地震岩相识别概率表征方法

储层岩相分布信息是油藏表征的重要参数,基于地震资料开展储层岩相识别通常具有较强的不确定性.传统方法仅获取唯一确定的岩相分布信息,无法解析反演结果的不确定性,增加了油藏评价的风险.本文引入基于概率统计的多步骤反演方法开展地震岩相识别,通过在其各个环节建立输入与输出参量的统计关系,然后融合各环节概率统计信息构建地震数据与储层岩相的条件概率关系以反演岩相分布概率信息.与传统方法相比,文中方法通过概率统计关系表征了地震岩相识别各个环节中地球物理响应关系的不确定性,并通过融合各环节概率信息实现了不确定性传递的数值模拟,最终反演的岩相概率信息能够客观准确地反映地震岩相识别结果的不确定性,为油藏评价及储层建模提供了重要参考信息.模型数据和实际资料应用验证了方法的有效性.

A probabilistic approach for seismic facies classification

The distribution information of reservoir facies is critical for reservoir characterization. However, because of the insufficiency of well-log data, reservoir facies classification in the early stage of oilfield exploration is mainly based on seismic data, which usually manifest strong uncertainty. Traditional method provides us with only a certain outcome of facies distribution that does not contain any uncertainty information of the inversion results. It increases the risk of reservoir characterization and decision-making in any petroleum reservoir.In order to assess the associated uncertainty of seismic facies classification, multistep inversion based on a probabilistic way is introduced in this study. We firstly built the statistical relationships between input and output parameters in each step of seismic facies classification, such as well-log facies definition, probabilistic scale change, and seismic inversion. Then, the probabilistic information of all steps was integrated in a Bayesian framework to compute the seismic facies probability. Furthermore, facies probability of oil shale that is the targeted facies in this case was evaluated by different threshold values. Experiments had been conducted on both synthetic and field data.Compared with the traditional method, the methodology in this paper takes the uncertainties in each step of seismic facies classification into account by a probabilistic multistep approach. The inverted facies probability contains not only the information of facies distributions in the target zone, but also the uncertainty information of seismic facies classification. It plays an important guide for reservoir characterization as well as modeling. By analyzing the probability of oil shale with different thresholds, the locations where have a high occurrence probability of oil shale were illustrated vividly. Seismic facies classification by the probabilistic multistep inversion brings much more information of facies distribution in the target zone than traditional method, which can only afford a certain outcome of facies distribution without any uncertainty information of inversion results. The methodology provides us a simple way to evaluate the uncertainty of seismic facies classification as well as the great value for risk management and optimal decision-making in the petroleum industry. However, the resolution of seismic facies probability is generally low, because the results of probabilistic multistep inversion share the same resolving power with seismic data. In the light of this problem, when the hard data is sufficient, the computed facies probability can be used as the conditional information in the reservoir modeling for acquiring a high-resolution outcome of facies distribution.