A new physics-informed method for the fracability evaluation of shale oil reservoirs
-
摘要: 储层可压性的准确评价是储层压裂设计和压后产能评估的重要前提。目前，采用岩石力学参数进行页岩可压裂性评价取得了较好的现场应用效果。因此，如何准确获取岩石力学参数成为至关重要的问题。通过建立一种基于物理信息约束的神经网络模型，该模型采用物理和数据双驱动，仅使用少量数据就能够实现岩石力学参数的准确预测。为验证模型性能的优异性，采用人工神经网络、随机森林和XGBoost模型与之进行对比。结果表明，物理信息约束的神经网络在少量数据下预测岩石力学参数的平均准确率高于95%，性能远优于其他模型。采用物理信息约束的神经网络预测得到弹性模量、泊松比、抗拉强度和断裂韧性4种岩石力学参数，基于岩石力学参数对储层可压性的影响，建立了基于脆性指数和力学参数的可压性评价方法。最后，以渤海湾盆地沧东凹陷K2段不同储层可压性为例进行验证。结果表明：研究区整体可压性较好，其中，纹层状混合质页岩可压裂指数高于0.7，可压性良好；纹层状长英质页岩、厚层状灰云质页岩和薄层灰云质页岩可压裂指数均处在0.4~0.7，可压性中等。评价结果与实际施工现场各储层日采油量进行对比，证实了可压性智能评价方法的可靠性，该方法可以推广至页岩储层可压性评价工作中。Abstract: The accurate evaluation of reservoir fracability is an essential prerequisite for the fracturing design and post-fracturing productivity evaluation of reservoirs. Rock mechanical parameters have been applied to the fracability evaluation of shales presently, exhibiting great field application performance. Accordingly, it is crucial to obtain accurate rock mechanical parameters. This study developed a physics-informed neural network (PINN) model. Driven by data and physical information, the PINN model can accurately predict rock mechanical parameters using only a small amount of data. To verify its performance, the PINN model was compared with the artificial neural network, random forest, and XGBoost models. The comparison results show that the PINN model yielded an average accuracy greater than 95%, outperforming other models. Using the PINN model, this study obtained four rock mechanical parameters, namely modulus of elasticity, Poisson's ratio, tensile strength, and fracture toughness. Given the influence of rock mechanical parameters on reservoir fracability, this study developed an evaluation method for reservoir fracability based on the brittleness index and mechanical parameters. This fracability evaluation method was applied to reservoirs in the K2 member in the Cangdong sag of the Bohai Bay Basin. The evaluation results indicate generally high fracability of the study area. Specifically, lamellar mixed shales showed a fracability index of higher than 0.7, indicating high fracability, while lamellar felsic shales and thickly and thinly laminated shales comprising calcareous and dolomitic rocks of equal amounts exhibited fracability indices of 0.4‒0.7, indicating moderate fracability. The comparison between the evaluation results and the daily oil production of various reservoirs at the construction site verified the reliability of the smart fracability evaluation method developed in this study. Therefore, this fracability evaluation method can be applied to the fracability evaluation of shale reservoirs.
-
Key words:
- shale oil researvoirs /
- rock mechanical parameter /
- fracability /
- machine learning /
- physics-informed
-
图 2 渤海湾沧东凹陷K2段储层可压性评价工作流程
Fig. 2 Workflow for the fracability evaluation of reservoirs in the K2 member of the Cangdong sag, Bohai Bay Basin
图 3 渤海湾沧东凹陷A井井位
Fig. 3 Map showing the location of well A in Cangdong sag, Bohai Bay Basin
图 4 渤海湾沧东凹陷A井3009~3214 m单井柱状图
Fig. 4 Single-well stratigraphic column of well A at a depth of 3009‒3214 m in the Cangdong sag, Bohai Bay Basin
图 5 测井数据、岩石矿物组分与岩石力学参数的Pearson相关系数
Fig. 5 Pearson correlation coefficients of logging data, rock mineral components, and rock mechanical parameters.
