基于深度学习的页岩气藏压裂缝网反演方法研究

doi:10.13809/j.cnki.cn32-1825/te.2025.01.018

摘要/Abstract

摘要：

页岩气储集层致密性强，非均质性显著，自然产量极低，必须采用水力压裂技术进行增产改造才能获得工业气流，而评估压裂作业成效及优化工艺参数的关键在于获取准确的压裂缝网参数。传统裂缝监测技术（如微地震监测）费用高昂，无法实现井区全覆盖监测，而数值模拟预测模型需要大量的工程地质参数，导致地质资料不完整或缺失井段预测效果不佳，亟须一种经济高效地获取缝网参数的新方法。为此，提出一种基于深度学习的页岩气藏压裂缝网反演方法，其核心是以现场施工压裂曲线数据为基础，对压裂曲线特征参数进行量化分析，以缝网参数的强相关性指标作为输入，以微地震监测缝网参数（包括缝网长度、宽度、高度、体积）作为目标输出，建立BP（误差反向传播）神经网络反演模型，实现压裂缝网参数精确反演。根据渝西地区页岩气井现场450段压裂曲线，对模型进行了训练和参数优化，测试集缝网参数反演结果平均相对误差低于15%，验证了这种新方法对页岩气藏压裂缝网反演的可行性。

关键词: 页岩气, 压裂曲线, 缝网参数预测, BP神经网络, 反演

Abstract:

Shale gas reservoirs are characterized by high compactness and significant heterogeneity, with naturally low production that necessitates hydraulic fracturing technology for enhanced productivity to achieve industrial gas flow. The key to evaluating the effectiveness of fracturing operations and optimizing process parameters lies in obtaining accurate fracture network parameters. Traditional fracture monitoring techniques, such as microseismic monitoring, are costly and cannot achieve full coverage monitoring of well areas. Numerical simulation prediction models require a large number of engineering geological parameters, leading to poor prediction effects for geological data that are incomplete or missing well sections. There is an urgent need for a new method that is economically efficient in obtaining fracture network parameters. To address this, a shale gas reservoir fracture network inversion method based on deep learning was proposed. The core of this method is to quantitatively analyze the fracturing curve characteristic parameters based on the site fracturing curve data, using strongly correlated indicators of fracture network parameters as inputs and microseismic monitoring fracture network parameters (including length, width, height, and volume) as target outputs. A back-propagation (BP) neural network inversion model was established to achieve accurate inversion of fracture network parameters. The model was trained and optimized using 450 fracturing curve segments from shale gas wells in western Chongqing, with the average relative error of fracture network parameter inversion results in the test set being below 15%, which verified the feasibility of this new method for inversion of shale gas reservoir fracture networks.

Key words: shale gas, fracturing curve, fracture network parameter prediction, BP neural network, inversion

中图分类号:

TE377

陈维铭,蒋琳,罗彤彤等. 基于深度学习的页岩气藏压裂缝网反演方法研究[J]. 油气藏评价与开发, 2025, 15(1): 142-151.

CHEN Weiming,JIANG Lin,LUO Tongtong, et al. Research on deep learning-based fracture network inversion method for shale gas reservoirs[J]. Petroleum Reservoir Evaluation and Development, 2025, 15(1): 142-151.

