吴思源^1,2,3,,
李守定^1,2,3,,,
陈冬⁵,
李晓^1,2,3,
杜爱民^2,3,4,
张莹^2,3,4
1. 中国科学院页岩气与地质工程重点实验室, 中国科学院地质与地球物理研究所, 北京 100029
2. 行星与地球科学学院, 中国科学院大学, 北京 100049
3. 中国科学院地球科学研究院, 北京 100029
4. 中国科学院深地资源装备技术工程实验室, 中国科学院地质与地球物理研究所, 北京 100029
5. 油气资源与探测国家重点实验室, 中国石油大学(北京), 北京 102249

基金项目: 中国科学院战略性先导科技专项A类（XDA14040400），国家自然科学基金资助项目（42090023），中国科学院科研仪器设备研制项目（YJKYYQ20190043），中国科学院重点部署项目（ZDRW-ZS-2021-3-1和ZDBS-LY-DQC003和KFZD-SW-422），中国科学院关键技术人才项目，中国科学院地质与地球物理研究所重点部署项目（IGGCAS-201903和SZJJ201901）资助


详细信息

作者简介: 吴思源, 女, 1995年生, 硕士, 主要从事地球物理人工智能研究.E-mail: wusiyuan181@mails.ucas.ac.cn

通讯作者: 李守定, 男, 1979年生, 正高级工程师, 博士生导师, 主要从事储层工程地质力学研究.E-mail: lsdlyh@mail.iggcas.ac.cn

中图分类号: P631

收稿日期:2020-11-18
修回日期:2021-06-15
上线日期:2021-11-10

An intelligent-while-drilling steering method of global closed-loop servo control

WU SiYuan^1,2,3,,
LI ShouDing^1,2,3,,,
CHEN Dong⁵,
LI Xiao^1,2,3,
DU AiMin^2,3,4,
ZHANG Ying^2,3,4
1. Key Laboratory of Shale Gas and Geoengineering, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China
2. College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing 100049, China
3. Innovation Academy for Earth Science, Chinese Academy of Sciences, Beijing 100029, China
4. CAS Engineering Laboratory for Deep Resources Equipment and Technology, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China
5. State Key Laboratory of Petroleum Resources and Prospecting, China University of Petroleum(Beijing), Beijing 102249, China


More Information

Corresponding author: LI ShouDing,E-mail:lsdlyh@mail.iggcas.ac.cn

MSC: P631

--> Received Date: 18 November 2020
Revised Date: 15 June 2021
Available Online: 10 November 2021


摘要
摘要:深层油气资源量巨大，是全球油气开发的重要方向.随着钻井朝着深层（>4500 m）和超深层（>6000 m）发展，地质条件更加复杂，深层钻井泥浆信号传输速率受限，井下随钻测井等数据传输延迟，增加了钻井事故的频率及钻出储层的风险.当前井场智能决策钻井的方法不适用，井下自主智能钻进是未来深层超深层高效钻进的发展方向.本文借鉴无人驾驶汽车的理论技术架构，提出了一种大闭环伺服控制随钻智能导向钻井方法，集旋转导向、地质导向、随钻地震、电磁前探、随钻测量、信号传输、自动钻机等技术于一体，利用"边钻边学"的人工智能评价与决策方法，智能识别钻头前方油气藏甜点，智能决策钻进方向和钻速，并利用大闭环伺服控制实现井下钻头的自主智能导向和钻进.大闭环伺服控制随钻智能导向钻井架构包括钻进感知、智能决策与大闭环控制3个部分.钻进感知部分通过随钻测井数据获取钻头定位信息、井周地层及钻头前方特性参数，智能决策部分依据钻进感知部分获取的信息通过人工智能决策模型修正轨道和优化钻进策略，大闭环控制部分根据智能决策指令调整钻进方向和速度.本文在钻进感知部分采用支持向量机算法利用随钻测井数据进行岩性智能识别，优选随机森林算法和长短期记忆（Long Short Term Memory，LSTM）循环神经网络对孔隙度、渗透率、饱和度和泥质含量进行评价.在智能决策部分优选随机森林算法对机械钻速进行预测与优化，均获得了高准确率.
关键词: 随钻测井/
人工智能/
深度学习/
机器学习/
地质导向/
智能钻井/
大闭环伺服控制

Abstract:The reserves of deep oil and gas resources are huge, making them significant for global oil and gas development. As the drilling depth goes towards to deep (>4500 m) and ultra-deep (>6000 m) formations, the geological conditions becomes more complex, the transmission rate of drilling mud signal is limited. The delay in the downhole logging while drilling data transmission would increase the risks of drilling accidents and drilling out of the reservoir. The current drilling site decision-making method is not applicable, and the downhole autonomous intelligent drilling will be an important direction of deep and ultra-deep drilling operations. Referring to the theoretical and technical framework of autonomous car, a global closed-loop servo control intelligent drilling method is proposed. This method integrates rotary steering, geosteering, seismic while drilling, far-field electromagnetic measurement, measurement while drilling, signal transmission, automatic drill rig and other technologies. After using the "learning while drilling" approach, the artificial intelligence evaluation and decision method, it is able to intelligent identification of sweet spot in front of drill bit, intelligent determination of drilling direction and rate of penetration, and allow the drill bit automatically navigate and drill downhole with the global closed-loop servo control. The global closed-loop servo control intelligent drilling system framework includes three parts: drilling perception, intelligent decision-making and global closed-loop control. The drilling perception part obtains the bit position and the characteristic parameters of formations around and in front of the well through the logging while drilling data. Based on the information obtained by the drilling perception part, the intelligent decision-making part uses the AI (Artificial Intelligence) decision-making model to update the well path and optimize the drilling strategy. The global closed-loop control part adjusts the drilling direction and rate of penetration according to the intelligent decision instruction. In the drilling perception part, support vector machine learning algorithm is used to intelligently identify lithology using logging while drilling data. Random forest algorithm and LSTM (Long Short Term Memory) recurrent neural network are used to evaluate porosity, permeability, saturation and shale content. The intelligent decision-making part uses the random forest algorithm to predict and optimize the rate of penetration. They all have achieved high accuracy.
Key words:Logging While Drilling (LWD)/
Artificial intelligence/
Deep learning/
Machine learning/
Geosteering/
Intelligent drilling/
Large closed-loop servo control



PDF全文下载地址:

http://www.geophy.cn/data/article/export-pdf?id=38cc6e3b-937a-434c-b3a0-48f0687ec9e4