南海天然气水合物智能识别方法与应用

田冬梅; 杨胜雄; 刘鑫; 李沅衡; 胡广; 曹荆亚; 周军明; 邓雨恬

doi:10.16562/j.cnki.0256-1492.2024092401

南海天然气水合物智能识别方法与应用

1.
南方海洋科学与工程广东省实验室（广州），广州 511458

2.
广州海洋地质调查局，广州 511458

基金项目: 国家自然科学基金国家重大科研仪器研制项目“海底地震与电磁同步探测系统关键技术及验证样机”（42327901）；国家自然科学基金项目“南海北部高富集天然气水合物储层特征与成藏控制机理研究”（U2244224）；广州市基础与应用基础研究项目“基于地震属性海域天然气水合物识别方法研究”（2023A04J0916）

详细信息

作者简介: 田冬梅（1995—），女，博士，主要从事海洋地球物理、天然气水合物识别计算研究，E-mail：dongmeitian@gmlab.ac.cn

通讯作者: 刘鑫（1983—），男，硕士，高级工程师，主要从事海洋地质环境相关研究，E-mail：lxin05@gmlab.ac.cn

中图分类号: P714

收稿日期: 2024-09-24

修回日期: 2024-12-23

刊出日期: 2024-12-28

Intelligent identification and application of gas hydrate in South China Sea

1.
Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou), Guangzhou 511458, China

2.
Guangzhou Marine Geological Survey, Guangzhou 511458, China

More Information

Corresponding author: LIU Xin, lxin05@gmlab.ac.cn

Received Date: 24 September 2024

Revised Date: 23 December 2024

Publish Date: 28 December 2024

摘要

摘要:
天然气水合物是一种重要的能源资源，具有能量高、储量大、分布广和埋藏浅等优势。准确地从地层中识别出天然气水合物储层是应用天然气水合物资源的必要前提。本文围绕水合物勘探识别的难点问题，结合海洋-地质-人工智能学科交叉技术，以具有显示度的地球物理属性参数为基础，研究并提出了有效的含水合物地层识别技术方法，在中国南海东沙海域研究区进行了方法的验证，选择了几种较为常用的机器学习算法，例如随机森林、Bagging、AdaBoost、和最近邻（KNN）算法，对水合物变化灵敏度较高的纵波速度和密度属性进行数据分析，通过训练优化不同算法模型参数，对比不同算法模型的识别分类效果。结果表明，这几种算法都能够较好地对地层中是否含有水合物进行区分，其中KNN性能最好，表明基于机器学习手段能够提高天然气水合物的识别精度和准确性。
- 水合物 /
- 识别 /
- 机器学习 /
- 地震属性 /
- 南海
Abstract:
Gas hydrate is an important ideal energy source, with advantages of high energy, large reserves, wide distribution, and shallow burial. Accurate identification of gas hydrate reservoirs and estimation of hydrate saturation are the prerequisite for the application of gas hydrate resources. This study focuses on the difficult issues of hydrate identification, combining the interdisciplinary technologies of oceanology, geology, and artificial intelligence. Effective methods of hydrate-bearing strata identification were proposed based on the geophysical attributes, and verified in the Dongsha area of South China Sea. Machine-learning algorithms were used to analyze whether the sediment contains gas hydrates. Several commonly used machine-learning algorithms were selected, including random forest, Bagging, AdaBoost, and KNN; and data were analyzed based on the P-wave velocity and density attributes that are more sensitive to hydrate existence. The parameters of different algorithms were trained and optimized, and the effects of different algorithms on the identification and classification were compared. All these algorithms could do good on whether there is hydrate in the sediment, of which KNN algorithm was shown the best. Therefore, machine-learning-based methods could improve the identification accuracy of gas hydrate.
- hydrate /
- identification /
- machine learning /
- seismic attributes /
- South China Sea

HTML全文

图 1 全球天然气水合物分布图（据文献[3]修改）

Figure 1.

下载: 全尺寸图片幻灯片

图 2 研究区位置

Figure 2.

下载: 全尺寸图片幻灯片

图 3 GMGS2-05、GMGS2-08和GMGS2-16站位纵波速度和密度测井曲线

Figure 3.

下载: 全尺寸图片幻灯片

图 4 AdaBoost算法对GMGS2-05、GMGS2-08和GMGS2-16纵波速度和密度数据训练识别水合物结果

Figure 4.

下载: 全尺寸图片幻灯片

图 5 随机森林算法对GMGS2-05、GMGS2-08和GMGS2-16站位纵波速度和密度数据训练识别水合物结果

Figure 5.

下载: 全尺寸图片幻灯片

图 6 Bagging算法对GMGS2-05、GMGS2-08和GMGS2-16站位纵波速度和密度数据训练识别水合物结果

Figure 6.

