In response to the limitations of current statistical downscaling methods in simulating extreme precipitation, uncertainty estimation was combined with downscaling, and a multi-site Neyman-Scott rectangular pulse(STNSRP) model, in conjunction with the perturbation change factors based on global climate models, was used to develop an STNSRP disturbance downscaling model for extreme precipitation prediction. The model was employed to predict precipitation in the Taihu Basin from 2026 to 2100. The results indicate that the STNSRP disturbance downscaling model effectively simulates extreme precipitation characteristics by using a stochastic point process, and it outperforms the Delta method and Neyman Scott rectangular pulse disturbance downscaling model in terms of simulation accuracy. The annual precipitation in the Taihu Basin is projected to increase from 2026 to 2100, with an increase of approximately 0.93 mm/a. The precipitation during the flood season will increase for most of the time, with the increase becoming more pronounced over time. However, the precipitation in January and from October to November will decrease. The precipitation increase in western mountainous areas will be greater than that in eastern plains, and the increase in Zhexi District will be greater than that in Huxi District. Additionally, the precipitation in the Taihu Basin will tend to become more concentrated, with both the intensity and frequency of extreme precipitation events increasing. It is suggested that the future defense against flood and drought disasters in the Taihu Basin should focus on strengthening the construction of flood control and drainage engineering systems, scientifically optimizing the scheduling of reservoirs and water resources allocation strategies, and improving the emergency response mechanism under extreme weather conditions.