石油地球物理勘探2025，Vol.60Issue(5)：1111-1123,13.DOI:10.13810/j.cnki.issn.1000-7210.20240347
基于双视图GraphSAGE的测井识别方法
Logging identification of thermal reservoirs in Southern Songliao Basin based on dual-view GraphSAGE
摘要
Abstract
With the depletion of conventional oil and gas resources and the increase of water content in oilfields in the east of China,geothermal energy development has become the key to the green and low-carbon transfor-mation of old oilfields,and thermal reservoir identification is the core of geothermal field research.The existing thermal reservoir identification algorithms fail to employ the hidden sample relationships between logging data as inputs for conducting training and tests,and a single view is insufficient for the extraction of depth sequence information and spatial features embedded in it.To this end,a thermal reservoir logging identification method based on dual-view GraphSAGE(dv-GraphSAGE)is proposed.Firstly,the depth distance map and feature similarity map are constructed by depth sequence and feature similarity,and then features are extracted by adopting GraphSAGE and the feature self-attention mechanism(FSAtt)to retain the information richness and complex associations of the views.Finally,the view features are fused by an adaptive feature fusion module and fed into a multilayer perceptron(MLP)network to achieve thermal reservoir identification.The experimental results of logging data from 30 geothermal wells show that the overall identification accuracy of the dv-Graph-SAGE model for the mudstone layer,dense layer,dry layer,oil layer,and water layer reaches 95.4%,of which the identification rate for the water layer is 96.9%.The experimental comparison results also indicate that dv-GraphSAGE has a better thermal reservoir identification effect,which provides a new idea for geothermal development of oilfields.关键词
地热/热储层识别/深度学习/双视图GraphSAGE/松辽盆地Key words
geothermal/thermal reservoir identification/deep learning/dv-GraphSAGE/Songliao Basin分类
天文与地球科学引用本文复制引用
田枫,唐莎莎,刘芳,刘宗堡,张庆斌,赵德利..基于双视图GraphSAGE的测井识别方法[J].石油地球物理勘探,2025,60(5):1111-1123,13.基金项目
本项研究受国家自然科学基金项目"基于储层构型理论解析的三角洲相致密砂岩储集性多级次表征研究"(42172161)、黑龙江省哲学社会科学研究规划项目"基于多维度学习者模型的网络学习资源个性化推荐方法研究"(22EDE389)和东北石油大学特色领域团队专项项目"领域软件工程与石油大数据团队"(2022TSTD-03)联合资助. （42172161）
