Abstract: This study introduces a Model Predictive Control (MPC) method to construct a MPC scheme for joint optimal operation of cascade reservoirs.in the downstream Jinshajiang River. The successive approximation dynamic programming algorithm is used to search the optimal solution within a closed-loop with the objective of maximizing power generation over a forecast period, and subsequently executes optimal decisions in open-loop MPC scheme. Five forecast accuracy scenarios are configured for sensitive analysis by synthesizing systematic and random error characteristics, the forecasting and operating of the cascade reservoirs are simulated across a complete hydrological year, i.e., water supply, main flood prevention, and impoundment periods, and compares with current joint operation scheme. Results demonstrate that: ① Smaller forecast errors or longer forecast horizons lead to higher total hydropower generation of cascade reservoirs. ② Compared with the joint operation scheme, the MPC scheme with a 3-day forecasting error variance can significantly reduce water spillage by 5.21 billion m³ (−9.1%) and increase power generation by 2.96 billion kWh (+1.5%). ③ The MPC scheme exhibits profound intra-annual adaptability, which generates more power during the main flood prevention and impoundment periods while reduces spillway discharge during the refill period. The proposed MPC scheme features favorable control performance and strong robustness, and provides a new technical approach for implementing rolling "forecasting—operating—decision-making" for reservoir groups.