Fracture Intelligent Identification Using Well Logs of Continental Shale Oil Reservoir of Fengcheng Formation in Mahu Sag, Junggar Basin
-
摘要: 准噶尔盆地玛湖凹陷北部斜坡区风城组陆相页岩油储层为多物源混合沉积，多种岩性频繁互层，岩石力学层厚度小，导致其裂缝尺度小，裂缝常规测井响应弱，识别难度大.针对页岩裂缝测井识别的难题，应用集成学习中的极端梯度提升树方法，通过深度挖掘裂缝信息与测井数据之间的非线性关系，将多个弱分类器集成强分类器，降低裂缝识别的不确定性，以提高裂缝的识别能力.该方法将岩心裂缝描述和井壁成像测井裂缝解释结果作为标签，常规测井信息作为模型训练的输入数据，在异常点筛查、SMOTE过采样处理和特征优选的基础上，通过网格搜索方法获得裂缝智能识别模型的最优超参数.通过与目前常用的支持向量机和逻辑回归等机器学习方法对比，极端梯度提升树具有比其他两种非线性机器学习方法更好的裂缝识别效果，测试集识别准确率可达90%.A1井风3段识别结果反映了该段裂缝较为发育，且模型对于裂缝段与非裂缝段都具有较好的识别效果，与岩心观察结果符合率较高.表明极端梯度提升树具有较好的裂缝识别能力，能够为玛湖凹陷陆相页岩油储层的裂缝智能识别提供有效手段.Abstract: The continental shale oil reservoir of Fengcheng Formation in the northern slope area of Mahu Sag, Junggar Basin is a mixed deposition of multiple provenances, with frequent interbedding of various lithologies and small thickness of rock mechanics layer, resulting in small fracture scale, weak conventional logging response of fractures and great difficulty in identification. Aiming at the logging identification of shale fractures, in this paper it applies the extreme gradient boosting (XGBoost) method in ensemble learning to deeply mine the nonlinear relationship between fracture information and logging data, integrates multiple weak classifiers into strong classifiers, reduces the uncertainty of fracture identification, and improves the accuracy of fracture identification. In this method, core fracture description and fracture interpretation results of image logging are used as labels, and conventional logging information is used as input data for model training. On the basis of outlier screening, SMOTE oversampling and feature optimization, the optimal hyperparameters of fracture intelligent identification model are obtained through grid search method. Compared with the commonly used machine learning methods such as support vector machine and logical regression, the XGBoost has better fracture identification ability than the other two nonlinear machine learning methods, and the accuracy of test set identification reaches 90%. The identification results of the P1f3 of Well A1 reflect that the fractures in this section are relatively developed, and the model has a good identification ability for both fractured and non-fractured sections, with a high coincidence rate with the core observation results. It can provide effective means for intelligent identification of fractures in continental shale oil reservoirs in Mahu Sag.
-
Key words:
- XGBoost /
- fracture intelligent identification /
- continental shale oil /
- Mahu Sag /
- Junggar Basin /
- petroleum geology
-
图 1 准噶尔盆地玛湖凹陷构造区划图
据支东明等（2021）修改. a. 准噶尔盆地构造区划分图；b.玛湖凹陷构造区划分图；c.风城组综合柱状图
Fig. 1. Tectonic division of Mahu Sag in Junggar Basin
图 3 基于极端梯度提升树的裂缝智能识别流程图
Fig. 3. Flow chart of intelligent fracture identification based on XGBoost
图 4 研究区风城组取心段裂缝照片
a. A1井，4 037.01 m，一组3条高角度剪切缝，被方解石半充填；b. A1井，4 026.28 m，两组高角度剪切缝，被方解石局部充填；c. A1井，4 027.73 m，低角度剪切缝被方解石全充填；d. A1井，4 031.04 m，高角度剪切缝被方解石半充填；e.A1井，4 036.88 m，高角度剪切缝无充填；f.A1井，4 033.60~4 033.35 m，破碎带与剪切缝发育段；g.A2井，4 616.60~4 619.60 m，成像测井中的高角度剪切裂缝
