地下水污染模拟模型的不确定性分析

为同时分析源汇项和水文地质参数不确定性对地下水污染数值模拟模型输出结果的影响,以抚顺市某煤矸石堆放场为研究实例展开研究.首先以硫酸根离子作为模拟因子,建立该场地地下水污染数值模拟模型.然后,分别采用局部灵敏度分析和全局灵敏度分析两种方法对模拟模型参数进行灵敏度分析并对比二者的结果,最终筛选出对模型输出结果影响较大的两个参数作为模型的随机参数.为减少反复调用模拟模型产生的计算负荷,分别对3口观测井应用克里格、核极限学习机、支持向量机和BP神经网络4种方法建立模拟模型的替代模型,根据这4种替代模型在不同井的拟合效果,为每口井优选出一个表现最好的替代模型,并利用优选出的替代模型完成蒙特卡洛随机模拟.最后,对随机模拟的结果进行统计分析与风险评估.结果表明,在置信度为80%时,1,2,3号三口井浓度的置信区间分别为:211.48~845.04mg/L,0~406.98mg/L,231.42~958.37mg/L.此外,依据《地下水质量标准》和各井的污染物浓度分布函数曲线得出:1号井和3号井的水质达标Ⅴ类水的概率分别为90.1%和93.1%,2号井达标Ⅲ类水的概率为80.7%,藉此为地下水资源管理和污染防治提供合理依据.

Uncertainty analysis of groundwater pollution simulation model

In order to simultaneously analyze the effect of source and sink items and hydrogeological parameters uncertainty on the output of groundwater pollution numerical simulation model, a gangue dump site in Fushun City was taken as an example to study. Firstly, the groundwater pollution numerical simulation model of the site was established with sulfate ion as the simulation factor. Then, local sensitivity analysis and global sensitivity analysis were used to analyze the sensitivity of the simulation model parameters, and the results of thetwowere compared. Finally, two parameters that have a great impact on the model output were selected as the random parameters of the model.In order to reduce the calculation load caused by repeatedly calling the simulation model, four methods of Kriging(KRG), kernel extremelearning machine(KELM), support vector machine(SVM) and BP neural network (BPNN) were used to establish the surrogate model of the simulation model for three observation wells respectively. According to the fitting effect of the four surrogate models in different wells, a surrogate model with the best performance was selected for each well, Monte Carlo stochastic simulation was completed by using the optimized surrogate model. Finally, statistical analysis and risk assessment were carried out on the results of random simulation. The results show that when the confidence was 80%, the confidence intervals of the concentration values of the well 1,2,3 were 211.48~845.04mg/L, 0~406.98mg/L, 231.42~958.37mg/L. In addition, according to the groundwater quality standard and the pollutant concentration distribution function curve of each well, the probability that the water quality of well 1 and well 3 met the class V water standard was 90.1% and 93.1% respectively, and the probability that well 2 met the class III water standard was 80.7%, so as to provide a reasonable basis for groundwater resource management and pollution prevention and control.