耦合敏感性分析与两阶段马尔科夫链蒙特卡洛算法的地下水污染溯源辨识
Coupled Sensitivity Analysis and Two-stage MCMC Algorithm for Groundwater Pollution Source Identification
李子乐 1安永凯 1闫雪嫚2
作者信息
- 1. 长安大学水利与环境学院,陕西西安 710054;长安大学旱区地下水文与生态效应教育部重点实验室,陕西西安 710054
- 2. 西北大学城市与环境学院,陕西西安 710127
- 折叠
摘要
为高精度地开展地下水污染溯源辨识,在对污染源参数进行敏感性分析的基础上,研究应用两阶段马尔科夫链蒙特卡洛(Markov Chain Monte Carlo,MCMC)算法辨识确定污染源参数;同时,探索应用多层感知机(Multi-layer Perceptron,MLP)方法构建地下水污染运移数值模型的代理模型,用以提高地下水污染溯源辨识的效率.为验证上述方法的有效性和可行性,开展了两个数值算例研究.结果表明:采用MLP方法构建的代理模型对地下水污染运移数值模型的逼近精度高,不仅能够有效提升地下水污染溯源辨识效率,还能保持良好的计算精度;所提出的耦合敏感性分析与两阶段MCMC算法能够显著提升低敏感性污染源参数的辨识精度.
Abstract
To achieve high-precision groundwater pollution source identification,the two-stage Markov Chain Monte Carlo(MCMC)algorithm was used to identify the pollution source parameters based on sensitivity analysis of pollution source parameters.At the same time,the surrogate model of the numerical model of groundwater pollution transport using multi-layer perceptron(MLP)method was explored to improve the efficiency of groundwater pollution source identification.Two numerical examples were implemented to verify the effectiveness and feasibility of the above methods.The results show that the surrogate models constructed by the MLP method has high approximation accuracy for the numerical model of groundwater pollution transport,which can not only effectively improve the efficiency of groundwater pollution source identification,but also maintain good calculation accuracy.The proposed coupled sensitivity analysis and two-stage MCMC algorithm can significantly improve the identification accuracy of pollution source parameters with low sensitivity.
关键词
地下水/污染运移/溯源辨识/数值模型/两阶段马尔科夫链蒙特卡洛算法/敏感性分析/多层感知机/代理模型Key words
groundwater/pollution transport/source identification/numerical model/two-stage MCMC algorithm/sensitivity analysis/MLP/surrogate model引用本文复制引用
基金项目
国家自然科学基金项目(42102287)
中国博士后科学基金项目(2020M683399)
陕西省自然科学基础研究计划项目(2023-JC-QN-0290)
出版年
2024
NETL
NSTL
万方数据