概念性水文模型参数自动优选方法的比较研究  

Comparative Study of Parameter Optimization Algorithms for Hydrological Model

作　　者：张洪刚[1] 王金星[1] 刘攀[1] 彭定志[1] 何新林[2] 

机构地区：[1]武汉大学水问题研究中心,湖北武昌430072 [2]石河子大学水利建筑工程学院,新疆石河子832003

出　　处：《石河子大学学报（自然科学版）》2002年第3期229-232,共4页Journal of Shihezi University(Natural Science)

基　　金：国家自然科学基金 (5 0 1790 2 6 );教育部跨世纪人才基金资助项目

摘　　要：介绍了 3种参数自动优选方法 :罗森布瑞克法 (Rosenbrock)、单纯形法 (Simplex)和基因法 (Genetic)。应用 4个流域的水文资料 ,以三水源新安江模型为例 ,分析了基因法参数L和IMAX对优选结果的影响 ;对这 3种优化方法的收敛速度、参数初值的选定以及不同组合情况的优选结果进行了比较研究。结果表明 ,基因法参数取IMAX =5 0 0 0 ,L =10时比较合理 ;3种优化方法各有优点 ,基因法不依赖于参数初值的选定 ,罗森布瑞克法收敛速度快 ,单纯形法精度较高 ,建议以基因法优选结果为初值 ,然后采用罗森布瑞克法 ,最后采用单纯形法 。Three optimization algorithms (Genetic, Rosenbrock and Simplex) for hydrological model calibration were introduced. Four basin rainfall runoff data and the Xinanjiang model were selected for this study. The value of genetic algorithm's parameter L and IMAX was analyzed and discussed. The converge speed, the initial parameter value and the different composition of the three algorithms were comparatively studied in detail. The results show that each algorithm has its own advantage, Genetic is insensitive to the initial parameter value, Rosenbrock is the fastest algorithm and Simplex has the highest calibration precision. It is suggested that the combination of Genetic, Rosenbrock and Simplex algorithms for rainfall runoff models calibration can obtain the best parameter values.

关 键 词：概念性水文模型 参数自动优选方法 比较研究 基因法 罗森布瑞克法 单纯形法 降水模拟 

分 类 号：P333.9[天文地球—水文科学]