<p>Many studies have focused on constructing scheduling models based on varying inflow data and applying optimization algorithms to enhance the benefits. However, these models are often characterized by a lack of adaptability to dynamic environments, making them challenging to apply in reservoir operations. For this issue, this study proposes a scheduling model that couples medium- and long-term runoff forecasts, which enhances adaptability as follows: (1) through variable weight multi-model combined forecast; and (2) by rolling and feedback corrections of forecasts and scheduling across various time scales. This study applied the model to the Huangjinxia Reservoir to evaluate its effectiveness. The results show that the variable-weight forecast model remains highly accurate even under varying environmental conditions and across different time scales, with the NSE consistently greater than 0.8. In addition, compared to the deterministic optimal scheduling model, the coupled forecast scheduling model significantly improved the model’s adaptability, with scheduling benefits increasing by 6.7% and 0.8%, respectively. The coupled medium- and long-term forecast scheduling model has slightly lower benefits than the coupled long-term forecast scheduling model; however, it exhibits higher adaptability in scheduling and is more aligned with the actual operational requirements. Finally, this model also allows for an adjustment of the scheduling scheme based on real-time conditions, further cementing its applicability of reservoir scheduling.</p>