测井技术2023，Vol.47Issue(6)：662-670,9.DOI:10.16489/j.issn.1004-1338.2023.06.003
基于CNN-GRU的复杂岩性识别方法研究与应用
Research and Application of Complex Lithology Identification Method Based on CNN-GRU
摘要
Abstract
In the exploration and development of oil and gas reservoirs,lithology identification is an important component of reservoir log evaluation.The Hailar basin is characterized by proximal sedimentation and transport deposition,with significant differences in rock composition and structure.The types of lithology are complex and diverse,mainly including fine sandy mudstone,siltstone,shale,tuffaceous sandstone,tuffaceous mudstone,oil-bearing coarse sandstone,oil-bearing fine sandstone,crystal detrital tuff conglomerate,sandstone conglomerate,and dense tuff conglomerate.Traditional identification methods have low accuracy in dealing with complex lithologies,which severely restricts the accuracy of reservoir logging interpretation.This study integrates convolutional neural networks with gated recurrent units(CNN-GRU)and selects six logging parameters,including sonic time difference,natural potential,natural gamma,density,and shallow and deep lateral resistivity,to train sample wells in the Hailar basin.A CNN-GRU model for identifying complex lithologies is constructed.The research results show that the average accuracy of the CNN-GRU model reaches 92.3%,with an improvement of 5.5%～10.0%compared to a single network.After applying this model to well A in the Hailar basin,the lithology identification conformity rate reaches 94.8%,which provides a reliable lithological basis for the accuracy of reservoir log interpretation.关键词
复杂岩性识别/海拉尔盆地/卷积神经网络/门控循环单元/长短时记忆神经网络Key words
complex lithology identification/Hailar basin/convolutional neural network/gated recurrent unit/long short-term memory neural network引用本文复制引用
张晓峰,庞春阳,胡锐,朱云峰,李红星..基于CNN-GRU的复杂岩性识别方法研究与应用[J].测井技术,2023,47(6):662-670,9.基金项目
核资源与环境国家重点实验室开放基金项目"致密层系井震结合计算三维TOC实现油铀兼探方法研究"(2020NRE27) （2020NRE27）
