高精度流动单元分类方法及应用  

High-precision classification method and application of flow unit

作　　者：王猛[1] 董宇 张志强[1] 刘志杰 刘海波[1] 周悦 WANG Meng;DONG Yu;ZHANG Zhiqiang;LIU Zhijie;LIU Haibo;ZHOU Yue(China Oilfield Services Limited,Langfang 065201,China)

机构地区：[1]中海油田服务股份有限公司,河北廊坊065201

出　　处：《断块油气田》2022年第1期89-94,共6页Fault-Block Oil & Gas Field

基　　金：国家科技重大专项课题“超低渗地层测试技术与装备”(2017ZX05019-004)。

摘　　要：西湖凹陷砂岩储层具有孔隙结构差、孔隙类型多样的特点,采用传统渗透率模型对其计算的结果,无法满足储层评价的精度要求。在岩心资料综合分析的基础上,根据流动单元分类建立的渗透率模型,虽然计算结果精度高,但利用测井参数计算的流动单元指数误差较大,不能准确划分流动单元类型。而基于机器学习算法的流动单元分类方法均为数据驱动,由于取心作业一般具有针对性,故某一类岩心数量会远远多于其他类,利用该算法预测时,其结果就会偏向数量多的类。针对传统流动单元划分方法存在的问题,文中提出了一种基于集成神经网络的流动单元分类方法。实际应用结果表明,利用集成神经网络分类模型划分的储层流动单元类型与实际岩心的流动单元类型吻合度高,具有良好的推广应用前景。The sandstone reservoir in Xihu Depression has the characteristics of poor pore structure and diverse pore types,the calculation results of traditional permeability models cannot meet the accuracy requirement of reservoir evaluation.Based on the comprehensive analysis of core data,the permeability model established by the classification of flow units has high calculation accuracy,but the error of the flow unit index calculated by logging parameters is large,and the type of flow units cannot be accurately classified.The flow unit classification methods based on machine learning algorithm are all data-driven.Because the coring operations is generally targeted,one type of cores will be much more than other types.The prediction results of machine learning model will be biased towards the types with larger number.Aiming at the problems of traditional flow unit classification methods,a flow unit classification method based on neural network ensemble learning is proposed.The practical application results show that the types of reservoir flow units classified by the neural network ensemble learning model is highly consistent with the actual core flow unit type,and the application prospects are good.

关 键 词：渗透率模型 流动单元分类 集成神经网络 西湖凹陷 

分 类 号：TE32[石油与天然气工程—油气田开发工程]