深度学习在元素测井中的应用  

Application of Deep Learning in Element Logging

作　　者：葛云龙 岳爱忠 宋森 张磊[3] 骆庆锋 何绪新 GE Yunlong;YUE Aizhong;SONG Sen;ZHANG Lei;LUO Qingfeng;HE Xuxin(Logging Technology Research Institute,China National Logging Corporation,Xi’an,Shaanxi 710077,China;Well Logging Technology Pilot Test Center,China National Petroleum Corporation,Xi’an,Shaanxi 710077,China;Southwest Branch,China National Logging Corporation,Chongqing 409912,China)

机构地区：[1]中国石油集团测井有限公司测井技术研究院,陕西西安710077 [2]中国石油天然气集团有限公司测井技术试验基地,陕西西安710077 [3]中国石油集团测井有限公司西南分公司,重庆409912

出　　处：《测井技术》2023年第4期452-456,共5页Well Logging Technology

基　　金：中国石油天然气集团有限公司科学研究与技术开发项目“175℃、200 h测井高性能中子管研究”(2020B-3714);中国石油集团测井有限公司项目“随钻孔隙度系列优化升级与集成配套”(CPL2021A02)。

摘　　要：小直径中子伽马仪器伽马能谱测量计数率低,采用常规解谱方法,元素干重计算误差较大。基于蒙特卡罗方法构建学习样本,并通过增加噪声的方式提高学习模型鲁棒性,利用深度学习对元素测井进行解谱、矿物含量与孔隙度计算。研究结果表明:仅用模拟数据构建的学习模型,当元素种类较少时,深度学习的解谱结果可准确计算元素干重。当元素干重准确时,深度学习能准确地得到矿物含量。基于深度学习,可以在元素测井的基础上获得岩石基质密度,现场实例显示,通过深度学习计算的孔隙度接近岩心孔隙度。The count rate of gamma spectrum measured by the small size element logging tool is low.Based on the conventional spectral analysis method,the calculation error of element dry weight is large.In this paper,learning samples are constructed based on Monte Carlo method,and the robustness of the learning model is improved by adding noise.Deep learning is used to analyze spectra and calculate the mineral content and porosity.The results show that the learning model constructed with only simulated data is accurate when there are fewer elements.When the element dry weight is accurate,deep learning can get the mineral content accurately.Based on the element logging,the rock matrix density can be obtained based on the deep learning.The field example shows the porosity calculated by the deep learning is close to the core porosity.

关 键 词：深度学习 元素测井 解谱 矿物含量 孔隙度 

分 类 号：P631.84[天文地球—地质矿产勘探]