潜江凹陷潜江组盐间页岩油岩石物理建模研究

潜江凹陷潜江组地层盐间页岩层矿物类型复杂,相关的岩石物理建模研究较少,因此仅利用常规测井资料进行地层参数解释的难度较大,后期甜点预测存在风险。首先,选择合适的有监督的机器学习方法解决利用常规资料难以准确求解多种矿物含量的问题,具体如下:①将LithoScanner岩性成果作为样本数据进行训练及预测,并建立多个矿物含量的预测模型;②将随机森林算法作为研究区较为准确的机器学习方法,并将矿物含量预测结果用来直接约束测井评价模型进行优化求解;③得到的测井解释成果与实验室X衍射成果数据基本一致,说明了该机器学习方法用于解决研究区复杂岩性问题时的可行性和优势。然后,由实验室岩石物理测试分析确定盐间页岩层大部分岩石为各向同性介质,饱油条件下其速度存在较为明显的频散现象,考虑到盐间页岩层矿物类型多且主要发育矿物的含量相对均等,故将以多种矿物为骨架的自洽各向同性介质模型作为矿物和干岩石混合模型,结合Boris全频段流体置换模型,建立了研究区盐间页岩层岩石物理模型及建模流程。最后得到的模型预测结果和实际测井数据吻合度高,验证了模型的准确性和可靠性,也为潜江组页岩油“甜点”的地震预测提供了科学依据。

Rock physics modeling of inter-salt shale oil in the Qianjiang formation of Qianjiang sag,China

The inter-salt shale stratum of the Qianjiang Formation in the Qianjiang sag is characterized by a frequent vertical alternation of mudstone,sandstone,carbonate,glauberite,and salt rocks of complex and diverse mineral types.The interpretation of petrophysical parameters only on the basis of conventional logging data is challenging.In this work,a supervised machine learning method was used to accurately calculate the content of various minerals.Litho-scanner data were selected for data training to obtain multiple models for mineral content prediction.A random forest algorithm was chosen as the machine learning method to be applied on the study area.The evaluation results can be taken as the input for the optimal logging interpretation model to further constrain and optimize the results of the interpretation.Subsequently,rock physics testing in the laboratory confirmed that most of the rocks in the inter-salt shale are isotropic,with great velocity dispersion under the condition of oil saturation.Considering the presence of many types of minerals in the inter-salt strata and that the main minerals are present in relatively equal proportion,a self-consistent isotropic medium model,which use multiple minerals for the rock matrix,was selected as the mixed model of minerals and dry rock.By combining this model with the Boris full-band fluid substitution model,the rock physics model of the inter-salt shale strata in the study area could be established.The prediction results using the model were consistent with the measured logging data,thereby verifying the effectiveness of the model.