[1]伊燕平,卢文喜,张耘,等.基于径向基函数神经网络的地下水数值模拟模型的替代模型研究[J].水土保持研究,2012,19(04):265-269.
　YI Yan-ping,LU Wen-xi,ZHANG Yun,et al.Study on the Surrogate Model of Groundwater Numerical Simulation Model Based on Radial Basis Function Neural Network[J].Research of Soil and Water Conservation,2012,19(04):265-269.
点击复制

基于径向基函数神经网络的地下水数值模拟模型的替代模型研究

《水土保持研究》[ISSN:1005-3409/CN:61-1272/P] 卷: 19 期数: 2012年04期 页码: 265-269 栏目: 出版日期: 2012-08-20
Title:
Study on the Surrogate Model of Groundwater Numerical Simulation Model Based on Radial Basis Function Neural Network
作者:
伊燕平1,2, 卢文喜2, 张耘1, 芦贵君3, 王大中3, 洪德法4
1. 吉林省宏利水土保持咨询有限公司, 长春 130033;
2. 吉林大学 环境与资源学院, 长春 130026;
3. 吉林省水土保持科学研究院, 长春 130033;
4. 长春市城乡规划设计研究院, 长春 130021
Author(s):
YI Yan-ping1,2, LU Wen-xi2, ZHANG Yun1, LU Gui-jun3, WANG Da-zhong3, HONG De-fa4
1. Hongli Soil and Water Conservation Consulting Co., Ltd. of Jilin Province, Changchun 130033, China;
2. College of Environment and Resources, Jilin University, Changchun 130026, China;
3. Jilin Institute of Soil and Water Conservation, Changchun 130033, China;
4. Changchun Institute of Urban Planning&Designing, Changchun 130021, China
关键词:
替代模型; 径向基函数神经网络; 拉丁超立方抽样; 金泉工业园区
Keywords:
surrogate model; radial basis function neural network; Latin Hypercube Sampling; Jinquan Industrial Park
分类号:
P641.8
摘要:
近年来提出的替代模型方法是一种连接数值模拟模型与优化模型的有效途径，替代模型质量的好坏取决于采样方法和替代模型种类。以金泉工业园区地下水水源地为研究区，基于拉丁超立方抽样方法，结合研究区地下水数值模拟模型，获取输入（抽水量）输出（水位降深）数据集，运用人工神经网络方法，建立径向基函数神经网络模型，作为地下水数值模拟模型的近似替代模型。经验证，径向基函数神经网络模型输出得到的水位降深均值与模拟模型计算结果的拟合平均相对误差为0.038；水位降深剩余标准差的拟合平均相对误差为0.042。拟合平均相对误差较小，表明径向基函数神经网络模型能够有效地替代地下水数值模拟模型，为日后替代模型的深入研究提供了科学依据。
Abstract:
In recent years, the surrogate model has become an effective way to connect the numerical simulation model and optimization model. The quality of surrogate model depends on sampling method and the type of surrogate model. In this paper, Jinquan Industrial Park groundwater sources was selected as the study area, the radial basis function neural network model was established as the surrogate model of groundwater numerical simulation model by combining with the groundwater numerical simulation model of the study area, accessing to input (pumping) output (water level drawdown) data set, and using artificial neural network based on Latin Hypercube Sampling techniques. It was proven that the mean water level drawdown output of RBF neural network model and simulation model results fit the average relative error was 0.038; water level drawdown of the remaining average relative standard deviation of the fitting error was 0.042. Fitting the average relative error was small, showing that the radial basis function neural network model can effectively replace groundwater numerical simulation model. The result of research provides a scientific basis for the future in-depth study of surrogate model.

参考文献/References:

[1] He L,Huang G H,Lu H W.A stochastic optimization model under modeling uncertainty and parameter certainty for groundwater remediation design-Part 1:model development[J].Journal of Hazardous Materials,2010,176(1/3):521-526.
[2] Qin X S,Huang G H,He L.Simulation and optimization technologies for petroleum waste management and remediation process control[J].Journal of Environmental Management,2009,90(1):54-76.
[3] 卢文喜.地下水系统的模拟预测和优化管理[M].北京:科学出版社,1999.
[4] Qin X S,Huang G H,Chakma A,et al.Simulation-based process optimization for surfactant-enhanced aquifer remediation at heterogeneous DNAPL-contaminated sites[J].Science of the Total Environment,2007,381(1/3):17-37.
[5] He L,Huang G H,Lu H W,et al.Optimization of surfactant-enhanced aquifer remediation for a laboratory BTEX system under parameter uncertainty[J].Environmental Science & Technology,2008,42(6):2009-2014.
[6] Qin X S,Huang G H,Chakma A,et al.Simulation-based process optimization for surfactant-enhanced aquifer remediation at heterogeneous DNAPL-contaminated sites[J].Science of the Total Environment,2007,381(1/3):17-37.
[7] Johnson V M,Rogers L L.Accuracy of neural network approximators in simulation-optimization[J].Journal of Water Resources Planning and Management (ASCE),2000,126(2):48-56.
[8] Yan S Q,Minsker B.Optimal groundwater remediation design using an adaptive neural network genetic algorithm[J].Water Resources Research,2006,42(5):407-420.
[9] 薛禹群.地下水动力学[M].北京:地质出版社,1997.
[10] McKay M D,Beckman R J,Conover W J.A comparison of three methods for selecting values of input variables in the analysis of output from acomputer code[J].Technometrics,1979,21(2):39-245.
[11] 熊学玉,顾炜.基于改进LHS方法的预应力混凝土结构长期性能概率分析[J].工程力学,2010,27(4):163-168.
[12] 吴振君,王水林,葛修润.LHS方法在边坡可靠度分析中的应用[J].岩土力学,31(4):1047-1054.
[13] Asim R,Sandeep G,Raymond M.A neural-network learning theory and a polynomial time RBF algorithm[J].IEEE Transactionson Neural Network,1997,18(6):1301-1313.
[14] Lin Faajeng,Wai Rongjong,Duan Rouyong.Fuzzy neural networks for identification and control of ultrasonic motor drive with LLCC resonant technique[J].IEEE Transactions on Industrial Electronics,1999,46(5):999-1011.
[15] 王洪斌,杨香兰,王洪瑞.一种改进的RBF神经网络学习算法[J].系统工程与电子技术,2002,24(6):103-105.

相似文献/References:

[1]郭皓,邢贞相,付强,等.基于密度参数K-均值算法的RBF网络及其在降水量预测中的应用[J].水土保持研究,2014,21(06):299.
　GUO Hao,XING Zhen-xiang,FU Qiang,et al.RBF Neural Network of K-Means Algorithm Based on Density Parameter and the Application to the Rainfall Forecasting[J].Research of Soil and Water Conservation,2014,21(04):299.

备注/Memo

收稿日期:2011-7-11;修回日期:2011-10-5
基金项目:国家自然科学基金资助项目（41072171）
作者简介:伊燕平（1984- ），女，吉林长春人，硕士，助理工程师，主要从事水土保持规划设计工作。E-mail：yiyanping-101@163.com。
通讯作者:卢文喜（1956- ），男，吉林德惠人，教授，主要从事地下水数值模拟与优化管理等研究。E-mail：luwenxi@jlu.edu.cn。

更新日期/Last Update: 1900-01-01