Lithology identification is regarded as a critical component of formation subdivision and reservoir evaluation during the process of oil and gas exploration and development. To address the challenges posed by high-dimensional non-linear characteristics of well-logging data and the imbalanced distribution of lithological samples under complex geological conditions, a novel identification method based on a data balancing strategy and an improved TabNet network was proposed. Firstly, BSMOTE(borderline-SMOTE) algorithm was introduced for data equalization to mitigate model training biases caused by the scarcity of minority lithological samples in the raw logging data. Secondly, an improved TabNet model integrated with ViT (vision transformer) was constructed. By embedding the ViT encoder module into the Attentive Transformer of the original TabNet architecture, a composite attention transformation module with global feature modeling capabilities was formed. Experimental results based on field data from the M block in Xinjiang demonstrate that the improved TabNet model, following BSMOTE-based data balancing, achieves a lithology identification accuracy of 92.79%. The comprehensive performance is found to be significantly superior to traditional machine learning methods, such as RF (random forest), as well as deep learning benchmarks like CNN (convolutional neural networks).
XIANYU Hao-fan, XIONG Jian, LIU Jing-yan, LIU Xiang-jun, LIANG Li-xi, DING Yi.
Well Logging Lithology Identification Method Based on Improved TabNet Algorithm[J].
Science Technology and Engineering, 2026, 26(30): 12993-13005 DOI:10.12404/j.issn.1671-1815.2600014
XuZhenhao, MaWen, LiShucai, et al. Lithology identification: method, research status and intelligent development trend[J]. Geological Review, 2022, 68(6): 2290-2304.
ChenXiujuan, FengZhentao, ZengFurong, et al. Development status of logging-based lithology identification technology for shale formations[J]. Xinjiang Petroleum Geology, 2024, 45(6): 742-752.
MaLongfei, XiaoHanmin, TaoJingwei, et al. Research and application of lithology classification based on deep learning[J]. Science Technology and Engineering, 2022, 22(7): 2609-2617.
DuanZhongyi, XiaoKun, YangYaxin, et al. Logging identification of borehole lithology of sandstone-type uranium deposit in Songliao Basin[J]. Progress in Geophysics, 2023, 38(6): 2490-2501.
LuoGongwei, AnXiaoping, YaoWeihua, et al. Application status and development trends of artificial intelligence in logging interpretation for unconventional oil and gas reservoirs[J]. Petroleum Science Bulletin, 2025, 10(5): 908-925.
LiXi, FanXiangyu, WangZhaofeng, et al. Logging lithology identification method research based on PSO-SVM: a case study of Paleozoic (Pz) reservoir in K oil field, South Turgay Basin, Kazakhstan[J]. Progress in Geophysics, 2022, 37(2): 617-626.
ZengRui, ShaoYanlin, HuangYu, et al. Intelligent identification of logging lithology based on deep learning[J]. Science Technology and Engineering, 2025, 25(30): 12804-12812.
ShangYazhou, ZhangZhaohui, XuDuonian, et al. Log-based lithology identification of volcanic rocks using random forest method: a case study of Carboniferous strata in the Dixi area, Junggar Basin[J]. Geophysical and Geochemical Exploration, 2024, 48(4): 1025-1036.
[17]
Al-khudafiA M, Al-SharifiH A, HamadaG M, et al. Evaluation of different tree-based machine learning approaches for formation lithology classification[C]// ARMA/DGS/SEG International Geomechanics Symposium. Al Khobar: ARMA, 2023. DOI: 10.56952/IGS-2023-0026.
GaoXiaoping, ZhuHongbo, ZhaoKai, et al. Rock identification method based on convolutional neural network-long short term memory artificial neural network-self attention mechanism[J]. Science Technology and Engineering, 2026, 26(2): 544-552.
SuFu, MaLei, LuoRenze, et al. Research and application of logging lithology identification based on improved multi-class twin support vector machine[J]. Progress in Geophysics, 2020, 35(1): 174-180.
SunJing, ZhaoJunlong, ZhangYuchen, et al. Lithology identification of tight sandstone reservoirs based on SSMO-SSA-LGBM algorithm[J]. Geological Bulletin of China, 2025, 44(5): 935-948.
YinQiong. Lithology identification method and application based on generative adversarial neural network[J]. Well Logging Technology, 2025, 49(1): 57-67.
TangBaiqiang, MengQingtao, YangLiang, et al. Intelligent prediction of fine-grained sedimentary lithofacies based on BSMOTE-SVM: a case study of the Member 1 of Qingshankou Formation in Songliao Basin[J]. Journal of Palaeogeography, 2025, 27(4): 937-949.
[32]
ArikS Ö, PfisterT. Tabnet: attentive interpretable tabular learning[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2021, 35(8): 6679-6687.
LiChaogang, JuNengpan, HeChaoyang, et al. Stepped landslide displacement prediction based on stratified sampling and attention mechanism[J]. Science Technology and Engineering, 2026, 26(2): 476-489.
SunJing, YouXincai, ZhangQuan, et al. Development characteristics and genesis of deep tight conglomerate reservoirs of Mahu area in Junggar Basin[J]. Natural Gas Geoscience, 2023, 34(2): 240-252.
[39]
LiP, MengJ B, LiJ, et al. Enhanced lithofacies classification of tight sandstone reservoirs using a hybrid CNN-GRU model with BSMOTE and heat kernel imputation[J]. Applied Geophysics, 2025, 22: 1141-1157.
[40]
HeY B, ShengK M, DuM L, et al. An interpretable deep learning framework using FCT-SMOTE and BO-TabNet algorithms for reservoir water sensitivity damage prediction[J]. Scientific Reports, 2025, 15(1): 18655.
MaYubin, LiuShiyou, CaoDanping, et al. Application of random forest algorithm based on Bayesian optimization for reservoirprediction[J]. Chinese Journal of Geophysics, 2025, 68(8): 3247-3257.
[48]
ZhangW G, WuC Z, ZhongH Y, et al. Prediction of undrained shear strength using extreme gradient boosting and random forest based on Bayesian optimization[J]. Geoscience Frontiers, 2021, 12(1): 469-477.