基于斑马算法优化支持向量回归机模型预测页岩地层压力

doi:10.12108/yxyqc.20240602

岩性油气藏 ›› 2024, Vol. 36 ›› Issue (6): 12–22.doi: 10.12108/yxyqc.20240602

基于斑马算法优化支持向量回归机模型预测页岩地层压力

赵军¹, 李勇¹, 文晓峰², 徐文远², 焦世祥¹

1. 西南石油大学地球科学与技术学院, 成都 610500;
2. 中国石油集团测井有限公司长庆分公司, 西安 710201

收稿日期:2024-05-17 修回日期:2024-07-11 出版日期:2024-11-01 发布日期:2024-11-04
第一作者:赵军（1970—）,男,博士,教授,主要从事测井解释与岩石物理研究。地址:（710201）四川省成都市新都区新都大道8号。Email:zhaojun_70@126.com
基金资助:
中国石油集团测井有限公司重点攻关项目“页岩油工程品质评价方法研究”（编号：ZYCJ-CQ-2023-JS-2337）资助。

Prediction of shale formation pore pressure based on Zebra Optimization Algorithm-optimized support vector regression

ZHAO Jun¹, LI Yong¹, WEN Xiaofeng², XU Wenyuan², JIAO Shixiang¹

1. School of Geoscience and Technology, Southwest Petroleum University, Chengdu 610500, China;
2. Changqing Branch, China Petroleum Group Logging Co., Ltd, Xi'an 710201, China

Received:2024-05-17 Revised:2024-07-11 Online:2024-11-01 Published:2024-11-04

摘要/Abstract

摘要： 针对陇东地区三叠系延长组7段（长7段）页岩孔隙结构复杂、非均质性强、地层压力预测精度较低等问题，提出了一种基于斑马算法优化支持向量回归机（ZOA-SVR）模型预测地层压力的方法，并在实际钻井中进行了应用，将预测结果与基于机器算法的模型和常规地层压力预测方法结果进行了对比。研究结果表明：①ZOA-SVR模型以实测地层压力数据为目标变量，优选与陇东地区长7段页岩地层压力数据关联度达到0.70以上的深度、声波时差、密度、补偿中子、自然伽马、深侧向电阻率、泥质含量等7个参数作为输入特征参数，设置训练样本数为40，交叉验证折数为5，初始化斑马种群数量为10，最大迭代次数为70，对惩罚因子和核参数进行优化并建模，参数优化后拟合优度指标R²达到0.942，模型预测的地层压力数据在训练集和测试集上的绝对误差均低于1 MPa，预测测试集地层压力数据与实测压力数据的平均相对误差为2.42%。②ZOA-SVR模型在研究区长7段地层压力预测中优势明显，比基于粒子群优化算法、灰狼算法和蚁群算法的模型具有更好的参数调节及优化能力，R²分别提高了0.209，0.327，0.142；比等效深度法、Eaton法、有效应力法预测的地层压力精度更高，相对误差分别降低了32.53%，15.31%，5.91%。③ZOA-SVR模型在实际钻井中的应用结果显示，研究区长7段地层压力在垂向上分布较稳定，泥页岩段的地层压力高于砂岩段，地层压力系数主要为0.80～0.90，整体上属于异常低压环境，与实际地层情况相符。

关键词: 页岩, 地层压力, 斑马优化算法, 支持向量回归机, 机器学习, 测井曲线, 长7段, 三叠系, 陇东地区

Abstract: To address the complexity and heterogeneity of the pore structure of seventh member of Triassic Yanchang Formation（Chang 7 member）in Longdong area，a method based on the Zebra Optimization Algorithm-optimized Support Vector Regression（ZOA-SVR）model was proposed. This method utilizes formation pressure test data and conventional well logging curves. It was applied in actual wells，and compared with other machine learning models，as well as conventional methods for predicting formation pressure. The results show that：（1）The ZOA-SVR model uses actual formation pressure data as target variables，selecting seven input feature parameter，such as depth， sonic transit time，density，compensated neutron，natural gamma，deep resistivity，and clay content，associated with shale formation pressure data correlation above 0.7 in the study area. The model was trained with 40 samples， validated with 5-fold cross-validation，and optimized with 10 initial Zebra populations and a maximum of 70 iterations. After optimizing penalty factors and kernel parameters，the model achieved a fit indicator R² of 0.942. Predicted formation pressure data had absolute errors below 1 MPa on both training and test sets，with an aver age relative error of 2.42% compared to measured data.（2）The ZOA-SVR model demonstrated significant advantages in predicting of Chang 7 Member formation pressure in the study area compared to models based on Particle Swarm Optimization，Grey Wolf Algorithm，and Ant Colony Algorithm. It shows a better parameter adjustment and optimization capabilities，with coefficient of determination increases of 0.209，0.327，and 0.142 respectively. It also exhibited higher accuracy in pressure prediction compared to Equivalent Depth Method，Eaton Method，and Effective Stress Method，reducing relative errors by 32.53%，15.31%，and 5.91% respectively. （3）Application of the ZOA-SVR model in actual wells indicated stable vertical distribution of chang 7 Member formation pressure in the study area. Pressure in mud shale sections is higher than the sandstone sections，with pressure coefficients mainly range from 0.80 to 0.90，indicating an overall abnormally low pressure environment consistent with actual formation conditions.

