代理模型参数率定方法在TOPKAPI模型中的应用  

Application of surrogate modeling parameter calibration method in TOPKAPI model

作　　者：汤岭 王海军 李致家[1] 黄迎春[1] 盛奕华 TANG Ling;WANG Haijun;LI Zhijia;HUANG Yingchun;SHENG Yihua(College of Hydrology and Water Resources,Hohai University,Nanjing 210098,China;Hydrology Center of Shandong Province,Jinan 250013,China)

机构地区：[1]河海大学水文水资源学院,江苏南京210098 [2]山东省水文中心,山东济南250013

出　　处：《河海大学学报（自然科学版）》2024年第1期21-28,共8页Journal of Hohai University(Natural Sciences)

基　　金：国家自然科学基金项目(52079035)。

摘　　要：为提高复杂水文模型参数优化效率,通过Morris参数敏感性分析确定敏感参数,随后将多目标自适应代理模型优化(MO-ASMO)算法应用在TOPKAPI模型的参数率定中,通过最小欧几里得距离筛选Pareto解集中的相对最优解,从解集分布和每场洪水模拟效果两个维度与传统多目标优化方法NSGA-Ⅱ、NSGA-Ⅲ进行比较。结果表明:在相同模型运行次数下,MO-ASMO相较于NSGA-Ⅱ和NSGA-Ⅲ具有更优的Pareto前沿;无论是率定期还是验证期,MO-ASMO算法的评价指标均表现较好,综合表现优于NSGA-Ⅱ、NSGA-Ⅲ算法,MO-ASMO算法有效提升了模型参数优化效率。In order to improve the efficiency of parameter optimization for complex hydrological models,sensitive parameters are determined through the Morris parameter sensitivity analysis.Subsequently,the multi-objective adaptive surrogate model optimization(MO-ASMO)algorithm is applied in the parameter calibration of the TOPKAPI model.The relative optimal solutions in the Pareto solution set are then selected based on the minimum Euclidean distance.The performance of MO-ASMO is compared with traditional multi-objective optimization methods,such as NSGA-Ⅱ and NSGA-Ⅲ,from two dimensions in terms of solution set distribution and simulation effectiveness of each flood event.The results indicate that,under the same number of model runs,MO-ASMO has a superior Pareto front compared to NSGA-Ⅱ and NSGA-Ⅲ.Both in the calibration and validation periods,the evaluation indicators of the MO-ASMO algorithm show better performance,overall surpassing the NSGA-Ⅱ and NSGA-Ⅲ algorithms,and as a result,the MO-ASMO algorithm effectively improves the efficiency of model parameter optimization.

关 键 词：TOPKAPI模型 代理模型 多目标算法 NSGA-Ⅱ NSGA-Ⅲ 洪水预报 临沂流域 

分 类 号：P338[天文地球—水文科学]