大连海事大学学报 >
2020 , Vol. 46 >Issue 2: 96 - 106
DOI: https://doi.org/10.16411/j.cnki.issn1006-7736.2020.02.012
分布式水文模型多参数优化与不确定性分析
收稿日期: 2019-10-18
修回日期: 2019-12-28
网络出版日期: 2019-12-28
基金资助
Multi-parameter optimization and uncertainty analysis of distributed hydrological models
Received date: 2019-10-18
Revised date: 2019-12-28
Online published: 2019-12-28
Supported by
水文模拟效率高、应用范围广，但参数的率定仍存在一定困难，尤其是分布式水文模型因数据集比较大，对参数优化与率定技术提出了更高的要求。为此，提出一种单纯形算法(Simplex algorithm)结合内尔米德(Nelder-Mead)搜索算法的优化方法，并以温德河流域为研究案例，采用分布式水文模型和Nelder-Mead单纯形法，分别考查多种目标函数并获得相应的纳什系数(NSE)、均方根误差(RMSE)等评估系数。结果表明：经寻优后得到的参数准确性较高，且均在合理物理意义的取值范围内，4次洪水模拟的NSE系数均达到0.9以上；以残差平方均值为目标函数所得到的NSE效率系数及RMSE系数均优于其它目标函数结果。此外，通过不确定性分析方法的进一步验证，优化参数和参数率定结果具备科学可靠性。
王琰 , 丁永生 , 孔乔 , 王一帆 , 陈胤名 . 分布式水文模型多参数优化与不确定性分析[J]. 大连海事大学学报, 2020 , 46(2) : 96 -106 . DOI: 10.16411/j.cnki.issn1006-7736.2020.02.012
Though higher efficiency and wider range application of hydrologic simulation, there are still some difficulties for parameter calibration, in particular, due to the large data set of distributed hydrologic model, higher requirements are being put forward for parameter optimization and calibration technology. In this paper, an optimization method of simplex algorithm combined with Nelder Mead search algorithm was proposed, and taking Wende River Basin as a study case, the distributed hydrological model and Nelder Mead simplex method were used to examine various objective functions respectively and obtain corresponding evaluation coefficients such as NSE and RMSE. The results show that the parameters obtained by optimization have higher accuracy, and within the range of reasonable physical meaning, the NSE coefficients of four flood simulation are all above 0.9, the NSE efficiency coefficients and RMSE coefficients obtained by taking the mean square of residual error as the objective function are better than those of other objective functions. In addition, through the further verification of uncertainty analysis method, the optimization parameters and parameter calibration results have scientific reliability.
/
| 〈 |
|
〉 |