图 6 随机森林算法工作流程
f为特征；P(c|f)为预测结果；Nj为决策树的个数。
Fig. 6 Workflow of the random forest algorithm
图 10 机器学习模型在预测岩石力学参数时性能对比
Fig. 10 Comparison of the performance of machine learning models in predicting rock mechanical parameters
图 11 机器学习模型对测试集数据的拟合能力沿井深变化
Fig. 11 Well depth-varying fitting ability of machine learning models for data in the test set
图 13 渤海湾沧东凹陷K2段不同井深可压性变化
Fig. 13 Fracability index at different well depths in the K2 member of the Cangdong sag, Bohai Bay Basin
图 14 不同储层可压性评价结果验证
Fig. 14 Verification of the fracability evaluation results of different reservoirs
表 1 4种岩石力学参数预测时输入参数的选取
Table 1 Input parameters for the prediction of four rock mechanical parameters
岩石力学参数 输入特征 井深 横波 纵波 黏土 方沸石 石英 长石 方解石 钙质 白云石 弹性模量 √ √ √ √ √ √ 泊松比 √ √ √ √ √ 抗拉强度 √ √ √ √ √ √ √ √ √ √ 断裂韧性 √ √ √ √ √ √ √ √ √ √ 
下载: 导出CSV
表 2 数据处理前与数据处理后对比
Table 2 Data pre and post processing
参数 处理前数据 处理后数据 井深/m (3 009.2~3 214.9)/3 121.6 (−0.98~1.39)/0.41 横波时差/(μs·ft−1) (76.1~217.3)/177.4 (−1.12~2.46)/0.67 纵波时差/(μs·ft−1) (75.5~122.5)/94.8 (−1.09~2.13)/0.52 黏土w/% (5~41)/17.1 (−2.73~3.97)/1.39 方沸石w/% (0~59)/15.2 (−1.37~3.93)/1.38 石英w/% (5~31)/16.8 (−2.61~3.43)/0.43 长石w/% (2~45)/17.2 (−1.36~3.67)/1.29 方解石w/% (0~35)/10.3 (−1.52~3.56)/1.03 钙质w/% (4~83)/31.6 (−1.57~3.04)/0.74 白云石w/% (0~83)/21.3 (−1.27~3.52)/1.12 弹性模量/GPa (10.2~32.3)/20.5 (−2.31~3.34)/0.59 泊松比 (0.16~0.32)/0.24 (−1.71~2.23)/0.25 抗拉强度/MPa (3.08~8.23)/5.52 (−1.85~3.15)/0.43 断裂韧性/
(MPa·m1/2)(0.14~0.35)/0.26 (−1.65~2.16)/0.32 注：表中(0~59)/15.2表示最小值~最大值/平均值，其他同。
下载: 导出CSV
表 3 4种机器学习模型预测岩石力学参数时的评价结果
Table 3 Evaluation results of four machine learning models for predicting rock mechanical parameters
岩石力学
参数模型 EMA ERMS R2 弹性模量 RF 1.62~3.27/2.92/1.65 1.73~5.35/3.06/3.62 0.72~0.93/0.86/0.21 XGBoost 1.68~2.25/1.97/0.57 1.74~2.67/2.14/0.93 0.87~0.97/0.91/0.10 ANN 2.16~3.04/2.85/0.88 2.22~3.56/2.93/1.34 0.83~0.89/0.87/0.06 PINN 1.46~1.72/1.57/0.26 1.52~1.86/1.64/0.34 0.92~0.98/0.96/0.06 泊松比 RF 0.016~0.084/0.043/0.068 0.025~0.096/0.052/0.071 0.71~0.94/0.87/0.23 XGBoost 0.025~0.046/0.032/0.021 0.027~0.049/0.039/0.022 0.86~0.95/0.90/0.09 ANN 0.049~0.078/0.059/0.029 0.055~0.085/0.064/0.030 0.81~0.90/0.86/0.09 PINN 0.009~0.025/0.012/0.016 0.017~0.036/0.021/0.019 0.92~0.98/0.95/0.06 抗拉强度 RF 0.52~1.27/0.75/0.75 0.63~1.35/0.79/0.72 0.76~0.92/0.85/0.16 XGBoost 0.51~0.65/0.57/0.14 0.56~0.77/0.65/0.21 0.84~0.94/0.90/0.10 ANN 0.56~0.82/0.69/0.26 0.62~0.89/0.73/0.27 0.82~0.89/0.84/0.07 PINN 0.25~0.58/0.37/0.33 0.32~0.62/0.44/0.30 0.90~0.97/0.95/0.07 断裂韧性 RF 0.026~0.094/0.045/0.068 0.029~0.099/0.052/0.070 0.74~0.95/0.89/0.21 XGBoost 0.021~0.036/0.028/0.015 0.027~0.047/0.036/0.020 0.85~0.95/0.92/0.10 ANN 0.041~0.062/0.053/0.021 0.045~0.075/0.062/0.030 0.83~0.92/0.86/0.09 PINN 0.007~0.024/0.011/0.017 0.011~0.026/0.015/0.015 0.91~0.99/0.97/0.08 注：1.62~3.27/2.92/1.65表示最小值~最大值/平均值/最值差距，其他同。
