Study on distribution of tight gas reservoirs in shallow-water delta fronts based on well-seismic intelligent prediction: a case study of the Permian Shanxi Formation in Yan’an Gas Field,Ordos Basin
2 State Key Laboratory of Petroleum Resources and Engineering,China University of Petroleum(Beijing),Beijing 102249,China
3 College of Geosciences,China University of Petroleum(Beijing),Beijing 102249,China
XU Zhenhua,born in 1992,holds a doctoral degree and the titles of associate professor and master’s supervisor. He is mainly engaged in research on oil and gas field development geology. E-mail: xuzhenhua@cup.edu.cn.
WANG Xiangzeng,born in 1968,holds a doctoral degree and the title of professor-level senior engineer,academician of the Chinese Academy of Engineering. He is mainly engaged in research on the exploration and development of low-permeability oil and gas fields. E-mail: sxycpcwxz@126.com.
The Permian Shanxi Formation in the Yan’an Gas Field of the Ordos Basin contains tight sandstone gas reservoirs within a shallow delta front. However,predicting sandbody distribution is challenging due to strong reservoir heterogeneity,large well spacing,and the signal-attenuating effects of the Quaternary loess plateau. As a result,the distribution patterns of effective gas-bearing reservoirs remain poorly understood. Using the S23 submember of the Shanxi Formation in the Gaojiahe area of the central Yan’an Gas Field as a case study,this research proposes an intelligent sandbody prediction method based on an adaptive weighting strategy. This approach integrates the relationship between amplitude-frequency characteristics and tuned thickness into an attention mechanism,establishing an adaptive weighting framework for multi-band seismic attributes. The weighting strategy prioritizes high-frequency attributes for identifying thin layers and low-frequency attributes for thick layers,effectively addressing sandbody prediction challenges in the presence of low-quality seismic data. Compared with existing methods,the proposed approach yields the most accurate sandbody thickness predictions across various ratios of learning wells to blind wells. For example,in a test involving 45 learning wells and 63 blind wells,the correlation coefficient(R) between predicted and actual sandbody thickness reached 0.85. Prediction results indicate that the S23 submember in the Gaojiahe area develops wide,banded sand bodies deposited in a shallow-water delta-front environment. Thick,coarse-grained sediments in the main distributary channels,along with the main body and inner edge of the mouth bar,are confined within the palaeo-geomorphological low of the main valley. Due to wave scouring,these intervals exhibit high quartz content and represent the principal effective gas reservoirs.
WANG Xiangzeng,born in 1968,holds a doctoral degree and the title of professor-level senior engineer,academician of the Chinese Academy of Engineering. He is mainly engaged in research on the exploration and development of low-permeability oil and gas fields. E-mail: sxycpcwxz@126.com.
"}, bioImg=null, bioContent=
