岩性油气藏 ›› 2025, Vol. 37 ›› Issue (6): 172–179.doi: 10.12108/yxyqc.20250616

• 石油工程与油气田开发 • 上一篇    

基于核岭回归多参数优化的CO2驱最小混相压力模型

李佳旻1,2, 张艺钟1,2, 张茂林1,2, 秦博文1,2, 杨宇新1,2   

  1. 1. 长江大学 非常规油气湖北省协同创新中心, 武汉 430100;
    2. 长江大学 石油工程学院, 武汉 430100
  • 收稿日期:2025-02-21 修回日期:2025-03-20 发布日期:2025-11-07
  • 第一作者:李佳旻（2000—），女，长江大学在读硕士研究生，研究方向为油气藏数值模拟与动态分析。地址：（430100）湖北省武汉市蔡甸区大学路111号长江大学武汉校区。Email：minlytz@163.com。
  • 通信作者: 张艺钟(1991—),女,博士,副教授,主要从事注气提高采收率、非常规油气开发及油气藏流体相态方面的工作。Email:yizhong-zhang@yangtzeu.edu.cn。
  • 基金资助:
    国家自然科学基金青年基金项目“考虑地层水与储层多孔介质的页岩气藏多组分吸附相态行为研究”（编号：52004032）资助。

Minimum miscible pressure model of CO2 flooding based on kernel ridgeregression multi-parameter optimization

LI Jiamin1,2, ZHANG Yizhong1,2, ZHANG Maolin1,2, QIN Bowen1,2, YANG Yuxin1,2   

  1. 1. Hubei Cooperative Innovation Center of Unconventional Oil & Gas, Yangtze University, Wuhan 430100, China;
    2. School of Petroleum Engineering, Yangtze University, Wuhan 430100, China
  • Received:2025-02-21 Revised:2025-03-20 Published:2025-11-07

PDF (PC)

39

摘要： 结合细管实验数据，采用灰色关联度法对影响CO2驱油效率的主控因素进行识别与权重赋值，利用核岭回归（KRR）算法对参数集进行训练，并用遗传算法与网格搜索法优化模型超参数，建立了最小混相压力（MMP）预测模型。研究结果表明：①影响CO2驱油的主控因素包括油藏温度、原油组分及注入气组成，纯CO2注入条件下关联度排序依次为：T > x（C2—C4） > M（C7+） > x（C5—C6） > x（CH4+N2）。含杂质CO2注入条件下，杂质类型与含量对MMP的影响程度排序为：x（N2） > x（C1） > x（C2—C4）inj > x（H2S）。②相较Ridge模型和ElasticNet模型，KRR模型预测精度更高、误差更小。其中，KRR-GA模型综合性能最优，其测试集总平均绝对百分比误差（EMAP）为4.11%，均方根误差（ERMS）为0.856 MPa，决定系数（R2）为0.981。③KRR-GA模型对重质原油油藏及常规黑油油藏表现出更优的适用性，而KRR-GS模型更适用于注入气中有较高H2S含量的轻质原油油藏。

关键词: 核岭回归(KRR), 最小混相压力(MMP), CO2驱油, 灰色关联度法, 遗传算法, 油藏温度, 原油组分, 注入气组成

Abstract: Based on slim tubes experiment data, the grey corelation method was employed to identify the main controlling factors affecting CO2 oil recovery efficiency and determine their weight assignment. Kernel ridge re‐gression(KRR)algorithm was used to train parameter sets, with model hyperparameters optimized through ge‐netic algorithm(GA)and grid search(GS), the minimum miscible pressure(MMP)prediction model was established. The results show that: (1) The primary factors influencing CO2 flooding are reservoir temperature, crude oil composition, and injection gas composition. Under pure CO2 injection conditions, the order of grey corelation grade as follows: T > x(C2—C4) > M(C7+) > x(C5—C6) > x(CH4+N2). Under the condition of CO2 injection containing impurities, the impact ranking of impurity type and content on MMP is: x(N2) > x(C1) > x(C2—C4)inj > x(H2S).(2) Compared with Ridge model and ElasticNet model, KRR model exhibits higher prediction accuracy and lower error. Among them, KRR-GA model demonstrates the best overall performance, with the mean absolute percentage error(EMAP) of 4.11%, root mean square error(ERMS) of 0.856 MPa, and coefficient of determination(R2) of 0.981 on the test set.(3) KRR-GA model demonstrates superior applicability for heavy crude oil reservoirs and conventional black oil reservoirs, while KRR-GS model is more suitable for light crude oil reservoirs with high H2S content in the injection gas.

