论文详情
加拿大阿萨巴斯卡地区油砂储层内的侧积砂层和泥岩隔夹层发育,井间非均质性强,储层内岩性的空间精确刻画对蒸汽辅助重力泄油(SAGD)开发井部署至关重要,而井孔储层段的岩性识别是进行空间岩性研究的基础。目前该地区井孔岩性划分主要依据钻井取心数据,成本很高,如果使用常规测井曲线就能准确识别岩性,则可以降低生产成本。以该地区Kinosis工区为研究对象,采用了基于贝叶斯概率模型无监督学习的测井岩相分析方法,选用常规测井曲线数据,在主成分分析(PCA)基础上进行聚类分析,得到井位处的垂向岩相分布。综合测井曲线、测井解释结果、岩心照片等地质资料,对岩相结果进行统计分析和标定,确定储层内每个岩相的岩性特征和地质特征。应用结果表明,无监督学习测井岩相分析技术充分利用数据之间的内在关系,无需提供先验的岩性模型,结果更为客观;通过标定,岩相识别结果与取心数据吻合率高,展示了一种利用常规测井曲线预测油砂储层井孔岩性的较为经济的研究方法。
The lateral sands and interlayers in the oil sand reservoirs in the Athabasca area of Canada are well developed.Substantial heterogeneity exists across the wells.A spatially accurate characterization of the reservoir lithology is critical for the deployment of steam-assisted gravity drainage wells and constitutes the basis for spatial lithology research.Currently,the lithostratigraphy of the wells in this area is mainly based on data from drilling cores,which are costly.If conventional logging curves can be used to accurately identify the lithology,the production costs can be reduced.In this study,the Kinosis block in the Athabasca area was taken as the research object.A logging lithofacies analysis method based on a Bayesian probability unsupervised learning model was adopted.The method consists of the following steps.First,principal component data analysis is performed on the selected conventional logging curve data.Then,multiple probability models are established according to the inherent probability distribution of the data.Finally,cluster analysis and Bayesian criterion discrimination are performed on all sample points in each model.Through iterative calculations,the system automatically learns the strengths and weaknesses of the discriminant model and outputs classification results that are considered optimal.During the process,the model parameters and the number of clusters are selected using unsupervised autonomous learning,which is completely data-driven and thus free from human bias.The lithologic and geological characteristics of each lithofacies in the reservoir are determined by statistical analysis and calibration of the lithofacies from geological data of the logging curves,logging interpretation curves,and core photographs.The application to the study site showed that the proposed method can exploit the inherent relationships among the input data without the need of an a priori lithological model,thereby providing more objective results.The lithofacies identified by the proposed method matched well with the core data after calibration.
国家科技重大专项“油砂SAGD开发地质油藏评价及方案优化技术”(2016ZX05031-003)资助。