To address cross-well distribution shift, unstable boundary prediction, and missed detections in thin interbedded intervals during reservoir identification from multi-well logging data, a DeepTabNet hybrid network method for fine-grained reservoir identification was proposed. Five well-log curves, namely gamma ray, deep resistivity, bulk density, neutron porosity, and sonic transit time, were uniformly selected as the input features, and vertical contextual information was introduced through robust normalization and sliding-window sample construction. In terms of model architecture, DeepTabNet preserves the step-wise sparse feature selection strategy of the original TabNet while incorporating feature embedding and multi-head self-attention to capture high-order nonlinear interactions across logging curves. In addition, a one-dimensional convolutional texture branch was constructed to extract locally continuous morphological features along the depth direction, and the two types of features were ultimately fused to output reservoir probabilities. Experiments were conducted on logging data from 16 wells and evaluated using well-based splitting and five-fold cross-validation. The results show that DeepTabNet achieves an accuracy of 0.905, an F1-score of 0.862, and a PR-AUC of 0.914 on the validation set, while obtaining an average accuracy of 0.899 and a PR-AUC of 0.905 in independent cross-well testing, consistently outperforming the rule-based method, random forest, BiLSTM, TCN, and GNN baselines. Ablation studies further demonstrate that multi-head self-attention, Sparsemax sparse gating, and the texture branch all make positive contributions to performance improvement. The results indicate that the proposed method can effectively enhance reservoir identification accuracy, boundary stability, and cross-well generalization, while also exhibiting good interpretability and practical engineering potential.
HUANG Jun-jie, LI Quan-hou.
DeepTabNet Hybrid Network Method for Fine-grained Reservoir Identification from Well Logging Data[J].
Science Technology and Engineering, 2026, 26(25): 10804-10816 DOI:10.12404/j.issn.1671-1815.2600238
LiuGuoqiang, GongRenbin, ShiYujiang, et al. Construction of well logging knowledge graph and intelligent identification method of hydrocarbon-bearing formation[J]. Petroleum Exploration and Development, 2022, 49(3): 502-512.
HanJian, LiJing, CaoZhimin, et al. Well log curve restoration based on LSTM multi-scale co-occurrence relationship mining[J]. Journal of Electronics & Information Technology, 2022, 44(7): 2559-2567.
PanShaowei, WangZhaoyang, ZhangYun, et al. Lithology identification based on LSTM neural networks completing log and hybrid optimized XGBoost[J]. Journal of China University of Petroleum (Edition of Natural Science), 2022, 46(3): 62-71.
WangMin, YangJinlu, WangXin, et al. Identification of shale lithofacies by well logs based on random forest algorithm[J]. Earth Science, 2023, 48(1): 130-142.
XuFangying, ZouYanhong, YiZhuowei, et al. ADASYN-CatBoost method for intelligent identification of logging lithology considering unbalanced data: a case study of Zhaoxian gold deposit in northwestern Jiaodong Peninsula[J]. Gold Science and Technology, 2023, 31(5): 721-735.
WangJintao, WenXiaotao, HeYilong, et al. Logging curve prediction based on a CNN-GRU neural network[J]. Geophysical Prospecting for Petroleum, 2022, 61(2): 276-285.
XieWenmin, OuyangYong, TianDeliang, et al. Mechanism of fracture network development for horizontal wells using alternating injection of variable viscosity fluids in shale reservoirs: Changqing Oilfield Chang 7 shale oil[J]. Science Technology and Engineering, 2025, 25(32): 13771-13777.
BaiRu, WangShiyu, ZhangLu, et al. Automatic formation identification method for logging curves based on MAE neural network[J]. Natural Gas Exploration and Development, 2024, 47(4): 63-71.
LuoJiang, XiongJian, GouYanhong, et al. Main controlling factors and productivity prediction in initial production stage of shale oil horizontal wells[J]. Science Technology and Engineering, 2026, 26(7): 2844-2856.
[23]
ArikS O, PfisterT. TabNet: attentive interpretable tabular learning[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Palo Alto: AAAI, 2021, 35(8): 6679-6687.
[24]
MartinsA F T, AstudilloR F. From softmax to sparsemax: a sparse model of attention and multi-label classification[C]// Proceedings of the 33rd International Conference on Machine Learning. New York: PMLR, 2016, 48: 1614-1623.
[25]
LinT Y, GoyalP, GirshickR, et al. Focal loss for dense object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2): 318-327.
[26]
MinX, QinP, ZhangF. Research and application of logging lithology identification for igneous reservoirs based on deep learning[J]. Journal of Applied Geophysics, 2020, 173: 103929.
[27]
JiangC, ZhangD, ChenS. Lithology identification from well-log curves via neural networks with additional geologic constraint[J]. Geophysics, 2021, 86(5): 313-324.
[28]
JiangC, ZhangD, ChenS. Handling missing data in well-log curves with a gated graph neural network[J]. Geophysics, 2023, 88(1): 13-30.
[29]
JiangC, ZhangD. Lithofacies identification from well-logging curves via integrating prior knowledge into deep learning[J]. Geophysics, 2024, 89(1): 31-41.
[30]
LiuJ J, LiuJ C. Integrating deep learning and logging data analytics for lithofacies classification and 3D modeling of tight sandstone reservoirs[J]. Geoscience Frontiers, 2022, 13(1): 101311.
[31]
MohammadiK, MousaviS H R, Hosseini-NasabS M, et al. Transforming petrophysical well-logs to images: leveraging deep learning for lithology recognition[J]. Earth Science Informatics, 2025, 18(2): 457-469.
[32]
DasS, SinghaD K, MandalP P, et al. Identification of lithofacies from well log data in the upper Assam Basin using machine learning techniques[J]. Acta Geophysica, 2024, 72(6): 3191-3210.
[33]
WangA, ZhaoS, XieK, et al. Attention mechanism-enhanced graph convolutional neural network for unbalanced lithology identification[J]. Scientific Reports, 2024, 14: 17319.
[34]
DaiC, SiX, WuX. FlexLogNet: a flexible deep learning-based well-log completion method of adaptively using what you have to predict what you are missing[J]. Computers & Geosciences, 2024, 191: 105666.
[35]
ShiY, LiaoJ, GanL, et al. Lithofacies prediction from well log data based on deep learning: a case study from southern Sichuan, China[J]. Applied Sciences, 2024, 14(18): 8195.
[36]
FengH, ZhangZ, ZhangC, et al. Sequence-variable attention temporal convolutional network for volcanic lithology identification based on well logs[J]. Journal of Petroleum Exploration and Production Technology, 2025, 15(1): 6.
[37]
XieD, LiuZ, WangF, et al. A transformer and LSTM-based approach for blind well lithology prediction[J]. Symmetry, 2024, 16(5): 616.
[38]
SaporettiC M, da FonsecaL G, PereiraE. A lithology identification approach based on machine learning with evolutionary parameter tuning[J]. IEEE Geoscience and Remote Sensing Letters, 2019, 16(12): 1819-1823.