基于随钻实时数据的井下托压多级智能识别方法

doi:10.3969/j.issn.2096-1693.2026.03.010

石油科学通报 ›› 2026, Vol. 11 ›› Issue (2): 518-532. doi: 10.3969/j.issn.2096-1693.2026.03.010

基于随钻实时数据的井下托压多级智能识别方法

詹家豪¹(), 李军¹^,²^,^*(), 柳贡慧¹^,³, 杨宏伟¹, 王超⁴, 王彪¹

¹ 中国石油大学(北京)石油工程学院，北京 102249
² 中国石油大学(北京)克拉玛依校区石油学院，克拉玛依 834000
³ 北京工业大学机械与能源工程学院，北京 100124
⁴ 长江大学机械工程学院，荆州 434023

收稿日期:2026-01-14 修回日期:2026-03-12 出版日期:2026-04-15 发布日期:2026-04-30
通讯作者: ^*李军(1971年—)，博士，教授，主要研究方向为欠平衡/控压钻井，井下工况人工智能识别、井筒完整性、射孔,完井、页岩气开发、岩石力学及其应用, lijun17792692628@163.com
作者简介:詹家豪(1999年—)，博士研究生，主要研究方向为油气工程信息化与智能化技术，zjh37730904@163.com。
基金资助:
国家重点研发计划项目“陆上超深油气井井喷防控关键技术装备及示范应用”(2023YFC3009200);国家自然科学基金重大科研仪器研制项目“钻井复杂工况井下实时智能识别系统研制”(52227804);国家自然科学基金青年科学基金项目“深井气侵井下原位实时识别与定量解释方法研究”(52304001);国家自然科学基金面上项目“超深复杂地层溢流智能识别与关井-压井一体化调控方法”(52474018)

Downhole multi-level weight transfer recognition based on MWD real-time data

ZHAN Jiahao¹(), LI Jun¹^,²^,^*(), LIU Gonghui¹^,³, YANG Hongwei¹, WANG Chao⁴, WANG Biao¹

¹ College of Petroleum Engineering, China University of Petroleum, Beijing 102249, China
² College of Petroleum Engineering, China University of Petroleum-Beijing at Karamay, Karamay 834000, China
³ College of Mechanical and Energy Engineering, Beijing University of Technology, Beijing 100124, China
⁴ School of Mechanical Engineering, Yangtze University, Jingzhou 434023, China

Received:2026-01-14 Revised:2026-03-12 Online:2026-04-15 Published:2026-04-30
Contact: ^*lijun17792692628@163.com

摘要/Abstract

摘要：

针对定向井滑动钻进工况下托压实时识别的难题，提出了一种基于随钻实时数据的托压多级智能识别方法。建立了基于钻压传递率的四级托压评价体系，通过分析井下钻压与振动信号的响应特征，从统计域、频域和时序演化三个维度构建多维特征空间，采用综合重要性评价策略筛选出10个核心特征。针对井下闭环控制对实时性和轻量化的严格要求，设计了浅层随机森林识别模型，通过类别权重方法处理样本不平衡问题，采用基于井号的数据划分策略确保模型泛化能力。基于西部某油田5口定向井的实测数据，模型在独立井测试集上达到90.2%的准确率和0.900的Macro-F1分数，完全托压召回率为87.6%。实现了模型在ARM Cortex-M4处理器上的实际部署，模型存储空间52 KB、推理时间355 ms，满足井下硬件的全部约束条件。通过可解释性分析验证了模型判别依据与托压物理机理的一致性。研究成果可直接应用于水力振荡器等井下主动控制装置的智能启停，将响应时间从传统地面控制的分钟级缩短至5 s以内，对提高钻井效率、降低井下风险具有重要工程价值。

关键词: 托压识别, 随钻测量, 轻量化模型, 随机森林, 嵌入式部署

Abstract:

To address the challenge of real-time weight transfer recognition in directional well sliding drilling, an intelligent multi-level identification method based on measurement-while-drilling (MWD) real-time data is proposed. A four-level weight transfer evaluation system based on weight-on-bit (WOB) transfer ratio is established. By analyzing the response characteristics of downhole WOB and vibration signals, a multi-dimensional feature space is constructed from statistical domain, frequency domain, and temporal evolution perspectives. A comprehensive importance evaluation strategy is adopted to select 10 core features. To meet the strict requirements of real-time performance and lightweight design for downhole closed-loop control, a shallow random forest recognition model is designed, utilizing class weight methods to handle sample imbalance and well-based data partitioning strategy to ensure model generalization capability. Based on measured data from 5 directional wells in a western oilfield, the model achieves 90.2% accuracy and 0.900 Macro-F1 score on the independent well test set, with a recall rate of 87.6% for complete weight transfer. The model is successfully deployed on an ARM Cortex-M4 processor with 52 KB storage space and 355 milliseconds inference time, meeting all downhole hardware constraints. The consistency between the model decision logic and weight transfer physical mechanism is verified through interpretability analysis. The research results can be directly applied to intelligent on-off control of downhole active control devices such as hydraulic oscillators, reducing response time from minute-level of traditional surface control to within 5 seconds, which has significant engineering value for improving drilling efficiency and reducing downhole risks.

Key words: weight transfer recognition, measurement while drilling, lightweight model, random forest, embedded deployment

中图分类号:

TE271

詹家豪, 李军, 柳贡慧, 杨宏伟, 王超, 王彪. 基于随钻实时数据的井下托压多级智能识别方法[J]. 石油科学通报, 2026, 11(2): 518-532.

ZHAN Jiahao, LI Jun, LIU Gonghui, YANG Hongwei, WANG Chao, WANG Biao. Downhole multi-level weight transfer recognition based on MWD real-time data[J]. Petroleum Science Bulletin, 2026, 11(2): 518-532.

https://sykxtb.cup.edu.cn/CN/Y2026/V11/I2/518

图/表 17

图1 水平井滑动钻进工况与测点布置示意图

Fig. 1 Schematic of horizontal well sliding drilling and sensor placement

表1 四级托压分级标准与工况特征

Table 1 Four-level weight transfer classification criteria and operational characteristics

托压等级	钻压传递率	钻进能力表现
正常钻进	0.6~1.0	机械钻速保持合理范围
轻度托压	0.4~0.6	机械钻速下降，但仍可维持连续钻进
重度托压	0.2~0.4	钻头轴向力明显不足，机械钻速显著下降
完全托压	0~0.2	无法维持有效钻进，机械钻速处于极低水平

图2 四级托压典型工况的钻压传递与振动响应特征对比

Fig. 2 Comparison of WOB transfer and vibration responses across four drilling conditions

表2 标注数据集中四级托压样本分布

Table 2 Sample distribution of four-level weight transfer in the labeled dataset

托压等级	样本数量	占比/%
正常钻进	77 319	58.7
轻度托压	28 056	21.3
重度托压	20 021	15.2
完全托压	6 322	4.8
合计	131 718	100.0

图3 数据处理与特征构建流程

Fig. 3 Data processing and feature engineering workflow

图4 特征筛选结果与重要性对比

Fig. 4 Feature selection results and importance comparison

表3 不同模型算法的资源需求对比

Table 3 Resource requirements comparison of different model algorithms

模型类型	内存需求/KB	推理时间/ms	模型大小/KB	嵌入式可部署性
支持向量机	120~200	800~1500	150~250	不适合
深度神经网络	>400	1200~2500	>400	不适合
XGBoost	>500	500~800	>500	不适合
随机森林(浅层)	40~60	300~500	40~60	适合

