摘要
为了解决基于测井数据对油气水层的实时识别这一技术难题,利用计算机科学与现代数学,结合随钻测井技术与机器学习算法进行油气水层的随钻识别。首先,对训练集数据进行相关性分析,剔除弱相关或冗余数据;其次,选择一对多支持向量机、一对一支持向量机以及随机森林算法分别建立油气水层分类识别模型,并使用网格搜索方法及10折交叉验证法对3种分类识别模型参数进行优选;最后,运用参数优选后的各分类识别模型,对随钻测井数据进行油气水层的识别。研究结果表明,3种分类识别模型对研究区块油气水层随钻识别的准确率均达到75%以上。在训练样本较少的情况下,优先选用一对一支持向量机分类识别模型进行油气水层的随钻识别。
In order to solve the technical difficulty of real-time identification of oil/gas and water layers based on logging data,based on computer science and modern mathematics,logging while drilling technology is combined with machine learning algorithm to identify the oil/gas and water layer while drilling.Firstly,the correlation analysis is performed on the data in training set to eliminate weak correlation or redundant data.Secondly,oil/gas and water layers recognition models are established using one-versus-rest support vectormachine,one-versus-one support vector machine,and random forest algorithm respectively,and the parameters of three recognition models are optimized using grid search method and 10-fold crossvalidation method.Finally,based on logging data while-drilling,the oil/gas and water layers are identified using the models whose parameters are optimized.The results show that the recognition accuracy of three models in the study area is higher than 75%.In the case of less training samples,the recognition model established using one-versus-one support vector machine has the best effect to the identification of oil/gas and water layers while drilling.
作者
孙健
李琪
陈明强
任龙
SUN Jian;LI Qi;CHEN Mingqiang;REN Long(College of Petroleum Engineering,China University of Petroleum (Beijing),Beijing 102249,China;College of Petroleum Engineering,Xi'an Shiyou University,Xi'an 710065,Shaanxi,China;Key Laboratory of Shaanxi Province for Oil and Gas Well and Reservoir Seepage and Rock Mechanics,Xi'an Shiyou University,Xi'an 710065,Shaanxi,China)
出处
《西安石油大学学报（自然科学版）》
CAS
北大核心
2019年第5期79-85,90,共8页
Journal of Xi’an Shiyou University（Natural Science Edition）
基金
国家自然科学基金青年科学基金项目“致密油体积压裂缝网形成及多重介质流固全耦合流动模拟”(51704235)
陕西省高校科协青年人才托举计划“非常规储层体积压裂缝网形成机制及扩展模拟”(20180417)
关键词
随钻油气水层识别
机器学习
支持向量机
随机森林算法
模型优选
identification of oil/gas and water layers while drilling
machine learning
support vector machine
random forest algorithm
optimal selection of models
作者简介
孙健(1990-),男,博士研究生,研究方向:储层地质建模、油藏数值模拟。E-mail:xjkelsj@163.com.
