松辽盆地东南隆起区上白垩统青山口组油页岩有机碳含量预测及效果评价

Prediction and effect evaluation of organic carbon content of oil shale in Upper Cretaceous Qingshankou Formation in Southeast Uplift of Songliao Basin

通过建立实测有机碳含量与测井资料之间的关系进行有机碳含量预测,可以克服有机碳含量分析实验取心不全面、测试数据少、测试费用高等缺点。通过分析松辽盆地东南隆起区上白垩统青山口组油页岩测井资料与实测有机碳含量关系,根据自然伽马、电阻率、声波时差、密度与实测有机碳含量的相关性,建立了东南隆起区青山口组油页岩改进ΔlgR模型、多元逐步回归模型和BP神经网络模型,定量预测了JFD⁃8井青山口组油页岩有机碳含量随深度变化关系,分析了3种模型的预测w(TOC)和实测w(TOC)的误差。研究认为:BP神经网络模型适用于数据量大的地层,受岩性变化和压实作用等因素的影响小,当油页岩层的w(TOC)变化范围大时更具优越性;多元逐步回归模型适用于测井曲线对油页岩层段响应好的地层,显著参数越多预测越精确;改进ΔlgR模型适用于岩性单一的地层,操作简便,但误差较大;勘探开发初期,当测井及地化资料不充足时建议使用改进ΔlgR模型和多元逐步回归模型,勘探开发后期,当测井及地化资料充足时建议使用多元逐步回归模型和BP神经网络模型。该研究成果为松辽盆地油页岩勘探开发提供了有力的技术支持。

By establishing relationship between measured organic carbon content and logging data,the prediction of organic carbon content can overcome shortcomings of incomplete coring,less tested data and high test cost in TOC analysis experiment.By means of analyzing relationship between logging data and measured organic carbon content of oil shale in Upper Cretaceous Qingshankou Formation in Southeast Uplift of Songliao Basin,and with the help of the correlations between gamma⁃ray,resistivity,acoustic,density and measured organic carbon content,improvedΔlgR model,multiple stepwise regression model and BP neural network model are established to quantitatively predict the variation of organic carbon content of oil shale in Qingshankou Formation of Well JFD⁃8 with depth,and errors between predicted w(TOC)s by three models and measured w(TOC)s are analyzed.The studies show that BP neural network model is suitable for the strata with a large amount of data,which is less af⁃fected by lithology change and compaction with more advantageous when the w(TOC)of oil shale layer varies in a large range;multivariate stepwise regression model is suitable for the formation with good logging response to the oil shale intervals,and more significant parameters get more accurate prediction;improvedΔlgR model is suitable for the formation with single lithology,and the operation is simple but with more errors.In the early stage of exploration and development when logging and geochemical data are insufficient,it is suggested to use improvedΔlgR model and multiple stepwise regression model;in the later stage when logging and geochemical data are sufficient,multi⁃ple stepwise regression model and BP neural network model are suggested.The research on logging prediction model of organic carbon content provides technical support for oil shale exploration and development in Songliao Basin.