2024年受梅雨期集中强降雨影响，新安江水库发生建库以来入库洪量最大的洪水，最大5 d和7 d洪量重现期达到50 a一遇。基于实战预报调度过程开展复盘分析，融合多源多模态气象降雨预报成果，分析历史典型洪水雨强特征，按最不利、中等和有利等不同条件分时段配置预报降雨时序，在延长预见期的同时降低降雨不确定性；应用新安江三水源模型构建全流域河系洪水预报，耦合水库调度形成一体化计算模型方案，依据雨情工情逐小时滚动预报，考虑水库最高水位、下游安全流量和干流洪水错峰调度等目标，通过正反迭代试算模拟215套预报调度成果，支撑10轮开关闸决策研判。复盘对比2024年和2020年2场大洪水，2024年降雨量和洪量均超过2020年，但2024年预报更精细、调度更早更主动、和兰江错峰更精准，可见水库预报调度一体化模型可有效支撑多目标调度风险控制，实现无重大灾情险情和无人员伤亡的“双零”系统性目标。本研究全过程总结复盘充分说明动态精准预报是科学调度的关键，多方案比对研判是精细调度的依据，“早动快动小动”有利于全流域防洪减灾。

[Objective] In 2024， the Xin’anjiang Reservoir experienced the largest inflow flood since its construction due to persistent and intense rainfall during the Meiyu season. The return periods of the maximum 5-day and 7-day inflow volumes both reached the 50-year level. This study conducts a post-event assessment based on the actual forecasting and operation process. [Methods] Multi-source and multi-model quantitative precipitation forecasts were assimilated to analyze the rainfall intensity characteristics of typical historical floods. Representative forecast rainfall time series were optimized under unfavorable， moderate， and favorable conditions with phased temporal distribution， extending forecast lead time while reducing rainfall uncertainty. The Xin’anjiang three-component runoff generation framework was applied to construct a basin-wide flood forecasting model， coupled with reservoir operation rules to form an integrated forecasting-operation modeling scheme. Hourly rolling forecasts were performed based on real-time rainfall and operational information. Considering multiple objectives such as controlling maximum reservoir level， ensuring downstream safety discharge， and managing flood-peak staggering with the mainstream， iterative forward-and-reverse scenario simulations generated 215 forecast scenarios， providing precise support for ten rounds of gate operation decisions. [Results] A comparative assessment of the two major floods in 2024 and 2020 shows that although the 2024 event involved larger rainfall and greater inflow volume， the forecasting was more refined， the regulation was earlier and more proactive， and the peak staggering with the Lan River was more precise. This demonstrates that the integrated forecasting-operation model can effectively support multi-objective risk control， achieving the systemic goal of “zero major disaster and zero casualty”. [Conclusion] The full-process summary highlights that dynamic and accurate forecasting is the key to scientific operation， comparative scenario analysis supports refined decision-making， and the principle of “early action， rapid response， and incremental adjustment” is more conducive to basin-wide flood risk reduction.