图/表 14

图1

图2

图3

表1

图4

图5

表2

表3

图6

图7

图8

图9

表4

表5

参考文献 29

[1]	蔡勋育, 周德华, 赵培荣, 等. 中国石化深层、常压页岩气勘探开发进展与展望[J]. 石油实验地质, 2023, 45(6): 1039-1049.
	CAI Xunyu, ZHOU Dehua, ZHAO Peirong, et al. Development progress and outlook of deep and normal pressure shale gas of SINOPEC[J]. Petroleum Geology & Experiment, 2023, 45(6): 1039-1049.
[2]	邹才能, 董大忠, 王社教, 等. 中国页岩气形成机理、地质特征及资源潜力[J]. 石油勘探与开发, 2010, 37(6): 641-653.
	ZOU Caineng, DONG Dazhong, WANG Shejiao, et al. Geological characteristics, formation mechanism and resource potential of shale gas in China[J]. Petroleum Exploration and Development, 2010, 37(6): 641-653.
[3]	郭旭升, 胡东风, 俞凌杰, 等. 页岩自封闭性与页岩气保存的微观机理研究[J]. 石油实验地质, 2023, 45(5): 821-831.
	GUO Xusheng, HU Dongfeng, YU Lingjie, et al. Study on the micro mechanism of shale self-sealing and shale gas preservation[J]. Petroleum Geology & Experiment, 2023, 45(5): 821-831.
[4]	李敏, 刘雅利, 冯动军, 等. 中国海相页岩气资源潜力及未来勘探方向[J]. 石油实验地质, 2023, 45(6): 1097-1108.
	LI Min, LIU Yali, FENG Dongjun, et al. Potential and future exploration direction of marine shale gas resources in China[J]. Petroleum Geology & Experiment, 2023, 45(6): 1097-1108.
[5]	赵金洲, 任岚, 沈骋, 等. 页岩气储层缝网压裂理论与技术研究新进展[J]. 天然气工业, 2018, 38(3): 1-14.
	ZHAO Jinzhou, REN Lan, SHEN Pin, et al. Latest research progresses in network fracturing theories and technologies for shale gas reservoirs[J]. Natural Gas Industry, 2018, 38(3): 1-14.
[6]	杨永华, 宋燕高, 王兴文, 等. 威荣页岩气田压裂实践与认识[J]. 石油实验地质, 2023, 45(6): 1143-1150.
	YANG Yonghua, SONG Yangao, WANG Xingwen, et al. Practice and understanding of fracturing in Weirong shale gas field[J]. Petroleum Geology & Experiment, 2023, 45(6): 1143-1150.
[7]	郭建春, 赵志红, 路千里, 等. 深层页岩缝网压裂关键力学理论研究进展[J]. 天然气工业, 2021, 41(1): 102-117.
	GUO Jianchun, ZHAO Zhihong, LU Qianli, et al. Research progress in key mechanical theories of deep shale network fracturing[J]. Natural Gas Industry, 2021, 41(1): 102-117.
[8]	杨学锋, 夏自强, 赵圣贤, 等. 深层页岩气水平井天然裂缝发育特征及其对精细开发的启示: 以川南泸州区块五峰组—龙马溪组为例[J]. 石油实验地质, 2024, 46(4): 735-747.
	YANG Xuefeng, XIA Ziqiang, ZHAO Shengxian, et al. Development characteristics of natural fractures in horizontal wells for deep shale gas and their implications for enhanced development: a case study of Wufeng-Longmaxi formations in Luzhou area, southern Sichuan Basin[J]. Petroleum Geology & Experiment, 2024, 46(4): 735-747.
[9]	ZHAO Haifeng, WANG Xiaohua, LIU Zhiyuan. Experimental investigation of hydraulic sand fracturing on fracture propagation under the influence of coal macrolithotypes in Hancheng Block, China[J]. Journal of Petroleum Science and Engineering, 2019, 175: 60-71.
[10]	赵金洲, 付永强, 王振华, 等. 页岩气水平井缝网压裂施工压力曲线的诊断识别方法[J]. 天然气工业, 2022, 42(2): 11-19.
	ZHAO Jinzhou, FU Yongqiang, WANG Zhenhua, et al. Study on diagnosis model of shale gas fracture network fracturing operation pressure curves[J]. Natural Gas Industry, 2022, 42(2): 11-19.
[11]	王益民, 李继庆, 万云强, 等. 基于返排数据计算页岩气井压裂有效体积的方法及应用[J]. 石油实验地质, 2023, 45(6): 1215-1220.
	WANG Yimin, LI Jiqing, WAN Yunqiang, et al. Method of estimating the effective fracture volume of shale gas wells using flowback data and its application[J]. Petroleum Geology & Experiment, 2023, 45(6): 1215-1220.
[12]	李宪文, 赵振峰, 李喆, 等. 随机分形压裂水平井缝网参数反演方法[J]. 断块油气田, 2019, 26(2): 205-209.
	LI Xianwen, ZHAO Zhenfeng, LI Zhe, et al. Stochastic fractal model for fracture network parameter inversion of fracturing horizontal well[J]. Fault-Block Oil & Gas Field, 2019, 26(2): 205-209.
[13]	易新民, 唐雪萍, 梁涛, 等. 利用测井资料预测判断水力压裂裂缝高度[J]. 西南石油大学学报(自然科学版), 2009, 31(5): 21-24.
	YI Xinmin, TANG Xueping, LIANG Tao, et al. Prediction and assessment of fracture height of hydraulic fracturing with logging date[J]. Journal of Southwest Petroleum University (Science & Technology Edition), 2009, 31(5): 21-24.
[14]	LI H T, LUO H W, XIANG Y W, et al. DTS-based hydraulic fracture identification and production profile interpretation method for horizontal shale gas wells[J]. Natural Gas Industry B, 2021, 8(5): 494-504.
[15]	尹丛彬, 李彦超, 王素兵, 等. 页岩压裂裂缝网络预测方法及其应用[J]. 天然气工业, 2017, 37(4): 60-68.
	YIN Congbin, LI Yanchao, WANG Subing, et al. Methodology of hydraulic fracture network prediction in shale reservoirs and its application[J]. Natural Gas Industry, 2017, 37(4): 60-68.
[16]	朱昭颖, 戴邓仅, 鲁伍支, 等. 基于水击效应的水平井压裂裂缝数值诊断[J]. 计算物理, 2023, 40(6): 735-741.
	ZHU Zhaoying, DAI Dengjin, LU Wuzhi, et al. Numerical fracture diagnosis method for horizontal well fracturing based on water hammer[J]. Chinese Journal of Computational Physics, 2023, 40(6): 735-741.
[17]	赵玉龙, 崔乾晨, 高上钧, 等. 基于停泵压降数据的页岩气井单段裂缝参数反演: 以长宁N209井区页岩气井为例[J]. 深圳大学学报(理工版), 2024, 41(1): 22-32.
	ZHAO Yulong, CUI Qianchen, GAO Shangyun, et al. The inversion of the single-stage hydraulic fracture parameters for shale gas well based on the shut-in pressure-drop data: A case of shale gas well in Changning N209 well area[J]. Journal of Shenzhen University(Science and Engineering), 2024, 41(1): 22-32.
[18]	于豪, 黄家强, 兰雪梅, 等. 川西北双鱼石地区栖霞组地震资料优化处理及裂缝预测技术应用[J]. 科学技术与工程, 2020, 20(22): 8933-8942.
	YU Hao, HUANG Jiaqiang, LAN Xuemei, et al. Application of seismic data optimal processing and fracture prediction in the Shuangyushi Block, Northwest Sichuan[J]. Science Technology and Engineering, 2020, 20(22): 8933-8942.
[19]	ZHANG D X, CHEN Y T, MENG J, et al. Synthetic well logs generation via Recurrent Neural Networks[J]. Petroleum Exploration and Development, 2018, 45(4): 629-639.
[20]	冯明洁. 含天然裂缝页岩水力压裂缝网的形貌演化及神经网络预测模型研究[D]. 重庆: 重庆大学, 2022.
	FENG Mingjie. Study on morphology evolution and neural network prediction model of hydraulic fracture network in shale with natural fracture[D]. Chongqing: Chongqing University, 2022.
[21]	常德宽, 雍学善, 王一惠, 等. 基于深度卷积神经网络的地震数据断层识别方法[J]. 石油地球物理勘探, 2021, 56(1): 1-8.
	CHANG Dekuan, YONG Xueshan, WANG Yihui, et al. Seismic fault interpretation based on deep convolutional neural networks[J]. Oil Geophysical Prospecting, 2021, 56(1): 1-8.
[22]	ZHANG M Z, LIU Z Y, JIANG Q, et al. Influence of natural caves on hydraulic fracturing pressure curves: Numerical modeling and ANNs[J]. Arabian Journal of Geosciences, 2021, 14(20): 1-22.
[23]	刘宝锋. 基于人工神经网络的超深井储层敏感性预测[D]. 青岛: 中国石油大学(华东), 2009.
	LIU Baofeng. The sensitivity prediction of the ultra-deep reservoir based on artificial neural network[D]. Qingdao: China University of Petroleum (East China), 2009.
[24]	LIU Y Y, MA X H, ZHANG X W, et al. A deep-learning-based prediction method of the estimated ultimate recovery(EUR) of shale gas wells[J]. Petroleum Science, 2021, 18(5): 1450-1464.
[25]	王兴文, 缪尉杰, 何新星, 等. 川南威荣气田深层页岩气工程技术进展[J]. 石油实验地质, 2023, 45(6): 1170-1177.
	WANG Xingwen, MIAO Weijie, HE Xinxing, et al. Progress in deep shale gas engineering technology in Weirong gas field in southern Sichuan[J]. Petroleum Geology & Experiment, 2023, 45(6): 1170-1177.
[26]	刘庆. 水平井体积压裂井底净压力计算及分析[D]. 北京: 中国石油大学(北京), 2017.
	LIU Qing. Net pressure calculation and analysis of horizontal well volume fracturing[D]. Beijing: China University of Petroleum(Beijing), 2017.
[27]	卞晓冰, 蒋延学, 贾长贵, 等. 基于施工曲线的页岩气井压后评估新方法[J]. 天然气工业, 2016, 36(2): 60-65.
	BIAN Xiaobing, JIANG Tingxue, JIA Changgui, et al. A new post-fracturing evaluation method for shale gas wells based on fracturing curves[J]. Natural Gas Industry, 2016, 36(2): 60-65.
[28]	彭勇民, 董世雄, 边瑞康, 等. 四川盆地东部页岩气水平井裂缝识别方法及应用[J]. 石油实验地质, 2023, 45(6): 1196-1203.
	PENG Yongmin, DONG Shixiong, BIAN Ruikang, et al. Method for identification of fractures in shale gas horizontal wells in eastern Sichuan Basin and its application[J]. Petroleum Geology & Experiment, 2023, 45(6): 1196-1203.
[29]	SACCENTI E, HENDRIKS M H W B, SMILDE A K. Corruption of the Pearson correlation coefficient by measurement error and its estimation, bias, and correction under different error models[J]. Scientific Reports, 2020, 10(1): 438.

