计算机技术与发展
計算機技術與髮展
계산궤기술여발전
COMPUTER TECHNOLOGY AND DEVELOPMENT
2015年
8期
175-178,183
,共5页
尚福华%姜萌%马楠%解红涛
尚福華%薑萌%馬楠%解紅濤
상복화%강맹%마남%해홍도
聚类分析%马氏距离%油田分类%权重
聚類分析%馬氏距離%油田分類%權重
취류분석%마씨거리%유전분류%권중
cluster analysis%Mahalanobis distance%oil classification%weight
针对大庆低渗透油田开发效果变差、开发储量品位降低、优选区块技术难度增大,导致原有的评价方法、优选技术、开采方式难以满足低渗透油田发展需要的问题,提出了应用聚类分析的方法对低渗透油田储层进行分类。首先简述聚类分析以及马氏距离的基本原理,并根据油田不同储层参数对储层采出贡献程度不同的特点,以及原有马氏距离在估计协方差矩阵方面难度大导致计算效率低的问题,提出了马氏距离聚类过程中估计协方差矩阵的迭代法。该方法充分考虑到了变量权重和样本类别的影响,对协方差矩阵的估计进行改进,能够在一定程度上减少权值确定上的主观因素并提高计算效率。以油田开发实际数据为例进行实验分类,结果表明该方法是有效可行的。
針對大慶低滲透油田開髮效果變差、開髮儲量品位降低、優選區塊技術難度增大,導緻原有的評價方法、優選技術、開採方式難以滿足低滲透油田髮展需要的問題,提齣瞭應用聚類分析的方法對低滲透油田儲層進行分類。首先簡述聚類分析以及馬氏距離的基本原理,併根據油田不同儲層參數對儲層採齣貢獻程度不同的特點,以及原有馬氏距離在估計協方差矩陣方麵難度大導緻計算效率低的問題,提齣瞭馬氏距離聚類過程中估計協方差矩陣的迭代法。該方法充分攷慮到瞭變量權重和樣本類彆的影響,對協方差矩陣的估計進行改進,能夠在一定程度上減少權值確定上的主觀因素併提高計算效率。以油田開髮實際數據為例進行實驗分類,結果錶明該方法是有效可行的。
침대대경저삼투유전개발효과변차、개발저량품위강저、우선구괴기술난도증대,도치원유적평개방법、우선기술、개채방식난이만족저삼투유전발전수요적문제,제출료응용취류분석적방법대저삼투유전저층진행분류。수선간술취류분석이급마씨거리적기본원리,병근거유전불동저층삼수대저층채출공헌정도불동적특점,이급원유마씨거리재고계협방차구진방면난도대도치계산효솔저적문제,제출료마씨거리취류과정중고계협방차구진적질대법。해방법충분고필도료변량권중화양본유별적영향,대협방차구진적고계진행개진,능구재일정정도상감소권치학정상적주관인소병제고계산효솔。이유전개발실제수거위례진행실험분류,결과표명해방법시유효가행적。
Aiming at the problem that original evaluation method,optimal technology and exploring way can’ t meet the development need in Daqing oilfield with low permeability,which caused by worse development effect,lower development grade,and the technology of pre-fer block is more hard,propose that applying the clustering analysis to classify the oil reservoir with low permeability. First,describe the basic principle of cluster analysis and Mahalanobis distance briefly,according to the characteristics that the oil of different reservoir param-eters have different degree of contribution on reservoir recovery and the problem of the original Mahalanobis distance has low calculation efficiency in evaluating covariance matrix,present an iteration method to estimate the covariance matrix of Mahalanobis distance during the cluster analysis. The weights of variables and categories of samples are taken into account,estimation of the covariance matrix is im-proved,the method can reduce the subjective factors on determining weights and improve the computational efficiency in a certain extent. With actual oilfield development data as an example for classification,the experimental results demonstrate the effectiveness of the meth-od.