图5 托压识别模型推理流程

Fig. 5 Inference workflow of weight transfer recognition model

图6 超参数优化的平行坐标图

Fig. 6 Parallel coordinates plot for hyperparameter optimization

图7 测试集混淆矩阵

Fig. 7 Confusion matrix on test set

表4 各类别识别性能指标

Table 4 Recognition performance metrics for each class

	正常钻进	轻度托压	重度托压	完全托压
精确率/%	98.0	85.2	81.6	95.2
召回率/%	95.5	88.5	89.2	87.6
F1分数/%	96.7	86.8	85.2	91.2

表5 不同方法的识别性能对比

Table 5 Recognition performance comparison of different methods

方法	准确率/%	Macro-F1
钻压传递率阈值规则方法	72.5	0.685
仅使用钻压特征的随机森林	82.3	0.798
XGBoost(100棵树、深度6)	91.8	0.910
本文方法(10特征随机森林)	90.2	0.900

图8 不同方法的多维性能对比雷达图

Fig. 8 Multi-dimensional performance comparison radar chart of different methods

图9 连续井段托压演化过程的特征与模型输出对比

Fig. 9 Comparison of feature evolution and model output during continuous weight transfer progression

图10 48 min起短时序列样本在随机森林中的决策过程可视化

Fig. 10 Visualization of random forest decision behavior for a short sequence of samples starting at 48 min

表6 模型在目标硬件平台上的资源消耗

Table 6 Resource consumption on target hardware platform

指标	测量值	硬件约束
模型大小/KB	52	<256
RAM峰值/KB	18	<256
推理时间/ms	355	<1000
系统运行周期/s	4.13	≤5

图11 托压识别模型在测量短节硬件平台上的部署测试

Fig. 11 Deployment testing of weight transfer recognition model on measurement sub hardware platform

