According to the 2024 drilling safety report of Tarim Oilfield, 31% of well control accidents are attributed to calculation deviations of downhole kick velocity. To address the challenge of time-varying nonlinear modeling in the monitoring of downhole fluid migration during oil and gas drilling, a hybrid intelligent model integrating feature engineering, physical constraints, and deep learning was proposed, and an optimized calculation framework for kick velocity driven by both physics and data was established. Deep purification of data was achieved by constructing a three-tier feature screening system: the noise filtering layer employed an improved Savitzky-Golay filter to suppress time-frequency domain noise; the physical correlation layer constructed a feature relevance tensor based on Darcy's Law to screen effective feature pairs; the difference amplification layer enhanced the distinguishability of key features using manifold learning, which increased the proportion of effective features to 79%. On this basis, a hybrid SENet-LSTM-Arps architecture was designed: the SENet channel attention mechanism was introduced to dynamically modulate channel weights; a bidirectional LSTM(long short-term memory) network was utilized to capture temporal dependencies; the Arps production decline model was embedded to construct a physical constraint layer, and the fusion of data-driven and knowledge-driven approaches was achieved via the Lagrange multiplier method. Experimental results based on the measured data from Well Duwa 103 in the Tarim Basin (with hyperparameter combinations determined by Bayesian optimization) show that: in the calculation of lag time, the coefficient of determination (R2) of this architecture reaches 0.992 5, the mean squared error (MSE) is as low as 0.837 1, and its performance is improved by two orders of magnitude compared with traditional methods; the synergistic effectiveness of the SENet module and the Arps constraint layer is verified by ablation experiments. The research results provide a full-link optimization paradigm of “feature engineering-physical constraints-temporal modeling” for the accurate calculation of kick velocity in complex reservoir environments. In engineering applications, the early warning time can be advanced to 15 minutes (a 65% reduction compared with traditional methods), and the well control early warning accuracy reaches 97.3%.
FU Xing-yu, HE Shi-ming, TANG Ming, WANG Chen, JIAN Zhi-kai, WANG Qing-yun.
Dual-driven Hybrid Intelligent Model Integrating Feature Engineering and Physical Constraints for Kick Velocity Calculation[J].
Science Technology and Engineering, 2026, 26(23): 9849-9865 DOI:10.12404/j.issn.1671-1815.2504916
油气钻井安全高效运行对国家能源战略实施至关重要,而井下油气上窜速度的精准计算仍是智能钻井发展的关键瓶颈。传统物理模型在裂缝性储层、非达西流等复杂工况下误差显著(>45%),难以适应非常规油气开发需求[1-2];数据驱动模型虽可捕捉非线性特征,却因缺乏物理约束导致虚假相关性学习,且对非等间隔数据敏感性(误差增幅40%~60%)。国际钻井承包商协会(International Association of Drilling Contractors,IADC)统计表明,全球17.3%的井控事故源于现有方法无法实时表征流体运移动态特性。据塔里木油田2024年钻井安全报告,低渗储层因上窜速度计算偏差导致的井控事故占比达31%,现有方法在实时性(>200 ms)与复杂工况适应性(误差>25%)方面亟待突破。传统迟到时间法因忽略钻具排替效应与井身结构影响[3-4],在深井中误差达30%,导致塔里木油田频繁采用提密度等保守措施,诱发井漏风险。近年虽有时空修正法[5]、体积当量法等改进方法,但未解决物理参数动态演化与特征噪声耦合问题。
如图1、图2所示为该滤波器在钻井压力信号去噪中的显著效果。通过对比原始信号与滤波后信号,可以看出标准Savitzky-Golay滤波、高斯滤波以及改进型Savitzky-Golay滤波的不同表现。其中,改进型Savitzky-Golay滤波在突变区域保留了更多细节特征,滤波残差相对较小。此外,从性能对比来看,改进型Savitzky-Golay滤波的均方根误差(root mean square error,RMSE)值较低,耗时虽稍长,但整体性能优于标准Savitzky-Golay滤波和高斯滤波,体现了其在压力信号去噪中的优越性和实用性。
$ \mu_{\text {trapezoid }}(x)=\left\{\begin{array}{ll}0, & x \leqslant a \text { 或 } x \geqslant d \\\frac{x-a}{c-a}, & a<x<b \\1, & b \leqslant x<c \\\frac{d-x}{d-c}, & c \leqslant x<d\end{array}\right.$
NouraN H, AllalZ, SalmanO, et al. Feature engineering for fault detection in petroleum transmission systems using adaptive tree-based architectures[J]. Journal of Petroleum Engineering Advances, 2025, 15: 101234.
