Automated Well-Log Interpretation Using Deep Neural Networks
DOI:
https://doi.org/10.63282/3050-9262.IJAIDSML-V2I4P109Keywords:
Well-Log Interpretation, Deep Learning, 1D-CNN, LSTM Networks, Lithofacies Classification, Sequential Modeling, Automated PetrophysicsAbstract
Automated well-log interpretation has become increasingly important as reservoirs grow more complex and multi-well datasets expand beyond the capacity of traditional manual workflows. The classical machine learning methods like SVMs, ANNs and ensemble models offered the first steps to automation although they had their drawbacks of manual feature engineering, lacked depth-dependent relations and inconsistent generalization. Deep learning (DL) systems, such as fully connected deep neural networks (DNNs), one-dimensional convolutional neural networks (1D-CNNs) and recurrent models, such as LSTMs, provide significant improvements, and thus learn hierarchical, nonlinear and sequential features directly upon raw log curves. The present review will summarize the main developments in the field of deep learning in terms of well-log interpretation and list the main studies of interest, as well as enumerate the advantages and drawbacks of the leading architecture types. Based on a mixed Volve and KGS data, the article describes a standardized workflow that includes the preprocessing, model training, and cross-well testing. It has been reported that CNNs and LSTMs perform better than DNNs and classical ML models by being more accurate, recognizing more thin beds, and generalizing to unseen wells. The remaining issues are small labeled datasets, inter-basin variability and lack of interpretability. The research paper ends by giving suggested future research directions that focus on semi-supervised learning, synthetic log augmentation, standardized benchmarks, and integration of core data to have more reliable and scalable automated interpretation.
References
[1] F. Segesman, S. Soloway, and M. Watson, “Well logging—The exploration of subsurface geology,” Proc. IRE, vol. 50, no. 11, pp. 2227–2243, 2007.
[2] D. N. Obiora, D. Gbenga, and G. Ogobiri, “Reservoir characterization and formation evaluation of a ‘Royal onshore field’, Southern Niger Delta using geophysical well log data,” J. Geol. Soc. India, vol. 87, no. 5, pp. 591–600, 2016.
[3] M. Ehsan and H. Gu, “An integrated approach for the identification of lithofacies and clay mineralogy through Neuro-Fuzzy, cross plot, and statistical analyses, from well log data,” J. Earth Syst. Sci., vol. 129, no. 1, p. 101, 2020.
[4] R. Akkurt et al., “Accelerating and enhancing petrophysical analysis with machine learning: A case study of an automated system for well log outlier detection and reconstruction,” in SPWLA Annu. Logging Symp., 2018.
[5] W. Chen et al., “Landslide spatial modeling: Introducing new ensembles of ANN, MaxEnt, and SVM machine learning techniques,” Geoderma, vol. 305, pp. 314–327, 2017.
[6] Z. Tariq et al., “A systematic review of data science and machine learning applications to the oil and gas industry,” J. Pet. Explor. Prod. Technol., vol. 11, no. 12, pp. 4339–4374, 2021.
[7] K. J. Bergen et al., “Machine learning for data-driven discovery in solid Earth geoscience,” Science, vol. 363, no. 6433, p. eaau0323, 2019.
[8] M. Tian, M. Miao, et al., “Deep learning assisted well log inversion for fracture identification,” Geophysical Prospecting, vol. 69, no. 2, pp. 419–433, 2021.
[9] J. Jeong et al., “Interpreting the subsurface lithofacies at high lithological resolution by integrating information from well-log data and rock-core digital images,” J. Geophys. Res.: Solid Earth, vol. 125, no. 2, p. e2019JB018204, 2020.
[10] M. Dentith et al., “Petrophysics and mineral exploration: a workflow for data analysis and a new interpretation framework,” Geophys. Prospect., vol. 68, no. 1, pp. 178–199, 2020.
[11] S. G. Olsen, Improved Petrophysical Evaluation of a Thinly Bedded Oil and Gas Bearing Reservoir, Using Triaxial Resistivity and Nuclear Magnetic Resonance Well Logging Measurements, M.S. thesis, NTNU, 2016.
[12] S. Chaki, A. Routray, W. K. Mohanty, and M. Jenamani, “A novel multiclass SVM-based framework to classify lithology from well logs: A real-world application,” ICSI–Advances in Swarm and Computational Intelligence, LNCS 9141, Springer, 2015.
[13] A. A. Konaté et al., “Machine learning interpretation of conventional well logs in crystalline rocks,” in Advances in Swarm and Computational Intelligence, vol. 9141, Springer, 2015, pp. —.
[14] S. Yu and J. Ma, “Deep learning for geophysics: Current and future trends,” Rev. Geophys., vol. 59, no. 3, p. e2021RG000742, 2021.
[15] E. M. Viggen et al., “Better automatic interpretation of cement evaluation logs through feature engineering,” SPE J., vol. 26, no. 5, pp. 2894–2913, 2021.
[16] D. Zhang, Y. Chen, and J. Meng, “Synthetic well logs generation via recurrent neural networks,” Pet. Explor. Dev., vol. 45, no. 4, pp. 629–639, 2018.
[17] D. L. Bergman et al., “Applying statistical learning to quantitative well log analysis,” in SPE Reservoir Characterisation and Simulation Conf., 2017.
[18] Pham, N., & Zabihi Naeini, E. (2019). Missing well log prediction using deep recurrent neural networks. In Proceedings of the 81st EAGE Conference and Exhibition (pp. 1–5). European Association of Geoscientists & Engineers. https://doi.org/10.3997/2214-4609.201901612.
[19] X. Cheng, L. Fan, and W. Gu, Comprehensive Practice of Sequence Stratigraphy Techniques and Methods, in Comprehensive Practice of Exploration and Evaluation Techniques in Complex Reservoirs. Singapore: Springer, 2019, pp. 1–101.
[20] A. T. Tunkiel, T. Wiktorski, and D. Sui, “Drilling dataset exploration, processing and interpretation using Volve field data,” in OMAE, vol. 84430, ASME, 2020.
[21] T. Saryerwinnie, Hydrogeological Study of a Municipal Groundwater Well Field in Central Kansas, 2017.
[22] K. Raterman, Y. Liu, and L. Warren, “Analysis of a drained rock volume: An Eagle Ford example,” in URTEC, 2019.
[23] A. Adeyemo, H. Wimmer, and L. M. Powell, “Effects of normalization techniques on logistic regression in data science,” J. Inf. Syst. Appl. Res., vol. 12, no. 2, pp. 37–—, 2019.
[24] Y.-D. Zhang et al., “Multiple sclerosis identification by convolutional neural network with dropout and parametric ReLU,” J. Comput. Sci., vol. 28, pp. 1–10, 2018.