Fig. 4. Photos of fractures in the core section of Fengcheng Formation in the study area
图 5 裂缝段与非裂缝段测井曲线交会图
Fig. 5. Cross plots of logging curves of fractured section and non-fractured section
图 7 选取特征个数与分类准确率的关系
Fig. 7. Relationship between the number of feature selections and classification accuracy
图 9 A1井极端梯度提升树裂缝智能识别结果与其他两种方法对比
Fig. 9. Comparison of the intelligent fracture identification results of XGBoost in Well A1 and the other two methods
图 10 孤立森林异常点检测结果直方图
Fig. 10. Histogram of detection results of isolated forest for abnormal data points
图 11 极端梯度提升树原始数据与过采样数据识别结果对比
Fig. 11. Comparison of identification results between original data and oversampled data using XGBoost
图 12 极端梯度提升树原始数据与过采样数据识别结果混淆矩阵对比
a.未经SMOTE算法处理；b.经过SMOTE算法处理
Fig. 12. Confusion matrix comparison between original data and oversampled data identification results of XGBoost
表 1 极端梯度提升树超参数选取及最优设定
Table 1. Hyperparameter selection and optimal setting of XGBoost
分类器 超参数 范围 最优设定 极端梯度提升树 伽马值 0~5 0.2 树的最大深度 1~50 27.0 叶节点样本权重之和最小值 1~25 1.0 
下载: 导出CSV
表 3 极端梯度提升树、支持向量机、逻辑回归测试集准确率、精确率、召回率和f1得分统计
Table 3. XGBoost, SVM, Logistic Regression testing set Accuracy, Precision, Recall and f1 score statistics
分类器 准确率（Ac） 精确率（Pr） 召回率（Re） f1得分（f1） 种类 极端梯度提升树 90% 93% 93% 93% 0（非裂缝） 84% 84% 84% 1（裂缝） 支持向量机 87% 88% 94% 91% 0（非裂缝） 83% 69% 76% 1（裂缝） 逻辑回归 68% 82% 71% 76% 0（非裂缝） 48% 63% 54% 1（裂缝）
下载: 导出CSV
-
Aghli, G., Soleimani, B., Moussavi-Harami, R., et al., 2016. Fractured Zones Detection Using Conventional Petrophysical Logs by Differentiation Method and Its Correlation with Image Logs. Journal of Petroleum Science and Engineering, 142: 152-162. https://doi.org/10.1016/j.petrol.2016.02.002 Bao, M. Y., Dong, S. Q., Zeng, L. B., et al., 2023. Artificial Intelligence Prediction Method for Tight Carbonate Reservoir Fracture Distribution Based on Seismic Attributes. Earth Science, 48(7): 2462-2474 (in Chinese with English abstract). Cao, J., Lei, D. W., Li, Y. W., et al., 2015. Ancient High-Quality Alkaline Lacustrine Source Rocks Discovered in the Lower Permian Fengcheng Formation, Junggar Basin. Acta Petrolei Sinica, 36(7): 781-790 (in Chinese with English abstract). Chen, G. H., Hu, C., Zeng, Y. L., et al., 2015. Logging Identification Method of Fillings in Fractures and Caverns in Carbonate Reservoir Based on BP Neural Network. Geophysical Prospecting for Petroleum, 54(1): 99-104 (in Chinese with English abstract). doi: 10.3969/j.issn.1000-1441.2015.01.014 Dong, S. Q., Wang, Z. Z., Zeng, L. B., 2016. Lithology Identification Using Kernel Fisher Discriminant Analysis with Well Logs. Journal of Petroleum Science and Engineering, 143: 95-102. https://doi.org/10.1016/j.petrol.2016.02.017 Dong, S. Q., Zeng, L. B., Lyu, W. Y., et al., 2020a. Fracture Identification by Semi-Supervised Learning Using Conventional Logs in Tight Sandstones of Ordos Basin, China. Journal of Natural Gas Science and Engineering, 76: 103131. https://doi.org/10.1016/j.jngse.2019.103131 Dong, S. Q., Zeng, L. B., Lyu, W. Y., et al., 2020b. Fracture Identification and Evaluation Using Conventional Logs in Tight Sandstones: