多元信息融合在地震属性储层预测中的应用

发布时间：2018-05-19 07:54

本文选题：塔河油田 + 地震属性　；参考：《成都理工大学》2015年硕士论文

【摘要】：塔河油田碳酸盐岩油藏埋藏较深,一般大于5300米,基岩不具备储集性,储集空间以溶洞、裂缝为主。在TH10前期研究中,主要以T74不整合面以下0~80m范围内的风化壳岩溶及部分垂直岩溶段为主体,而针对更深层的岩溶及溶洞体很少涉及。然而钻井等资料表明,在目前研究认识的含油缝洞底界依然存在可开发的洞穴型储集体。因此,加大中深部洞穴的储层预测研究,能够深化认识岩溶发育规律,拓展中深部储集体的资源领域,为塔河油田的产建转型提供地质依据。地震属性分析技术在储层预测中应用较广,然而利用常规地震属性进行储层预测时效果较差、精度较低,预测结果常常具有多解性。而信息融合方法可以解决多个属性之间的相关性等矛盾,产生的融合属性比单一的属性数据更准确更有效。因此,通过将信息融合技术应用于地震属性储层预测中,可以更好检测储层溶洞体,同时对解决多解性问题具有重要的作用。本论文主要研究了主成分分析、核主成分分析、模糊C均值聚类、核模糊C均值聚类等四种信息融合算法的基本原理与计算步骤,并将其应用到TH10中深部“串珠状”溶洞储集体的检测中,并主要取得了方法、软件、应用等三个方面的研究成果。(1)方法成果方面。主成分分析是一种较好的信息融合方法,本质上是一种线性方法,适用于属性间相关性较强的情形;核主成分分析的非线性处理能力明显强于主成分分析,在信息融合效果上也具有更好的检测能力;模糊C均值聚类是一种基于模糊数学的聚类分析融合方法,非线性处理能力不强;对于非线性关系较强的属性数据,核模糊C均值算法能解决线性空间中不能被线性分割的问题,因此比模糊C均值聚类的融合效果更好。(2)软件成果方面。本文主要基于Visual C++6.0平台,对提取出的地震属性数据分别进行四种信息融合算法的代码实现。函数模块在储层预测中效果较好,具有一定的应用开发价值。(3)应用成果方面。本文应用成果包括地质模型、连井剖面、研究区域平面三个部分:(a)地质模型上的成果。本文建立了多套不同的溶洞地质模型,通过褶积得到正演结果,然后分析了地震响应特征。当溶洞规模较小时,地震波不能识别出溶洞;随着溶洞规模的增大,地震响应强度也随之增大;当充填物的速度减小时,地震响应强度增强;在纵向可分辨情况下,溶洞间隔增大,地震响应强度也随之增大。(b)连井剖面上的应用成果。将融合方法应用于连井剖面,可以知道信息融合参数能准确检测出溶洞的位置和规模,且在一定程度上能检测出溶洞体的流体属性。从融合效果上看,核主成分分析的检测能力比主成分分析强,核模糊聚类也比模糊聚类强。(c)研究区域平面上的应用成果。以TH10鹰山组二段为研究目标,在精细层位追踪后提取了多种属性参数,用信息融合方法进行地震属性融合,获得目的层溶洞平面分布特征:融合平面上储层溶洞体主要呈短轴状散乱分布,总体呈“东多西少,北多南少”的特征。综合生产开发成果提出,“中等融合检测值、溶洞集中且平面不连通”的地区,为有利储层和可能的油气富集区。本次论文采用多元信息数据融合方法在该研究区首次进行了储层溶洞检测应用,取得较好的预测效果,具有一定的创新性,对实际生产也具有一定的指导意义。
[Abstract]:The carbonate reservoir in Tahe oilfield is deep buried, generally more than 5300 meters, and the bedrock is not possessed of reservoir property. The reservoir space is mainly composed of karst caves and cracks. In the early study of TH10, the main body of the weathered crust karst and some vertical karst segments in the 0~80m range below the T74 unconformities is mainly, but it is rarely involved in the deeper karst and karst cave bodies. Drilling and other data show that there is still a developing cave type reservoir in the bottom boundary of the oil bearing seam. Therefore, increasing the reservoir prediction research in the middle and deep parts can deepen the understanding of the law of karst development, expand the resources of the middle and deep reservoirs, and provide geological basis for the transformation of Tahe oilfield. Analysis technology is widely used in reservoir prediction, however, the effect of reservoir prediction is poor, precision is low, and the prediction results often have multiple solutions. The information fusion method can solve the contradiction between multiple attributes, and the fusion is more accurate and effective than the single attribute data. By applying the information fusion technology to seismic attribute reservoir prediction, it can better detect the reservoir cavern and play an important role in solving the problem of multi solution. This paper mainly studies the basic principles and plans of four information fusion algorithms, such as principal component analysis, kernel principal component analysis, fuzzy C means clustering, and kernel fuzzy C mean clustering. It is applied to the detection of the deep "bead like" cave storage in TH10, and the main results are obtained in three aspects, such as method, software, application and so on. (1) the method results. The principal component analysis is a better information fusion method, which is essentially a linear method, which is suitable for the strong relationship among the attributes. The nonlinear processing ability of the kernel principal component analysis is stronger than the principal component analysis and has better detection ability in the information fusion effect. Fuzzy C means clustering is a clustering analysis fusion method based on fuzzy mathematics, and the nonlinear processing ability is not strong. For the attribute data with strong non linear relation, the kernel fuzzy C mean algorithm can be used. Solving the problem that linear space can not be segmented linearly, so the fusion effect is better than that of fuzzy C means clustering. (2) in the aspect of software achievement. This paper mainly based on the Visual C++6.0 platform, carries out the code realization of four information fusion algorithms for the extracted seismic attribute data. The function module has a good effect in the reservoir prediction, and has a good effect. Application development value. (3) application results. The application results include geological model, well section and three parts of regional plane: (a) geological model. In this paper, a number of different karst cave geological models are established, and the results are obtained by convolution, and then the characteristics of seismic response are analyzed. Seismic waves can not identify the karst cave; with the increase of the cave scale, the seismic response strength increases, and the seismic response strength increases when the velocity of the filling decreases. In the case of longitudinal resolution, the interval of the cave increases and the seismic response strength increases. (b) the application results on the well section are applied to the cross section, It can be known that the information fusion parameters can accurately detect the location and size of the cave, and to some extent can detect the fluid properties of the cavern. From the fusion effect, the detection ability of the nuclear principal component analysis is stronger than the principal component analysis, and the kernel fuzzy clustering is stronger than the fuzzy clustering. (c) the application results on the regional plane are studied. TH10 Yingshan is used. The two section of the group is the research target. After the tracking of the fine layer, a variety of attribute parameters are extracted. The information fusion method is used to fuse the seismic attributes and obtain the characteristics of the plane distribution of the target cave. The reservoir bodies in the fusion plane are mainly in the short axis and scattered distribution, and the overall production and development results are "East and West, and the north is more and more South". It is proposed that "medium fusion detection value, concentration of karst cave and unconnected plane" is a favorable reservoir and possible oil and gas enrichment area. This paper uses multiple information data fusion method in this study area for the first time to carry out the application of reservoir karst cave detection in this study area, which has obtained good prediction effect, and has certain innovation and is also of practical production. A certain guiding significance.
【学位授予单位】：成都理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P631.4;P618.13

【参考文献】