Key words: kernel ridge regression(KRR), minimum miscible pressure, CO2 flooding, grey correlationmethod, genetic algorithm, reservoir temperature, crude oil components, injection gas composition

中图分类号: 

  • TE341
[1] DAI Bo, WANG Leifei, ZHUANG Jian, et al. Experiment ofminimum miscibie pressure of CO2 flooding in ultra-low permeability reservoir[J]. Lithologic Reservoirs, 2020, 32(2): 129-133. 代波, 王磊飞, 庄建, 等. 超低渗透油藏CO2驱最小混相压力实验[J]. 岩性油气藏, 2020, 32(2): 129-133.
[2] CUI Chuanzhi, LI Jing, WU Zhongwei. Simulation of microscopic seepage characteristics of CO2 immiscible flooding under the effect of diffusion and adsorption[J]. Lithologic Reservoirs, 2024, 36(6): 181-188. 崔传智, 李静, 吴忠维. 扩散吸附作用下CO2非混相驱微观渗流特征模拟[J]. 岩性油气藏, 2024, 36(6): 181-188.
[3] ELSHARKAWY A M, POETTMANN F H, CHRISTIANSENR L. Measuring CO2 minimum miscibility pressures: Slim-tube orrising-bubble method[J]. Energy & fuels, 1996, 10(2): 443-449.
[4] YANG Fulin, YU Peng. Technical standard of minimum miscible flooding pressure determination with slim tube experiment [J]. Special Oil & Gas Reservoirs, 2019, 26(6): 118-122. 杨付林, 喻鹏. 细管实验确定最小混相驱压力技术标准[J]. 特种油气藏, 2019, 26(6): 118-122.
[5] YE Anping, GUO Ping, WANG Shaoping, et al. Determinationof minimum miscibility pressure for CO2 flooding by using PRequation of state[J]. Lithologic Reservoirs, 2012, 24(6): 125-128. 叶安平, 郭平, 王绍平, 等. 利用PR状态方程确定CO2驱最小混相压力[J]. 岩性油气藏, 2012, 24(6): 125-128.
[6] ZHAO Yuejun, SONG Kaoping, FAN Guangjuan. Research onminimum miscible pressure of supercritical carbon dioxide andcrude oil system under reservoir condition[J]. Journal of Dalian University of Technology, 2017, 57(2): 119-125. 赵跃军, 宋考平, 范广娟. 储层条件下超临界二氧化碳与原油体系最小混相压力研究[J]. 大连理工大学学报, 2017, 57 (2): 119-125.
[7] ZHANG Hui. Study on methods for determining minimum miscibility pressure[J]. Petrochemical Technology, 2016, 23(3): 7-8. 张慧. 最小混相压力确定方法研究[J]. 石化技术, 2016, 23 (3): 7-8.
[8] GUI Jinyong, LI Shengjun, GAO Jianhu, et al. A random forestsprediction method for gas saturation based on feature variableextension[J]. Lithologic Reservoirs, 2024, 36(2): 65-75. 桂金咏, 李胜军, 高建虎, 等. 基于特征变量扩展的含气饱和度随机森林预测方法[J]. 岩性油气藏, 2024, 36(2): 65-75.
[9] CHEN Guangying, LIANG Zhiwu, FU Kaiyun, et al. Predictionof minimum miscibility pressure between carbon dioxide andcrude oil using artificial neural networks[J]. Chemical Industry & Engineering Progress, 2011, 30(Suppl 1): 526. 陈光莹, 梁志武, 符开云, 等. 人工神经网络法预测二氧化碳-原油最小混相压力[J]. 化工进展, 2011, 30(增刊1): 526.
[10] YAO Jian. Experimental measurement and modeling study ofminimum miscibility pressure between CO2 and crude oil[D]. Chengdu: Southwest Petroleum University, 2019. 姚健. CO2-原油最小混相压力实验测定及模型研究[D]. 成都: 西南石油大学, 2019.
[11] LIN He, DU Jinling, XU Gang, et al. The application of random forest algorithm in predicting the casing deformation ofhydraulic fracturing[J]. Lithologic Reservoirs, 2025, 37(3): 185-193. 林鹤, 杜金玲, 徐刚, 等. 随机森林算法在水力压裂套管变形预测中的应用[J]. 岩性油气藏, 2025, 37(3): 185-193.
[12] QIN Bowen, CAI Xulong, NI Peng, et al. Prediction of the minimum miscibility pressure for CO2 flooding based on a physicalinformation neural network algorithm[J]. Measurement Science & Technology, 2024, 35: 1-14.
[13] WANG Han, BAI Hongkun, WANG Shiqian, et al. Regionalcarbon emission prediction method based on combined ensemblelearning model[J]. Power Demand Side Management, 2023, 25 (4): 55-59. 王涵, 白宏坤, 王世谦, 等. 基于组合集成学习模型的区域碳排放预测方法研究[J]. 电力需求侧管理, 2023, 25(4): 55-59.
[14] CHEN Guangying. Simulation prediction and experimental determination of CO2-oil minimum miscibility pressure[D]. Changsha: Hunan University, 2016. 陈光莹. CO2与原油最小混相压力模拟预测及实验测定[D]. 长沙: 湖南大学, 2016.
