【摘要】 传统的水库调度单点最优决策常基于动态规划系列算法求解，据此改进得到的区间近优决策虽然提高了调度灵活性，但是受限于径流多样性与目标维度。采用随机径流模拟器扩展径流输入，构建梯级水库发电联合调度模型并结合“参数-模拟-优化”框架，对累积目标（多年平均发电量）、最值目标（最小出力）和阈值目标（出力保证率）进行稳健性优化，应用于金沙江下游梯级水库并探讨多个方案下的优化发电出力区间。结果表明：虽同为发电效益，最小出力目标与多年平均发电量、出力保证率目标依旧存在相互制约的关系；基于maximin指标优化可以提高最值目标的稳健性；折中方案在权衡多个目标的基础上可增加多年平均发电量93.2亿kW·h(+4.96%)，采用20%～80%分位数归纳折中方案的坝上水位柔性决策区间，可指导金沙江下游梯级水库发电联合调度运行。更多还原

【Abstract】 The traditional single-point optimal decision for reservoir operation is often based on dynamic programming algorithms. These methods yield near-optimal interval decisions that enhance operating flexibility,yet their effectiveness is curtailed by the variability of runoff and the complexity of the objectives involved. A random runoff simulator was used to extend the runoff input, and the joint reservoir operation model for power generation of cascade reservoirs was built. Combined with the "parameter-simulation-optimization" framework,the robust optimization was performed to maximize the cumulative objective(multi-year average annual power generation), the minimum objective(minimum output) and the threshold objective(output guarantee rate). The model was applied to the cascade reservoirs in the lower reaches of Jinsha River, and the optimal generation output intervals under multiple schemes were discussed. The results show that there exist mutual restraining relationships among the objectives of the multi-year average power generation, the guaranteed rate, and the minimum output, although they are power generation benefits. Optimizing based on maximin index can improve the robustness of the minimum objective, while the compromise scheme can increase the multi-year average annual generation capacity by 9.32 billion kWh(+4.96%) on the basis of weighing multiple objectives. The flexible decision interval of water level of the compromise scheme is concluded by using 20%-80% quantiles,which can guide the joint power generation operation of cascade reservoirs in the lower reaches of Jinsha River.更多还原