A Case Study in the Ordos Basin, China. Energy Geoscience, 1(3-4): 115-123. https://doi.org/10.1016/j.engeos.2020.06.003 Dong, S. Q., Zeng, L. B., Che, X. H., et al, 2023. Application of Artificial Intelligence in Fracture Identification Using Well Logs in Tight Reservoirs. Earth Science, 48(7): 2443-2461 (in Chinese with English abstract). Fei, S. P., Hassan, M. A., He, Z. H., et al., 2021. Assessment of Ensemble Learning to Predict Wheat Grain Yield Based on UAV-Multispectral Reflectance. Remote Sensing, 13(12): 2338. https://doi.org/10.3390/rs13122338 Ghosh, S., Galvis-Portilla, H. A., Klockow, C. M., et al., 2018. An Application of Outcrop Analogues to Understanding the Origin and Abundance of Natural Fractures in the Woodford Shale. Journal of Petroleum Science and Engineering, 164: 623-639. https://doi.org/10.1016/j.petrol.2017.11.073 Gong, L., Wang, J., Gao, S., et al., 2021. Characterization, Controlling Factors and Evolution of Fracture Effectiveness in Shale Oil Reservoirs. Journal of Petroleum Science and Engineering, 203: 108655. https://doi.org/10.1016/j.petrol.2021.108655 Gu, Y. F., Zhang, D. Y., Lin, Y. B., et al., 2021. Data-Driven Lithology Prediction for Tight Sandstone Reservoirs Based on New Ensemble Learning of Conventional Logs: A Demonstration of a Yanchang Member, Ordos Basin. Journal of Petroleum Science and Engineering, 207: 109292. https://doi.org/10.1016/j.petrol.2021.109292 Guo, X. S., 2022. Discussion and Research Direction of Future Onshore Oil and Gas Exploration in China. Earth Science, 47(10): 3511-3523 (in Chinese with English abstract). Ja'Fari, A., Kadkhodaie-Ilkhchi, A., Sharghi, Y., et al., 2012. Fracture Density Estimation from Petrophysical Log Data Using the Adaptive Neuro-Fuzzy Inference System. Journal of Geophysics and Engineering, 9(1): 105-114. https://doi.org/10.1088/1742-2132/9/1/013 Jin, Z. J., Zhu, R. K., Liang, X. P., et al., 2021. Several Issues Worthy of Attention in Current Lacustrine Shale Oil Exploration and Development. Petroleum Exploration and Development, 48(6): 1276-1287 (in Chinese with English abstract). Lei, D. W., Chen, G. Q., Liu, H. L., et al., 2017. Study on the Forming Conditions and Exploration Fields of the Mahu Giant Oil (Gas) Province, Junggar Basin. Acta Geologica Sinica, 91(7): 1604-1619 (in Chinese with English abstract). doi: 10.3969/j.issn.0001-5717.2017.07.012 Li, T. Y., Wang, R. H., Wang, Z. Z., et al., 2018. Prediction of Fracture Density Using Genetic Algorithm Support Vector Machine Based on Acoustic Logging Data. Geophysics, 83(2): D49-D60. https://doi.org/10.1190/geo2017-0229.1 Liu, G. P., Zeng, L. B., Sun, G. Q., et al., 2020a. Natural Fractures in Tight Gas Volcanic Reservoirs and Their Influences on Production in the Xujiaweizi Depression, Songliao Basin, China. AAPG Bulletin, 104(10): 2099-2123. https://doi.org/10.1306/05122017169 Liu, G. P., Zeng, L. B., Wang, X. J., et al., 2020b. Natural Fractures in Deep Tight Gas Sandstone Reservoirs in the Thrust Belt of the Southern Junggar Basin, Northwestern China. Interpretation, 8(4): SP81-SP93. https://doi.org/10.1190/int-2020-0051.1 