基于核Fisher判别的复杂储层岩性识别研究

复杂储层中多种岩性均可作为储层,不同岩性的物性特征差异较大,分岩性解释复杂储层物性是求准物性较为有效的一种方法,但是不同岩性的测井特征相近,常规线性分类方法识别效果不理想,因为复杂储层的岩石识别中非线性分类特征占较大比例。针对这一问题,本文将Fisher判别分析(FDA)做核推广,形成核Fisher判别分析(KFDA),进一步利用Fisher判别中未提取的非线性信息,通过升维获得更多的非线性分类特征,然后再通过降维来提取利于岩性分类的特征。文章通过实验对核Fisher在数据预处理、关键参数的选取等方面进行了详细介绍,并将核Fisher方法与其它分类方法进行比较,验证了核Fisher方法的岩性识别能力,而对于不同岩性间的差异相似关系,造成岩性识别精度低的情况,提出了分级核Fisher判别分析的思路,研究证明利用分级核Fisher判别分析的思路可进一步提高岩性的识别精度。

Lithologic identification of complex reservoirs based on kernel Fisher discriminant analysis

Many kinds of lithology can be used as reservoirs in complex reservoirs, The physical characteristics of different lithologies are different, It is an effective method to explain the physical property of complex reservoirs by lithology, But the logging characteristics of different lithologies are similar, The conventional linear classification method is not ideal, because the nonlinear classification characteristics of the complex reservoir rock recognition. To solve this problem, the Fisher discriminant analysis (FDA) for kernel promotion, discriminant analysis, kernel Fisher (KFDA) by using the nonlinear information without further extraction of Fisher discriminant, nonlinear classification features more by ascended, and then by dimension reduction of feature extraction for rock classification. Through the experiment of kernel Fisher in data preprocessing, the key parameters are introduced in detail, and the kernel Fisher method is compared with other classification methods, to verify the ability of kernel Fisher lithology identification method, and the differences between different lithological similarity caused by lithology identification problem of low accuracy, is proposed classification of kernel Fisher discriminant analysis method, discriminant analysis method research that can further improve the recognition accuracy by using lithologic classification of kernel Fisher.