WANG Xiangzeng,born in 1968,holds a doctoral degree and the title of professor-level senior engineer,academician of the Chinese Academy of Engineering. He is mainly engaged in research on the exploration and development of low-permeability oil and gas fields. E-mail: sxycpcwxz@126.com.
致密气作为一种非常规天然气,赋存于低孔隙度(<10%)、低渗透率(<0.1×10-3 μm2)且含气饱和度低(<60%)的储集层中,通常需要进行压裂改造才能实现商业化开采(Sharif,2007;邹才能等,2009;戴金星等,2012)。近年来,随着水平钻井和水力压裂技术的不断发展,致密气已成为重要的非常规能源类型(Law and Curtis,2002;贾承造等,2012;赵政璋和杜金虎,2012;Fic and Pedersen,2013;邹才能等,2013),在中国天然气储量和产量中的占比分别超过了1/3与1/4(贾爱林等,2025)。
上古生界二叠系山西组是鄂尔多斯盆地致密砂岩气的主要产层之一,主要发育河流—三角洲砂体(郭艳琴等,2021;Li et al., 2021)。前人研究认为,苏里格与大牛地气田山西组主要发育河流相与浅水三角洲平原砂体,并针对其储集层构型与储集层质量开展过大量研究(Yang et al., 2008;Dou et al., 2022;Li et al., 2022;Wan et al., 2022;Wu et al., 2022)。延安气田山西组沉积期邻近湖盆中心,发育浅水三角洲前缘砂体,其规模小、非均质性强,有效储集层分布十分复杂(王若谷等,2021;Wang et al., 2022)。已有学者对此开展过山西组内各个段级次的宏观沉积相分布及成岩作用研究(Feng et al., 2016;王香增等,2018;王若谷等,2021),但针对亚段级次的精细有效储集层分布规律尚不清楚,其核心难点在于砂体精细预测难度大,原因有三: (1)砂体规模小、厚度薄、非均质性强; (2)井网井距大,平均大于1 km;(3)第四系厚层黄土覆盖导致地震波能量衰减(王大兴等,2017;Chen et al., 2018;王香增等,2018;王若谷等,2021)。
延安气田位于中国陕西省境内,构造位置处于鄂尔多斯盆地东南缘的伊陕斜坡带(图 1-a)。该区现今构造平缓,坡度小于1°,构造稳定性较强(王香增等,2018)。气田已探明的产气层段包括石炭系本溪组、二叠系山西组及下石盒子组(图 1-b)(Zhao et al., 2014;李浩,2015)。山西组作为主要的致密砂岩产气层段,可细分为山1段与山2段,进一步可划分为6个亚段,其中,山23亚段的产能最高,为文中关注的研究层位(图 1-b)。
高家河地区位于延安气田中部(图 1-c),在山2段沉积时期邻近湖盆中心,主要发育浅水三角洲前缘沉积,物源来自于北部(图 1-d)(王香增和周进松,2017;Wang et al., 2022)。有学者提出,山23亚段沉积时期发育海退背景下的浅水三角洲前缘沉积(于兴河等,2017),沉积物经历了多期海岸带的改造与分选作用(Zhu et al., 2008)。该层段主要发育石英砂岩和岩屑石英砂岩,储集空间以次生溶孔和残余粒间孔为主(Wang et al., 2022),物性相对较好,平均孔隙度为5.32%,平均渗透率为0.11×10-3 μm2,这也是该层位产能较高的主要原因。
前文已述,研究区山23亚段砂体厚度横向变化较快、井距较大、地震主频较低,砂体预测难度大(图 2-d)。为提升对这类非均质性强、厚度较薄的砂体的高精度预测,前人提出了地震分频属性智能融合方法。该方法基于不同频段地震波对砂体厚度响应差异的原理,通过数学运算对多频段地震属性进行融合,增强地震属性对砂体厚度的解释能力(Dorrington and Link,2004;Chopra and Marfurt,2005)。当前主流智能融合方法主要包括支持向量机、随机森林、神经网络以及集成学习等(Xie et al.,2023;Ali et al., 2024;刘磊等,2024;Zhen et al., 2024)。这些方法在砂体定量预测及厚层砂体边界刻画方面取得了一定进展,但其普遍采用全局固定权重策略,即通过优化训练集整体获得固定的属性权重系数(Liu et al., 2024;Zhen et al., 2024),该策略难以动态适配层内砂体分布的快速变化,无法实现高低频段信息的自适应融合,进而导致不同频段地震属性在不同砂体厚度区域的特征信息未能被充分挖掘,故针对厚度横向变化较快的砂体,在低品质地震资料情况下,砂体预测效果不佳。
与文中基于自适应权重策略的深度神经网络模型相比,以往采用的智能融合模型(如支持向量回归(SVR)、深度神经网络(DNN)、多层感知器(MLP)和K近邻算法(KNN)等)均依赖静态权重分配机制(Anifowose et al., 2019;Li et al., 2019;岳大力等, 2022;Xin et al., 2024)。通过对比这些已有模型,作者进一步评估了基于自适应权重策略的深度神经网络模型的性能。在不同训练集—测试集比例下,基于自适应权重策略的深度神经网络模型均表现出更高的预测精度和更好的砂岩厚度空间表征能力,具体体现为与盲井实测砂体厚度间具有更高的相关系数(图 14),且基于自适应权重策略的深度神经网络模型能够同时有效反映薄层和厚层砂体的分布特征(图 8)。同时,在模型的相关评价指标MAE和RMSE上,基于全局固定权重的模型,如KNN在训练集和测试集上的MAE值分别为8.62和8.24,RMSE值分别为10.88和11.24。而DW-DNN模型在训练集和测试集上的MAE值分别为3.11和3.17,RMSE值分别为5.22和5.47,明显优于其他模型(图 15)。
[DaiJ X, NiY Y, WuX Q. 2012. Tight gas in China and its significance in exploration and exploitation. Petroleum Exploration and Development, 39(3): 277-284]
[GuoY Q, ZhaoL S, GuoB C, FeiS X, LiW H, ZhangQ, YuanZ, MaY, HeZ Q, LiB Q. 2021. Sedimentary characteristics of the Lower Permian in Ordos Basin and its adjacent areas. Journal of Palaeogeography(Chinese Edition), 23(1): 65-80]
[JiaA L, GuoZ, HanJ C. 2025. Efficient development technologies and application effects of tight sandstone gas in Ordos Basin. Acta Petrolei Sinica, 46(1): 255-264]
[JiaC Z, ZhengM, ZhangY F. 2012. Unconventional hydrocarbon resources in China and the prospect of exploration and development. Petroleum Exploration and Development, 39(2): 129-136]
[11]
李浩. 2015. 鄂尔多斯盆地中部上古生界烃源岩研究. 西北大学博士学位论文.
