基于逆序贯模拟及相关概率场方法的非高斯渗透系数场反演

发布时间：2018-03-19 08:59

本文选题：逆模拟　切入点：正态变换　出处：《中国地质大学(北京)》2016年博士论文　论文类型：学位论文

【摘要】：集合卡尔曼滤波(EnKF)已被证明是一种非常高效的能够刻画满足非线性状态方程和高斯分布条件下的水文渗透系数场的逆方法。但是它不足之处是在于其难于刻画非高斯参数场。本论文的研究主题就是如何合理的刻画非高斯水文渗透系数场。首先,我们提出了一种新的随机逆方法-逆序贯模拟(iSS)。iSS借鉴了序贯模拟算法和正态卡尔曼滤波算法中的某些概念。该方法通过渗透系数场实现集合和承压水位场实现集合来计算两者之间的非平稳交叉协方差,之后再利用序贯模拟来生成参数变量实现。我们运用正态转换技术来确保变量呈边缘高斯分布。之后,基于渗透系数条件点数据和承压水位条件点数据,用标准多变量序贯高斯条件模拟来更新渗透系数场实现。设定正态集合卡尔曼滤波(NS-EnKF)的作为参考基准,研究结果表明iSS是能够合理的更新逆条件下的非高斯渗透系数场实现,对比iSS和NS-EnKF这两种方法更新结果,可知这两种方法更新的渗透系数场实现的质量相仿。其次,我们进一步研究了Hu et al.,2013提出的逆方法,并在此基础上对其进行了改进。Hu et al.,2013是利用EnKF直接更新非相关的均匀随机场(在序贯模拟中,通过这些均匀随机数据,从局部条件边缘分布中取值),不同于Hu et al.,2013的想法,新提出的改进方法是用相关均匀随机场作为参数对象地,类似于概率场模拟法(Froidevaux,1993)中均匀随机场的运用。这新旧两种方法研究对比结果表明,在获取渗透系数场空间结构特征和减少实现不确定性方面,新的改进方法比原先的方法要好很多。
[Abstract]:The ensemble Kalman filter has been proved to be a very efficient inverse method for characterizing the hydrological permeability coefficient field under the condition of nonlinear equation of state and Gao Si distribution. However, its shortcoming is that it is difficult to describe the non-permeability coefficient field. Gao Si parameter field. The research topic of this paper is how to describe the non-#china_person1# hydrological permeability coefficient field reasonably. First of all, In this paper, we propose a new stochastic inverse method-inverse sequential simulation. ISS uses some concepts of sequential simulation algorithm and normal Kalman filter algorithm for reference. This method realizes set by permeability coefficient field and sets by water level field under pressure. To calculate the nonstationary cross covariance between the two, Then we use sequential simulation to generate parameter variables. We use normal transformation technology to ensure that variables are distributed on the edge of Gao Si. Then, based on the data of permeability coefficient condition point and pressure water level condition point, we use normal transformation technology to ensure that the variables are distributed on the edge of Gao Si. The standard multivariable sequential Gao Si condition simulation is used to update the permeability coefficient field. The normal set Kalman filter (NS-EnKF) is set as the reference datum. The research results show that iSS can reasonably update the non-#china_person1# permeability coefficient field under the inverse condition. By comparing the results of iSS and NS-EnKF, we can see that the quality of permeability field of these two methods is similar. Secondly, we further study the inverse method proposed by Hu et al.2013. On the basis of this, we improve it. Hu et al. 2013 is to use EnKF to update the uncorrelated uniform random field directly. (in sequential simulation, through these uniform random data, we can get the value from the local conditional edge distribution, which is different from Hu et al. 2013. The improved method is to use the correlation uniform random field as the parameter object, similar to the application of the uniform random field in the probability field simulation method. The new improved method is much better than the original one in obtaining the spatial structure characteristics of the permeability coefficient field and reducing the uncertainty of the implementation.
【学位授予单位】：中国地质大学(北京)
【学位级别】：博士
【学位授予年份】：2016
【分类号】：P641.2

【参考文献】