Reconstruction of Well Logging Based on Adversarial Autoencoders and Transformers
SPE Journal · 10.09.2025 · 30 (09) · 5177-5188
DOI: 10.2118/228325-pa ↗Kurz gefasst
Fachartikel in der Zeitschrift SPE Journal, online erschienen am 25.06.2025. Verfasst von Fuhao Zhao und 8 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 34 Literaturangaben.
Abstract (Originalsprache)
Well-logging data are fundamental for obtaining accurate subsurface information and enhancing resource exploration. However, missing or incomplete logging curves are common due to factors such as limited logging conditions or instrument failures. Existing reconstruction methods often assume that data from training and test wells share identical distributions, overlooking the impact of original feature distribution differences on prediction performance. To address these challenges, we propose a transformer framework with an adversarial autoencoder (AAE-Transformer) based on a domain adaptation method. The reconstruction of an acoustic (AC) logging curve is used as an example to verify the model’s validity. Initially, a traditional bidirectional long short-term memory (Bi-LSTM) model is used to verify the effectiveness of the AAE model. The maximum mean discrepancy (MMD) values between wells confirm that the AAE reduces domain differences by aligning features from the training and test wells in the latent space. Integrating the AAE model significantly enhances the robustness of the Bi-LSTM, reducing discrepancies in root mean square error (RMSE) and mean absolute error (MAE) between the training and test sets. Experiment results show the importance of domain adaptation in well-logging reconstruction. To further improve prediction performance, a transformer structure is developed as a regressor, leveraging its capability to process long sequential data efficiently. The proposed AAE-Transformer framework demonstrates superior performance in reconstructing high-precision logging curves, as validated by the reconstruction of the AC curve in the test well.
Zitieren
APA
Zhao, F., Huang, W., Zhao, Z., Gong, W., Wang, H., Zhu, L., Wang, W., Bai, R., & Zhu, Y. (2025). Reconstruction of Well Logging Based on Adversarial Autoencoders and Transformers. SPE Journal, 30(09), 5177-5188. https://doi.org/10.2118/228325-pa
Vancouver
Zhao F, Huang W, Zhao Z, Gong W, Wang H, Zhu L, et al. Reconstruction of Well Logging Based on Adversarial Autoencoders and Transformers. SPE Journal. 2025;30(09):5177-5188. doi:10.2118/228325-pa
BibTeX
@article{Zhao2025,
author = {Zhao, Fuhao and Huang, Weian and Zhao, Zhiqiang and Gong, Wuzhen and Wang, Haochen and Zhu, Lixin and Wang, Weidong and Bai, Ruiyu and Zhu, Yuqing},
title = {Reconstruction of Well Logging Based on Adversarial Autoencoders and Transformers},
journal = {SPE Journal},
year = {2025},
volume = {30},
number = {09},
pages = {5177--5188},
doi = {10.2118/228325-pa}
}
EUVIMED