Key words: shale, formation pressure, zebra optimization algorithm, support vector regression, machine lear-ning, well logging curves, Chang 7 member, Triassic, Longdong area

中图分类号:

P618.13

赵军, 李勇, 文晓峰, 徐文远, 焦世祥. 基于斑马算法优化支持向量回归机模型预测页岩地层压力[J]. 岩性油气藏, 2024, 36(6): 12–22.

ZHAO Jun, LI Yong, WEN Xiaofeng, XU Wenyuan, JIAO Shixiang. Prediction of shale formation pore pressure based on Zebra Optimization Algorithm-optimized support vector regression[J]. Lithologic Reservoirs, 2024, 36(6): 12–22.

参考文献

[1] 瞿建华，王泽胜，任本兵，等. 准噶尔盆地环玛湖斜坡区异常高压成因机理分析及压力预测方法[J]. 岩性油气藏，2014， 26（5）:36-39. QU Jianhua，WANG Zesheng，REN Benbing，et al. Genetic mechanism analysis and prediction method of abnormal high pressure in Mahu slope area，Junggar Basin[J]. Lithologic Reservoirs，2014，26（5）:36-39.
[2] 肖玲，陈曦，雷宁，等. 鄂尔多斯盆地合水地区三叠系长7段页岩油储层特征及主控因素[J]. 岩性油气藏，2023，35（2）: 80-93. XIAO Ling，CHEN Xi，LEI Ning，et al. Characteristics and main controlling factors of shale oil reservoirs of Triassic Chang 7 member in Heshui area，Ordos Basin[J]. Lithologic Reservoirs， 2023，35（2）:80-93.
[3] 杨韬政，刘成林，田继先，等. 柴达木盆地大风山凸起地层压力预测及成因分析[J].岩性油气藏，2023，35（1）:96-107. YANG Taozheng，LIU Chenglin，TIAN Jixian，et al. Prediction and genesis of formation pressure in Dafengshan uplift，Qa idam Basin[J]. Lithologic Reservoirs，2023，35（1）:96-107.
[4] BRUCE B. An introduction to this special section:Pore pressure [J]. The Leading Edge，2002，21（2）:169-177.
[5] HOTTMANN C E，JOHNSON R K. Estimation of formation pressures from log-derived shale properties[J]. Journal of Pe troleum Technology，1965，17（6）:717-722.
[6] EATON B A. The effect of overburden stress on geopressure prediction from well logs[J]. Journal of Petroleum Technology， 1972，24（8）:929-934.
[7] 李玉凤，孙炜，何巍巍，等. 基于叠前反演的泥页岩地层压力预测方法[J]. 岩性油气藏，2019，31（1）:113-121. LI Yufeng，SUN Wei，HE Weiwei，et al. Prediction method of shale formation pressure based on pre-stack inversion[J]. Lithologic Reservoirs，2019，31（1）:113-121.
[8] BOWERS G L. Pore pressure estimation from velocity data: Accounting for overpressure mechanisms besides under com paction[J]. SPE Drilling & Completion，1995，10（2）:89-95.
[9] 樊洪海. 利用声速检测欠压实泥岩异常高压的简易方法与应用[J]. 石油钻探技术，2001，29（5）:9-11. FAN Honghai. A simple pore pressure estimation method for a disequilibrium compaction shale using sonic velocity[J]. Petro leum Drilling Techniques，2001，29（5）:9-11.
[10] 杨进. 地层压力BP网络预测方法研究及其应用[J]. 石油钻采工艺，2003，25（3）:13-15. YANG Jin. Study on formation pressure prediction with BP net work and its application[J]. Oil Drilling & Production Technol ogy，2003，25（3）:13-15.
[11] HARIS A，SITORUS R J，RIYANTO A. Pore pressure predic tion using probabilistic neural network:Case study of South Sumatra Basin[J]. IOP Conference Series:Earth and Environ mental Science，2017，62（1）:012021.
[12] AHMED A，ELKATATNY S，ALI A，et al. New model for pore pressure prediction while drilling using artificial neural net works[J]. Arabian Journal for Science & Engineering，2019，44（6）:6079-6088.
[13] ABDULMALEK A S，MAHMOUD A A，ELKATATNY S，et al. Prediction of pore and fracture pressures using support vec tor machine[R]. Virtual:International Petroleum Technology Conference，2019.