[15] PAN Yi, ZHAO Qiuxia, SUN Lei, et al. Prediction model ofminimum miscible pressure in CO2 flooding[J]. PetroleumReservoir Evaluation and Development, 2022, 12(5): 748-753. 潘毅, 赵秋霞, 孙雷, 等. CO2驱最小混相压力预测模型研究[J]. 油气藏评价与开发, 2022, 12(5): 748-753.
[16] ZHENG Xu, LEI Yuan, DING Wengang, et al. Research on prediction method of the minimum miscible pressure for carbon dioxide flooding in low permeability reservoirs[J]. Contemporary Chemical Industry, 2021, 50(11): 2636-2639. 郑旭, 雷源, 丁文刚, 等. 低渗油藏二氧化碳驱最小混相压力预测方法研究[J]. 当代化工, 2021, 50(11): 2636-2639.
[17] SPENCE A S J R, WATKINS R W. The effect of microscopiccore heterogeneity on miscible flood residual oil saturation[R]. Dallas, SPE Annual Technical Conference and Exhibition, 1980.
[18] ALSTON R, KOKOLIS G, JAMES C. CO2 minimum miscibility pressure: A correlation for impure CO2 streams and live oilsystems[J]. Society of Petroleum Engineers Journal, 1985, 25 (2): 268-274.
[19] EAKIN B, MITCH F. Measurement and correlation of miscibility pressures of reservoir oils[R]. Houston, SPE Annual Technical Conference and Exhibition, 1988.
[20] METCALFE R S. Effects of impurities on minimum miscibility pressures and minimum enrichment levels for CO2 and richgas displacements[J]. Society of Petroleum Engineers Journal, 1982, 22(2): 219-225.
[21] BON J, EMERA M K, SARMA H K. An experimental studyand genetic algorithm(GA)correlation to explore the effect ofnC5 on impure CO2 minimum miscibility pressure(MMP)[R]. Adelaide, SPE Asia Pacific Oil & Gas Conference and Exhibition, 2006.
[22] LIU Hao, LIU Jiahao. Sensitivity analysis of CO2 huff and puffparameters based on ultra-low permeability reservoir[J]. Advances in Fine Petrochemicals, 2023, 24(1): 24-28. 刘浩, 刘嘉豪. 基于特低渗油藏注CO2吞吐参数敏感性分析[J]. 精细石油化工进展, 2023, 24(1): 24-28.
[23] SUN Jing, LIU Dehua, ZHANG Liang, et al. Grey correlationanalysis of the influencing factors on production decline in lowpermeability reservoirs[J]. Special Oil & Gas Reservoirs, 2012, 19(2): 90-93. 孙敬, 刘德华, 张亮, 等. 低渗透油藏递减影响因素的灰色关联分析[J]. 特种油气藏, 2012, 19(2): 90-93.
[24] QIAN Fuyuan. Study on the influence of formation oil betweenCO2 and minimum miscibility pressure[D]. Daqing: NortheastPetroleum University, 2020. 钱福源. CO2与地层油最小混相压力影响因素的研究[D]. 大庆: 东北石油大学, 2020.
[25] HAN Bo, ZHAI Zhiwei, YU Weidong, et al. Dynamic analysisof minimum miscibility pressure during CO2 flooding reservoirsand its influencing factors[J]. Unconventional Oil & Gas, 2022, 9(1): 98-104. 韩波, 翟志伟, 于伟东, 等. 油藏CO2驱过程中最小混相压力的动态变化及其影响因素分析[J]. 非常规油气, 2022, 9(1): 98-104.
[26] LONG Zhenyu, WANG Changquan, SHI Lihong, et al. Studyon CO2 solubility model in formation water based on kernelridge regression algorithm[J]. Journal of Xi'an Shiyou University(Natural Science Edition), 2023, 38(1): 95-101. 龙震宇, 王长权, 石立红, 等. 基于核岭回归算法的地层水中CO2溶解度模型研究[J]. 西安石油大学学报(自然科学版), 2023, 38(1): 95-101.
[27] LI Hu, PU Chunsheng, WU Feipeng. Prediction of minimummiscibility pressure in CO2 flooding based on general regressionneural network[J]. Lithologic Reservoirs, 2012, 24(1): 108-111. 李虎, 蒲春生, 吴飞鹏. 基于广义回归神经网络的CO2驱最小混相压力预测[J]. 岩性油气藏, 2012, 24(1): 108-111.
[1] 马力, 欧阳传湘, 谭钲扬, 王长权, 宋岩, 林飞. 低渗透油藏CO2驱中后期提效方法研究[J]. 岩性油气藏, 2018, 30(2): 139-145.
[2] 杨红, 王宏, 南宇峰, 屈亚宁, 梁凯强, 江绍静. 油藏CO2驱油提高采收率适宜性评价[J]. 岩性油气藏, 2017, 29(3): 140-146.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!

陇ICP备17006252号
版权所有© 2018 中国石油集团西北地质研究所有限公司《岩性油气藏》编辑部

地址：甘肃省兰州市城关区雁儿湾路535号(陇ICP备17006252号)    电话：0931-8686158;8686058;8686288    邮箱：yxyqc@petrochina.com.cn

甘公网安备 62010202002827号

本系统由北京玛格泰克科技发展有限公司设计开发