含水层渗透系数预测及不确定性分析耦合模型
|
摘要
本研究旨在精细计算冲洪积平原地区的渗透系数,并为进一步建立溶质运移模型提供基础数据。通过建立人工神经网络(ANN)与通用似然不确定估计法(GLUE)的耦合模型对含水层渗透系数进行预测,并对模型参数的不确定性进行分析。利用马尔可夫蒙特卡洛采样法(MCMC)取代常见的通用似然不确定性估计方法中的蒙特卡洛法(MC),将其与人工神经网络技术耦合,以150个典型粒度组分样本作为输入数据,构建研究区含水层渗透系数预测及不确定性分析的GLUE-ANN模型。通过对华北平原典型地区实例研究,验证该方法具有较好的采样效率和寻优性能。计算结果表明,与渗透系数的实测值相比较,GLUE-ANN模型的相对误差介于1.55%~23.53%之间,模型的计算精度满足地下水资源评价的要求。通过模型参数的后验分布得出参数全局最优值所在的区域,表明模型能够更合理地反映水文地质参数的不确定性。
This study aims at fine calculation of alluvial-proluvial plain region and providing fundamentaldata for construction of solute transport model in the further research. Hydraulic conductivities of aquifersin the study area are predicted through establishing a coupling model between artificial neural network(ANN) and generalized likelihood uncertainty estimation(GLUE),and uncertainty of the model parametersis analyzed. Markov Chain Monte Carlo(MCMC) was used to replace Monte Carlo(MC) in commonGLUE,and coupled it with artificial neural network technology,an overall model of aquifer hydraulic con-ductivity prediction and its uncertainty analysis(GLUE-ANN) was built by using 150 typical grain-sizefraction samples as input data. Via case study in a typical area of North China Plain the study corroboratesa better sampling efficiency and optimization capability; compared to measured values of hydraulic conductivi-ty,relative errors of the GLUE-ANN model are between 1.55 % and 23.53 %,the calculation precision ofthe model meets the requirements of groundwater resources assessment. By posterior distributions of the mod-el parameters,the areas of parameter global optimum are obtained,which indicates the model is capableof reasonably reflecting parameter uncertainty of hydrogeological model.
This study aims at fine calculation of alluvial-proluvial plain region and providing fundamentaldata for construction of solute transport model in the further research. Hydraulic conductivities of aquifersin the study area are predicted through establishing a coupling model between artificial neural network(ANN) and generalized likelihood uncertainty estimation(GLUE),and uncertainty of the model parametersis analyzed. Markov Chain Monte Carlo(MCMC) was used to replace Monte Carlo(MC) in commonGLUE,and coupled it with artificial neural network technology,an overall model of aquifer hydraulic con-ductivity prediction and its uncertainty analysis(GLUE-ANN) was built by using 150 typical grain-sizefraction samples as input data. Via case study in a typical area of North China Plain the study corroboratesa better sampling efficiency and optimization capability; compared to measured values of hydraulic conductivi-ty,relative errors of the GLUE-ANN model are between 1.55 % and 23.53 %,the calculation precision ofthe model meets the requirements of groundwater resources assessment. By posterior distributions of the mod-el parameters,the areas of parameter global optimum are obtained,which indicates the model is capableof reasonably reflecting parameter uncertainty of hydrogeological model.
引文
[1]Smiles D E,Youngs E G.A multiple-well method for determining the hydraulic conductivity of a saturated soil in situ[J].Journal of Hydrology,1963,1(4):279-287.
[2]David E,Daniel.Measurement of hydraulic conductivity of unsaturated soils with thermocouple psychrometers[J].Soil Science Society of America Journal,1982,46(6):1125-1129.
[3]蒋中明,陈胜宏,冯树荣,等.高压条件下岩体渗透系数取值方法研究[J].水利学报,2010,41(10):1228-1232.
[4]Koekkoeke J W,Booltink H.Neural network models to predict soil water retention[J].European Journal of Soil Science,1999,50(3):489-495.
[5]董佩.潜水面动态蒸发的砂柱实验研究[D].北京:中国地质大学,2010.
[6]Salarashayeri A F,Siosemarde M.Prediction of soil hydraulic conductivity from particle-size distribution[J].World Academy of Science,Engineering and Technology,2012,61:454-457.
[7]Mahmoud S,Alyamani,Zekai S.Determination of hydraulic conductivity from complete grain-size distribution curves[J].Ground Water,1993,31(4):551-555.
[8]樊贵盛,邢日县,张明斌.不同级配砂砾石介质渗透系数的试验研究[J].太原理工大学学报,2012,43(3):373-378.
[9]马荣,石建省,刘继朝,等.基于含水层渗透系数研究的云-Markov模型的建立与应用[J].水利学报,2012,43(7):767-777.
[10]David C W,Asce F.Goodbye,Hazen;Hello,Kozeny-Carman[J].Journal of geotechnical and geoenvironmental engineering,2003,129(11):1054-1056.
[11]Justine O.Evalauation of empirical formulae for determination of hydraulic conductivity based on grain-size analysis[J].Journal of American Science,2007,3(3):54-60.
[12]Russel G S.Correlations of permeability and grain size[J].Ground Water,1989,27(5):633-636.
[13]Awad H S,Bassam A M.A computer program to calculate hydraulic conductivity from grain size data in Saudi Arabia[J].International Journal of Water Resources Development,2001,17(2):237-246.
[14]徐鹏,邱淑霞,姜舟婷,等.各向同性多孔介质中Kozeny-Carman常数的分形分析[J].重庆大学学报,2011,34(4):78-81.
