基于K-均值聚类与贝叶斯判别的储层定量评价——以大安油田泉四段储层为例
闫 明,张云峰,李易霖

东北石油大学地球科学学院,黑龙江大庆 163318

储层评价; K-均值聚类; 贝叶斯判别; 低渗储层; 分级评价; 数学地质; 泉四段储层; 大安油田

Quantitative reservoir evaluation based on K-means cluster analysis and Bayes discriminant analysis: a case study on reservoir in the 4th member of Quantou Formation in Daan oilfield
Yan Ming, Zhang Yunfeng, and Li Yilin

College of Earth Sciences, Northeast Petroleum University, Daqing 163318, Heilongjiang Province, P.R.China

reservoir evaluation; K-means cluster; Bayes discriminant; low permeability reservoir; classification evaluation; mathematical geology; the 4th member of Quantou Formation; Daan oilfield

DOI: 10.3724/SP.J.1249.2016.02211

备注

以大安油田泉四段砂岩储层为研究对象,运用K-均值聚类分析贝叶斯判别确定孔隙度、渗透率、最大孔隙半径、平均孔隙半径、分选系数、最大汞饱和度和排驱压力7个特征参数,建立特征参数与储层类别的定量判别评价,并以此标准进行其他储层样品的判别分类.结果表明,研究区储层共分为3类,Ⅰ类到Ⅲ类储层物性逐渐变差,孔隙半径、分选系数逐渐减小,排驱压力逐渐增加,且利用贝叶斯判别可以快速判定储层样品类别; 与Q型聚类分析和判别函数法、层次分析法对比发现,3种方法分类、判别结果相近,说明运用K-均值聚类分析和贝叶斯判别分析进行储层分类评价不仅有效,而且具有主观影响小和定量化程度高等特点,对于低渗透油藏勘探开发具有指导意义.

Based on the reservoir evaluation in the 4th member of the Quantou Formation in the Daan oilfield, we choose the porosity, the permeability, the maximal pore radium, the average pore radium, the sorting coefficient, the maximal mercury saturation and the displacement pressure as characteristic parameters for classification, and the K-means cluster analysis as the method to evaluate and classify reservoirs. We apply the Bayes discriminant analysis to establish the quantitative relationship between the characteristic parameters and the type of reservoir. The results indicate that the reservoir in the area can be divided into three types. The physical property gradually worsens, the pore radium and the sorting coefficient decrease, and the displacement pressure increases from type I to type III. It is a fast way to classify the reservoirs by using Bayes discriminant analysis, which gives a similar classification result to Q type cluster analysis and analytic hierarchy process. It illustrates that it is effective to evaluate and classify reservoirs by integrating the K-means cluster analysis and the Bayes discriminant analysis. The proposed method possesses a smaller error of subjective judgment and a stronger degree of quantitative characterization, and it provides significant guidance for the exploration and development of low permeability reservoirs.

·