节点文献

多尺度地震资料正反演方法研究

Method Research of Multiscale Seismic Data Modeling and Inversion

【作者】 曹丹平

【导师】 印兴耀;

【作者基本信息】 中国石油大学 , 地质资源与地质工程, 2008, 博士

【摘要】 油藏地球物理是地球物理技术面向油气藏开发生产综合应用的新型学科,油藏地球物理的内涵包括油藏描述与油藏管理,利用地球物理方法进行油藏圈定和描述,并在油藏开发开采过程中监测油藏变化。油藏地球物理依托于高精度三维地震、井中地震、时延地震、多波多分量地震等多种地震技术的发展,其中高精度三维地震与其它技术的综合应用是油藏地球物理的核心技术。开展多尺度地震资料正反演方法研究,提高多尺度地震资料协同解决油藏地质问题的能力;基于流变学理论研究高精度粘弹性波动方程正演模拟方法,准确认识地质体在不同观测系统下的多尺度地震响应特征;利用贝叶斯反演理论建立多尺度地震资料联合反演方法,减小地震反演多解性、提高反演分辨率,充分发挥地震资料为油田开发服务的作用。地震波在实际介质中的固有传播衰减特征很复杂,导致在理论上出现了多种描述粘弹性介质的衰减模型,但大部分用于粘弹性介质波动方程时间域正演模拟的衰减模型对其常Q特征都存在较大的近似。基于流变学理论研究,开展了标准线性体、Kelvin等多种衰减模型对粘弹性介质波动方程常Q正演模拟的适应性研究,分析了能够用于精确描述粘弹性介质中地震波衰减规律的多种广义流变学模型特征,并证明了广义流变模型之间具有等价性。分析了不同阶次广义Maxwell流变模型对粘弹性介质常Q衰减特征的拟合程度,完善了利用广义流变模型开展时间域粘弹性介质波动方程常Q正演模拟的基础,并将模型特征与常用的Futterman等衰减模型特征进行比较。对多种衰减模型开展了一维平面波正演模拟分析,解释了由于模型不同所引起的地震波传播衰减特征差异,指出了广义流变模型在波动方程时间域正演模拟中的优势。将广义流变模型引入到时间域粘弹性介质波动方程常Q正演模拟过程中,并分别针对广义Maxwell流变模型和广义标准线性体两种模型特征进行分析,对前者考虑了两种非弹性函数定义方式来建立对应的波动方程,分别采用弹性模量和松弛模量来描述基于广义标准线性体模型的波动方程,重新定义记忆变量使得通过两种广义流变模型所建立的波动方程表达式得到统一。完整的给出了基于两种模量表示的各种广义流变模型所对应的位移-应力、位移、位移-速度-应力和速度-应力方程,指出了早期波动方程由于采用松弛模量进行表示所引起的误解。基于五阶广义Maxwell流变模型建立了高精度粘弹性介质高阶交错网格有限差分波动方程常Q正演模拟表达式,采用PML边界条件有效的消除人工边界反射,讨论了模拟过程中的稳定性条件和震源问题,将有限差分正演模拟结果与二维均匀粘弹性介质中的解析解进行对比,分析了高精度粘弹性介质波动方程有限差分数值模拟方法的精度。采用高精度粘弹性介质常Q波动方程正演模拟方法开展模型响应特征分析认识到地震波在粘弹性介质中具有较强的衰减频散特征,随着Q值的减小和传播距离的增大,地震波产生振幅衰减、相位改变、主频降低等现象。针对不同的观测系统开展正演模拟,通过波场、共炮集及偏移成像剖面特征进行对比分析可知,地面地震、VSP和井间地震三种多尺度地震资料之间既有联系也有差异,联系在于多尺度地震资料是相同地下地质体在相应观测方式下的地震波响应特征;差异在于资料所包含的波场信息丰富程度不一样、反映地质体信息的范围不一样、分辨地质体细节特征的能力也不一样。加入近地表强衰减层的影响表明,多尺度地震资料之间存在一定的相位差异,进一步认识了实际多尺度地震资料之间既具有一致性又存在一定差异的特征。基于贝叶斯反演理论,通过似然函数有机的将地面地震、VSP和井间地震三种多尺度地震资料进行整合,建立起基于修正柯西分布的多尺度地震资料联合反演方法体系,并完善了联合反演流程。利用一维模型验证了联合反演算法的正确性,加噪后通过卡方分布测试了联合反演参数的合理选择范围,并采用二维多尺度正演模拟资料验证了联合反演方法的有效性及对多尺度资料的适应性。对联合反演方法进行退化建立了时间域单一尺度地震资料反演方法;针对连井井间地震剖面的特征重点强调了高精度深时转换、井地时深关系一致性及加强横向约束等改进策略,完善了深度域井间地震资料的反演方法和流程。将多尺度地震资料反演方法对胜利油田垦71区块开展了应用研究,利用深度域井间地震反演方法获得连井井间地震波阻抗反演剖面,从该反演结果中提取的虚拟井能够有效的匹配地面地震资料反射特征。虚拟井约束下的井地联合反演只能够提高地面地震反演精度,而基于贝叶斯理论框架建立的多尺度地震资料联合反演方法则能够有效的提高地面地震资料反演结果的分辨率,实现多尺度地震资料特征的融合。

