基于机器学习高含水致密砂岩气藏气水层差异分布样式厘定

Determination of gas-water layers and differential distribution patterns in high-water-saturation tight sandstone gas reservoirs based on machine learning

摘要

川中广安须家河组四段为典型高含水致密砂岩气藏,砂体物性差且非均质性强,气水关系复杂且气井普遍产水,常规储层参数评价及预测方法适用性差,使得气水层识别难度大且气水分布规律不清。综合利用钻测井、岩心及试验测试和油气生产资料,应用机器学习误差反向传播神经网络理论方法建立非线性测井岩性、物性和含气性解释模型,并对气、水层分布样式进行厘定。结果表明:模型岩性解释综合准确率达94.7%;孔隙度±1.5%绝对误差范围内准确率95.6%;渗透率半个数量级内准确率95.5%;含水饱和度±8%相对误差范围内准确率94.1%,均符合油气精细勘探开发要求,模型可信度高;气层判别准确率达96%;须家河组四段高含水致密砂岩以岩性、构造-岩性复合气藏为特征,气水差异分布样式可划分为孤立型、正常型、倒置型、中气边水型以及断层封堵型5类,粗粒优势物性相带匹配局部构造高部位气、水分异程度最高,也是气层发育最有利目标。

Abstract

The Xu4 member in the Guang̍an area of the Central Sichuan Basin is a typical tight sandstone gas reservoir characterized by high water saturation, poor reservoir properties, and strong heterogeneity. Complex gas-water relationships and frequent water production from wells present significant challenges for reservoir evaluation and gas-water discrimination. Conventional reservoir parameter evaluation and prediction methods show limited applicability, resulting in difficulties in identifying gas and water layers and constraining gas-water distribution patterns. To address these issues, this study integrates drilling, logging, core, experimental, and production data to establish nonlinear logging interpretation models for lithology, reservoir properties, and gas-bearing characteristics using a machine learning-based BP neural network approach. These models are further applied to characterize gas-water layers and their distribution patterns. The results show that the interpretation results achieve a lithology prediction accuracy of 94.7%, porosity prediction accuracy of 95.6% within an absolute error of ±1.5%, permeability prediction accuracy of 95.5% within half an order of magnitude, and water saturation prediction accuracy of 94.1% within a relative error of ±8%. Regional gas-water identification achieves an overall gas-layer discrimination accuracy of 96%. The gas-water distribution patterns of the high-water-saturation tight sandstone reservoirs in the Xu4 member can be classified into five types, namely isolated, normal, inverted, center-gas-edge-water, and fault-sealed patterns. The most favorable gas-bearing targets are developed in coarse-grained, high-quality reservoir facies belts associated with local structural highs, where buoyancy and reservoir quality jointly enhance gas-water differentiation.

关键词

神经网络 / 高含水致密砂岩气藏 / 气水层识别 / 气水分布样式 / 须家河组四段

Key words

neural network / high water saturation tight reservoir / gas-water layer identification / gas-water distribution pattern / Xu4 member

参考文献

[1] 李剑, 魏国齐, 谢增业, 等 . 中国致密砂岩大气田成藏机理与主控因素:以鄂尔多斯盆地和四川盆地为例[J]. 石油学报, 2013, 34(增1): 14-28.

[2] LI Jian, WEI Guoqi, XIE Zengye, et al. Accumulation mechanism and main controlling factors of large tight sandstone gas fields in China: cases study on Ordos Basin and Sichuan Basin[J]. Acta Petrolei Sinica, 2013, 34(sup1): 14-28.

[3] 李建忠, 郭彬程, 郑民, 等 . 中国致密砂岩气主要类型、地质特征与资源潜力[J]. 天然气地球科学, 2012, 23(4): 607-615.

[4] LI Jianzhong, GUO Bincheng, ZHENG Min, et al. Main types, geological features and resource potential of tight sandstone gas in China[J]. Natural Gas Geoscience, 2012, 23(4): 607-615.

[5] 邹才能, 陶士振, 吴松涛, 等 . 四川盆地“气大庆”建设成效及天然气千亿立方米产能基地发展战略[J]. 天然气工业, 2025, 45(12): 26-39.

[6] ZOU Caineng, TAO Shizhen, WU Songtao, et al. Construction achievements of "Gas Daqing" in the Sichuan Basin and development strategies for a natural gas productivity base with a capacity of 100 billion cubic meters[J]. Natural Gas Industry, 2025, 45(12): 26-39.