参考文献 30

[1]	LEI Q, XU Y, CAI B, et al. Progress and prospects of horizontal well fracturing technology for shale oil and gas reservoirs[J]. Petroleum Exploration and Development, 2022, 49(1): 191-199. doi: 10.1016/S1876-3804(22)60015-6
[2]	WANG X Y, NI H J, SHOR R, et al. A model-based optimization and control method of slide drilling operations[J]. Journal of Petroleum Science and Engineering, 2021, 198: 108203. doi: 10.1016/j.petrol.2020.108203 URL
[3]	MA Z C, ZHANG H B, HU H, et al. An analytical model for quantitative evaluation of friction drag in directional sliding drilling[J]. Scientific Reports, 2025, 15(1): 17906. doi: 10.1038/s41598-025-03171-9
[4]	王正, 宋先知, 李洪松, 等. 基于1dCNN-BiGRU和注意力机制的钻井工况智能识别方法[J]. 石油科学通报, 2025, 10(05): 926-940.
	[WANG Z, SONG X Z, LI H S, et al. Intelligent drilling condition recognition method based on 1D CNN-BiGRU and attention mechanism[J]. Petroleum Science Bulletin, 2025, 10(5): 926-940.]
[5]	田媛, 戴杰. 智能钻井技术现状与发展趋势探讨[J]. 石化技术, 2023, 30(4): 256-258.
	[TIAN Y, DAI J. Current status and development trends of intelligent drilling technology[J]. Petrochemical Technology, 2023, 30(4): 256-258.]
[6]	卢运虎, 金衍, 王汉青, 等. 井漏风险层位钻前智能识别方法研究[J]. 石油科学通报, 2024, 9(4): 574-585.
	[LU Y H, JIN Y, WANG H Q, et al. Intelligent pre-drilling identification method for lost circulation risk formations[J]. Petroleum Science Bulletin, 2024, 9(4): 574-585.]
[7]	CAO W P, MEI D Y, GUO Y M. Deep learning approach to prediction of drill-bit torque in directional drilling sliding mode: Energy saving[J]. Measurement, 2025, 250: 117144. doi: 10.1016/j.measurement.2025.117144 URL
[8]	LIU Y, MA T S, CHEN P, et al. Method and apparatus for monitoring of downhole dynamic drag and torque of drill-string in horizontal wells[J]. Journal of Petroleum Science and Engineering, 2018, 164: 320-332. doi: 10.1016/j.petrol.2018.01.077 URL
[9]	CAI Z W, ZHAO R L, HUANG Z M, et al. Research on the development trends of measurement while drilling (MWD) technology in oil and gas drilling[J]. Petroleum Science, 2025. Available online 11 December 2025. DOI: 10.1016/j.petsci.2025.12.013.(该文当前为JournalPre-proof,暂未见正式卷期页)
[10]	黄文君, 高德利. 大位移井规律性阻卡力学机理与控制措施研究[J]. 石油科学通报, 2020, 5(1): 49-57.
	[HUANG W J, GAO D L. Mechanical mechanisms and control measures of regular stuck pipe in extended-reach wells[J]. Petroleum Science Bulletin, 2020, 5(1): 49-57.]
[11]	SHI X L, HUANG W J, GAO D L. Mechanical behavior of drillstring with drag reduction oscillators and its effects on sliding drilling limits[J]. Petroleum Science, 2021, 18(6): 1689-1697. doi: 10.1016/j.petsci.2021.09.007 URL
[12]	WU S Y, LI S D, CHEN D, et al. An intelligent-while-drilling steering method of global closed-loop servo control[J]. Chinese Journal of Geophysics, 2021, 64(11): 4215-4227.
[13]	张菲菲, 张聪, 王茜, 等. 井眼数字孪生与钻井智能决策技术及展望[J/OL]. 石油科学通报, 1-29[2026-01-13].
	[ZHANG F F, ZHANG C, WANG Q, et al. Digital twin of wellbore and intelligent drilling decision-making technology: Current status and prospects[J/OL]. Petroleum Science Bulletin, 2026: 1-29.]
[14]	XIAO S Y, ZHAO X D, WANG X Z, et al. A neural network-based real-time casing collar recognition system for downhole instruments[J]. Electronics, 2026, 15(5): 1046. doi: 10.3390/electronics15051046 URL
[15]	WANG X M, XU Z P, FU Y K, et al. A pressure wave recognition and prediction method for intelligent sliding sleeve downlink communication systems based on LSTM[J]. Energies, 2025, 18(16): 4384. doi: 10.3390/en18164384 URL
[16]	LU C D, WU M, CHEN L F, et al. An event-triggered approach to torsional vibration control of drill-string system using measurement-while-drilling data[J]. Control Engineering Practice, 2021, 106: 104668. doi: 10.1016/j.conengprac.2020.104668 URL
[17]	狄勤丰, 杨赫源, 王文昌, 等. 钻柱动力学研究进展及发展趋势[J]. 石油科学通报, 2024, 9(2): 224-239.
