摘要
Abstract
The reservoir of Block Sudong41-33 in Sulige Gas Field has the characteristics of low porosity,low permeability and high heterogeneity.The carbonate reservoir in Lower Ordovician Majiagou Formation were subjected to multi-stage multi-type construction,sedimentary and other effects,which makes the lithology complex and diverse,and thus the accurate identification of lithology has become a difficult problem of development in this area.In recent years,more and more at tention has been focused on the use of decision tree method in machine learning in the field of geoscience,especially in lithology prediction.Based on the data of well logging and the analysis of lithological parameters,six kinds of well logging parameters that are sensitive to lithology were selected,which includes acoustic time difference (AC),natural gamma ray (GR),photoelectric absorption cross section index (PE),density(DEN),deep lateral resistivity (RLLD)and compensated neutron(CNL).Through the analysis of the six well logging parameters,a multi classifier was constructed based on decision tree method,and the information of lithology and rock characteristics were fused.Compared with the lithologic data of well logging,the recognition accuracy is over 80％.When compared with the Naive Bayesian,the accuracy of lithology recognition is improved by 13％.关键词
复杂碳酸盐岩/测井岩性识别/决策树方法/机器学习方法/苏里格气田Key words
complex carbonate rock/lithology identification of well logging/decision tree method/machine learning method/Sulige Gas Field分类
能源科技