[2]
JiangJ, ZhouG, JiH, et al. Hyper-spectral feature fusion for real-time mud property prediction in intelligent drilling anomalies[J]. Journal of Petroleum Science and Engineering, 2024, 239: 112987.
LiJiwei, LiuGonghui, LiJun, et al. Accurate caculation method for upward channeling velocity of oil and gas[J]. Science Technology and Engineering, 2014, 14(22): 180-184.
SunXiaobo, WangXianguang, ShenHaidong. A real-time calculation algorithm for on-site post-effect gas up-channeling velocity[J]. Oil Field Equipment, 2024, 53(11): 1-8.
ChenZhi, LiKairong, RenYongli. A spatio-temporal correction method for calculating the upward channeling velocity of post-effect oil and gas[J]. Petroleum Drilling Techniques, 2024, 52(6): 34-42.
[9]
OnahE, EzeJ U, AbdulraheemS A, et al. Unsupervised feature engineering for differentiated hydrocarbon migration prediction[J]. Journal of Petroleum Science and Engineering, 2025, 241: 112345.
WangHao. Cuttings concentration distribution and lag time calculation in deepwater drilling[J]. Science Technology and Engineering, 2012, 12(32): 8660-8662.
[12]
GelenA M, BaruaD P, TasciI, et al. Order transition pattern feature engineering for drilling fluid dynamic modeling[J]. Journal of Natural Gas Science and Engineering, 2025, 78: 103567.
LiuJinzi, DuWenying, DongChi, et al. Intelligent classification and identification of complex lithology reservoir rock samples based on color space and deep residual network ResNet-50[J]. Science Technology and Engineering, 2023, 23(29): 12628-12637.
[15]
SaxenaV, SinghU, SinhaK L. Multi-scale feature engineering for landslide susceptibility mapping in petroleum basins[J]. Journal of Petroleum Exploration and Production Technology, 2025, 15(3): 1-15.
HuangGuangqing. Production decline analysis and decline rate characterization formula in ultra-high water cut period[J]. Science Technology and Engineering, 2019, 19(15): 99-104.
HuJinqiu, ZhangLaibin, HuYangbai, et al. Safety hazard identification in oil and gas drilling and production sites based on YOLOv5s-SimAM-ASFF and development of intelligent wearable equipment[J]. Science Technology and Engineering, 2025, 25(16): 7004-7012.
CuiJinmin, WangJianguo, LiZhiqiang. Research on feature engineering for microfacies identification of oil and gas reservoirs based on HRNet[J]. Acta Petrolei Sinica, 2025, 46(2): 256-268.
LiMan, GaoJing, WenXin. Hybrid architecture design and edge computing optimization for intelligent oilfield platforms[J]. Petroleum Machinery, 2025, 53(3): 155-164.
[26]
AliO A, ElmarghanyR M, HamedM A, et al. Physics-informed neural networks for hydrocarbon migration modeling[J]. Petroleum Science, 2025, 23(4): 456-472.
[27]
ChhillarS R. Hybrid cloud-fog architecture for real-time drilling data processing[J]. Journal of Petroleum Engineering and Technology, 2025, 23(2): 14-28.
XieShushan, ZhengMohong, LuoHao. An intelligent routing architecture for oil and gas transmission based on probabilistic graphical models[J]. Natural Gas Industry, 2025, 45(3): 120-128.
WeiJiangang, WuXinzhen, ZhengYi, et al. Prediction of concrete-filled steel tubular shaft wall bearing capacity based on SHAP interpretation[J]. Chinese Journal of Geotechnical Engineering, 2025, 47(5): 890-899.
HanLi, WuYunxuan, LiuLijun, et al. Hybrid architecture design of an intelligent production system for drilling workshops[J]. Intelligent Manufacturing, 2025(1): 50-53.
LiJiwei, MaChunlin, HuangZhen. Inversion of formation oil and gas up-channeling velocity based on displacement back-calculation[J]. Drilling & Production Technology, 2024, 47(6): 12-16.
JiYuanwu, DiDuolin, GuoHong. Accurate calculation of oil and gas up-channeling velocity based on seepage theory[J]. Acta Petrolei Sinica (Petroleum Processing), 2024, 40(5): 1123-1132.
XiongWenxue, HuanZhipeng, ZhangShiming. A multi-parameter coupling calculation model for measuring oil and gas up-channeling by short tripping[J]. Natural Gas Exploration and Development, 2024, 47(4): 56-63.
JiangRui, LiXuecong, ChenZhi. Sensitivity analysis of influencing factors in the calculation of oil and gas up-channeling velocity[J]. Petroleum Engineering Construction, 2024, 50(6): 45-52.
PDF
(20103KB)
