基于动态规划与模拟退火算法的河-湖-梯级泵站系统水资源优化配置研究  

Research on optimal water resources allocation of river-lake-pumping stations system by Dynamic Programming and Simulated Annealing approach

作　　者：侍翰生[1] 程吉林[1] 方红远[1] 陆小伟[2] 

机构地区：[1]扬州大学水利科学与工程学院,江苏扬州225009 [2]江苏省水利勘测设计研究院有限公司,江苏扬州225009

出　　处：《水利学报》2013年第1期91-96,共6页Journal of Hydraulic Engineering

基　　金：国家自然科学基金项目(60974099);教育部博士点基金项目(20093250110002);江苏省属高校自然科学重大项目(09KJA570001);江苏省水利科技重点项目(2010060);江苏省普通高校研究生科研创新计划项目(CX8Z12-0901)

摘　　要：提出了具有多决策变量的河-湖-梯级泵站系统水资源优化配置动态规划数学模型,以各区间缺水量(弃水量)的平方和最小为目标函数,各阶段的湖泊蓄水量为状态变量,各阶段抽水入湖量、湖泊放水量为决策变量。以南水北调长江-洪泽湖段调水工程为例,针对状态变量离散点少、决策变量可行域大且离散点多的特点,采用基于动态规划与模拟退火相结合的混合算法(DP-SA)对该模型进行计算,结果表明:采用该模型进行优化调度,不但可以提高供水保证程度,而且可以减少系统的总抽水量,降低供水系统运行成本;通过与动态规划逐次逼近法(DPSA)进行比较,可以得出混合算法在求解此类问题方面具有计算结果好、收敛速度快等优点。A dynamic programming model with multiple decision variables for optimal water resources allocation of river-lake-pumping stations system is proposed, in which minimum sum of squares of water shortage （discarded water） in each section is set as the objective function, the water storage of the lake of each stage is defined as the state variable, and the water volume to be pumped into the lake and the water release from the lake are expressed as decision variables. The Yangtze River-Hongze Lake section of South-to-North Water Transfer Project is taken as a case study. According to the characteristics that the discrete values of the state variable are relatively less while the feasible solution space of each decision variable is large and its more discrete values, this paper applies a hybrid method （DP-SA） using dynamic programming and simulated annealing combined by the authors to calculate the model. The application results show that using the proposed model to make optimal operation not only can improve the water-supply guaranteed rate, but also can decrease the water volume to be pumped which can reduce the operation cost of the water supply system. In comparison to dynamic programming with successive approximations （DPSA）, we can obtain that the proposed hybrid method has better calculation results and faster convergence rate in solving this kind of problem.

关 键 词：动态规划 模拟退火 水资源 优化 

分 类 号：TV213.9[水利工程—水文学及水资源]