成都理工大学学报（自然科学版）2025，Vol.52Issue(5)：931-950,20.DOI:10.12474/cdlgzrkx.2025012401
多源数据与地质逻辑融合的深层页岩储层构造裂缝测井智能识别
Intelligent logging identification of structural fractures in deep shale reservoirs based on the fusion of multisource data and geological logic
摘要
Abstract
Structural fractures in deep shale reservoirs exert obvious controlling effects on shale gas enrichment and gas production.Their accurate identification holds important guiding significance for the development,and subsequent exploration,of shale gas.In this paper,based on drilling,logging,well-logging,and production data of shale gas in the Longmaxi Formation in southeastern Sichuan Basin,the response characteristics of multisource data to natural fractures are identified via a combination of machine learning and fracture response information.An intelligent identification model of structural fractures,including their developmental characteristics,is constructed,and a quantitative well-logging evaluation method for assessing the effectiveness of structural fracture networks was developed.The results show that four types of structural fractures have developed in the shale of the Longmaxi Formation,which are dominated by high-angle,vertical,tensile fractures and shear fractures.When the core fracture density exceeds 12 pieces per meter in organic-rich shale and eight pieces per meter in organic-bearing shale,the drilling time and gas logging responses become more pronounced.Cross-plot analysis and Pearson correlation coefficient analysis indicated that the acoustic transit time,neutron porosity,deep-to-shallow laterolog resistivity ratio,compressional waves,shear waves,and Stoneley waves exhibit high sensitivity to the density and dip of structural fractures in the shale reservoir.By optimizing a tabular data,deep learning architecture model called TabNet,we achieved identification accuracies exceeding 93%for identifying the fracture scale and 80%for fracture occurrence within organic-rich shale.Compared with other approaches,the model presented here demonstrates the best performance in identifying high-angle,vertical tensile and shear fractures within the lower part of the first member of the Longmaxi Formation.Considering the fractures'geometric parameters and petrophysical properties,a quantitative evaluation index for fracture network effectiveness was established.It is considered that the better the effectiveness of the fracture network,the higher the enrichment degree of shale gas,and the easier it is to form complex fracture networks after transformation.These results can further promote the fine evaluation of deep shale reservoir deposits and the optimization design of fracturing transformation.关键词
构造裂缝/测井识别/智能算法/深层页岩储层/龙马溪组/川东南Key words
structural fractures/logging identification/intelligent algorithm/deep shale reservoir/Longmaxi Formation/southeastern Sichuan Basin分类
能源科技引用本文复制引用
徐碧兰,王濡岳,何建华,吴炎峰,李丹,曹峰,蒋睿,李勇,李可赛,邓虎成..多源数据与地质逻辑融合的深层页岩储层构造裂缝测井智能识别[J].成都理工大学学报（自然科学版）,2025,52(5):931-950,20.基金项目
国家自然科学基金青年基金项目(42402148) （42402148）
页岩油气富集机理与高效开发全国重点实验室开发基金项目(ZC0699) （ZC0699）
宁夏回族自治区非常规天然气勘查开发创新团队科研项目(2022BSB03105) （2022BSB03105）
宁夏回族自治区重点研发计划项目(2025BEGD2022). （2025BEGD2022）
