水文模型参数多目标优化率定及目标函数组合对优化结果的影响  

Study on Multi-objective Calibration of Hydrological Model and Effect of Objective Functions Combination on Optimization Results

作　　者：郭俊[1] 周建中[1] 邹强[1] 宋利祥[1] 张勇传[1] 

机构地区：[1]华中科技大学水电与数字化工程学院,湖北武汉430074

出　　处：《四川大学学报（工程科学版）》2011年第6期58-63,共6页Journal of Sichuan University (Engineering Science Edition)

基　　金：国家"973"计划重点基础研究发展计划资助项目(2007CB714107);高等学校博士学科点专项科研基金资助项目(20100142110012);水利部公益性行业科研专项资助项目(201001080);国家科技支撑计划项目(2008BAB29B08)

摘　　要：针对水文模型多目标优化中目标函数组合难以合理构造与选取的难题,基于多目标智能优化算法的优化结果,提出了一种简便的不同目标组合优化结果性能比较方法,并以一种日径流预报模型为实例,构造了3组目标函数组合,采用经典多目标优化算法NSGA-II优化率定模型参数,在模型参数优化率定结果的基础上,应用作者提出的目标组合优化结果比较方法,分析了不同目标组合对模型参数优化结果性能的影响,并尝试找到一种能权衡各种目标性能的目标组合。结果表明,提出的目标组合优化结果比较方法可以有效分析与比较不同目标组合对优化结果的影响,能够为水文预报人员合理选取与构造目标组合提供科学的数量依据,具有良好的工程应用前景。When the parameter optimization of hydrological models is done under the framework of multiobjective,how to select or construct proper combination of objective functions is one of the most important issues.To solve this problem,one simple and convenient method for the comparison of the influence of different objective function combinations on the optimization results,which were obtained by multi-objective evolutionary algorithms,was proposed.A hydrological model for daily streamflow forecasting was taken as an example.Three objective function combinations were constructed,and the multiobjective algorithm NSGA-II was employed to optimize the hydrological model.Based on the optimization results,the proposed method was applied to analyze the performance of model calibration with different combinations of objective functions.It was also expected to find a combination of objective functions which can well balance different objective functions.The results showed that the proposed method is effective to analyze the performance of different combinations of objective functions,can provide more comprehensive evidence for decision making in hydrological forecasting,and has good prospects in practical engineering.

关 键 词：水文模型参数率定 径流预测 多目标 目标组合 

分 类 号：TV121[水利工程—水文学及水资源]