In view of the limitation that the traditional historical frequency matching method (Hist-FMM) cannot effectively correct the positional deviation of precipitation areas, an improved frequency matching correction scheme based on low-level circulation similarity (Sim-FMM) is proposed, which uses wind direction and speed at 925 hPa as dynamic constraints. By analyzing the coupling relationship between precipitation and the wind field at 925 hPa, together with sensitivity experiments, the optimal parameter scheme for Sim-FMM was established. For general precipitation correction, a wide tolerance window (30° for wind direction and 4 m·s?1 for wind speed) without seasonal constraints is adopted to ensure sample representativeness. For heavy rainfall correction, a narrower tolerance window (10° for wind direction and 1 m·s?1 for wind speed) combined with seasonal constraints is applied to balance circulation similarity and climatological characteristics. Based on surface observations and ECMWF forecast data from December 2023 to May 2026, the correction performance of Sim-FMM was evaluated. The results show that Sim-FMM outperforms Hist-FMM in precipitation correction. For winter rainfall events, Sim-FMM effectively reduces widespread false alarms by matching analogous circulation samples, increasing the rain/no-rain accuracy by 1.9% relative to Hist-FMM. For local heavy rainfall events in spring and summer, Sim-FMM alleviates the systematic underestimation of rainstorm magnitude, improving the rainstorm threat score (TS) by 8.4% relative to Hist-FMM during 2024-2026. Nevertheless, Sim-FMM performs slightly worse for autumn heavy rainfall, which may be attributed to the highly concentrated seasonal features of autumn rainfall over Hainan Island and the inability of the regional mean wind field at a single time to fully depict the heterogeneous structures of complex systems such as typhoons and shear lines.