通用人工智能发展现状及在油气行业应用探索  

The Current Status of General Artificial Intelligence Development and Exploratory Applications in the Oil and Gas Industry

作　　者：窦宏恩[1,2] 李彦辉 张蕾 王洪亮 曾倩 吴雨婷 DOU Hongen;LI Yanhui;ZHANG Lei;WANG Hongliang;ZENG Qian;WU Yuting(PetroChina Research Institute of Petroleum Exploration&Development,Beijing 100083,China;Artificial Intelligence Technology R&D Center for Exploration and Development,CNPC,Beijing 100083,China;Beijing Shulian Wanxiang Network Technology Co.,Ltd.,Beijing 102400,China;College of Engineering,Peking University,Beijing 100871,China;China University of Petroleum(Beijing),Beijing 102200,China)

机构地区：[1]中国石油勘探开发研究院,北京市1000083 [2]中国石油勘探开发人工智能技术研发中心 [3]北京数链万象网络科技有限公司 [4]北京大学工学院 [5]中国石油大学(北京)

出　　处：《钻采工艺》2025年第1期10-20,共11页Drilling & Production Technology

基　　金：国家重点研发计划项目“战略性资源开发区风险评估应用示范”(编号:2022YFF0801204);中国石油集团公司前瞻性基础性技术攻关项目“油气勘探开发人工智能关键技术研究”(编号:2023DJ84)联合资助。

摘　　要：从OpenAI发布GPT-4o以来,通用人工智能发展从纯文本交互对话发展到了图像和语音等多模态交互模式,人工智能发展进入了“智能体”发展阶段,国内外各行各业出现了研发各类通用人工智能的风潮,出现了千模大战的局面。国际大油公司也不例外,一是以英国石油、壳牌和道达尔能源等在公司部署了微软Office 365在后台捆绑的OpenAI最新GPT-4o的Copilot;二是以埃克森美孚及阿联酋石油公司等企业允许员工直接在公司内使用GPT-4处理和分析大量数据;三是以斯伦贝谢、亚马逊网络公司等服务型公司正在研发油气领域的大语言模型LLM。文章对石油行业的油井动态分析、抽油机井示功图分析、岩心图像分析、核磁测井曲线分析、抽油泵泵筒与柱塞的摩擦力公式推导等需要专业人员做的工作让通用人工智能软件平台去完成,结果表明:GPT-4/GPT-4o能够正确地解决石油行业各种不同难度级别的专业问题,证明了目前GPT-4/GPT-4o这样具有“世界大脑”的智能体可以解决专业领域的特定复杂问题。Since the release of GPT-4o by OpenAI,the development of general artificial intelligence has evolved from pure text-based interactions to modes involving images and voice.This marks the entry of artificial intelligence into the"agent"development stage,sparking a trend across various industries,both domestically and internationally,to research and develop various types of general artificial intelligence,leading to a"thousand models"competition.Major international oil companies are no exception.Firstly,companies like BP,Shell,and Total Energy have deployed Microsoft Office 365 integrated with OpenAI's latest GPT-4o Copilot in their operations.Secondly,companies like ExxonMobil and the UAE's oil company have allowed employees to directly use GPT-4 to process and analyze large amounts of data.Thirdly,service companies such as Schlumberger and Amazon Web Services are developing large language models(LLMs)for the oil and gas sector.This article discusses various tasks traditionally performed by professionals in the petroleum industry—such as dynamic analysis of oil wells,analysis of pumping unit performance graphs,core image analysis,nuclear magnetic logging curve analysis,and the derivation of friction force formulas for pump barrels and plungers—and proposes that these tasks be handled by general artificial intelligence software platforms.The results indicate that only GPT-4/GPT-4o can accurately address various specialized problems in the petroleum industry,demonstrating that agents like GPT-4/GPT-4o,which function as a"world brain,"can solve specific complex problems in specialized fields.

关 键 词：通用人工智能 GPT-4/GPT-4o AI智能体 油气行业 大模型 

分 类 号：F42[经济管理—产业经济]