, and the signal-to-noise ratio becomes positive only when the number of Bender configurations is no fewer than 22. After filtering, the accuracy improvement of polar-orbit formation and Bender formation becomes marginal when their numbers exceed 17 and 5, respectively. Multi-polar formations are better suited to Swenson plus Gaussian filtering, although this combination underestimates signal amplitude. In contrast, Bender formations are better matched with PnMl de-striping combined with weak Gaussian or fan filtering, which can reduce errors while better preserving spatial structures and amplitude variations. Conclusions： Multi-satellite joint inversion can significantly improve the recovery accuracy of regional water storage changes, but appropriate post-processing methods are still required.