[15]黑肯S.神经网络原理(原书第二版)[M].叶世伟,史忠植译.北京:机械工业出版社,2004.
[16]Erzin Y,Gumaste S D,Gupta A K,et al.Artificial Neural Network(ANN)models for determining hydraulic con-——527ductivity of compacted fine-grained soil[J].Canadian Geotechnical Journal,2009,46:955-968.
[17]Park H I.Development of neural network model to estimate the permeability coefficient of soils[J].Marine Georesources&Geotechnology,2011,29(4):267-278.
[18]Das S K,Samui P,Sabat A K.Prediction of field hydraulic conductivity of clay liners using an artificial neural network and support vector machine[J].International Journal of Geomechanics,2012,12(5):606-611.
[19]Keith B.A manifesto for the equifinality thesis[J].Journal of Hydrology,2006,320:18-36.
[20]茆诗松.贝叶斯统计[M].北京:中国统计出版社,1999.
[21]陆乐,吴吉春,陈景雅.基于贝叶斯方法的水文地质参数识别[J].水文地质工程地质,2008,58(5):58-62.
[22]Smith A F M,Robert G O.Bayesian computation via the Gibbs sampler and related Markov chain Monte Carlo methods[J].Journal of Royal Statistical Society Series B,1993,55:3-23.
[23]Keith B,Andrew B.The future of distributed models:model calibration and uncertainty prediction[J].Hydrological Processes,1992,6:279-298.
[24]龚光鲁,钱敏平.应用随机过程教程及其在算法与智能计算中的应用[M].北京:清华大学出版社,2003.
[25]Haario H,Saksman E,Tamminen J.Componentwise adaptation for high dimensional MCMC[J].Computational Statistics,2005,20(2):265-273.
[2]David E,Daniel.Measurement of hydraulic conductivity of unsaturated soils with thermocouple psychrometers[J].Soil Science Society of America Journal,1982,46(6):1125-1129.
[3]蒋中明,陈胜宏,冯树荣,等.高压条件下岩体渗透系数取值方法研究[J].水利学报,2010,41(10):1228-1232.
[4]Koekkoeke J W,Booltink H.Neural network models to predict soil water retention[J].European Journal of Soil Science,1999,50(3):489-495.
[5]董佩.潜水面动态蒸发的砂柱实验研究[D].北京:中国地质大学,2010.
[6]Salarashayeri A F,Siosemarde M.Prediction of soil hydraulic conductivity from particle-size distribution[J].World Academy of Science,Engineering and Technology,2012,61:454-457.
[7]Mahmoud S,Alyamani,Zekai S.Determination of hydraulic conductivity from complete grain-size distribution curves[J].Ground Water,1993,31(4):551-555.
[8]樊贵盛,邢日县,张明斌.不同级配砂砾石介质渗透系数的试验研究[J].太原理工大学学报,2012,43(3):373-378.
[9]马荣,石建省,刘继朝,等.基于含水层渗透系数研究的云-Markov模型的建立与应用[J].水利学报,2012,43(7):767-777.
[10]David C W,Asce F.Goodbye,Hazen;Hello,Kozeny-Carman[J].Journal of geotechnical and geoenvironmental engineering,2003,129(11):1054-1056.
[11]Justine O.Evalauation of empirical formulae for determination of hydraulic conductivity based on grain-size analysis[J].Journal of American Science,2007,3(3):54-60.
[12]Russel G S.Correlations of permeability and grain size[J].Ground Water,1989,27(5):633-636.
[13]Awad H S,Bassam A M.A computer program to calculate hydraulic conductivity from grain size data in Saudi Arabia[J].International Journal of Water Resources Development,2001,17(2):237-246.
[14]徐鹏,邱淑霞,姜舟婷,等.各向同性多孔介质中Kozeny-Carman常数的分形分析[J].重庆大学学报,2011,34(4):78-81.
[15]黑肯S.神经网络原理(原书第二版)[M].叶世伟,史忠植译.北京:机械工业出版社,2004.
[16]Erzin Y,Gumaste S D,Gupta A K,et al.Artificial Neural Network(ANN)models for determining hydraulic con-——527ductivity of compacted fine-grained soil[J].Canadian Geotechnical Journal,2009,46:955-968.
[17]Park H I.Development of neural network model to estimate the permeability coefficient of soils[J].Marine Georesources&Geotechnology,2011,29(4):267-278.
[18]Das S K,Samui P,Sabat A K.Prediction of field hydraulic conductivity of clay liners using an artificial neural network and support vector machine[J].International Journal of Geomechanics,2012,12(5):606-611.
[19]Keith B.A manifesto for the equifinality thesis[J].Journal of Hydrology,2006,320:18-36.
[20]茆诗松.贝叶斯统计[M].北京:中国统计出版社,1999.
[21]陆乐,吴吉春,陈景雅.基于贝叶斯方法的水文地质参数识别[J].水文地质工程地质,2008,58(5):58-62.
[22]Smith A F M,Robert G O.Bayesian computation via the Gibbs sampler and related Markov chain Monte Carlo methods[J].Journal of Royal Statistical Society Series B,1993,55:3-23.
[23]Keith B,Andrew B.The future of distributed models:model calibration and uncertainty prediction[J].Hydrological Processes,1992,6:279-298.
[24]龚光鲁,钱敏平.应用随机过程教程及其在算法与智能计算中的应用[M].北京:清华大学出版社,2003.
[25]Haario H,Saksman E,Tamminen J.Componentwise adaptation for high dimensional MCMC[J].Computational Statistics,2005,20(2):265-273.