威远地区早古生代筇竹寺组页岩储层有机碳预测方法研究  

Organic carbon prediction method research for shale reservoir of early paleozoic Qiongzhusi formation in Weiyuan area

作　　者：刘成 李俊翔 张庆 杨亚东 周昕 吴朝容[1] 李勇[1] 张兵[1] 李金玺[1] 韩建辉 LIU Cheng;LI Junxiang;ZHANG Qing;YANG Yadong;ZHOU Xin;WU Chaorong;LI Yong;ZHANG Bin;LI Jinxi;HAN Jianhui(School of Geophysics, Chengdu University of technology, Chengdu 610059,China;Shale gas project management department of CNPC Chuanqing Drilling EngineeringCompany Limited, Chengdu 610051,China)

机构地区：[1]成都理工大学地球物理学院,成都610059 [2]中国石油川庆钻探工程有限公司页岩气项目经理部,成都610051

出　　处：《物探化探计算技术》2021年第6期705-714,共10页Computing Techniques For Geophysical and Geochemical Exploration

基　　金：国家自然科学基金(41774095)。

摘　　要：总有机碳含量(TOC)是评价页岩生烃能力的关键性指标,岩心样品测试仅能获得离散的TOC含量,且成本较高。基于页岩在常规测井上的响应特征,运用多元回归分析法、拓展ΔlgR法、RBF神经网络法对威远地区筇竹寺组页岩TOC含量进行预测。结果表明:①优质页岩测井曲线响应特征通常表现为“四高两低一扩”,即高自然伽马、高电阻率、高声波时差、高中子、低密度、低光电吸收截面指数、井径扩径,同时测井曲线响应特征还受矿物成分的影响;②ΔlgR法运用于较深页岩储层的TOC预测时,需根据研究区具体地质情况进行合理的改进;③相比多元回归分析法和拓展ΔlgR法,RBF神经网络法能通过空间变换较精确的描述有机碳与测井参数的非线性关系,使预测的有机碳与实测值拟合度高,预测效果最好,为研究区最佳有机碳预测方法。Content of the total organic carbon(TOC)is a key index to evaluate the hydrocarbon production capacity of shale.Core sample testing can only obtain discrete TOC content and is costly.In this paper,based on the response characteristics of shale on conventional logs,multiple regression analysis,extendedΔlgR method and RBF neural network method were applied to predict the TOC content of shale in the Qiongzhusi Formation in Weiyuan area.The results show that:(1)the response characteristics of high-quality shale logging curves are usually characterized by"four highs,two lows and one dilation",that is,high natural gamma,high resistivity,high acoustic time difference,high neutron,low density,low photoelectric absorption cross-section index and well diameter dilation,and the response characteristics are also influenced by mineral composition;(2)theΔlgR method needs to be further improved according to the specific geological conditions of the study area when applied to the TOC prediction of deeper shale reservoirs;(3)compared with the multiple regression analysis method and the extendedΔlgR method,the RBF neural network method can describe the nonlinear relationship between organic carbon and logging parameters more accurately through spatial transformation,so that the predicted organic carbon fits the measured values well and has the best prediction effect.This is the best organic carbon prediction method in the study area.

关 键 词：筇竹寺组页岩 TOC含量预测 多元回归法 拓展ΔlgR法 RBF神经网络法 

分 类 号：P631.4[天文地球—地质矿产勘探]