摘要
针对复杂岩性碳酸盐岩储层原有岩性识别方法精度较低、泛化能力不足、结果不稳定等问题,提出基于粗糙集-随机森林算法的复杂岩性识别方法。利用邻域粗糙集的属性约简选取岩性敏感曲线,在不影响岩性识别基础上将不必要曲线删除,能有效去除冗余信息;其次将筛选出的曲线作为随机森林模型输入,建立粗糙集-随机森林算法的岩性识别模型。通过对某区块502块岩心数据处理,该模型岩性判别率稳定到88.3%,比Fisher判别、Bayes判别等方法精度高,且实现简单,有较强泛化能力。该方法可作为复杂岩性储层岩性识别方法,为复杂岩性储层的勘探开发提供帮助。
In view of the problems of the low precision of the original identifying methods,insufficient generalized ability,unstable results and so on for the complex-lithology carbonate reservoir,the identifying method for the complex lithology was proposed based on rough set-random forest algorism. With the help of the attributes of the neighborhood rough set,the lithology sensitive curves were reduced and chosen,on the basis of the non-effects on the lithology identification,the unnecessary curves were cut out to effectively delete the redundant information; and then the chosen curves were input into the random forest model and the lithology identifying model for the rough set-random forest algorism was established. Through the core data processing of 502 samples in a block,the identified ratio of the lithology for the model has been stabilized to 88. 3% i. e. is significantly higher than other discriminating methods（ Fisher and Bayes） and furthermore the task is easy to realized and the model possesses much stronger ca-pacity of the generalization. The method can be regarded as the lithology identifying approach for the complex-lithology reservoir and has provided the help for the exploration and development of this kind of the reservoir.
出处
《大庆石油地质与开发》
CAS
CSCD
北大核心
2017年第6期127-133,共7页
Petroleum Geology & Oilfield Development in Daqing
基金
国家自然科学基金项目"页岩油储层岩石物理特性数值模拟研究"(41504094)
"致密气储层岩石导电机理研究及饱和度评价"(41404084)
"十三五"国家科技重大专项"复杂碳酸盐岩储层测井评价关键技术研究与应用"(2017ZX05032003-005)
关键词
复杂岩性储层
碳酸盐岩
岩性识别
邻域粗糙集
随机森林
complex-lithology reservoir
carbonate rock
lithologic identification
neighborhood rough set
random forest
