Permeability prediction of tight sandstone reservoirs under few-shot and imbalanced conditions: A generative data enhancement and bidirectional attention fusion framework

Abstract

Permeability prediction in tight sandstone reservoirs is strongly influenced by diagenesis, multiscale pore structures and multiphase flow effects, resulting in pronounced spatial heterogeneity. Traditional empirical models based on homogenization assumptions suffer from inherent physical limitations, making their predictive accuracy insufficient for practical applications. Although existing data-driven deep learning methods offer advantages in nonlinear modeling, their performance is constrained by few-shot samples, imbalanced distributions, low-density information and the absence of physical constraints, which hinder the extraction of effective representations of inter-features and lead to biased learning and physical deviations. To address these challenges, this study proposes a framework including a data enhancement strategy based on high-order feature construction for quality improving and adversarial generation for quantity expanding, and a deep learning model termed DAFM-PI, which integrates bidirectional attention fusion with physical information constraints. Experimental results show that the test accuracy R² of DAFM-PI improves from 0.444 before enhancement to 0.969 after that, representing a 52.2% improvement. Furthermore, compared with conventional POR-LR, ResNet, and Transformer Encoder models, DAFM-PI achieves accuracy improvements of 61.9%, 18.1%, and 6.2%, demonstrating its effectiveness in enhancing both prediction accuracy and generalization for permeability prediction in tight sandstone reservoirs.

Keywords

Tight sandstone reservoirs / Well logs / Permeability prediction / Data enhancement / Attention fusion

References

[1] C. Zou, G. Zhang, Z. Yang, S. Tao, L. Hou, R. Zhu, X. Yuan, Q. Ran, D. Li, Z. Wang, Geological concepts, characteristics, resource potential and key techniques of unconventional hydrocarbon: on unconventional petroleum geology, Petrol. Explor. Dev. 40 (2013) 385-399, https://doi.org/10.11698/PED.2013.04.01.

[2] C. Zou, Y. Ding, Y. Lu, X. Liu, J. Chen, X. Wang, Z. Yang, B. Cai, Z. Yang, C. He, Z. Wang, Y. Luo, Concept, technology and practice of "man-made reservoirs" development, Petrol. Explor. Dev. 44 (2017) 144-154, https://doi.org/10.11698/PED.2017.01.18.

[3] C. Zou, S. Li, B. Xiong, Y. Chen, G. Zhang, X. Xie, H. Liu, F. Ma, Y. Liang, K. Zhu, C. Guan, S. Pan, M. Hou, Y. Yuan, S. Luo, Connotation, pathway and significance of carbon neutrality "super energy system": a case study of the ordos basin, nw China, Petrol. Explor. Dev. 51 (2024) 924-936, https://doi.org/10.11698/PED.20240021.

[4] P. Yu, Z. Tang, Application of bayesian theory to predict permeability of reservoir, J. Cent. South Univ. 46 (2015), https://doi.org/10.11817/j.issn.1672-7207.2015.03.02.

[5] H. Zhang, D. Yang, Y. Wu, Classification of reservoir permeability prediction technology based on fisher discrimination, Geol. Sci. Technol. Inf. 36 (2017) 242-246, https://doi.org/10.19509/i.cnki.dzkq.2017.0132.

[6] L. Gan, Y. Wang, X. Luo, M. Zhang, X. Li, X. Dai, H. Yang, A permeability prediction method based on pore structure and lithofacies, Petrol. Explor. Dev. 46 (2019) 883-890, https://doi.org/10.11698/PED.2019.05.07.

[7] P. Lu, H. Wang, C. Gao, Y. Hao, F. Tan, J. Liu, W. Liu, Y. Bai, J. Gao, Research progress of permeability prediction technology for tight sandstone reservoirs, Prog. Geophys. 37 (2022) 2428-2438, https://doi.org/10.6038/pg2022FF0236.

[8] F. Xiong, L. Gan, W. Sun, H. Yang, Characterization of reservoir permeability and analysis of influencing factors in fracture-pore media, Chin. J. Geophys. 64 (2021) 279-288, https://doi.org/10.6038/cjg2021N0175.

[9] A. Alvarez, R. Missagia, R. Brito da Silva, A. Carrasquilla, Non-linear regression for permeability prediction and hydraulic flow units' characterization from core and well logs of a pre-salt carbonate reservoir in the santos basin, Brazil, J. Appl. Geophys. 241 (2025) 105865, https://doi.org/10.1016/j.jappgeo.2025.105865.

[10] X. Zhao, Y. Zhong, P. Li, Rtg-gnn: a novel rock topology-guided approach for permeability prediction using graph neural networks, Geoenergy Sci. Eng. 243 (2024) 213358, https://doi.org/10.1016/j.geoen.2024.213358.

[11] M. Matinkia, R. Hashami, M. Mehrad, M.R. Hajsaeedi, A. Velayati, Prediction of permeability from well logs using a new hybrid machine learning algorithm, Petroleum 9 (2023) 108-123, https://doi.org/10.1016/j.petlm.2022.03.003.