	[DI Q F, YANG H Y, WANG W C, et al. Progress and development trends in drill-string dynamics research[J]. Petroleum Science Bulletin, 2024, 9(2): 224-239.]
[18]	KUANG Y C, ZHANG T, HE J B, et al. WOB fuzzy control and experimental study on horizontal well experimental setup[J]. Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, 2025, 239(4): 1102-1117. doi: 10.1177/09544062241288966 URL
[19]	SUN Z F, GENG D X, GUO H L, et al. Introducing transversal vibration in twist drilling: Material removal mechanisms and surface integrity[J]. Journal of Materials Processing Technology, 2024, 325: 118296. doi: 10.1016/j.jmatprotec.2024.118296 URL
[20]	XU B L, LIU S Y, LI H S. Drill string’s axial force transfer law in slide directional drilling in underground coal mine[J]. Tunnelling and Underground Space Technology, 2022, 130: 104701. doi: 10.1016/j.tust.2022.104701 URL
[21]	张菲菲, 崔亚辉, 于琛, 等. 基于机器学习的钻井工况识别技术现状及发展[J]. 长江大学学报(自然科学版), 2023, 20(4): 53-65, 143.
	[ZHANG F F, CUI Y H, YU C, et al. Current status and development of machine learning-based drilling condition recognition technologies[J]. Journal of Yangtze University (Natural Science Edition), 2023, 20(4): 53-65, 143.]
[22]	FLEGNER P, KAČUR J, DURDÁN M, et al. Evaluation of the acceleration vibration signal for aggregates of the horizontal drilling stand[J]. Applied Sciences, 2022, 12(8): 3984. doi: 10.3390/app12083984 URL
[23]	BENSCHOP N D, ZEWOTIR T, NAIDOO R N, et al. A new data-standardization procedure for comprehensive outlier detection in correlated meteorological sensor data[J]. Advances in Statistical Climatology, Meteorology and Oceanography, 2025, 11(2): 133-158. doi: 10.5194/ascmo-11-133-2025 URL
[24]	PINSKY E, KLAWANSKY S. MAD (about median) vs. quantile-based alternatives for classical standard deviation, skewness, and kurtosis[J]. Frontiers in Applied Mathematics and Statistics, 2023, 9: 1206537. doi: 10.3389/fams.2023.1206537 URL
[25]	李占山, 刘兆赓. 基于XGBoost的特征选择算法[J]. 通信学报, 2019, 40(10): 101-108. doi: 10.11959/j.issn.1000-436x.2019154
	[LI Z S, LIU Z G. Feature selection algorithm based on XGBoost[J]. Journal on Communications, 2019, 40(10): 101-108.] doi: 10.11959/j.issn.1000-436x.2019154
[26]	NAMJOO E, O’CONNOR A N, BUCKLEY J, et al. GlassBoost: A lightweight and explainable classification framework for tabular datasets[J]. Applied Sciences, 2025, 15(12): 6931. doi: 10.3390/app15126931 URL
[27]	BISWAS S, GRUNDLINGH N, BOARDMAN J, et al. A target permutation test for statistical significance of feature importance in differentiable models[J]. Electronics, 2025, 14(3): 571. doi: 10.3390/electronics14030571 URL
[28]	WANG X Y, KEE T. Integrating Shapley value and least core attribution for robust explainable AI in rent prediction[J]. Buildings, 2025, 15(17): 3133. doi: 10.3390/buildings15173133 URL
[29]	STACUL A, PASTAFIGLIA D, DI GIOVANNI A D, et al. A hardware system with ARM-based data processing for nano satellites[J]. International Journal of Reconfigurable and Embedded Systems, 2020, 9(2): 102.
[30]	AZAD M, NEHAL T H, MOSHKOV M. A novel ensemble learning method using majority based voting of multiple selective decision trees[J]. Computing, 2025, 107(1): 42. doi: 10.1007/s00607-024-01394-8

选择文件类型/文献管理软件名称

选择包含的内容

基于随钻实时数据的井下托压多级智能识别方法

Downhole multi-level weight transfer recognition based on MWD real-time data

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 17

参考文献 30

相关文章 3

编辑推荐

Metrics

本文评价

[1]	李志晔；刘铮；张卫卫；杨学奇；敖威；雷胜兰. 频变泊松阻抗在深水重力流砂体含气性预测中的应用[J]. , 2023, 8(6): 755-766.
[2]	苏义脑；窦修荣；高文凯；刘珂. 油气井随钻测量技术发展思考与展望[J]. , 2023, 8(5): 535-554.
[3]	徐凯；孙赞东. 基于随机森林方法的地震插值方法研究[J]. , 2018, 3(1): 22-31.

模态框（Modal）标题

选择文件类型/文献管理软件名称

选择包含的内容

基于随钻实时数据的井下托压多级智能识别方法

Downhole multi-level weight transfer recognition based on MWD real-time data

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 17

参考文献 30

相关文章 3

编辑推荐

Metrics

本文评价