中文摘要：

水库群调度中发电与供水、生态等目标既非完全对立也非完全协调,其优化调度决策非常复杂。本文以长江上游15座大型水库群联合调度为背景,研究建立了包含发电、河道外供水和河道内生态用水等目标的非线性优化调度模型,以逐步优化算法（POA）为基础,引入优化窗口（L）和滑动距离（l）两个参数,提出了扩展型逐步优化算法（E-POA）以提高非线性模型的求解效率与效能。经与线性全局优化策略相比,E-POA方法有效、解集合理。长江上游15座大型水库长系列联合多目标调度计算结果表明,水库群调度在发电、供水、生态等目标存在不对称博弈关系,河道外供水与河道内生态用水竞争强烈,发电与河道内供水竞争较弱。当优化计算的总发电量为总设计发电量的96.3%时,系统的发电量、河道外供水和河道内生态用水目标可达成较好的协调统一。多年平均条件下,15座水库群联合调度的总期望发电量为2625.8～2 700.1亿kW·h,略小于总设计发电量。

英文摘要：

In a multi-reservoir system,the operation objectives,such as hydropower generation,water supply and ecological protection are neither fully opposite nor fully coordinated,and the optimal joint operation decision of reservoirs is very complex. With the background of joint operation of 15 reservoirs in the Upper Yangtze River Basin,a nonlinear multiobjective optimization model is developed to coordinate hydropower generation,off-stream water use and in-stream water use in the paper. Based on the principle of POA（Progressive Optimality Algorithm）,the extended POA（E-POA）is proposed to accelerate the efficiency and effectiveness of solving nonlinear optimization model by introducing two parameters,i.e. the optimiza-tion window（L）and the sliding distance（l）. The operation model is solved by nonlinear progressive optimization strategy and compared with the results from linear global optimization strategy. The results indicate thatthe nonlinear progressive optimization strategy is acceptable and the set of solutions is reasonable. The re-sults also show the asymmetry trade-offs among hydropower generation,water supply and ecological protection. It illustrates that the off-stream water use intensely compete with in-stream ecological water use,but the water use of hydropower generation softly compete with in-stream ecological water use. If the averageannual hydropower capacity of the system approximates to 96.3 % of the original designed value,thesethree objectives can be balanced well,and the expected average annual hydropower capacity of the systemis about 262.58-270.01BKH （Billion Kilowatt Hours）,slightly less than the original designed value of the system.