基于储层分类的支持向量机渗透率预测

Permeability Prediction Using Support Vector Machine Based on Reservoir Classification

研究区孔隙度、渗透率分布范围广、非均质性强,在进行储层渗透率求取时存在较大误差。根据取心物性资料、测井资料,选用流动层带指标I_(FZ)划分方法将取心井储层划分Ⅰ、Ⅱ、Ⅲ、Ⅳ类,并建立每类储层的储层类型预测模型。根据与渗透率有关的测井属性变量,利用支持向量机技术对每类储层进行训练学习,分别建立各类储层的渗透率预测模型。对研究区取心井测试样本渗透率进行预测,与常规的统计方法以及分类前的支持向量机预测模型相比,分类后的模型预测精度有了明显提高,为研究区的储层评价提供了一种有效的研究途径。

The porosity and permeability widely distribute and heterogeneity is strong in the studied area, so, there is a big error of permeability calculation. Combining with core data and log data, the coring well reservoirs are divided into four types by flow zone index （IFz）, moreover, prediction model of every reservoir is established. Based on this, the support vector machine technology is used to learn and train every type of reservoir samples from coring well. The permeability prediction model of every reservoir is established by the technology. Permeability of another coring well is calculated with the models which have been established. Permeability prediction accuracy is obviously improved compared with conventional statistical method and prediction model of support vector machine before classification, which provides an Key words：