[12] Y. Tian, G. Wang, H. Li, Y. Huang, F. Zhao, Y. Guo, J. Gao, J. Lai, A novel deep learning method based on 2-d cnns and grus for permeability prediction of tight sandstone, Geoenergy Sci. Eng. 238 (2024) 212851, https://doi.org/10.1016/j.geoen.2024.212851.

[13] A. Al-AbdulJabbar, K. Al-Azani, S. Elkatatny, Estimation of reservoir porosity from drilling parameters using artificial neural networks, Petrophysics 61 (2020) 318-330, https://doi.org/10.30632/PJV61N3-2020a5.

[14] C. Cheng, P. Li, Y. Chen, Y. Ye, Y. Gao, L. Zhang, Research progress of reservoir logging evaluation based on machine learning, Prog. Geophys. 37 (2022) 164-177, https://doi.org/10.6038/pg2022FF0031.

[15] Q. Wang, M. Tan, Y. Shi, G. Li, X. Cheng, W. Luo, Prediction of relative permeability and calculation of water cut of tight sandstone reservoir based on radial basis function neural network, Oil Geophys. Prospect. 55 (2020) 864-872, https://doi.org/10.13810/j.cnki.issn.1000-7210.2020.04.018.

[16] L. Zhu, C. Zhang, X. He, Z. Wu, X. Zhou, S. Di, Y. Li, Permeability prediction of tight sandstone reservoir based on improved bpnn and t2 full-spectrum, Geophys. Prospect. Pet. 56 (2017) 727-734, https://doi.org/10.3969/i.issn.1000-1441.2017.05.013.

[17] Y. Gu, D. Zhang, Z. Bao, Permeability prediction using pso-xgboost based on logging data, Oil Geophys. Prospect. 56 (2021) 26-37, https://doi.org/10.13810/j.cnki.issn.1000-7210.2021.01.003.

[18] Y. Zhang, C. Zhang, K. Sun, W. Yang, M. Wang, Logging interpretation model on complex carbonate reservoir permeability based on hybrid simulated annealing-genetic algorithm-random forest algorithm, Petroleum Geology and Recovery Efficiency 29 (2022), https://doi.org/10.13673/j.cnki.cn37-1359/te.2022.01.007.

[19] S. Luo, Y. Zhao, H. Xiao, W. Chen, G. He, L. Zhang, C. Du, Fracture parameter optimization of tight sandstone gas reservoirs based on the hybrid optimization algorithm and deep neural network model, Nat. Gas. Ind. 44 (2024) 140-151, https://doi.org/10.3787/i.issn.1000-0976.2024.09.013.

[20] M.A. Ahmadi, Z. Chen, Comparison of machine learning methods for estimating permeability and porosity of oil reservoirs via petro-physical logs, Petroleum 5 (2019) 271-284, https://doi.org/10.1016/j.petlm.2018.06.002.

[21] P. Xu, H. Zhou, X. Liu, L. Chen, C. Xiong, F. Lyu, J. Zhou, J. Liu, Permeability prediction using logging data in a heterogeneous carbonate reservoir: a new self-adaptive predictor, Geoenergy Sci. Eng. 224 (2023) 211635, https://doi.org/10.1016/j.geoen.2023.211635.

[22] H. Wang, B. Zhao, J. Liu, L. Tao, C. Gao, W. Liao, Density logging curve reconstruction method based on tcn-bigru with multi-head attention mechanism, Prog. Geophys. 40 (2025) 592-604, https://doi.org/10.6038/pg2025II0114.

[23] F. Huo, Q. Li, H. Dong, Y. Chen, Prediction method of reservoir evaluation parameters based on improved stacking fusion model, Prog. Geophys. 40 (2025) 691-704, https://doi.org/10.6038/pg2025II0072.

[24] B. Zeng, H. Xiao, Z. Hao, H. Liu, Reservoir porosity prediction using the vae-bigru-attention model: an example study of middle-low permeability sandstone reservoir, Prog. Geophys. 40 (2025) 658-669, https://doi.org/10.6038/pg2025II0097.

[25] A.E. Pedrol, P.A. Nara, S.A. Marcio, A transformer-based approach for anomaly detection in intelligent well completions, Petrol. Explor. Dev. 52 (2025) 1-12, https://doi.org/10.11698/PED.20240715.

[26] M. Masroor, M. Emami Niri, M.H. Sharifinasab, A multiple-input deep residual convolutional neural network for reservoir permeability prediction, Geoenergy Sci. Eng. 222 (2023) 211420, https://doi.org/10.1016/j.geoen.2023.211420.

[27] J. Wang, J. Cao, J. You, J. Liu, X. Zhou, Prediction of reservoir porosity, permeability, and saturation based on gated recurrent unit neural network, Geophys. Prospect. Pet. 59 (2020) 616-627, https://doi.org/10.3969/i.issn.1000-1441.2020.04.012.

[28] W. Yang, C. Zhang, M. Yang, Y. Zhang, M. Wang, K. Sun, Permeability logging evaluation of carbonate reservoirs in oilfield h of Iraq based on long short-term memory recurrent neural network, Pet. Geol. Oilfield Dev. Daqing 41 (2022), https://doi.org/10.19597/J.ISSN.1000-3754.202105003.