[7] 郭旭升, 魏志红, 王威, 等 . 四川盆地油气资源特征与天然气高质量上产展望[J]. 天然气工业, 2025, 45(12): 40-52.

[8] GUO Xusheng, WEI Zhihong, WANG Wei, et al. Characteristics of oil and gas resources in the Sichuan Basin and prospects for high-quality gas production growth[J]. Natural Gas Industry, 2025, 45(12): 40-52.

[9] 贾承造, 赵文, 宋岩, 等 . 常规-致密过渡型砂岩气地质特征与成藏机制:以四川盆地须家河组-沙溪庙组为例[J]. 天然气工业, 2025, 45(12): 53-65.

[10] JIA Chengzao, ZHAO Wen, SONG Yan, et al. Geological characteristics and hydrocarbon accumulation mechanisms of conventional-tight transitional sandstone gas reservoirs: a case study of the Xujiahe Formation and Shaximiao Formation in the Sichuan Basin[J]. Natural Gas Industry, 2025, 45(12): 53-65.

[11] 罗瑀峰, 厚刚福, 刘占国, 等 . 四川盆地大安探区须家河组致密气成藏条件与勘探领域[J]. 海相油气地质, 2025, 30(2): 157-166.

[12] LUO Yufeng, HOU Gangfu, LIU Zhanguo, et al. The formation conditions and exploration fields of tight gas of Xujiahe Formation in Da̍an exploration area, Sichuan Basin[J]. Marine Origin Petroleum Geology, 2025, 30(2): 157-166.

[13] 郭迎春, 庞雄奇, 陈冬霞, 等 . 致密砂岩气成藏研究进展及值得关注的几个问题[J]. 石油与天然气地质, 2013, 34(6): 717-724.

[14] GUO Yingchun, PANG Xiongqi, CHEN Dongxia, et al. Progress of research on hydrocarbon accumulation of tight sand gas and several issues for concerns[J]. Oil & Gas Geology, 2013, 34(6): 717-724.

[15] 李舟波 . 利用测井方法识别复杂油气储层的流体性质[J]. 石油与天然气地质, 2004, 25(4): 356-362.

[16] LI Zhoubo. Problem and advance on the fluid typing for complex reservoirs using well log[J]. Oil & Gas Geology, 2004, 25(4): 356-362.

[17] 朱文卿, 高树生, 沈杰, 等 . 致密砂岩气藏气水层识别新方法:以苏76井区为例[J]. 天然气地球科学, 2019, 30(11): 1629-1638.

[18] ZHU Wenqing, GAO Shusheng, SHEN Jie, et al. A new method for identifying gas-water layer in tight sandstone gas reservoirs: case study of Su76 well area[J]. Natural Gas Geoscience, 2019, 30(11): 1629-1638.

[19] 杨华, 张本健, 陈明江, 等 . 裂缝性碳酸盐岩储层气水识别的测井解释新途径:以四川盆地九龙山地区中二叠统为例[J]. 天然气工业, 2021, 41(7): 56-62.

[20] YANG Hua, ZHANG Benjian, CHEN Mingjiang, et al. A new log interpretation method for identifying water and gas layers in fractured carbonate reservoirs: a case study of Middle Permian in Jiulongshan area of northwestern Sichuan Basin[J]. Natural Gas Industry, 2021, 41(7): 56-62.

[21] 肖玉峰, 张审琴, 陈鹏, 等 . 柴达木盆地深层天然气测井识别方法研究[J]. 测井技术, 2015, 39(3): 305-309.

[22] XIAO Yufeng, ZHANG Shenqin, CHEN Peng, et al. Method study on deep gas logging identification in Qaidam Basin[J]. Well Logging Technology, 2015, 39(3): 305-309.

[23] 李浩, 孙兵, 魏修平, 等 . 松南气田火山岩储层测井解释研究[J]. 地球物理学进展, 2012, 27(5): 2033-2042.

[24] LI Hao, SUN Bing, WEI Xiuping, et al. Logging interpretation of volcanic rock reservoir in Songnan gas field[J]. Progress in Geophysics, 2012, 27(5): 2033-2042.

[25] 田冷, 何顺利 . 基于改进人工神经网络的气水层识别技术[J]. 测井技术, 2009, 33(5): 449-452.