数据类型	平均值	最大值	最小值
前置液单位体积上升压力/ [（MPa·min）/m³]	1.24	2.87	0.37
破裂压力/MPa	99.60	114.52	88.04
前置液压力下降斜率/（MPa/min）	-4.30	-1.32	-7.15
前置液下降压差/MPa	6.46	29.52	1.25
压裂液排量/（m³/min）	17.42	19.00	15.63
延伸压力/MPa	98.18	104.00	93.91
压裂时间/min	167.60	235.01	88.93
总砂量/m³	337.46	440.44	320.11
暂堵生效瞬时上升压力/MPa	8.17	14.69	2.36
暂堵前后平均上升压力/MPa	5.95	16.34	2.31
低黏占比/%	0.95	1.00	0.95
波动周期/min	2.88	5.27	1.19
波动振幅/MPa	2.55	4.78	1.18
波动次数	2.26	4.00	1.00
替挤峰值/MPa	105.61	113.05	93.62
停泵斜率	-7.04	-1.96	-14.10
裂缝闭合压力/MPa	71.77	81.18	65.02
拟线性流斜率	-0.39	-0.01	-0.85
缝网长度/m	399.17	475.00	284.00
缝网宽度/m	99.22	125.00	80.00
缝网高度/m	93.91	124.00	70.00
缝网体积/10⁴ m³	296.12	505.00	243.25

