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:浏览次   下载 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于双向长短期记忆网络的流体高精度识别新方法
周雪晴1,2,张占松1,2,朱林奇3,4,张超谟1,2
(1.长江大学油气资源与勘探技术教育部重点实验室,湖北武汉 430100;2.长江大学地球物理与石油资源学院,湖北武汉 430100;3.中国科学院深海科学与工程研究所,海南三亚 572000;4.青岛海洋科学与技术国家实验室海洋地质过程与环境功能实验室,山东青岛 266237)
摘要:
碳酸盐岩储层的储集空间类型多样、储层性质复杂,导致流体的测井响应受到强非均质性的影响,给流体识别工作带来极大困难。针对该问题,提出基于测井序列信息的双向长短期记忆网络(Bi-LSTM)流体识别模型,从测井响应特征差异性分析及相似性分析两方面出发,确定敏感曲线,结合Bi-LSTM网络的输入要求,建立流体识别样本库,并获得基于Bi-LSTM的流体识别模型。应用该方法对鄂尔多斯盆地马家沟组进行流体识别,与单向LSTM模型及其他3类机器学习算法预测结果进行对比。结果表明:基于Bi-LSTM的流体识别模型流体识别的符合率从82.7%提高到91.5%,取得较好的应用效果;该模型既能充分利用井下对应深度测井曲线的响应值,又能兼顾测井曲线随深度的变化趋势和前后关联,最大程度避免储层纵向非均质性带来的影响,提高流体识别能力。
关键词:  流体识别  双向长短期记忆网络  碳酸盐岩  测井序列
DOI:10.3969/j.issn.1673-5005.2021.01.008
分类号::P 631.8
文献标识码:A
基金项目:
A new method for high-precision fluid identification in bidirectional long short-term memory network
ZHOU Xueqing1,2, ZHANG Zhansong1,2, ZHU Linqi3,4, ZHANG Chaomo1,2
(1.Key Laboratory of Exploration Technologies for Oil and Gas Resources (Yangtze University), Ministry of Education, Wuhan 430100,China;2.Hubei Cooperative Innovation Center of Unconventional Oil and Gas (Yangtze University), Wuhan 430100, China;3.Institute of Deep-sea Science and Engineering, Chinese Academy of Science, Sanya 572000, China;4.Laboratory for Marine Geology, Qingdao National Laboratory for Marine Science and Technology, Qingdao 266237, China)
Abstract:
Due to the diversity of space types and complex reservoir properties of carbonate reservoir, the logging response of fluid is affected by strong heterogeneity, which causes great difficulties in fluid identification. Aiming at this problem of fluid identification in complex carbonate reservoir, the bidirectional long short-term memory network (Bi-LSTM) fluid identification model based on logging sequence information was proposed. From the analysis of the differences of logging response characteristics and the similarity analysis of characteristic curves, the sensitivity curve was determined. Combined with the input requirements of the Bi-LSTM, the fluid identification sample database was established, and the fluid identification model was obtained. This method was applied to identify Majiagou Formation in Ordos Basin. Compared with the predictions of the LSTM model and other three types of machine learning algorithms, the accuracy of fluid identification increased from 82.7% to 91.5%. The model can not only make full use of the logging response value of the corresponding depth, but also take into account the changing trend and correlation of the logging curve with depth in ways that avoid the influence of vertical heterogeneity of reservoir and therefore improve fluid identification ability.
Key words:  fluid identification  bidirectional long short-term memory network  carbonate rock  logging response sequences
版权所有 中国石油大学学报(自然科学版)编辑部 Copyright©2008 All Rights Reserved
主管单位:中华人民共和国教育部 主办单位:中国石油大学(华东)
地址: 青岛市黄岛区长江西路66号中国石油大学期刊社 邮编:266580 电话:0532-86983553 E-mail: journal@upc.edu.cn
本系统由:北京勤云科技发展有限公司设计