With the implementation of source controlling and pollution intercepting in Taihu Basin, sediment has become a non-negligible pollution source in Lake Taihu. To simulate endogenous release dynamically, a diagenesis model based on EFDC model was established taking ammonia nitrogen and nitrate nitrogen as the water quality targets. The Latin hypercube sampling (LHS) was adopted to permute 200 combinations of 18 diagenesis model parameters, and the statistical method of probability distribution was applied to analyze uncertainty, similarly for standard rank stepwise regression method to identify sensitive parameters. The results showed that the nitrogen concentration was characterized by the spatial features that the uncertainty was greater in Meiliang Bay and the northwest lake area, and the temporal features that the uncertainty was largest in summer and then was in spring and winter. The uncertainty increased with the rise of the background concentration of water quality. The sensitive parameters for ammonia nitrogen were diffusion coefficient in porewater and reaction velocity for nitrification whose contribution rates to the uncertainty of ammonia nitrogen were 41.68% and 37.82% respectively. The sensitive parameters for nitrate nitrogen were diffusion coefficient in porewater and reaction velocity for denitrification in the aerobic layer, the contribution rates to the uncertainty of nitrate nitrogen were 29.15% and 42.34% respectively. The predominantly sensitive parameters were mainly related to nitrification, denitrification and diffusion process at sediment-water interface. With the growth of algae from dormancy to aggregation, the uncertainty of simulation results increased, and dominantly sensitive parameters turned from biochemical parameters to hydrodynamic parameters. The research discerned the key physicochemical processes and pivotal parameters in endogenous release in Lake Taihu, which can be references for further researches on the endogenous release of other nutrient such as carbon and silicon, and for the laboratory simulation qualitatively of endogenous release in large shallow lakes.