基于BP神经网络替代模型的地下水污染随机模拟

Random Simulation of Groundwater Pollution Based on BP Neural Network Substitution Model

为分析水文地质参数的不确定性对地下水数值模拟模型输出结果的影响,针对假想算例进行研究。首先建立地下水数值模拟模型。运用灵敏度分析法筛选出对模拟模型输出结果影响较大的参数作为随机变量,为减少反复调用模拟模型产生的计算负荷,分别采用克里格法和BP神经网络方法建立模拟模型的替代模型并比较二者的精度,并选择精度较高的BP神经网络替代模型进行蒙特卡罗模拟。其中,采用BP神经网络方法建立替代模型时,使用三分法算法快速地确定了使替代模型误差最小的隐含层节点数。最后,对随机模拟的结果进行统计分析与区间估计,评价了地下水污染风险。结果表明:在置信水平为90%时,三口观测井浓度的置信区间分别为367.48~415.67、205.12~230.33、118.85~132.82 mg/L。结合风险评估,计算出研究区内一号/二号/三号观测井地下水遭受污染的风险分别为0.66、0.60、0.58,籍此为地下水污染防治提供科学依据。

In order to analyze the influence of the uncertainty of hydrogeological parameters on the output of the groundwater numerical simulation model,this paper studies a hypothetical example. First,a groundwater numerical simulation model is established. Then sensitivity analysis is used to screen out the parameters that have a greater impact on the output of the simulation model. As a random variable,in order to reduce the computational load caused by repeatedly calling the simulation model,the Kriging method and the BP neural network method are used to establish the replacement model of the simulation model and the accuracy levels of the two are compared,and the BP neural network replacement model with higher accuracy is selected to perform Monte Carlo simulation. Among them,when the BP neural network method is used to establish the replacement model,the third-point algorithm is used to quickly determine the number of hidden layer nodes that minimize the error of the replacement model. Finally,the results of the random simulation are statistically analyzed and the interval estimation and evaluation of the risk of groundwater pollution are compared. The results show that when the confidence level is 90%,the confidence intervals for the concentration of the three observation wells are 367.48~415.67,205.12~230.33 and 118.85~132.82 mg/L. Combined with the risk assessment,it is calculated that the risk of groundwater pollution in observation wells No.1,No.2 and No.3 in the study area are0.66,0.60,and 0.58,which provides a scientific basis for groundwater pollution prevention and control.