下载: 导出CSV
-
[1] 李玉伟,龙敏,汤继周,等. 考虑裂尖塑性区影响的水力压裂缝高计算模型[J]. 石油勘探与开发,2020,47(1):175−185.LI Yuwei,LONG Min,TANG Jizhou,et al. A hydraulic fracture height mathematical model considering the influence of plastic region at fracture tip[J]. Petroleum Exploration and Development,2020,47(1):175−185. [2] MAO Shaowen,ZHANG Zhuo,CHUN T,et al. Field–scale numerical investigation of proppant transport among multicluster hydraulic fractures[J]. SPE Journal,2021,26(1):307−323.. doi: 10.2118/203834-PA [3] MAO Shaowen,SIDDHAMSHETTY P,ZHANG Zhuo,et al. Impact of proppant pumping schedule on well production for slickwater fracturing[J]. SPE Journal,2020,26(1):342−358. [4] 刘合,黄有泉,蔡萌,等. 松辽盆地古龙页岩油储集层压裂改造工艺实践与发展建议[J]. 石油勘探与开发,2023,50(3):603−612.. doi: 10.1016/S1876-3804(23)60413-6LIU He,HUANG Youquan,CAI Meng,et al. Practice and development suggestions of hydraulic fracturing technology in the Gulong shale oil reservoirs of Songliao Basin,NE China[J]. Petroleum Exploration and Development,2023,50(3):603−612.. doi: 10.1016/S1876-3804(23)60413-6 [5] MENG Siwei,LI Dongxu,LIU Xin,et al. Study on dynamic fracture growth mechanism of continental shale under compression failure[J]. Gas Science and Engineering,2023,114:204983.. doi: 10.1016/j.jgsce.2023.204983 [6] 李玉伟,彭根博,陈勉,等. CO2泡沫压裂井筒气–液–固三相流动模型[J]. 石油学报,2022,43(3):386−398.LI Yuwei,PENG Genbo,CHEN Mian,et al. Gas–liquid–solid three phase flow model of CO2 foam fracturing in wellbore[J]. Acta Petrolei Sinica,2022,43(3):386−398. [7] LIU Xin,MENG Siwei,LIANG Zhengzhao,et al. Microscale crack propagation in shale samples using focused ion beam scanning electron microscopy and three–dimensional numerical modeling[J]. Petroleum Science,2022,13(4):23. [8] 张哲豪,李新,赵建斌,等. 页岩油储层岩石物理实验技术现状及发展[J]. 测井技术,2022,46(6):656−663.ZHANG Zhehao,LI Xin,ZHAO Jianbin,et al. Current situation and development of petrophysical experiment technology in shale oil reservoir[J]. Well Logging Technology,2022,46(6):656−663. [9] NIKRACESH P E. Computer–aided analysis of mechanical systems[M]. Prentice–Hall, Inc. , 1988, 2(2): 11. [10] ALAJMI B N, AHMED K H, ADAM G P, et al. Modular multilevel inverter with maximum power point tracking for grid connected photovoltaic application[C]//IEEE International Symposium on Industrial Electronics, 2011: 2057–2062. [11] RUSE C M, AHMADOV J, LIU Ning, et al. An integrated analytics and machine learning solution for predicting the anisotropic static geomechanical properties of the Tuscaloosa Marine Shale[C]//2021 SPE/AAPG/SEG Unconventional Resources Technology Conference, 2021. [12] ZHOU Jian,LI Xibing,MITRI H S. Comparative performance of six supervised learning methods for the