Considering the impact of the uncertainty of the initial water level caused by the dynamic control of the operating water level during the flood season on subsequent flood control scheduling, a multi-objective flood control risk analysis method for reservoir groups based on the dynamic control of the operating water level during the flood season was proposed. And Monte Carlo method was used to quantitatively analyze the synergistic effect of hydrological and meteorological forecasting errors, systematically characterizing the uncertainty of the starting water level. A multi-objective flood control optimization scheduling model for reservoir groups considering upstream and downstream flood control target conflicts was constructed by incorporating the uncertainty of the starting water level into flood control risk sources. A multi-level risk index system of “reservoir-point-surface” was proposed, revealing the risk feedback mechanism in stochastic environments. The results of case study in the Dongpi River Basin indicate that the uncertainty of the starting water level cannot be ignored in the short-term flood control process. There is a clear competitive relationship between upstream and downstream flood control targets, and the system risk rate shows marginal benefit characteristics with the selection of non inferior solutions. The coupling effect of uncertainty between flood forecasting and water level adjustment reduces the risk rate of the reservoir group system by 8%, indicating a certain synergistic offsetting mechanism among heterogeneous risk sources.