Luo, G., Xiao, L. Z., Shi, Y. Q., et al., 2022. Machine Learning for Reservoir Fluid Identification with Logs. Petroleum Science Bulletin, 7(1): 24-33 (in Chinese with English abstract). Shi, G. R., 2008. Superiorities of Support Vector Machine in Fracture Prediction and Gassiness Evaluation. Petroleum Exploration and Development, 35(5): 588-594 (in Chinese with English abstract). doi: 10.1016/S1876-3804(09)60091-4 Song, X. Z., Yao, X. Z., Li, G. S., et al., 2022. A Novel Method to Calculate Formation Pressure Based on the LSTM-BP Neural Network. Petroleum Science Bulletin, 7(1): 12-23 (in Chinese with English abstract). Tang, Y., Guo, W. J., Wang, X. T., et al., 2019. A New Breakthrough in Exploration of Large Conglomerate Oil Province in Mahu Sag and Its Implications. Xinjiang Petroleum Geology, 40(2): 127-137 (in Chinese with English abstract). Tokhmechi, B., Memarian, H., Noubari, H. A., et al., 2009. A Novel Approach Proposed for Fractured Zone Detection Using Petrophysical Logs. Journal of Geophysics and Engineering, 6(4): 365-373. https://doi.org/10.1088/1742-2132/6/4/004 Torlay, L., Perrone-Bertolotti, M., Thomas, E., et al., 2017. Machine Learning-XGBoost Analysis of Language Networks to Classify Patients with Epilepsy. Brain Informatics, 4(3): 159-169. https://doi.org/10.1007/s40708-017-0065-7 Xie, Y. X., Zhu, C. Y., Zhou, W., et al., 2018. Evaluation of Machine Learning Methods for Formation Lithology Identification: A Comparison of Tuning Processes and Model Performances. Journal of Petroleum Science and Engineering, 160: 182-193. https://doi.org/10.1016/j.petrol.2017.10.028 Xue, Y. C., Cheng, L. S., Mou, J. Y., et al., 2014. A New Fracture Prediction Method by Combining Genetic Algorithm with Neural Network in Low-Permeability Reservoirs. Journal of Petroleum Science and Engineering, 121: 159-166. https://doi.org/10.1016/j.petrol.2014.06.033 Yang, X., Wang, Z. Z., Zhou, Z. Y., et al., 2019. Lithology Classification of Acidic Volcanic Rocks Based on Parameter-Optimized AdaBoost Algorithm. Acta Petrolei Sinica, 40(4): 457-467 (in Chinese with English abstract). Zeng, L. B., 2010. Microfracturing in the Upper Triassic Sichuan Basin Tight-Gas Sandstones: Tectonic, Overpressure, and Diagenetic Origins. AAPG Bulletin, 94(12): 1811-1825. https://doi.org/10.1306/06301009191 Zeng, L. B., Jiang, J. W., Yang, Y. L., 2010. Fractures in the Low Porosity and Ultra-Low Permeability Glutenite Reservoirs: A Case Study of the Late Eocene Hetaoyuan Formation in the Anpeng Oilfield, Nanxiang Basin, China. Marine and Petroleum Geology, 27(7): 1642-1650. https://doi.org/10.1016/j.marpetgeo.2010.03.009 Zeng, L. B., Lyu, W. Y., Li, J. A., et al., 2016. Natural Fractures and Their Influence on Shale Gas Enrichment in Sichuan Basin, China. Journal of Natural Gas Science and Engineering, 30: 1-9. https://doi.org/10.1016/j.jngse.2015.11.048 Zeng, L. B., Lyu, W. Y., Zhang, Y. Z., et al., 2021. The Effect of Multi-Scale Faults and Fractures on Oil Enrichment and Production in Tight Sandstone Reservoirs: A Case Study in the