Page 111 - 2023年第54卷第10期
P. 111
[ 5] 顾文龙,卢文喜,张宇,等.基于贝叶斯推理与改进的 MCMC方法反演地下水污染源释放历史[J].水利
学报,2016,47(6):772 - 779.
[ 6] HUSSAINM F,BARTONRR,JOSHISB.Metamodeling:Radialbasisfunctions,versuspolynomials[J].Eu
ropeanJournalofOperationalResearch ,2002,138(1):142 - 154.
[ 7] 高学平,朱洪涛,闫晨丹,等.基于 RBF代理模型的调水过程优化研究[J].水利学报,2019,50(4):439 - 447.
[ 8] YOONH,JUNS,HYUNY,etal.Acomparativestudyofartificialneuralnetworksandsupportvectormachines
forpredictinggroundwaterlevelsinacoastalaquifer [J].JournalofHydrology,2011,396(1 - 2):128 - 138.
[ 9] 余红玲,王晓玲,王成,等.贝叶斯框架下大坝渗流参数反演组合代理 模 型 研 究 [J].水 利 学 报,2022,
53(3):306 - 315,324.
[10] 梁浩,黄生志,孟二浩,等.基于多种混合模型的径流预测研究[J].水利学报,2020,51(1):112 - 125.
[11] LALOYE,ROGIERSB,VRUGTJA,etal.Efficientposteriorexplorationofahigh - dimensionalgroundwater
modelfrom two - stageMarkovchainMonteCarlosimulationandpolynomialchaosexpansion [J].WaterResources
Research,2013,49(5):2664 - 2682.
[12] FIENENM N,MASTERSONJP,PLANTNG,etal.Bridginggroundwatermodelsanddecisionsupportwitha
Bayesiannetwork[J].WaterResourcesResearch,2013,49(10):6459 - 6473.
[13] 张江江.地下水污染源解析的贝叶斯监测设计与参数反演方法[D].杭州:浙江大学,2017.
[14] 孙博闻,杨晰縮,暴柱,等.基于数据同化的深水湖库水温中短期预报[J].水利学报,2022,53(12):1445 - 1455.
[15] 傅甜甜.EnKF方法在北京市平谷区地下水数值模拟中的应用研究[D].北京:中国地质大学(北京),2021.
[16] 陈冲,张伟,邢庆辉,等.基于数据同化的地下水模型不确定性分析[J].冰川冻土,2022,44(6):1912 - 1924.
[17] 周念清,张瑞城,江思珉,等.ES - MDA算法融合 ERT数据联合反演地下水污染源与含水层参数[J].南
水北调与水利科技(中英文),2022,20(3):478 - 486.
[18] 张将伟,卢文喜,曲延光,等.基于 MonteCarlo方法的地表水地下水耦合模拟模型不确定分析[J].水利
学报,2018,49(10):1254 - 1264.
[19] BAKKERM,POSTV, LANGEVIN C D, etal.ScriptingMODFLOW ModelDevelopmentUsingPythonand
FloPy[J].Groundwater,2016,54(5):733 - 739.
[20] 李典庆,蒋水华,周创兵,等.考虑参数空间变异性的边坡可靠度分析非侵入式随机有限元法[J].岩土
工程学报,2013,35(8):1413 - 1422.
[21] GALLAGERRG.DiscreteStochasticProcesses[M].SpringerNewYork,2013.
Inversionofheterogeneousparameterfieldandextractionamountof
groundwaternumericalmodelbasedonMCMCandES - MDAmethods
LIUYongda,CHENXi,GAOMan,MENGXiangbo,LIUWeihan,HUANGRichao
(InstituteofSurface - EarthSystemScience,SchoolofEarthSystemScience,TianjinUniversity,Tianjin 300072,China)
Abstract:MarkovchainMonteCarlo(MCMC)methodandEnsembleSmootherMultipleDataAssimilation(ES -
MDA )methodshavebeenwidelyusedingroundwaterparameterinversioninrecentyears,buttheaccuracyandcom
putationalefficiencyofthree - dimensionalmulti - layerheterogeneousaquiferparameterinversionarestilllackingcom
parativeanalysis.Tothisend ,weconstructacasestudyofunderwaterandmultilayerconfinedwateraquiferswith
heterogeneousparameterfieldsbasedonKarhunen - Loeveexpansion,andestablishanumericalgroundwatermodel
andanalternativemodelbasedonKrigingmethodtosimulatethevariationofwaterheadinlayersofaquifers.The
permeabilitycoefficientandrecoveryofaquiferobtainedfrom theMCMCbasedonthealternativemodel,thetwo -
stageMCMCbasedonthecombinationofthealternativemodelandthenumericalmodel,andtheES - MDAmethod
arediscussed.Theresultsshowthatthetwo - stageMCMCinversionparameterhasthehighestaccuracyandtheES -
MDAmethodhasthehighestcomputationalefficiencyintheinversionofheterogeneousparametersandminingquanti
ty.Thisstudyprovidesreferencefortheselectionofparameterinversionmethodofgroundwaternumericalmodel.
Keywords:MCMCalgorithm;ES - MDAalgorithm;Surrogatemodel;groundwaterparameters;numericalsimu
lationofgroundwater
(责任编辑:耿庆斋)
2
— 1 4 7 —