基于替代模型的非点源污染模拟不确定性分析  

Uncertainty Analysis of Non-Point Source Pollution Simulation Based on Surrogate Model

作　　者：马梦蝶 李传奇[1] 崔佳伟 杨幸子 王德振 MA Mengdie;LI Chuanqi;CUI Jiawei;YANG Xingzi;WANG Dezhen(School of Civil Engineering,Shandong University,Jinan 250061,China;Zhongnan Engineering Corporation Limited,Power China,Changsha 410000,China)

机构地区：[1]山东大学土建与水利学院,山东济南250061 [2]中国电建集团中南勘测设计研究院有限公司,湖南长沙410000

出　　处：《人民黄河》2019年第6期66-70,共5页Yellow River

基　　金：全球环境基金(GEF)赠款淮河流域海域污染防治项目(C-3-C-1)

摘　　要：为探究非点源污染机理模型参数的不确定性对模型精度和模拟结果的影响,以广利河流域为研究区,运用小波神经网络建立研究区非点源污染模拟SWAT模型的替代模型,解决使用蒙特卡罗方法进行不确定性分析时运算量大的问题,从而实现对非点源污染模拟模型参数的不确定性分析。研究结果表明:小波神经网络模型不仅能够很好地代替SWAT非点源污染模拟模型,而且能大幅度减少蒙特卡罗方法的计算时间;研究区氨氮和总磷输出结果在90%置信水平下的置信区间分别为4.02×10^4~25.85×10^4、9.01×10^3~28.38×10^3 kg/a,相比总磷,氨氮输出结果的离散程度更高,不确定性程度更大。In order to understand the influence of the model parameters uncertainty to the model accuracy and simulation results of the non-point source pollution(NPS),taking Guangli River basin as a study area,this paper used wavelet neural network to establish surrogate model of SWAT model to reduce computational load of the Monte Carlo method and realized parameter uncertainty analysis of the non-point source pollution.The study result shows that the wavelet neural network can well replace the SWAT model of the study area,which can significantly reduce the computational load.At a confidence level of 90%,the output of ammonia nitrogen(NH3-N)4.02×10^4 to 25.85×10^4and total phosphorus(TP)ranged from 9.01×10^3 to 28.38×10^3,and compared with the TP,the output of the NH3-N shows greater uncertainty.

关 键 词：不确定性分析 非点源污染模拟 蒙特卡罗方法 拉丁超立方抽样方法 小波神经网络 广利河流域 

分 类 号：X522[环境科学与工程—环境工程]