[14] ANDRIAN D，ROSID M S，SEPTYANDY M R. Pore pressure prediction using Anfis method on well and seismic data field “Ayah”[J]. IOP Conference Series:Materials Science and En gineering，2020，854（1）:012-041.
[15] FARSI M，MOHAMADIAN N，GHORBANI H，et al. Predict ing formation pore-pressure from well-log data with hybrid ma chine learning optimization algorithms[J]. Natural Resources Research，2021，30（5）:3455-3481.
[16] PICHITA B，THANANYA R，PAKAWAT P，et al. Pore pressure estimation by using machine learning model[R]. Virtual:Inter national Petroleum Technology Conference，2021.
[17] 杨琼. 基于机器学习算法的地层压力智能预测及参数优选[D]. 北京:中国石油大学（北京），2024. YANG Qiong. Intelligent prediction of formation pressure and parameter optimization based on machine learning algorithm [D]. Beijing:China University of Petroleum（Beijing），2024.
[18] 文志刚，罗雨舒，刘江艳，等. 陇东地区三叠系长7段页岩油储层孔隙结构特征及成因机制[J]. 岩性油气藏，2022，34（6）:47-59. WEN Zhigang，LUO Yushu，LIU Jiangyan，et al. Pore structure characteristics and genetic mechanism of Triassic Chang 7 shale oil reservoir in Longdong area[J]. Lithologic Reservoirs， 2022，34（6）:47-59.
[19] 沈健. 鄂尔多斯盆地陇东地区致密砂岩储层碳酸盐胶结物特征及成因机理[J]. 岩性油气藏，2020，32（2）:24-32. SHEN Jian. Carbonate cementation characteristics and genetic mechanism of tight sandstone reservoirs in Longdong area，Or dos Basin[J]. Lithologic Reservoirs，2020，32（2）:24-32.
[20] 曹江骏，王茜，王刘伟，等. 鄂尔多斯盆地合水地区三叠系长7段夹层型页岩油储层特征及主控因素[J]. 岩性油气藏， 2024，36（3）:158-171. CAO Jiangjun，WANG Xi，WANG Liuwei，et al. Characteris tics and main controlling factors of interbedded shale oil reser voirs of Triassic Chang 7 member in Heshui area，Ordos Basin [J]. Lithologic Reservoirs，2024，36（3）:158-171.
[21] 张才利，刘新社，杨亚娟，等. 鄂尔多斯盆地长庆油田油气勘探历程与启示[J]. 新疆石油地质，2021，42（3）:253-263. ZHANG Caili，LIU Xinshe，YANG Yajuan，et al. Petroleum ex ploration history and enlightenment of Changqing oil field in Ordos Basin[J]. Xinjiang Petroleum Geology，2021，42（3）: 253-263.
[22] 李士祥. 鄂尔多斯盆地中生界异常低压成因及对成藏的影响[D]. 成都:成都理工大学，2019. LI Shixiang. Effects on petroleum accumulation and genesis of the Mesozoic abnormal low pressure in Ordos Basin[D]. Chengdu: Chengdu University of Technology，2019.
[23] 宋宣毅，刘月田，马晶，等. 基于灰狼算法优化的支持向量机产能预测[J]. 岩性油气藏，2020，32（2）:134-140. SONG Xuanyi，LIU Yuetian，MA Jing，et al. Productivity fore cast based on support vector machine optimized by grey wolf optimizer [J]. Lithologic Reservoirs，2020，32（2）:134-140.
[24] TROJOVSKÁ E，DEHGHANI M TROJOVSKÝ P. Zebra opti mization algorithm:A new bio-inspired optimization algorithm for solving optimization algorithm[J]. IEEE Access，2022，（10）: 49445-49473.
[25] 周庆. 基于灰色主成分的聚合物驱操作成本预测方法[J]. 岩性油气藏，2012，24（5）:116-119. ZHOU Qing. Prediction method of polymer flooding operation cost based on gray principal component analysis[J]. Lithologic Reservoirs，2012，24（5）:116-119.
[26] 刘明明，王全，马收，等. 基于混合粒子群算法的煤层气井位优化方法[J]. 岩性油气藏，2020，32（6）:164-171. LIU Mingming，WANG Quan，MA Shou，et al. Well placement optimization of coalbed methane based on hybrid particle swarm optimization algorithm[J]. Lithologic Reservoirs，2020， 32（6）:164-171.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

