【作者】 张冰；

【导师】 刘财；

【作者基本信息】 吉林大学 ， 固体地球物理学， 2018， 博士

【摘要】 页岩已经成为非常规油气勘探的研究热点。为页岩储层描述提供准确丰富的参数有重要意义。使用地震数据准确地识别岩性和流体属性是储层描述的基础,但不同岩性存在参数重叠问题。基于贝叶斯理论的岩性识别可解决该问题。考虑到地下介质的连续性,可使用马尔可夫(Markov)随机链和随机场提供先验信息,提高岩性识别的准确性。使用地震数据进行储层参数反演则更加复杂,需要构建岩石物理模型来建立储层物性参数和各向异性参数间的联系,然后使用地震反演得到的弹性参数反演储层参数。但地震数据存在噪音,岩石物理模型无法准确地描述地下介质,反演过程存在着不确定性。本文通过统计岩石物理模型建立储层物性参数与弹性参数的定量关系,使用测井数据及井中岩石物理反演结果作为先验信息,将地震阻抗数据定量解释为储层物性参数、各向异性参数的空间分布。反演过程在贝叶斯框架下求得储层参数的后验概率密度函数,并从中得到参数的最优估计值及其不确定性的定量描述。在此过程中综合考虑了岩石物理模型对复杂地下介质的描述偏差和地震数据中噪声对反演不确定性的影响。在求取最大后验概率过程中使用模拟退火优化粒子群算法以提高收敛速度和计算准确性。将统计岩石物理技术应用于龙马溪组页岩气储层,得到储层泥质含量、压实指数、孔隙度、裂缝密度等物性参数,以及各向异性参数的空间分布及相应的不确定性估计,为页岩气储层的定量描述提供依据。针对含裂缝的页岩储层复杂的矿物组分与微观结构,本文首先介绍了应用自相容等效介质模型与Chapman多尺度孔隙系统模型建立的裂缝型页岩双孔隙系统岩石物理模型。考虑到页岩压实和裂缝对各向异性的影响,又构建了使用页岩压实模型和Chapman模型的各向异性页岩岩石物理模型,以准确地模拟页岩各向异性。作为贝叶斯理论的扩展,粒子滤波方法是一种递推贝叶斯估计方法。粒子滤波适用于非线性、非高斯问题,目前在参数估计、目标跟踪等领域得到广泛应用,在地球物理领域应用较少。本文提出了基于岩石物理模型和粒子滤波方法的横波速度预测方法。建立了适用于页岩横波速度预测的粒子滤波系统模型,使用该模型对实际测井资料进行计算,并对算法提升和参数选择做出分析。与传统反演方法对比,粒子滤波方法具有对先验准确性要求低、反演速度快、反演精度高、可提升空间大的优点。粒子滤波方法也被应用到页岩各向异性参数反演中。作为对比,首先使用了模拟退火优化粒子群算法及双孔隙系统模型,在反演过程中加入平滑约束项以减少多解性。然后建立了基于岩石物理模型和粒子滤波方法的页岩储层物性参数及各向异性参数反演方法,通过加入未知参数的先验信息,避免了多解性,使结果更加准确。将这两种方法同样应用于龙马溪组页岩气储层,反演得到的物性参数和各向异性参数与已有研究结果相一致,能为页岩储层评价提供多元化的信息。更多还原

【Abstract】 Shale has become an area of interest for unconventional hydrocarbon exploration.It is important to provide accurate and abundant parameters for shale reservoir characterization.The identification of lithology and fluid property by using seismic data is essential for reservoir characterization.However,parameters for different lithology always have overlap.The Bayesian theory is an available method to solve this problem.Besides,underground medium has continuity.The Markov chain and random filed can provide prior information of the lithology,which will improve the identification of lithology based on the Bayesian theory.Reservoir parameter inversion by using seismic data is more complicated.At first,a rock physics model is necessary to link the reservoir petrophysical parameters and anisotropy parameters.Then,the reservoir parameter inversion is processed by using elastic parameters which are obtained from seismic data.But there is uncertainty during the inversion,including noises of seismic data and errors of rock physics model.We build a stochastic rock physics model to link the reservoir petrophysical parameters and elastic parameters,then get the prior information from well data and well inversion result.Finally,the P-and S-wave impedances are translated into reservoir petrophysical parameter and anisotropy parameter sections.In the inversion process,the posterior probability distribution function(PDF)is achieved by using the Bayesian theory.Then the estimation of parameters and uncertainty of the inversion can be calculated from the posterior PDF.The uncertainty from seismic data and rock physics model are all taken account.The SA-PSO algorithm which combines the simulated annealing method and the particle swarm optimization method is used to get the maximum posterior probability.The algorithm is efficient and accurate in the optimization process.The stochastic inversion workflow is used in the Longmaxi shale gas formation to get the sections of clay content,clay lamination,porosity,fracture density,anisotropy parameters and uncertainty of the result.The estimated parameters can be used for better characterizations of shale gas reservoirs.In view of the complex mineral composition and microstructure of fractured shale reservoirs,we introduce a two-pore system model which combines the self-consistent effective media theory and Chapman’s model to account for the fractured shale.Considering the effects of shale compaction and fracture on anisotropy,an anisotropic shale rock physics model is constructed by combing the shale compaction model and Chapman’s model to simulate the anisotropy of shale accurately.As an extension of Bayesian theory,particle filter algorithm is a recursive Bayesian estimation method.Particle filter is applicable to nonlinear and non-Gaussian problems.It has been used in many fields,such as parameter estimation and target tracking.However,the algorithm is not very popular in geophysics.We present a shear wave velocity prediction method based on the rock physics model and particle filter.A particle filter system model for shale shear wave velocity prediction was established.The model was used in real well data,and analysis on algorithm improvement and parameter selection was made.Compared with the traditional method,particle filter has lower requirement for the accuracy of prior information,and better speed and accuracy.What’s more,the algorithm has potential to improve.Particle filter has also been used in anisotropy parameter inversion of shale.As a comparison,we first use the two-pore system model and the SA-PSO algorithm.In the inversion process,a smooth constraint is used to reduce the multiple solutions.Then,the inversion method of shale reservoir petrophysical parameters and anisotropic parameters based on rock physics model and particle filter is established.By using the prior information of unknown parameters,multiple solutions are avoided and the results become more accurate.These two methods are also used in the Longmaxi shale gas formation.The inversion result of reservoir petrophysical parameters and anisotropy parameters are consistent with former researches,which can provide abundant information for shale reservoir characterization.更多还原