[29] R. Shao, L. Xiao, G. Liao, Y. Shi, J. Zhou, G. Li, X. Hou, Multitask learning based reservoir parameters prediction with geophysical logs, Chin. J. Geophys. 65 (2022) 1883-1895, https://doi.org/10.6038/cjg2022P0177.

[30] M. Wang, X. Guo, H. Tang, S. Zhang, T. Yang, G. Zhong, Prediction case of core parameters of shale gas reservoirs through deep transformer transfer learning, Chin. J. Geophys. 66 (2023) 2592-2610, https://doi.org/10.6038/cjg2022Q0054.

[31] H. Liu, Y. Ren, X. Li, Y. Deng, Y. Wang, Q. Cao, J. Du, Z. Lin, W. Wang, Research status and application of artificial intelligence large models in the oil and gas industry, Petrol. Explor. Dev. 51 (2024) 910-923, https://doi.org/10.11698/PED.20240254.

[32] Z. Chen, Y. Zhang, J. Li, G. Hui, Y. Sun, Y. Li, Y. Chen, D. Zhang, Artificial intelligence large model for logging curve reconstruction, Petrol. Explor. Dev. 52 (2025) 744-756, https://doi.org/10.11698/PED.20240761.

[33] D. Zhang, Y. Chen, J. Meng, Synthetic well logs generation via recurrent neural networks, Petrol. Explor. Dev. 45 (2018) 598-607, https://doi.org/10.11698/PED.2018.04.06.

[34] G.E. Karniadakis, I.G. Kevrekidis, L. Lu, P. Perdikaris, S. Wang, L. Yang, Physics-informed machine learning, Nat. Rev. Phys. 3 (2021) 422-440, https://doi.org/10.1038/s42254-021-00314-5.

[35] P. Chiu, J. Wong, C. Ooi, M. Dao, Y. Ong, Can-pinn: a fast physics-informed neural network based on coupled-automatic-numerical differentiation method, Comput. Methods Appl. Mech. Eng. 395 (2022) 114909, https://doi.org/10.1016/j.cma.2022.114909.

[36] W. Shan, G. Xie, Q. Lv, X. Yang, Y. Yang, Physics-informed machine learning for solving partial differential equations in porous media, Adv. Geo-Energy Res. 7 (2023) 279-291, https://doi.org/10.46690/ager.2023.04.02.

[37] L. Kuang, H. Liu, Y. Ren, K. Luo, M. Shi, J. Su, X. Li, Application and development trend of artificial intelligence in petroleum exploration and development, Petrol. Explor. Dev. 48 (2021) 1-11, https://doi.org/10.11698/PED.2021.01.01.

[38] S. Dong, X. Yang, T. Xu, L. Zeng, S. Chen, L. Wang, Y. Niu, G. Xiong, Semi-supervised neural network for complex lithofacies identification using well logs, IEEE Trans. Geosci. Rem. Sens. 62 (2024) 1-19, https://doi.org/10.1109/TGRS.2024.3450103.

[39] S. Dong, L. Zeng, W. Lyu, D. Xia, G. Liu, Y. Wu, X. Du, Fracture identification and evaluation using conventional logs in tight sandstones: a case study in the ordos basin, China, Energy Geoscience 1 (2020) 115-123, https://doi.org/10.1016/j.engeos.2020.06.003.

[40] S. Dong, Z. Zhong, X. Cui, L. Zeng, X. Yang, J. Liu, Y. Sun, J. Hao, A deep kernel method for lithofacies identification using conventional well logs, Pet. Sci. 20 (2023) 1411-1428, https://doi.org/10.1016/j.petsci.2022.11.027.

[41] Y. Zhu, X. Li, L. Liao, D. Feng, Y. Mao, J. Li, K. Ye, H. Zhang, Experimental study of starting pressure gradient in carbonate gas reservoirs, Geotech. Geol. Eng. 43 (2024) 22, https://doi.org/10.1007/s10706-024-03004-y.

[42] G. Xie, Q. Lv, W. Shan, Y. Yang, X. Yang, Intelligent modeling with physics-informed machine learning for petroleum engineering problems, Adv. Geo-Energy Res. 8 (2023) 71-75, https://doi.org/10.46690/ager.2023.02.01.

[43] H. Song, J. Zhu, C. Wei, J. Wang, S. Du, C. Xie, Data-driven physics-informed interpolation evolution combining historical-predicted knowledge for remaining oil distribution prediction, J. Petrol. Sci. Eng. 217 (2022) 110795, https://doi.org/10.1016/j.petrol.2022.110795.

[44] Z. Zhang, A physics-informed deep convolutional neural network for simulating and predicting transient darcy flows in heterogeneous reservoirs without labeled data, J. Petrol. Sci. Eng. 211 (2022) 110179, https://doi.org/10.1016/j.petrol.2022.110179.

[45] P. Pothana, K. Ling, Physics-integrated neural networks for improved mineral volumes and porosity estimation from geophysical well logs, Energy Geoscience 6 (2025) 100410, https://doi.org/10.1016/j.engeos.2025.100410.