首页 | 本学科首页   官方微博 | 高级检索  
     

储层粒度神经网络预测模型研究
引用本文:王利华,楼一珊,马晓勇,程福山,陈宇. 储层粒度神经网络预测模型研究[J]. 西南石油大学学报(自然科学版), 2016, 38(1): 53. DOI: 10.11885/j.issn.1674-5086.2014.01.09.02
作者姓名:王利华  楼一珊  马晓勇  程福山  陈宇
作者单位:1. 油气资源与勘探技术教育部重点实验室·长江大学,湖北荆州4340232. 中国石化胜利石油工程有限公司,山东东营2570643. 北京锦辉博泰科技有限公司,北京昌平102249
基金项目:国家科技重大专项(2008ZX05056 002 03;2008ZX05024 003 01)
摘    要:国内外多年的研究表明,储层粒度特征值(d50)、非均质系数(d40=d90)是防砂设计的基础。常规获取粒度分布范围的方法主要有激光粒度测试法(LDA)与筛析法(SA),两种方法均需要通过岩芯粒度测试来获取数据,而在制定开发井的完井防砂措施时往往没有实际开采层位的岩芯,只能参照探井粒度数据进行设计,从而导致较大的误差。针对该问题,从测井的角度出发,开展了储层粒度与多种测井曲线的响应关系的研究,采用神经网络技术,建立了探井伽马、密度测井项与实测粒度特征值三者样本库,训练出满足工程需要的学习网络,进而结合开发井测井资料,获得了整个粒度纵向分布剖面,为防砂分层设计提供准确的基础数据支撑。目前,该方法在中国海上多个油田的分层防砂优化设计中获得了成功应用,预测误差可控制在10% 以内。

关 键 词:分层防砂  粒度特征值  神经网络  伽马密度测井  样本库  

Research on Neural Network Prediction Model of Reservoir Particle Size
WANG Lihua,LOU Yishan,MA Xiaoyong,CHENG Fushan,CHEN Yu. Research on Neural Network Prediction Model of Reservoir Particle Size[J]. Journal of Southwest Petroleum University(Seience & Technology Edition), 2016, 38(1): 53. DOI: 10.11885/j.issn.1674-5086.2014.01.09.02
Authors:WANG Lihua  LOU Yishan  MA Xiaoyong  CHENG Fushan  CHEN Yu
Affiliation:1. MOE key Laboratory of Exploration Technologies for Oil and Gas Resources,Yangtze University,Jingzhou,Hubei 434023,China;2. Shengli Oilfield Service Corporation,SINOPEC,Dongying,Shandong 257064,China;3. Jinhuibotai-Tech. CO. Ltd.,Changping,Beijing 102249,China
Abstract:According to researches at home and abroad,sand control design is based on reservoir particle size characteristicvalue. LDA and SA are the conventional methods used to analyze particle size distributions. Both methods requires data throughcore particle size testing. But sand control design can only use test well data,because no core at actual producing position canbe used when sand control measure is established,which can result in major errors. This article elaborates the relevance aboutmedian grain size and gamma ray logging or density logging through researches on reservoir particle size and variety of logcurve response relation. And then through establishing sample pool of gamma ray logging or density logging and characteristicvalue,and by neural network technology we trained learning network satisfing engineering requirements. Then the particlesize longitudinal distribution profile can be established according to development well logging data. This profile supplied databasis for sand control layering design. At present,this method has been successfully used in several Chinese offshore oil fieldin sand control optimization with errors below 10%.
Keywords:layered sand control  characteristic value of particle size  neural network  gamma and density logging  sample
pool
  
本文献已被 CNKI 万方数据 等数据库收录!
点击此处可从《西南石油大学学报(自然科学版)》浏览原始摘要信息
点击此处可从《西南石油大学学报(自然科学版)》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号