基于斑马算法优化支持向量回归机模型预测页岩地层压力

Prediction of shale formation pore pressure based on Zebra Optimization Algorithm-optimized support vector regression

PDF (PC)

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

Metrics

本文评价

推荐阅读 10

[1]	白玉彬, 李梦瑶, 朱涛, 赵靖舟, 任海姣, 吴伟涛, 吴和源. 玛湖凹陷二叠系风城组烃源岩地球化学特征及页岩油“甜点”评价[J]. 岩性油气藏, 2024, 36(6): 110-121.
[2]	洪智宾, 吴嘉, 方朋, 余进洋, 伍正宇, 于佳琦. 纳米限域下页岩中可溶有机质的非均质性及页岩油赋存状态[J]. 岩性油气藏, 2024, 36(6): 160-168.
[3]	尹虎, 屈红军, 孙晓晗, 杨博, 张磊岗, 朱荣幸. 鄂尔多斯盆地东南部三叠系长7油层组深水沉积特征及演化规律[J]. 岩性油气藏, 2024, 36(5): 145-155.
[4]	王子昕, 柳广弟, 袁光杰, 杨恒林, 付利, 王元, 陈刚, 张恒. 鄂尔多斯盆地庆城地区三叠系长7段烃源岩特征及控藏作用[J]. 岩性油气藏, 2024, 36(5): 133-144.
[5]	杨学锋, 赵圣贤, 刘勇, 刘绍军, 夏自强, 徐飞, 范存辉, 李雨桐. 四川盆地宁西地区奥陶系五峰组—志留系龙马溪组页岩气富集主控因素[J]. 岩性油气藏, 2024, 36(5): 99-110.
[6]	闫建平, 来思俣, 郭伟, 石学文, 廖茂杰, 唐洪明, 胡钦红, 黄毅. 页岩气井地质工程套管变形类型及影响因素研究进展[J]. 岩性油气藏, 2024, 36(5): 1-14.
[7]	包汉勇, 赵帅, 张莉, 刘皓天. 川东红星地区中上二叠统页岩气勘探成果及方向展望[J]. 岩性油气藏, 2024, 36(4): 12-24.
[8]	申有义, 王凯峰, 唐书恒, 张松航, 郗兆栋, 杨晓东. 沁水盆地榆社—武乡区块二叠系煤系页岩储层地质建模及“甜点”预测[J]. 岩性油气藏, 2024, 36(4): 98-108.
[9]	邹连松, 徐文礼, 梁西文, 刘皓天, 周坤, 霍飞, 周林, 文华国. 川东地区下侏罗统自流井组东岳庙段泥页岩沉积特征及物质来源[J]. 岩性油气藏, 2024, 36(4): 122-135.
[10]	朱彪, 邹妞妞, 张大权, 杜威, 陈祎. 黔北凤冈地区下寒武统牛蹄塘组页岩孔隙结构特征及油气地质意义[J]. 岩性油气藏, 2024, 36(4): 147-158.
[11]	牟蜚声, 尹相东, 胡琮, 张海峰, 陈世加, 代林锋, 陆奕帆. 鄂尔多斯盆地陕北地区三叠系长7段致密油分布特征及控制因素[J]. 岩性油气藏, 2024, 36(4): 71-84.
[12]	徐田录, 吴承美, 张金凤, 曹爱琼, 张腾. 吉木萨尔凹陷二叠系芦草沟组页岩油储层天然裂缝特征与压裂模拟[J]. 岩性油气藏, 2024, 36(4): 35-43.
[13]	计玉冰, 郭冰如, 梅珏, 尹志军, 邹辰. 四川盆地南缘昭通示范区罗布向斜志留系龙马溪组页岩储层裂缝建模[J]. 岩性油气藏, 2024, 36(3): 137-145.
[14]	何文渊, 赵莹, 钟建华, 孙宁亮. 松辽盆地古龙凹陷白垩系青山口组页岩油储层中微米孔缝特征及油气意义[J]. 岩性油气藏, 2024, 36(3): 1-18.
[15]	程静, 闫建平, 宋东江, 廖茂杰, 郭伟, 丁明海, 罗光东, 刘延梅. 川南长宁地区奥陶系五峰组—志留系龙马溪组页岩气储层低电阻率响应特征及主控因素[J]. 岩性油气藏, 2024, 36(3): 31-39.