[26] TIAN Leng, HE Shunli. Gas and water layers identification based on improved neural network model[J]. Well Logging Technology, 2009, 33(5): 449-452.

[27] IMAMVERDIYEV Y, SUKHOSTAT L . Lithological facies classification using deep convolutional neural network[J]. Journal of Petroleum Science and Engineering, 2019, 174: 216-228.

[28] KOESHIDAYATULLAH A, MORSILLI M, LEHRMANN D J, et al. Fully automated carbonate petrography using deep convolutional neural networks[J]. Marine and Petroleum Geology, 2020, 122: 104687.

[29] 刘军, 钟洁, 倪振, 等 . 基于机器学习的低含油饱和度砂岩储层参数预测:以准噶尔盆地夏子街油田夏77井区下克拉玛依组为例[J]. 石油实验地质, 2024, 46(5): 1123-1134.

[30] LIU Jun, ZHONG Jie, NI Zhen, et al. Machine learning-based prediction of low oil saturation sandstone reservoir parameters: a case study of Lower Karamay Formation in Xia77 well block of Xiazijie Oilfield, Junggar Basin[J]. Petroleum Geology and Experiment, 2024, 46(5): 1123-1134.

[31] 樊茹, 邵辉, 贾爱林, 等 . 川中广安须家河气藏气、水分布[J]. 石油与天然气地质, 2009, 30(6): 732-739.

[32] FAN Ru, SHAO Hui, JIA Ailin, et al. Gas and water distribution in gas reservoirs of the Xujiahe Formation, Guang̍an area, the central Sichuan Basin[J]. Oil & Gas Geology, 2009, 30(6): 732-739.

[33] 兰朝利, 张君峰, 何顺利, 等 . 低渗低阻产水气藏气水层识别:以广安气田须家河组为例[J]. 地球科学, 2010, 35(6): 1053-1059.

[34] LAN Chaoli, ZHANG Junfeng, HE Shunli, et al. Gas and water pay recognition of water-producing reservoir with low permeability and resistivity: a case from the Xujiahe Formation in Guang̍an gasfield of Sichuan Basin, southwestern China[J]. Earth Science, 2010, 35(6): 1053-1059.

[35] 罗超, 贾爱林, 何东博, 等 . 四川盆地广安气田须四段、须六段致密砂岩气藏气水分布对比[J]. 天然气地球科学, 2016, 27(2): 359-370, 376.

[36] LUO Chao, JIA Ailin, HE Dongbo, et al. Comparison study on the distribution of gas and water in Xu-4 and Xu-6 Formations tight gas sandstone reservoir of Guang̍an Gasfield, Sichuan Basin[J]. Natural Gas Geoscience, 2016, 27(2): 359-370, 376.

[37] 田冷, 何顺利, 刘胜军, 等 . 广安地区须家河组气藏气水分布特征[J]. 天然气工业, 2009, 29(6): 23-26.

[38] TIAN Leng, HE Shunli, LIU Shengjun, et al. Features of gas and water distribution in the Xujiahe formation gas reservoir of Guang̍an area[J]. Natural Gas Industry, 2009, 29(6): 23-26.

[39] 车国琼, 龚昌明, 汪楠, 等 . 广安地区须家河组气藏成藏条件[J]. 天然气工业, 2007, 27(6): 1-5.

[40] CHE Guoqiong, GONG Changming, WANG Nan, et al. Gas accumulation condition in the Xujiahe Group of Guang̍an area[J]. Natural Gas Industry, 2007, 27(6): 1-5.

[41] 兰朝利, 田冷, 王建国, 等 . 广安气田气水分布控制机理[J]. 西安石油大学学报(自然科学版), 2013, 28(5): 1-8.

[42] LAN Chaoli, TIAN Leng, WANG Jianguo, et al. Control mechanism of gas-water distribution in Guang̍an Gasfield[J]. Journal of Xi̍an Shiyou University (Natural Science Edition), 2013, 28(5): 1-8.

[43] 田冷, 刘倩, 佘治成, 等 . 广安地区有水气藏压力系统划分与控制原因分析[J]. 油气井测试, 2013, 22(5): 10-13, 75.

[44] TIAN Leng, LIU Qian, SHE Zhicheng, et al. Division of pressure system and analysis of controlling reason for water-driven gas reservoir of the Guang̍an district[J]. Well Testing, 2013, 22(5): 10-13, 75.

