基于BP神经网络的钻井轨迹滚动预测
顾乐成1,4 曹万鹏2 王灶红1 魏秦文3 梅丹阳3

1四川工业科技学院，四川省德阳，618000；

2新天绿色能源股份有限公司，河北省石家庄，050051；

3北华航天工业学院，河北廊坊，065000；

4廊坊川越科技有限公司，河北廊坊065000；

摘要：准确预测钻进轨迹对提高钻进效率较为重要，但影响钻进轨迹的因素较多，且井内工况复杂，难以通过力学模型进行预测。目前常用的几何预测法精度不足，传统神经网络预测方法需要的训练样本量大，对不同设计类型的轨迹适应性不高，为此，提出了一种基于时间序列的BP神经网络滚动预测模型，采用移动窗口的方式控制训练样本量，对轨迹延伸方向的参数进行预测，使用某定向井部分实钻数据进行模型验证，并与几何预测模型、支持向量机模型进行对比，结果显示，滚动预测的井斜角平均绝对误差分别较圆柱螺线法、自然参数法、支持向量机分别降低33.2%与15.2%，方位角平均绝对误差较圆柱螺线法、自然参数法、支持向量机分别降低91.7%、74.8%、17.5%。表明该模型精度较高，具备实时预测能力，可为轨迹预测控制提供一定的参考。

关键词：井眼轨迹；实时预测；滚动预测；神经网络

参考文献

[1]马玉凤,袁野.井眼轨迹预测及井眼轨迹三维可视化系统开发[J].微型电脑应用,2017,33(08):65-67+71.

Ma Yufeng,Yuan Ye.Development of wellbore trajectory prediction and 3D visualization system[J].Microcomputer Application,2017,33(08):65-67+71

[2]苗在强.旋转导向钻井工具稳斜钻进模式自动调控方法研究[D].中国石油大学(华东),2018.DOI:10.27644/d.cnki.gsydu.2018.000637.

Miao Zaiqiang.Research on Automatic Control Method for Stable Angle Drilling Mode of Rotating Directional Drilling Tools[D].China University of Petroleum(East China),2018.DOI10.27644/d.cnki.gsydu.2018.000637.

[3]王舸.推靠式旋转导向工具造斜率分析软件平台开发[D].中国石油大学(北京),2021.DOI:10.27643/d.cnki.gsybu.2021.001580.

Wang Ge.Development of software platform for slope analysis of push-pull rotary guidance tools[D].China University of Petroleum(Beijing),2021.DOI:10.27643/dcnki.gsybu.2021-001580.

[4]谢鑫,唐玉华,徐浩,等.水平段钻进井眼轨迹的预测方法[J].中国石油大学胜利学院学报,2022,36(04):62-66.

Xie Xin,Tang Yuhua,Xu Hao,et al.Prediction method for horizontal drilling wellbore trajectory[J].Journal of Shengli College,China University of Petroleum,2022,36(04):62-66.

[5]Jung,Tae Joon,Yeon Hwi Jeong,and Younggy Shin."Simulation of directional drilling by dynamic finite element method."Journal of Mechanical Science and Technology 36.7(2022):3239-3250.

[6]李臻,宋先知,李根生,等.基于双输入序列到序列模型的井眼轨迹实时智能预测方法[J].石油钻采工艺,2023,45(04):393-403.DOI:10.13639/j.odpt.202212019.

Li Zhen,Song Xianhe,Li Gensheng,et al.Real time intelligent prediction method for wellbore trajectory based on dual input sequence to sequence model[J].Petroleum Drilling and Production Technology,2023,45(04):393-403. DOI:10.13639/j.odpt.202212019.

[7]Tunkiel,Andrzej T.,Dan Sui,and Tomasz Wiktorski."Training-while-drilling approach to inclination prediction in directional drilling utilizing recurrent neural networks."Journal of Petroleum Science and Engineering 196(2021):108128.

[8]Wang,Lu,et al."Drilling trajectory survey technology based on 3D RISS with a single fiber optic gyroscope."Optik 203(2020):163971.

[9]Gao,Yi,Na Wang,and Yihao Ma."L2-SSA-LSTM prediction model of steering drilling wellbore trajectory."IEEE Access(2023).

[10]Ni,Qingjian,et al."A Novel Trajectory Feature-Boosting Network for Trajectory Prediction."Entropy 25.7(2023):1100.

[11]Huang,Meng,et al."Application of long short-term memory network for wellbore trajectory prediction."Petroleum Science and Technology(2023):1-20.

[12]Jeong,Jiho,et al."Multi-objective optimization of drilling trajectory considering buckling risk."Applied Sciences12.4(2022):1829.

[13]王申.非直井井眼轨道修正设计方法研究[D].东北石油大学,2023.DOI:10.26995/d.cnki.gdqsc.2023.000466.

Wang Shen.Research on Design Method for Non Vertical Well Trajectory Correction[D].NortheastUniversity of Petroleum,2023.DOI:10.26995/d.cnki.gdqsc.2023.000466.

[14]刘颖.煤层气井眼轨迹三维可视化建模研究[D].中国石油大学(华东),2017.DOI:10.27644/d.cnki.gsydu.2017.000341.

Liu Ying.Research on three-dimensional visualization modeling of coalbed methane wellbore trajectory[D].China University of Petroleum(East China),2017.DOI:10.27644/d.cnki.gsydu.2017.000341.

[15]王潘涛.井眼轨迹预测与待钻轨道优化设计方法研究[D].中国石油大学(北京),2022.DOI:10.27643/d.cnki.gsybu.2022.000421.

Wang Pantao.Research on wellbore trajectory prediction and optimization design method for drilling trajectory[D].China University of Petroleum (Beijing),2022.DOI:10.27643/d.cnki.gsybu.2022.000421.

[16]王延江,杨培杰,史清江,等.一种基于支撑向量机学习预测井眼轨迹的新方法[J].石油学报,2005(05): 102-105.

Wang Yanjiang,Yang Peijie,Shi Qingjiang,et al. A new method for predicting wellbore trajectory based on support vector machine learning[J].Journal of Petroleum,2005(05):102-105.