基于集合卡尔曼滤波的多相流模型参数估计——以室内二维砂箱中重质非水相污染物入渗为例

摘要

重质非水相有机污染物(DNAPL)泄漏到地下后,其运移与分布特征受渗透率非均质性影响显著.为刻画DNAPL污染源区结构特征,需进行参数估计以描述水文地质参数的非均质性.本研究构建了基于集合卡尔曼滤波方法(EnKF)与多相流运移模型的同化方案,通过融合DNAPL饱和度观测数据推估非均质介质渗透率空间分布.通过二维砂箱实际与理想算例,验证了同化方法的推估效果,并探讨了不同因素对同化的影响.研究结果表明:基于EnKF方法同化饱和度观测资料可有效地推估非均质渗透率场;参数推估精度随观测时空密度的增大而提高;观测点位置分布对同化效果有所影响,布置在污染集中区域的观测数据对于参数估计具有较高的数据价值.%The migration behavior and distribution of dense non-aqueous phase liquid(DNAPL)in subsurface are greatly influenced by geological heterogeneity.To fully understand the DNAPL source-zone architecture, parameter estimation is needed to characterize the permeability heterogeneity in multiphase flow simulation.The data assimilation method based on ensemble Kalman filter (EnKF) is applied to solve this parameter estimation problem.The performances of EnKF are investigated and compared by applying EnKF to a real-world and a synthetic DNAPL infiltration experiment in a two-dimensional laboratory-scale sandbox.The factors to control the performance of data assimilation in multiphase flow are also discussed.The results showed that the EnKF method can effectively estimate multiphase model parameters via DNAPL saturation observations.With the increased sampling density in spatial and time scale, EnKF exhibits a promotion of computational accuracy.Especially, EnKF method can produce satisfactory estimation when increasing the spatial sampling density in the high DNAPL-saturation region.