摘要
随着全球能源领域“数智”及人工智能时代的到来,油气勘探面临着前所未有的机遇和挑战。深度学习技术作为人工智能领域的重要分支,在油气勘探中应用场景广泛,对其综合分析可为解决新时代油气勘探中复杂问题提供新的思路和方法。为此,通过深度学习技术在地震勘探、测井、岩石薄片鉴定、油藏地质建模、油气大模型等勘探领域中的应用综述,重点阐述了卷积神经网络(CNN)及其变体在地震勘探中的应用,分析了其优势与局限性,并根据目前深度学习技术面临的挑战,指出了油气勘探领域对大模型的探索方向和应用潜力。研究结果表明:①深度学习技术已广泛应用于地震资料解释、测井分析、油藏评价等油气勘探领域,以CNN为代表的深度学习方法在去噪、速度建模、构造解释、地震反演等地震资料的处理与解释方面展现出巨大的应用潜力;②深度学习技术在测井评价、岩石薄片鉴定、油藏地质建模、油气大模型等任务中不仅能有效地提升勘探效率和精度,还能从复杂数据中发现新的规律,提出对油气勘探中非线性问题的解决方案;③深度学习技术在训练数据的质量及代表性、数据集的整合和共享、技术合作与交流等方面还存在问题与挑战。结论认为,基于大数据的深度学习技术将是未来油气勘探的主要技术手段,应建立一套完善的数据管理框架,注重数据标准化和质量控制,创新或持续优化现有模型,加大数据整合与共享,注重地质复杂性和非结构化解释等方面工作,以上工作将有助于推动油气地质勘探领域的科技进步和数智化发展。
With the emergence of the"digital intelligence"and artificial intelligence era in the field of global energy sector,oil and gas exploration is confronted with unprecedented challenges and opportunities.Deep learning technology,as a pivotal branch in the domain of artificial intelligence,has been extensively applied across a diverse range of scenarios in oil and gas exploration.A comprehensive analysis of its applications is capable of offering novel ideas and methodologies for addressing the intricate problems encountered in oil and gas exploration during the new era.After reviewing the applications of deep learning technology in exploration domains including seismic exploration,logging,rock thin section analysis,reservoir geological modeling,and large-scale oil and gas models,this paper places a particular emphasis on elaborating the applications of convolutional neural network(CNN)and its variants in seismic exploration,and analyzes their advantages and limitations.And in light of the current challenges to deep learning technology,the exploration directions and application potentials of large models within the field of oil and gas exploration are clarified.The following results are obtained.First,deep learning technology has been widely applied in the field of oil and gas exploration such as seismic data interpretation,logging analysis,and reservoir evaluation.Deep learning approaches represented by CNN have manifested significant application potentials in the processing and interpretation of seismic data,such as denoising,velocity modeling,structure interpretation,and seismic inversion.Second,in tasks such as logging evaluation,rock thin section analysis,reservoir geological modeling,and large-scale oil and gas models,deep learning technology can not only effectively enhance the exploration efficiency and precision,but also uncover new patterns from complex data to provide solutions to the nonlinear problems present in oil and gas exploration.Third,difficulties and challenges still remain in deep learning technology such as the quality and representativeness of training data,the integration and sharing of datasets,as well as technological cooperation and communication.In conclusion,deep learning technology based on big data will serve as the primary technical means in future oil and gas exploration.It is necessary to establish a complete set of data management framework,pay attention to data standardization and quality control,innovate and continuously optimize existing models,strengthen data integration and sharing,and concentrate on geological complexity and unstructured interpretation.All these endeavors will be conducive to facilitate the scientific and technological advancement and digital-intelligent development in the field of petroleum geological exploration.
作者
于强
王宝江
张禄明
田涛
高志亮
任战利
畅伟
YU Qiang;WANG Baojiang;ZHANG Luming;TIAN Tao;GAO Zhiliang;REN Zhanli;CHANG Wei(School of Earth Science and Resources,Chang'an University,Xi'an,Shaanxi 710054,China;Research Institute of Wisdom Oil&Gas Field of Chang'an University,Xi'an,Shaanxi 710054,China;Key Laboratory of Liangshan Agriculture Digital Transformationof Sichuan Provincial Higher Education Institutions,Xichang,Sichuan 615013,China;School of Information Technology,XichangUniversity,Xichang,Sichuan 615013,China;School of Mechanical and Electrical Engineering,Xichang University,Xichang,Sichuan 615013,China;Shaanxi Coal Geology Group Co.,Ltd.,Xi'an,Shaanxi 710048,China;Northwest University,Xi'an,Shaanxi 710069,China)
出处
《天然气工业》
北大核心
2025年第5期43-56,共14页
Natural Gas Industry
基金
国家自然科学基金项目“羌塘盆地构造热体制与烃源岩热演化时空差异”(编号:42241204)
“热年代学及多种古地温温标约束下银额叠合盆地苏红图坳陷上古生界热演化史恢复”(编号:42272152)
西昌学院博士科研启动项目“基于地震信息处理的断缝识别技术研究”(编号:YBZ202138)
校企合作项目“镇巴地区页岩气藏三维地质模型构建与分析”(编号:YQZC-FW-2023-050)
延长油田股份公司施工项目“吴起采油厂胜利山智慧油气田建设二期”(编号:YT5323GST0004)。
关键词
深度学习
油气勘探
地震资料处理
测井数据解释
油藏地质建模
Deep learning
Oil and gas exploration
Seismic data processing
Well logging data interpretation
Reservoir geological modeling