[45] 唐生寿, 杨斌, 靳九龙, 等 . 基于机器学习的总有机碳含量测井预测方法对比研究[J]. 测井技术, 2024, 48(4): 428-437.

[46] TANG Shengshou, YANG Bin, JIN Jiulong, et al. Comparative study on total organic carbon content logging prediction method based on machine learning[J]. Well Logging Technology, 2024, 48(4): 428-437.

[47] 刘英杰, 黄传炎, 岳家恒, 等 . 陆相湖盆层序地层格架内有机质发育及控制因素分析:以中上扬子建南地区侏罗系东岳庙段为例[J]. 天然气地球科学, 2017, 28(6): 930-938.

[48] LIU Yingjie, HUANG Chuanyan, YUE Jiaheng, et al. Analysis of organic matter characteristics and their controlling factors in the sequence stratigraphic framework: case study of Jurassic Dongyuemiao member of the Ziliujin Formation in Jiannan area, Upper and Middle Yangtze Region[J]. Natural Gas Geoscience, 2017, 28(6): 930-938.

[49] LEITE E P, VIDAL A C . 3D porosity prediction from seismic inversion and neural networks[J]. Computers & Geosciences, 2011, 37(8): 1174-1180.

[50] 王猛, 董宇, 蔡军, 等 . 基于BP神经网络的储层渗透率预测及质量评价方法[J]. 地球物理学进展, 2023, 38(1): 321-327.

[51] WANG Meng, DONG Yu, CAI Jun, et al. Method for reservoir permeability prediction and quality evaluation based on BP neural network[J]. Progress in Geophysics, 2023, 38(1): 321-327.

[52] AMAEFULE J O, ALTUNBAY M, TIAB D, et al. Enhanced reservoir description: using core and log data to identify hydraulic (flow) units and predict permeability in uncored intervals/wells[R]. SPE 26436-MS, 1993.

[53] SHAN L, CAO L, GUO B . Identification of flow units using the joint of WT and LSSVM based on FZI in a heterogeneous carbonate reservoir[J]. Journal of Petroleum Science and Engineering, 2018, 161: 219-230.

[54] 毛晨飞, 高衍武, 肖华, 等 . 基于流动单元划分的砂砾岩储层渗透率校正方法[J]. 测井技术, 2023, 47(4): 486-494.

[55] MAO Chenfei, GAO Yanwu, XIAO Hua, et al. Permeability correction method of sandy conglomerate reservoir based on flow unit division[J]. Well Logging Technology, 2023, 47(4): 486-494.

[56] 魏斌, 陈建文, 郑浚茂, 等 . 应用储层流动单元研究高含水油田剩余油分布[J]. 地学前缘, 2000, 7(4): 403-410.

[57] WEI Bin, CHEN Jianwen, ZHENG Junmao, et al. Utilization of reservoir flow unit to study remaining oil distribution in high water containing oilfield[J]. Earth Science Frontiers, 2000, 7(4): 403-410.

[58] 彭秋, 车国琼, 李顺, 等 . 川中广安须四段低孔渗砂岩有效储层物性下限厘定[J]. 油气地质与采收率, 2025, 32(5): 69-80.

[59] PENG Qiu, CHE Guoqiong, LI Shun, et al. Lower limit determination of effective reservoir physical properties of Xu4 member sandstones with low porosity and permeability in Guang̍an area, central Sichuan[J]. Petroleum Geology and Recovery Efficiency, 2025, 32(5): 69-80.

[60] 庞雄奇, 姜振学, 黄捍东, 等 . 叠复连续油气藏成因机制、发育模式及分布预测[J]. 石油学报, 2014, 35(5): 795-828.

[61] PANG Xiongqi, JIANG Zhenxue, HUANG Handong, et al. Formation mechanisms, distribution models, and prediction of superimposed, continuous hydrocarbon reservoirs[J]. Acta Petrolei Sinica, 2014, 35(5): 795-828.

基金资助

中国石油大学(北京)克拉玛依校区科研启动基金(XQZX20230010)

克拉玛依市科技计划项目(2024hjcxrc0076)

新疆维吾尔自治区自然科学基金(2022D01B41)

新疆维吾尔自治区重点研发计划项目(2024B01017-1)