输入参数	缝网长度模型	缝网宽度模型	缝网高度模型	缝网体积模型
前置液单位体积上升压力		○	○
破裂压力	○	○		○
前置液压力下降斜率			○
前置液下降压差		○		○
压裂液排量	○
延伸压力		○	○	○
压裂时间	○	○		○
总砂量	○	○	○	○
低黏占比	○	○		○
暂堵生效瞬时上升压力	○
暂堵前后平均上升压力	○	○
波动周期	○
波动振幅	○
波动次数
替挤峰值			○
停泵斜率	○	○		○
闭合压力				○
拟线性流斜率	○	○	○

模型参数	缝网长度模型	缝网宽度模型	缝网高度模型	缝网体积模型
输入层节点数	11	10	6	8
输出层节点数	1	1	1	1
隐含层数	3	3	3	3
隐含层节点数	8	7	6	7
激活函数	purelin(x) tansig(x)
训练方法	Levenberg-Marquardt(trainlm)
最大训练次数	10 000
学习率	0.001
训练精度	0.000 1

评价指标	缝网长度模型	缝网宽度模型	缝网高度模型
训练集平均相对误差/%	6.68	7.84	9.11
预测集平均相对误差/%	8.19	9.72	10.61
训练集平均绝对误差/m	27.61	5.69	4.86
预测集平均绝对误差/m	33.89	7.43	6.09
训练集均方根误差/m	37.09	7.40	7.44
预测集均方根误差/m	43.03	9.29	7.37
训练集相对误差大于10%的测点比例/%	21.65	31.80	34.62
预测集相对误差大于10%的测点比例/%	37.67	40.16	52.48
训练集相对误差大于15%的测点比例/%	11.87	13.32	21.03
预测集相对误差大于15%的测点比例/%	12.14	25.26	26.95

评价指标	缝网体积模型
训练集平均相对误差/%	11.39
预测集平均相对误差/%	13.56
训练集平均绝对误差/10⁴ m³	9.51
预测集平均绝对误差/10⁴ m³	10.74
训练集均方根误差/10⁴ m³	12.09
预测集均方根误差/10⁴ m³	12.02
训练集相对误差大于10%的测点比例/%	39.27
预测集相对误差大于10%的测点比例/%	57.13
训练集相对误差大于15%的测点比例/%	25.68
预测集相对误差大于15%的测点比例/%	31.60