[12]
[LiH. 2015. Research of source rocks in upper Paleozoic,central Ordos Basin. Doctoral dissertation of Northwest University]
[LiH H, YueD L, LiW, GuoC C, LiX, LyuM. 2023. Identification of point bar and abandoned channel of meandering river by spectral decomposition inversion based on machine learning. Oil Geophysical Prospecting, 58(2): 358-368]
[LiM R, ShiY H, FanL Y, DaiX D, JingX Y, ZhangY. 2024. Comparison of main reservoir characteristics between deep coal-rock gas of the No.8 coal seam of the Upper Paleozoic Benxi Formation and tight sand gas reservoirs,Ordos Basin. Oil & Gas Geology, 45(6): 1590-1604]
[ShanX. 2016. Marine to lacustrine deltaic evolution and the deltaic system in Member 2 of Shanxi Formation-Member 8 of Shihezi Formation,southeastern Ordos Basin. Doctoral dissertation of China University of Geosciences(Beijing) ]
[WangR G, ZhouJ S, DuY H, LiW H. 2021. Deposition evolution model of the Carboniferous-Permian in Yan’an Gas Field,the southeastern Ordos Basin. Scientia Geologica Sinica, 56(4): 1088-1105]
[WangX Z, ZhouJ S. 2017. Provenance system and sedimentary evolution model of the second Member of Lower Permian Shanxi Fm in the southeastern Ordos Basin. Natural Gas Industry, 37(11): 9-17]
[WangX Z, QiaoX Y, MiN Z, WangR G. 2018. Technologies for the benefit development of low-permeability tight sandstone gas reservoirs in the Yan’an Gas Field,Ordos Basin. Natural Gas Industry, 38(11): 43-51]
[YuT J, QuanM, LuoW Q, JinH, WanY P, TangM M, ChenG. 2024. Semi-quantitative prediction of reservoir thickness by 2D seismic multi-attributes: taking Shan-2 reservoir in the southeast of Ordos Basin as an example. Journal of Xi’an Shiyou University(Natural Science Edition), 39(5): 50-58,70 ]
[ZouC N, TaoS Z, HouL H, et al. 2013. Unconventional Petroleum Geology. Beijing: Geological Publishing House,127-167]
[45]
AliM, ChangxingyueH, WeiN, JiangR, ZhuP M, HaoZ, HussainW, AshrafU. 2024. Optimizing seismic-based reservoir property prediction: a synthetic data-driven approach using convolutional neural networks and transfer learning with real data integration. Artificial Intelligence Review, 58(9): 31.
[46]
AnifowoseF, AbdulraheemA, Al-ShuhailA. 2019. A parametric study of machine learning techniques in petroleum reservoir permeability prediction by integrating seismic attributes and wireline data. Journal of Petroleum Science and Engineering, 176: 762-774.
[47]
ChenS E, FanS H, WangX Z, WangR P, LiuY L, YangL C, NingX H, LiR X. 2018. Neotectonic movement in the southern margin of the Ordos Block inferred from the Qianhe River terraces near the north of the Qinghai-Tibet Plateau. Geological Journal, 53: 274-281.
[48]
ChopraS, MarfurtK J. 2005. Seismic attributes: a historical perspective. Geophysics, 70(5): 3SO-28SO.
[49]
DongG D, LiuX S, YinL L, FuX X, SongJ Y, ZhaoX H, PeiW C. 2023. Development characteristics and genetic analysis of dense limestone reservoirs in the Taiyuan Formation of the Ordos Basin,China. Journal of Natural Gas Geoscience, 8: 307-317.
[50]
DorringtonK P, LinkC A. 2004. Genetic-algorithm/neural-network approach to seismic attribute selection for well-log prediction. Geophysics, 69(1): 212-221.
[51]
DouM J, LiS H, LeiT, RenG L, LiX H, GuoY, FengW J, ZhangX H. 2022. Hierarchical multiple-point geostatistical modeling method and application based on braided river reservoir architecture. Minerals,12: 1398.
