摘要
低场核磁共振技术已被广泛应用于石油勘探与开发领域,在储层评价、产能预测等方面发挥着重要作用。但低场核磁共振信号极其微弱、信噪比低,导致核磁共振弛豫谱信号重叠、流体组分定量评价困难。因此,开展相对应的低信噪比核磁共振数据处理方法研究,对提升核磁共振测井技术在储层评价中的应用十分重要。随着人工智能技术的飞跃发展,越来越多学者提出将机器学习与行业融合,充分利用已有数据提高生产效率。该文首先系统归纳了机器学习在核磁共振测井中的应用及发展现状,其次分析归纳了机器学习在核磁共振测井数据处理中的研究,分为提高核磁共振测井信号的信噪比、提高反演弛豫谱的分辨率和提升弛豫谱的解释应用精度这3个应用方向,以准确完成流体组分划分和定量计算。最后基于充分调研和综合分析,对机器学习在核磁共振测井数据处理的应用发展提出了思考。
Low-field nuclear magnetic resonance(NMR) technology has been widely used in petroleum engineering,which plays a critical role in reservoir evaluation and production prediction.However,the extremely weak signal and low signal-to-noise ratio(SNR) of low-field NMR leads to overlapping signals in the NMR relaxation spectra and difficulties in the quantitative evaluation of fluid components.Therefore,it is very important to develop novel and practical NMR data processing methods to improve the application effects of NMR logging technology.With the rapid development of artificial intelligence technology,many scholars have proposed machine learning methods to improve the industry's productivity.Firstly,this paper summarized the application and development of machine learning used in NMR logging.Secondly,the progress of machine learning methods applied in NMR logging data processing are analyzed,which are divided into three aspects including SNR enhancement,spectra resolution improvement,and quantitative fluid identification improvement.Finally,the future development of machine learning applied NMR logging data processing is summarized and recommended.
作者
罗刚
罗嗣慧
肖立志
傅少庆
张家伟
邵蓉波
LUO Gang;LUO Sihui;XIAO Lizhi;FU Shaoqing;ZHANG Jiawei;SHAO Rongbo(National Key Laboratory of Oil and Gas Resources and Engineering,China University of Petroleum(Beijing),Beijing 102249,China;College of Carbon Neutral Energy,China University of Petroleum(Beijing),Beijing 102249,China;College of Information Science and Engineering/College of Artificial Intelligence,China University of Petroleum(Beijing),Beijing 102249,China;Well Logging Technology Institute,China National Logging Corporation,Beijing 102206,China;Well Logging Technology Pilot Test Center,China National Petroleum Corporation,Xi’an,Shaanxi 710077,China)
出处
《测井技术》
CAS
2023年第6期643-652,共10页
Well Logging Technology
基金
国家自然科学基金“方位扫描核磁共振探测新方法与实验验证”(42204106)
中国石油大学(北京)拔尖人才科研基金“超临界CO_(2)储层物性变化规律与核磁表征方法研究”(2462023BJRC002)。
关键词
核磁共振测井
机器学习
深度学习
数据处理
解释与应用
nuclear magnetic resonance logging
machine learning
deep learning
data processing
interpretation and application
