地震属性在沉积相预测中的方法研究—序贯随机模式识别

Seismic attributes in facies prediction—sequential stochastic pattern recognition methods

沉积相揭示了目的层段的沉积环境、储集岩成因及其分布规律.通过沉积相的研究,油藏工程师对储层的认识更加细致.对于钻井资料较少的研究区域,如何用地震信息研究沉积相是一个值得探讨的问题.传统的沉积相预测方法主要包括地质统计学和模式识别等方法.首先,地质统计学方法使用井点数据进行插值,考虑到数据空间变异性特征,但是忽略地震数据.其次,模式识别多变量预测方法通过井、震多变量信息建立识别模式,但是识别建立过程中忽略空间数据的结构特征.针对传统方法在识别过程中存在的问题,本文以苏里格气田苏10区块盒8下段为例,提出一种综合考虑地震属性和空间变异性特征的序贯随机模式识别方法,将此方法应用到沉积相预测过程中.首先提取多种地震属性,然后利用降维和属性优选等方法筛选出对沉积相较敏感的地震属性,利用变差函数分析工具进行空间结构性分析,最后通过井、震结合建立识别模式开展沉积相研究.研究实例表明,相比最邻近法和序贯指示模拟,其沉积相预测精度分别提高24%和8%.因此,综合空间结构特征和多变量信息融合识别沉积相,使得平面沉积微相展布研究更为精确.

Sediment facies reveal the sediment environment of the interest interval,the origin and distribution rule of the reservoir.Through the research of sediment facies,the reservoir engineers can know about the reservoirs more detailed. For the region with scarce drilling data,it is questionable how to study the sediment facies with seismic information. The traditional methods of sedimentary facies prediction include geostatistical methods and pattern recognition methods, etc. Firstly, the former methods use interpolation with well data. The process takes into account of the variability of spatial data,but ignores the abundant seismic data.Secondly,the latter methods establish the recognition mode with multi-variable which integrates into the well data and seismic information. But the process ignores the feature of spatial structure.Considering the problems of two methods in the process of recognition,the paper takes the example of Su10 area in Sulige Gas Field and proposes a new prediction methods called sequential stochastic pattern recognition, which synthetically considers the seismic attributes and then applies it into the sediment facies prediction. The process is as follows. Firstly,a variety of seismic attributes are extracted and then the methods of reducing the dimensions of attributes and attribute optimization are used to filter out the seismic attributes which is sensitive to the sedimentary facies. We can also use the tool of variogram to analyze the spatial structure. Finally,it can build the recognition pattern through the integration of well points and seismic and then carry out the research to predict the sedimentary facies. Compared to the pattern recognition methods and geostatistical methods, the prediction accuracy increased by 24% and 8%. Therefore the study shows that it can effectively enhance the prediction accuracy with the integration of spatial structure and multi-variable.