• 首页关于本刊投稿须知期刊订阅编委会期刊合作English
引用本文:
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 30次   下载 24 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于非监督学习的水力压裂裂缝平面识别
马孜卓1,2, 郑忆康1,2, 薛清峰1,2, 翟鸿宇3, 雷兴林4
1.中国科学院地质与地球物理研究所, 中国科学院油气资源研究院重点实验室, 北京 100029;2.中国科学院地球科学研究院, 北京 100029;3.中国地震局地球物理研究所, 北京 100081;4.日本产业技术综合研究所, 筑波 3058567
摘要:
在水力压裂施工中,如何有效获取压裂过程中产生的裂缝形态以及裂缝的动态扩展过程一直是困扰学术界和工业界的问题。目前,常规利用微地震事件定位结果进行分析的方法存在需要人工干预、散点信息表示能力不足等问题;采用数值模拟分析的方法往往因复杂的地下介质情况而引入计算偏差。本文基于非监督学习算法,通过提取微地震事件的空间和时间信息,实现对裂缝平面的识别以及裂缝网络拓展路径的分析;并通过引入水力压裂岩石物理实验,利用实际监测获得的声发射数据以及对应的真实破裂情况的CT扫描数据,检验方法的可行性。最终结果表明,本文所提方法对主断裂有较好的识别效果,识别结果与CT扫描的真实结果吻合性较好。
关键词:  非监督学习  聚类分析  裂缝平面识别  水力压裂  声发射
DOI:
分类号:P315
基金项目:中国科学院前沿科学重点研究项目(QYZDY-SSW-DQC009)资助
Hydraulic Fracture Identification Based on Unsupervised Learning
Ma Zizhuo1,2, Zheng Yikang1,2, Xue Qingfeng1,2, Zhai Hongyu3, Lei Xinglin4
1.Key Laboratory of Petroleum Resource Research, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China;2.Institution of Earth Science, Chinese Academy of Sciences, Beijing 100029, China;3.Institute of Geophysics, China Earthquake Administration, Beijing 100081, China;4.National Institute of Advanced Industrial Science and Technology, Tsukuba 3058567, Japan
Abstract:
For hydraulic fracturing,how to effectively obtain the fracture pattern generated during the fracturing process and the dynamic expansion process of fractures has been a problem that plagues the academic and industrial circles. At present,conventional methods usually use the obtained microseismic event localization results for qualitative manual analysis or numerical simulation analysis with various simulation tools. The method of using the microseismic event localization result for analysis has the problem of requiring manual intervention and insufficient capability of information representation. With the numerical simulation method,there is often a deviation between the calculation and the actual problem due to the complexity of the underground medium. In attempt of solving above problem,in this paper we apply the data-driven method,by extracting the spatial and temporal information of microseismic events,to realize the identification of the crack plane and the fracture network path. At the same time,the feasibility of our method was tested by introducing hydraulic fracturing rock experiments in which comparing the measured acoustic emission data with the corresponding CT scan data of the actual rupture conditions. The results show that our method has a good recognition on the main crack,and the algorithm result is in good agreement with the real data from the CT scan.
Key words:  Unsupervised learning  Clustering analysis  Fracture plane identification  Hydraulic fracturing  Acoustic emission