development of models of hard rock pillar stability prediction[J]. Natural Hazards,2015,79(1):291−316.. doi: 10.1007/s11069-015-1842-3 [13] CAO Jing,GAO Juncheng,RAD H N,et al. A novel systematic and evolved approach based on XGBoost–firefly algorithm to predict Young’s modulus and unconfined compressive strength of rock[J]. Engineering with Computers,2021,5(38):3829−3845. [14] DEHGHAN S,SATTARI G,CHEHREH C S,et al. Prediction of uniaxial compressive strength and modulus of elasticity for Travertine samples using regression and artificial neural networks[J]. Mining Science and Technology,2010,20(1):41−46. [15] TARIQ Z, ELKATATNY S, MAHMOUD M, et al. A new artificial intelligence based empirical correlation to predict sonic travel time[C]//International Petroleum Technology Conference, 2016. [16] 曾晓华,孟迪,彭文丰,等. 北部湾盆地涠西南凹陷W油田储层综合评价研究[J]. 石油科学通报,2023,8(1):20−31.ZENG Xiaohua,MENG Di,PENG Wenfeng,et al. Comprehensive evaluation of reservoirs in W oilfield in the southwest depression of Beibu Gulf Basin[J]. Petroleum Science Bulletin,2023,8(1):20−31. [17] XIE Chiyu,DU Shuyi,WANG Jiulong,et al. Intelligent modeling with physics–informed machine learning for petroleum engineering problems[J]. Advances in Geo–Energy Research,2023,8(2):71−75. [18] 徐梓辰,金衍,刘晓敏. 水基页岩抑制剂烷基糖苷季铵盐的页岩强度维持机理研究[J]. 石油科学通报,2020,5(1):67−77.XU Zichen,JIN Yan,LIU Xiaomin. Study of inhibition performance and the mechanism of action of alkyl glucoside quaternary ammonium salt as a new shale inhibitor[J]. Petroleum Science Bulletin,2020,5(1):67−77. [19] 周立宏,赵贤正,柴公权,等. 陆相页岩油效益勘探开发关键技术与工程实践:以渤海湾盆地沧东凹陷古近系孔二段为例[J]. 石油勘探与开发,2020,47(5):1059−1066.ZHOU Lihong,ZHAO Xianzheng,CHAI Gongquan,et al. Key exploration & development technologies and engineering practice of continental shale oil:A case study of Member 2 of Paleogene Kongdian Formation in Cangdong Sag,Bohai Bay Basin,East China[J]. Petroleum Exploration and Development,2020,47(5):1059−1066. [20] 陈立雄,董兴蒙. 基于改进人工蜂群的高精度纵波慢度提取方法[J]. 测井技术,2022,46(6):664−668.CHEN Lixiong,DONG Xingmeng. High–precision compression wave slowness extraction based on improved artificial bee colony[J]. Well Logging Technology,2022,46(6):664−668. [21] 王斌,邓继新,刘喜武,等. 矿物组分对龙马溪组页岩动、静态弹性特征的影响[J]. 地球物理学报,2019,62(12):4833−4845.. doi: 10.6038/cjg2019M0471WANG Bin,DENG Jixin,LIU Xiwu,et al. The influence of rock composition on dynamic and static elastic properties of Longmaxi Formation shales[J]. Chinese Journal of Geophysics,2019,62(12):4833−4845.. doi: 10.6038/cjg2019M0471 [22] 赵斌,王芝银,伍锦鹏. 矿物成分和细观结构与岩石材料力学性质的关系[J]. 煤田地质与勘探,2013,41(3):59−63.ZHAO Bin,WANG Zhiyin,WU Jinpeng. Relation between mineralogical composition and microstructure to the mechanical properties of rock materials[J]. Coal Geology & Exploration,2013,41(3):59−63. [23] 石林,史璨,田中兰,等. 