下载: 全尺寸图片幻灯片

图 7 KNN算法对GMGS2-05、GMGS2-08和GMGS2-16站位纵波速度和密度数据训练识别水合物结果

Figure 7.

下载: 全尺寸图片幻灯片

图 8 不同算法模型的ROC曲线

Figure 8.

下载: 全尺寸图片幻灯片

表 1 评价结果指标意义

Table 1. Significance of evaluation results indicators

	预测为真	预测为负
实际为真	TP	FN
实际为负	FP	TN

下载: 导出CSV

表 2 每个模型的超参数调整

Table 2. Hyperparameter tuning for each algorithm

监督模型	超参数	调整范围	最优参数
AdaBoost	N estimators	5～25	15
AdaBoost	Learning rate	0.1～0.9	0.5
RF	max leaf nodes	5～45	10
RF	n estimators	0～150	30
Bagging	n estimators	10～500	100
Bagging	Max samples	50～500	150
KNN	N neighbors	5～200	10
KNN	weights	uniform/distance	uniform

下载: 导出CSV

表 3 各算法评价指标计算结果

Table 3. Calculation results of evaluation indicators of each algorithm

Supervision algorithm	Accuracy	Precision	Recall	F₁ 值	AUC
AdaBoost	0.9171	0.8625	0.4726	0.6106	0.9146
Random forest	0.9388	0.8857	0.6370	0.7410	0.9278
Bagging	0.9341	0.8725	0.6096	0.7177	0.9449
KNN	0.9379	0.8636	0.6507	0.7422	0.9580

下载: 导出CSV

参考文献(28)