【Abstract】 Reservoir geophysics is a new discipline that geophysics technique towards the integrated application of oilfield development and production, the reservoir description and reservoir management is included. The geophysics technique is adapted to determin the reservoir, descript the reservoir and monitor the change in the production. It is relying on the development of 3-D high-precision seismic technique, borehole seismic technique, time-lapse seismic technique and multi-wave/multi-component technique, the integrated research of high-precision 3-D seismic and other methods is the key technology of reservoir geophysics. Research the modeling and inversion method of multiscale seismic data is improving the capacity of seismic data to resolve the geology problem. Develop the high precision seismic modeling method of wave-equation in viscoelastic media, determine the mulitscale seismic response characteristics of geology with different observation system. Build up the multiscale seismic data joint inversion method based on the Bayes theory, overcome the multiple solution of inversion problem and improve the resolution of seismic data inversion, then service for the oilfield development with seismic data.The nature attenuation law of seismic propagating in actual media is very complicated. many attenuation model of attenuation model is present in theory. The approximation to Q is obvious to most models in seismic wave-equation modeling at viscoelastic media in time domain. Based on the rheology theory, the adaptability of SLS (Standard Linear Solid), Kelvin and other attenuation model is researched in viscoelastic seismic modeling. The generalized rheology models that can be use to accurately describe the seismic wave propagating and attenuation characteristics have been analysied, the equivalence of generalize Maxwell model and generalized SLS model is proofed. The fitting degree to the constant Q of different order generalized Maxwell model is analysized and compared with the Futterman model and other models, the foundation is perfecing to modeling the constant Q characteristics in viscoelastic media. One dimension plane wave modeling is done with these attenuation models, the difference of seismic wave propagation and attenuation characteristic can be inteprated with these results, the advantage of generalized rheology model is clear.The generalized rheology model is used in wave-equation seismic modeling at viscoelastic media in time domain. For the generalized Maxwell model, two kinds of unelastic function is adapted to derive the wave-equation. The unelastic modulus and relaxed modulus have been used to build up the wave-equation for generalized SLS model. The unified expression between these two kinds of generalized rheology model after the memory variable is redefined. The expression of displacement equation, the displacement-stress equation, the displacement-velocity-stress equation and the velocity-stress equation is presented according different modulus form.With the equation expansion or variable transform, the commom viscoelastic wave equation in articles can be get, the disadvantage of early equation expressed by relaxed modulus is clear. The high-order staggered grid finite difference scheme of high-precision viscoelastic wave equation based on five-order generalized Maxwell model is present. It is the high-precision wave-equation modeling method to constant Q in viscoelastic media. The PML boundary condition is used to absorb the artificial boundary reflection, the stability condition and source problem is discussed, the high-order finite difference scheme is accurate enough according to the comparison with the analytical resolution of 2D homogeneous viscoelastic media.According the high-precision viscoelastic modeling, the attenuation and dispersion characterstic of seismic wave is obvious in viscoelastic media. With the Q decrease and the propagation distance increase, the amplitude is attenuated. the phase is lagged. the dominant frequency is decreased. Modeling with different observation system and compare with the wave field, common shot gather and migrated section characteristic, the relationship and difference among multiscale seismic data can be recognized. The relationship is multiscale seimic data is the seismic wave reflection response of the same geological body under different observation way. The difference is represented at the wave field information richness, the geological range and the detail characteristic in the multiscale seismic data. When the near-surface strong attenuation layer is considered, the phase difference is obvious between surface seismic data and others, the relationship and difference among multiscale seismic data can be inteprated.Based on the Bayesian theorem, the new likelihood fundction is integrated with surface, vsp and crosswell seismic data, the multiscale seismic data joint inversion framework is built with modified Cauchy distribution and the inversion flow is improved. The one-dimension model testing shows that the method is right, the chi-square distribution is used to select the parameter range of the joint inversion with gauss noise, the joint inversion flow is valid to the two-dimension multiscale seismic data modeling according to the wave equation. For the singlescale seismic data, the crosswell seismic data inversion workflow in depth domain is improved. The high-precision interpolation function was adapted in depth-time transformation, the reflection characteristic consistence between surface seismic data and crosswell seismic data is emphasized, the transverse constraint is strengthened according to the characteristic of multi-crosswell seismic data.The multiscale sesmic data inversion method is tested in Ken71 area of Shengli oilfield. The reasonable inversion result of multi-crosswell seismic data is get, the pseudo well exracted from it can mathch the characteristics of surface seismic data well. The constraint inversion with pseudo well can improve the precision of result. The multiscale seismic joint inversion method can reduce the multiplicity of solution, improve the inversion resolution, the multiscale data can be matched with each other according to the joint inversion result.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络