中石油页岩气开发中的几个岩石力学问题[J]. 石油科学通报,2019,4(3):223−232.SHI Lin,SHI Can,TIAN Zhonglan,et al. Several rock mechanics problems in the development of shale gas in PetroChina[J]. Petroleum Science Bulletin,2019,4(3):223−232. [24] ZHANG Yihuai,LEBEDEV M,AL–KYASERI A,et al. Characterization of nanoscale rock mechanical properties and microstructures of a Chinese sub–bituminous coal[J]. Journal of Natural Gas Science & Engineering,2018,52:106−116. [25] 李曦宁,李剑平,沈金松,等. 基于电成像测井的碳酸盐岩储层孔隙结构识别新方法[J]. 测井技术,2022,46(6):689−695.LI Xining,LI Jianping,SHEN Jinsong,et al. A novel identification method of pore structure in carbonate reservoirs based on electric imaging logging[J]. Well Logging Technology,2022,46(6):689−695. [26] BREIMAN L. Random forests[J]. Machine Learning,2001,45(1):5−32.. doi: 10.1023/A:1010933404324 [27] CHEN Tianqi,HE Tong. Xgboost:Extreme gradient boosting[J]. R package version 0. 4–2,2015,1(4):1−4. [28] HE Houtian,GAO Shangce,JIN Ting,et al. A seasonal–trend decomposition–based dendritic neuron model for financial time series prediction[J]. Applied Soft Computing,2021,108:107488.. doi: 10.1016/j.asoc.2021.107488 [29] 金衍,陈勉,张旭东. 利用测井资料预测深部地层岩石断裂韧性[J]. 岩石力学与工程学报,2001,20(4):454−456.. doi: 10.3321/j.issn:1000-6915.2001.04.007JIN Yan,CHEN Mian,ZHANG Xudong. Determination of fracture toughness for deep well rock with geophysical logging data[J]. Chinese Journal of Rock Mechanics and Engineering,2001,20(4):454−456.. doi: 10.3321/j.issn:1000-6915.2001.04.007 [30] 满轲,周宏伟. 不同赋存深度岩石的动态断裂韧性与拉伸强度研究[J]. 岩石力学与工程学报,2010,29(8):1657−1663.MAN Ke,ZHOU Hongwei. Research on dynamic fracture toughness and tensile strength of rock at different depths[J]. Chinese Journal of Rock Mechanics and Engineering,2010,29(8):1657−1663. [31] SLOTA–VALIM M. Static and dynamic elastic properties,the cause of the difference and conversion methods–case study[J]. Nafta–Gaz,2015,71(11):816−826. [32] RICKMAN R, MULLEN M, PETRE E, et al. A practical use of shale petrophysics for stimulation design optimization: All shale plays are not clones of the Barnett Shale[C]//SPE Annual Technical Conference and Exhibition, 2008. [33] 袁俊亮,邓金根,张定宇,等. 页岩气储层可压裂性评价技术[J]. 石油学报,2013,34(3):523−527.YUAN Junliang,DENG Jingen,ZHANG Dingyu,et al. Fracability evaluation of shale–gas reservoirs[J]. Acta Petrolei Sinica,2013,34(3):523−527. [34] 韩文中,赵贤正,金凤鸣,等. 渤海湾盆地沧东凹陷孔二段湖相页岩油甜点评价与勘探实践[J]. 石油勘探与开发,2021,48(4):777−786.HAN Wenzhong,ZHAO Xianzheng,JIN Fengming,et al. Sweet spots evaluation and exploration of lacustrine shale oil of the second member of Paleogene Kongdian Formation in Cangdong Sag,Bohai Bay Basin[J]. Petroleum Exploration and Development,2021,48(4):777−786. -
下载:
点击查看大图
图(14) / 表(3)
计量
- 文章访问数: 142
- HTML全文浏览量: 19
- PDF下载量: 32
- 被引次数: 0