[1]	Sloan E D Jr, Koh C A. Clathrate Hydrates of Natural Gases[M]. 3rd ed. Boca Raton: CRC Press, 2007.
[2]	Chong Z R, Yang S H B, Babu P, et al. Review of natural gas hydrates as an energy resource: Prospects and challenges[J]. Applied Energy, 2016, 162:1633-1652. doi: 10.1016/j.apenergy.2014.12.061
[3]	Collett T S, Johnson A H, Knapp C C, et al. Natural gas hydrates-A review[M]//Collett T, Johnson A, Knapp C, et al. Natural Gas Hydrates—Energy Resource Potential and Associated Geologic Hazards. American Association of Petroleum Geologists, 2009, 89: 146-219.
[4]	Boswell R, Hancock S, Yamamoto K, et al. Natural gas hydrates: Status of potential as an energy resource[M]//Letcher T M. Future Energy: Improved, Sustainable and Clean Options for our Planet. 3rd ed. New York: Elsevier, 2020: 111-135.
[5]	Yang S X, Liang J Q, Lei Y, et al. GMGS4 gas hydrate drilling expedition in the South China Sea[J]. Fire in the Ice, 2017, 17(1):7-11.
[6]	雷裕红, 宋颖睿, 张立宽, 等. 海洋天然气水合物成藏系统研究进展及发展方向[J]. 石油学报, 2021, 42(6):801-820 LEI Yuhong, SONG Yingrui, ZHANG Likuan, et al. Research progress and development direction of reservoir-forming system of marine gas hydrates[J]. Acta Petrolei Sinica, 2021, 42(6):801-820.]
[7]	Zhang G X, Liang J Q, Lu J A, et al. Geological features, controlling factors and potential prospects of the gas hydrate occurrence in the east part of the Pearl River Mouth Basin, South China Sea[J]. Marine and Petroleum Geology, 2015, 67:356-367. doi: 10.1016/j.marpetgeo.2015.05.021
[8]	Liang J Q, Zhang W, Lu J A, et al. Geological occurrence and accumulation mechanism of natural gas hydrates in the eastern Qiongdongnan Basin of the South China Sea: Insights from Site GMGS5-W9-2018[J]. Marine Geology, 2019, 418:106042. doi: 10.1016/j.margeo.2019.106042
[9]	Sha Z B, Liang J Q, Zhang G X, et al. A seepage gas hydrate system in northern South China Sea: Seismic and well log interpretations[J]. Marine Geology, 2015, 366:69-78. doi: 10.1016/j.margeo.2015.04.006
[10]	Milkov A V. Global estimates of hydrate-bound gas in marine sediments: how much is really out there?[J]. Earth-Science Reviews, 2004, 66(3-4):183-197. doi: 10.1016/j.earscirev.2003.11.002
[11]	勾丽敏, 张金华, 王嘉玮. 海洋天然气水合物地震识别方法研究进展[J]. 地球物理学进展, 2017, 32(6):2626-2635 GOU Limin, ZHANG Jinhua, WANG Jiawei. Progress in seismic identification approach of marine gas hydrate[J]. Progress in Geophysics, 2017, 32(6):2626-2635.]
[12]	宋海斌, 张岭, 江为为, 等. 海洋天然气水合物的地球物理研究(Ⅲ): 似海底反射[J]. 地球物理学进展, 2003, 18(2):182-187 SONG Haibin, ZHANG Ling, JIANG Weiwei, et al. Geophysical researches on marine gas hydrates (Ⅲ): bottom simulating reflections[J]. Progress in Geophysics, 2003, 18(2):182-187.]
[13]	White R S. Gas hydrate layers trapping free gas in the Gulf of Oman[J]. Earth and Planetary Science Letters, 1979, 42(1):114-120. doi: 10.1016/0012-821X(79)90196-1
[14]	Stoll R D, Ewing J, Bryan G M. Anomalous wave velocities in sediments containing gas hydrates[J]. Journal of Geophysical Research, 1971, 76(8):2090-2094. doi: 10.1029/JB076i008p02090
[15]	Liu X W, He J, Sun Q L. Gas hydrate identification from △Vp/△Vs[C]//Beijing 2009 International Geophysical Conference & Exposition. Beijing: Society of Exploration Geophysicists, 2009: 150.
[16]	Tian D M, Liu X W. Identification of gas hydrate based on velocity cross plot analysis[J]. Marine Geophysical Research, 2021, 42(2):11. doi: 10.1007/s11001-021-09431-3
[17]	Chen J Z, You J C, Wei J T, et al. Interpreting XGBoost predictions for shear-wave velocity using SHAP: Insights into gas hydrate morphology and saturation[J]. Fuel, 2024, 364:131145. doi: 10.1016/j.fuel.2024.131145
[18]	Luo Y L, Zhang G L, Liang G W, et al. Limited-label multiscale deep-learning multihorizon tracking[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62:5918814.
[19]	王迪, 张益明, 张繁昌, 等. 利用先验信息约束的深度学习方法定量预测致密砂岩“甜点”[J]. 石油地球物理勘探, 2023, 58(1):65-74 WANG Di, ZHANG Yiming, ZHANG Fanchang, et al. Quantitative prediction of tight sandstone sweet spots based on deep learning method with prior information constraints[J]. Oil Geophysical Prospecting, 2023, 58(1):65-74.]
[20]	Bai Y, Tan M J. Dynamic committee machine with fuzzy-c-means clustering for total organic carbon content prediction from wireline logs[J]. Computers & Geosciences, 2021, 146:104626.
[21]	陈钢花, 张寓侠, 王军, 等. 双向长短时记忆神经网络在滩坝砂储层岩性识别中的应用[J]. 测井技术, 2023, 47(3):319-325 CHEN Ganghua, ZHANG Yuxia, WANG Jun, et al. Application of BiLSTM in lithology identification of beach-bar sand reservoir[J]. Well Logging Technology, 2023, 47(3):319-325.]
[22]	Chen Y J, Dunn K J, Liu X W, et al. New method for estimating gas hydrate saturation in the Shenhu area[J]. Geophysics, 2014, 79(5):IM11-IM22. doi: 10.1190/geo2013-0264.1
[23]	Zhu X Y, Liu T, Ma S, et al. Morphology identification of gas hydrate based on a machine learning method and its applications on saturation estimation[J]. Geophysical Journal International, 2023, 234(2):1307-1325. doi: 10.1093/gji/ggad133
[24]	杨笑, 王志章, 周子勇, 等. 基于参数优化AdaBoost算法的酸性火山岩岩性分类[J]. 石油学报, 2019, 40(4):457-467 YANG Xiao, WANG Zhizhang, ZHOU Ziyong, et al. Lithology classification of acidic volcanic rocks based on parameter-optimized AdaBoost algorithm[J]. Acta Petrolei Sinica, 2019, 40(4):457-467.]
[25]	Breiman L. Random forests[J]. Machine Learning, 2001, 45(1):5-23. doi: 10.1023/A:1010933404324
[26]	任涛, 林梦楠, 陈宏峰, 等. 基于Bagging集成学习算法的地震事件性质识别分类[J]. 地球物理学报, 2019, 62(1):383-392 REN Tao, LIN Mengnan, CHEN Hongfeng, et al. Seismic event classification based on Bagging ensemble learning algorithm[J]. Chinese Journal of Geophysics, 2019, 62(1):383-392.]
[27]	Bayar B, Stamm M C. Constrained convolutional neural networks: a new approach towards general purpose image manipulation detection[J]. IEEE Transactions on Information Forensics and Security, 2018, 13(11):2691-2706. doi: 10.1109/TIFS.2018.2825953
[28]	Nguyen B, Morell C, De Baets B. Large-scale distance metric learning for k-nearest neighbors regression[J]. Neurocomputing, 2016, 214:805-814. doi: 10.1016/j.neucom.2016.07.005