基于神经网络的聚类分析在储层流动单元划分中的应用  

THE APPLICATION OF CLUSTER ANALYSIS BASED ON NEURAL NETWORK METHODS IN IDENTIFICATION RESERVOIR FLOW UNIT

作　　者：孙致学[1] 姚军[1] 孙治雷[2,3] 卢涛[4] 唐乐平[4] 杨勇[4] 韩继超[1] 

机构地区：[1]中国石油大学石油工程学院,山东青岛266555 [2]国土资源部海洋油气资源和环境地质重点实验室,山东青岛266071 [3]青岛海洋地质研究所,山东青岛266071 [4]长庆油田分公司勘探开发研究院,陕西西安710021

出　　处：《物探与化探》2011年第3期349-353,共5页Geophysical and Geochemical Exploration

基　　金：国家重大科技专项"鄂尔多斯盆地大型岩性地层油气藏勘探开发示范工程"(2008ZX050443-16)

摘　　要：以苏里格气田盒8段为研究对象,在细分层和精细沉积学研究基础上,通过对关键井详细研究,以流动层段指标为储层流动单元划分标准,将目的层分为3类流动单元。通过相关性分析结合专家经验,从诸多电性、物性、岩性等参数中优选出表征流动单元的10个特征变量作为预测模型的输入,应用基于神经网络算法的聚类分析方法建立储层流动单元非线性识别模型。通过对其他关键井的回判预测表明,建立的流动单元预测模型可以更全面地考虑各类地质因素与流动单元之间的结构性复杂映射关系,气井产能与流动单元具有较高的对应关系,为气田精细描述与开发井网的优化部署提供可靠基础。Identification and evaluation of the reservoir flow units is an important aspect of geological research of development mature oilfield.In this paper,take Sulige Gasfield He8 Group as an example,based on the sub-layer and detailed sediment study determined the characteristics of 10 variables to identify and rank reservoir flow units.Based on neural network algorithm of cluster analysis method to identify reservoir flow units of non-linear model.The established the neural network model can be more comprehensive consideration of various geological factors and the flow units.Using the established model,we can identify flow units for the well without boring,which can better solve the mature oilfield reservoir description abundance of information and accuracy requirements.Quarry applications indicates that in view of Chinese oil reservoir＇s main causes of continental origin,carry out reservoir flow units quantitative evaluation to matured oilfield have greater practical value.

关 键 词：流动单元 神经网络 聚类分析 盒8段 

分 类 号：P631.4[天文地球—地质矿产勘探] TE111.3[天文地球—地质学]