Abstract: This paper presents a method for real-time dynamic control of flood limited water level for the Qing River cascade reservoirs. In this method, we developed a weather forecasting model for forecasting the flood process in future three days, then an optimal-operation-curve model using a genetic algorithm for optimization of long-term reservoir operation and a binary Copula-SAR stochastic model to simulate the long-term behaviors of runoff. Finally, a real-time optimal operation model was constructed using an aggregation-disaggregation approach to dynamic control flood limited water level for cascade reservoirs and considering the rules for coupling long-term and short-term operations, and an optimal algorithm POA was adopted to solve the model. The results show that the real-time dynamic control with long-term operation rules can generate extra power outputs of 0.059 and 0.314 billion kW·h (increases by 6.2% and 45.1%) relative to those without coupling long-term operation rules and in design operation, respectively.