基于深度学习的致密气储层矿物含量计算方法研究  

Research on Mineral Content Calculation Method for Tight Gas Reservoirs Based on Deep Learning

作　　者：李立冬 LI Li-dong(COSL Shanghai Branch,Shanghai 200335,China)

机构地区：[1]中海油服上海分公司,上海200335

出　　处：《价值工程》2024年第10期137-141,共5页Value Engineering

摘　　要：致密气储层非均质性强,储层中矿物类型多、岩性变化快,常规的测井法以及岩心鉴定法无法有效地进行全井段矿物含量计算,进而影响储层岩性识别、压裂方式优选和压裂液体系优化。利用测井参数结合岩心鉴定结果,建立各种矿物含量的深度学习计算模型,通过不同测井曲线组合方式,进行学习和验证,最终优选出不同矿物含量的计算参数为:黏土含量和伊蒙混层采用伽马、密度、中子、声波曲线、光电界面指数模型计算;石英含量采用自然伽马、中子、密度、声波、电阻、光电截面指数模型计算;方解石含量采用密度、中子、声波、电阻率模型计算;通过对53口井的验证,最终准确率为90%,可以为致密气储层矿物含量计算提供借鉴。Tight gas reservoirs have strong heterogeneity,multiple types of minerals in the reservoir,and fast changes in lithology.Conventional logging and core identification methods cannot effectively calculate the mineral content of the entire well section,which has a significant impact on reservoir lithology identification,fracturing method optimization,and fracturing fluid system optimization.By combining logging parameters with core identification results,a deep learning calculation model for various mineral contents is established.Through different combination methods of logging curves,learning and verification are carried out,and the optimal calculation parameters for different mineral contents are:clay content and Yimeng mixed layer are calculated using gamma,density,neutron,acoustic curve,and photoelectric interface index.The quartz content is calculated using natural gamma rays,neutrons,density,sound waves,resistance,and photoelectric cross-section index;The content of calcite is calculated using density,neutrons,sound waves,and electrical resistivity.Through verification of 33 wells,the final accuracy is 90.1%,which can provide reference for the calculation of mineral content in tight gas reservoirs.

关 键 词：致密气 矿物含量 深度学习 测井曲线 岩心鉴定 

分 类 号：P618.13[天文地球—矿床学]