[52]
FengZ Q, LiuD, HuangS P, GongD Y, PengW L. 2016. Geochemical characteristics and genesis of natural gas in the Yan’an Gas Field,Ordos Basin,China. Organic Geochemistry, 102: 67-76.
[53]
FicJ, PedersenP K. 2013. Reservoir characterization of a “tight”oil reservoir,the Middle Jurassic Upper Shaunavon Member in the Whitemud and Eastbrook pools,SW Saskatchewan. Marine and Petroleum Geology, 44: 41-59.
[54]
LawB E, CurtisJ B. 2002. Introduction to unconventional petroleum systems. AAPG Bulletin, 86(11): 1851-1852.
[55]
LeCunY, BengioY, HintonG. 2015. Deep learning. Nature, 521(7553): 436-444
[56]
LiW, YueD L, WangW F, WangW R, WuS H, LiJ, ChenD P. 2019. Fusing multiple frequency-decomposed seismic attributes with machine learning for thickness prediction and sedimentary facies interpretation in fluvial reservoirs. Journal of Petroleum Science and Engineering, 177: 1087-1102.
[57]
LiX H, SuY L, RenG L. 2022. Research on architecture of multi-type sand bodies inside distributary channel of delta plain: a case study in the Daniudi Gas Field, Ordos Basin. Arabian Journal of Geosciences,15: 1753.
[58]
LiY, FanA P, YangR C, SunY P, LenhardtN. 2021. Sedimentary facies control on sandstone reservoir properties: a case study from the Permian Shanxi Formation in the southern Ordos Basin,Central China. Marine and Petroleum Geology,129: 105083.
[59]
LiuZ B, ZhuJ M, TianB. 2024. A novel seismic inversion method based on multiple attributes and machine learning for hydrocarbon reservoir prediction in Bohai Bay Basin,Eastern China. Frontiers in Earth Science, 12: 1498164.
[60]
SharifA. 2007. Tight gas resources in Western Australia. Petroleum in Western Australia,September 2007. Perth: Department of Mines and Petroleum.
[61]
WanQ, FanA P, YangR C, LenhardtN. 2022. Diagenetic mineralogy and its effect on the reservoir properties of the sandstones of the Permian of S 120 block(Sulige Gas Field),Ordos Basin,NW China. Journal of Palaeogeography, 11(3): 360-386.
[62]
WangX Z, DuY H, ZhangL, QiaoX Y, LiangQ S, SunP. 2022. The characteristics of the tight gas reservoir with strong heterogeneity and its 3D recovery technology in Yan’an Gas Field,China. Interpretation, 10(4): T865-T880.
[63]
WuW T, ZhaoJ Z, WeiX S, WangY, ZhangJ, WuH Y, LiJ. 2022. Evaluation of gas-rich “sweet-spot”and controlling factors of gas-water distribution in tight sandstone gas provinces: an example from the Permian He8 Member in Sulige Gas Province,central Ordos Basin,Northern China. Journal of Asian Earth Sciences,227: 105098.
[64]
XieP F, HouJ G, DuanD P. 2023. A novel genetic inversion workflow based on spectral decomposition and convolutional neural networks for sand prediction in Xihu sag of East China Sea. Geoenergy Science and Engineering,231: 212331.
[65]
XinC P, ZhouW, ZhangL, QiaoX Y, WangY K, XiaoY. 2024. Calculation model of drainage radius of single-layer/multi-layer commingled gas production wells in a closed constant-volume gas reservoir and its application. Applied Sciences,14: 1873.
[66]
YangH, FuJ H, WeiX S, LiuX S. 2008. Sulige Field in the Ordos Basin: geological setting,field discovery and tight gas reservoirs. Marine and Petroleum Geology, 25(4-5): 387-400.
[67]
ZengH L, BackusM M. 2005. Interpretive advantages of 90°-phase wavelets: Part 1: Modeling. Geophysics, 70(3): C7-C15.
[68]
ZhaoJ Z, ZhangW Z, LiJ, CaoQ, FanY F. 2014. Genesis of tight sand gas in the Ordos Basin,China. Organic Geochemistry, 74: 76-84.
[69]
ZhenY, ZhangA, ZhaoX M, GeJ W, ZhaoZ, YangC C. 2024. Prediction of deep low permeability sandstone seismic reservoir based on CBAM-CNN. Geoenergy Science and Engineering,242: 213241.
[70]
ZhuH T, ChenK Y, LiuK Y, HeS. 2008. A sequence stratigraphic model for reservoir sand-body distribution in the Lower Permian Shanxi Formation in the Ordos Basin,northern China. Marine and Petroleum Geology, 25(8): 731-743.