基于分布式光纤声波监测数据和机器学习的井筒流体类型识别方法  

Identification of fluid type in wellbore based on distributed acoustic sensing data and machine learning

作　　者：刘均荣[1,2] 韩艳慧 王哲 刘明 LIU Junrong;HAN Yanhui;WANG Zhe;LIU Ming(School of Petroleum Engineering in China University of Petroleum(East China),Qingdao 266580,China;National Engineering Research Center of Oil&Gas Drilling and Completion Technology,Qingdao 266580,China;Research Institute of Petroleum Engineering Technology,SINOPEC Shengli Oilfield Company,Dongying 257000,China)

机构地区：[1]中国石油大学(华东)石油工程学院,山东青岛266580 [2]油气钻完井技术国家工程研究中心,山东青岛266580 [3]中国石油化工股份有限公司胜利油田分公司石油工程技术研究院,山东东营257000

出　　处：《中国石油大学学报（自然科学版）》2023年第3期107-114,共8页Journal of China University of Petroleum(Edition of Natural Science)

基　　金：国家自然科学基金项目(52174048);中国石油化工股份有限公司重点科技项目(P21052)。

摘　　要：基于建立的实尺寸水平井筒流体流动光纤监测模拟试验系统,开展油气水流动模拟试验,获得不同流体流动情况下的分布式光纤声波监测(DAS)数据;利用小波时间散射变换和短时傅里叶变换对DAS数据进行处理,提取流体流动特征,建立基于机器学习的井筒流体类型识别分类方法。结果表明:融合低方差散射特征和短时时频特征数据能提高流体类型识别的准确率,并且随机森林算法的识别结果优于BP神经网络和决策树;该方法为井筒多相流体流动识别提供了一种新的技术手段。In this study,a large number of oil,gas and water flow experiments were carried out under different fluid flow conditions using a full-sized horizontal wellbore simulation set-up with a distributed acoustic sensing(DAS)system.The wavelet time scattering transform and short-time Fourier transform were used to process the DAS data,and the fluid flow characteristics were extracted.Then,an novel identification and classification method of wellbore fluid type was established with machine learning.The results show that the accuracy of fluid type identification can be improved by combining low variance scattering featured data and short time frequency featured data,and the identification results by the random forest algorithm are better than that of the BP neural network and decision tree.The results of this research can provide a new means for wellbore multiphase fluid flow identification.

关 键 词：分布式光纤声波监测 声纹数据 特征提取 机器学习 流体类型 分类识别 

分 类 号：TE34[石油与天然气工程—油气田开发工程]