基于核Fisher判别的复杂储层岩性识别

发布时间：2018-01-27 12:03

本文关键词： 复杂储层岩性识别核Fisher判别分级核Fisher　出处：《科学技术与工程》2017年31期 　论文类型：期刊论文

【摘要】：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。
[Abstract]:Many kinds of lithology can be used as reservoir in complex reservoir. The physical characteristics of different lithology are quite different. Lithologic interpretation of complex reservoir is a more effective method to obtain accurate physical properties, but the logging characteristics of different lithology are similar. The recognition effect of conventional linear classification method is not satisfactory, because the nonlinear classification characteristics of complex reservoir rock is a large proportion. In order to solve this problem. In this paper, we generalize Fisher discriminant analysis to form kernel Fisher discriminant analysis, and make further use of the nonlinear information not extracted from Fisher discriminant. More nonlinear classification features are obtained by increasing the dimension, and then the features of lithologic classification are extracted by reducing the dimension. In this paper, the kernel Fisher is preprocessed by experiments. The selection of key parameters is introduced in detail, and the core Fisher method is compared with other classification methods to verify the lithologic recognition ability of the kernel Fisher method. For the difference and similarity relationship between different lithology, resulting in low lithology recognition accuracy, the idea of classified kernel Fisher discriminant analysis is put forward. It is proved that the classification kernel Fisher discriminant analysis can further improve the accuracy of lithology recognition.
【作者单位】：中国石油化工股份有限公司石油勘探开发研究院;
【分类号】：P618.13;P631.81
【正文快照】： 引用格式:王鹏,胡向阳,魏水建.基于核Fisher判别的复杂储层岩性识别[J].科学技术与工程,2017,17(31):29—34Wang Peng,Hu Xiangyang,Wei Shuijian.Lithologic identification ofcomplex reservoirs based on kernel Fisher discriminant analysis[J].Sci-ence Technology and

【相似文献】