基于支持向量机的辫状河测井沉积微相识别

辫状河沉积环境下的单井沉积微相解释工作易受其繁琐性与主观性影响。提出一种基于测井曲线针对辫状河沉积环境的沉积微相自动识别方法,用于单井沉积微相解释,为后续储层精细描述以及地质建模提供依据。以工区内测井曲线为基础,通过对地层单元内的测井数据进行分析,得到统计参数作为支持向量机训练的输入参数,输出对应的沉积微相解释结果。使用鄂尔多斯盆地某区块1 944组沉积微相样本数据对支持向量机进行训练,验证集由另外648组样本数据组成。2组数据集被选用于测试训练后的支持向量机在沉积微相自动识别方面的应用效果,其中第1组由648组样本数据组成,来自与训练集验证集同一区域,用于测试支持向量机对沉积微相识别效果;另外一组数据集由816组来自于不同区域的样本数据组成,用于测试方法的泛化性。结果表明该方法对2组测试集分类的正确率分别能够达到95.4%和93.1%。该方法在单井沉积微相自动识别方面具有足够的准确性与可靠性。

Classification of Borehole Braided River Sedimentary Microfacies Based on Support Vector Machine

The complexity and exhaustive feature of sedimentary microfacies interpretation work is effected a lot by subjectivity . An automatic classification approach is proposed for a well sedimentary microfacies interpretation supporting reservoir characterization and providing basis for geology modeling .First ,the well log interval measured data should be divided into different units based on data analysis and threshold segmentation technology . T hen , a support vector machine will be trained to build the relationship between features extracting from log data and the different sedimentary facies .1 944 samples of region A from Ordos basin are chosen to train the support vector machine ,and other 648 samples consist the verification set to generalize training process .At last ,to test the stability and accuracy of support vector machine under training ,648 samples from region A and 816 samples from another region B consisting two test sets are chosen for this aim .The result shows these two test sets from different regions could get 95.4% and 93.1% of accuracy respectively .So ,the availability and reliability of this approach are proven to be effective .