水库调度决策单调性与动态规划算法改进

在灌溉、供水等调度中,水库经济效益随供水量增加常呈现"边际效用递减"的特征。本文在"边际效用递减"(即水库调度目标函数为凹函数)的假设下分析得到两阶段水库调度中本时段泄水量、下一时段余留水量与本时段初水库蓄水量间的单调增加关系,将这一单调关系由两阶段推广到多阶段水库调度并进一步提出动态规划算法改进:搜索域缩减算法和邻域搜索算法。在供水调度研究案例中,搜索域缩减算法和邻域搜索算法能将动态规划计算时间分别缩减50.4%和92.1%。进一步分析发电调度中水库水头、下泄水量对总发电量的耦合影响,讨论搜索域缩减算法和邻域搜索算法对发电调度的适用性,并将这两种算法在陡岭子水库发电调度中进行应用;计算结果表明搜索域缩减算法、邻域搜索算法能为发电调度提供有效的参考方案。

Monotonicity in operation decision of reservoir and improved dynamic programming algorithm

In irrigation and water supply reservoir operation, economic utility usually exhibits diminishing marginal utility properties. Based on a concavity assumption for reservoir utility function (i.e., diminishing marginal utility), this study analyzes the monotonic relationships between reservoir released water volume, storage carry-over volume and current reservoir storage, extends the monotonic relationships from two-stage to multi-stage reservoir operation, and applies these relationships to improve the dynamic programming algorithm. In the water supply case study, the improved dynamic programming algorithm can obtain the optimal solution and reduce the computation time of dynamic programming, in the hydropower case study, the improved algorithm can also obtain optimal solution and reduce computation time.