Southwestern Ordos Basin, China. Frontiers in Earth Science, 9: 1-12. https://doi.org/10.3389/feart.2021.664629 Zhi, D. M., Song, Y., He, W. J., et al., 2019. Geological Characteristics, Resource Potential and Exploration Direction of Shale Oil in Middle-Lower Permian, Junggar Basin. Xinjiang Petroleum Geology, 40(4): 389-401 (in Chinese with English abstract). Zhi, D. M., Tang, Y., He, W. J., et al., 2021. Orderly Coexistence and Accumulation Models of Conventional and Unconventional Hydrocarbons in Lower Permian Fengcheng Formation, Mahu Sag, Junggar Basin. Petroleum Exploration and Development, 48(1): 38-51 (in Chinese with English abstract). Zhou, H., Jiang, Y., 2023. Improved Isolation Forest Method Based on High Contrast Subspace. Application Research of Computers, 40(2): 388-393 (in Chinese with English abstract). Zou, C. N., Yang, Z., Li, G. X., et al., 2022. Why can China Realize the Continental Shale Oil Revolution? Earth Science, 47(10): 3860-3863 (in Chinese with English abstract). 鲍明阳, 董少群, 曾联波, 等, 2023. 基于地震属性的致密碳酸盐岩储层裂缝分布的人工智能预测方法. 地球科学, 48(7): 2462-2474. http://www.cnki.com.cn/Article/CJFDTotal-DQKX20220814002.htm 曹剑, 雷德文, 李玉文, 等, 2015. 古老碱湖优质烃源岩: 准噶尔盆地下二叠统风城组. 石油学报, 36(7): 781-790. https://www.cnki.com.cn/Article/CJFDTOTAL-SYXB201906002.htm 陈钢花, 胡琮, 曾亚丽, 等, 2015. 基于BP神经网络的碳酸盐岩储层缝洞充填物测井识别方法. 石油物探, 54(1): 99-104. https://www.cnki.com.cn/Article/CJFDTOTAL-SYWT201501015.htm 董少群, 曾联波, 车小花, 等, 2023. 人工智能在致密储层裂缝测井识别中的应用. 地球科学, 48(7): 2443-2461. https://www.cnki.com.cn/Article/CJFDTOTAL-DQKX202307002.htm 郭旭升, 2022. 我国陆上未来油气勘探领域探讨与攻关方向. 地球科学, 47(10): 3511-3523. doi: 10.3799/dqkx.2022.873 金之钧, 朱如凯, 梁新平, 等, 2021. 当前陆相页岩油勘探开发值得关注的几个问题. 石油勘探与开发, 48(6): 1276-1287. https://www.cnki.com.cn/Article/CJFDTOTAL-SKYK202106021.htm 雷德文, 陈刚强, 刘海磊, 等, 2017. 准噶尔盆地玛湖凹陷大油(气)区形成条件与勘探方向研究. 地质学报, 91(7): 1604-1619. https://www.cnki.com.cn/Article/CJFDTOTAL-DZXE201707013.htm 罗刚, 肖立志, 史燕青, 等, 2022. 基于机器学习的致密储层流体识别方法研究. 石油科学通报, 7(1): 24-33. https://www.cnki.com.cn/Article/CJFDTOTAL-SYKE202201003.htm 石广仁, 2008. 支持向量机在裂缝预测及含气性评价应用中的优越性. 石油勘探与开发, 35(5): 588-594. https://www.cnki.com.cn/Article/CJFDTOTAL-SKYK200805012.htm 宋先知, 姚学喆, 李根生, 等, 2022. 基于LSTM-BP神经网络的地层孔隙压力计算方法. 石油科学通报, 7(1): 12-23. https://www.cnki.com.cn/Article/CJFDTOTAL-SYKE202201002.htm 唐勇, 郭文建, 王霞田, 等, 2019. 玛湖凹陷砾岩大油区勘探新突破及启示. 新疆石油地质, 40(2): 127-137. https://www.cnki.com.cn/Article/CJFDTOTAL-XJSD201902001.htm 杨笑, 王志章, 周子勇, 等, 2019. 基于参数优化AdaBoost算法的酸性火山岩岩性分类. 石油学报, 40(4): 457-467. https://www.cnki.com.cn/Article/CJFDTOTAL-SYXB201904007.htm 支东明, 宋永, 何文军, 等, 2019. 准噶尔盆地中‒下二叠统页岩油地质特征、资源潜力及勘探方向. 新疆石油地质, 40(4): 389-401. https://www.cnki.com.cn/Article/CJFDTOTAL-XJSD201904002.htm 支东明, 唐勇, 何文军, 等, 2021. 准噶尔盆地玛湖凹陷风城组常规‒非常规油气有序共生与全油气系统成藏模式. 石油勘探与开发, 48(1): 38-51. https://www.cnki.com.cn/Article/CJFDTOTAL-SKYK202101006.htm 周杭, 蒋瑜, 2023. 基于高对比度子空间的改进孤立森林方法. 计算机应用研究, 40(2): 388-393. https://www.cnki.com.cn/Article/CJFDTOTAL-JSYJ202302012.htm 邹才能, 杨智, 李国欣, 等, 2022. 中国为什么可以实现陆相"页岩油革命"? 地球科学, 47(10): 3860-3863. doi: 10.3799/dqkx.2022.841 -
点击查看大图
图(12) / 表(3)
计量
- 文章访问数: 417
- HTML全文浏览量: 1039
- PDF下载量: 59
- 被引次数: 0
