From Automated Well Stratigraphic Interpretation to Its Seismic-Constrained Spatial Mapping Using Multitask Neural Network: Implications for Reservoir Property Modeling
Interpretation · 20.09.2026
DOI: 10.1190/int-2024-0165 ↗Kurz gefasst
Fachartikel in der Zeitschrift Interpretation, online erschienen am 20.09.2026. Verfasst von Yue Dai und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref). Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
The processes of stratigraphic subsurface identification and correlation form the basis for reservoir modeling. Manual correlation often relies on expert experience, making it challenging and time-consuming, particularly for mature fields with many wells. Current automatic interpretation methods predominantly focus on larger-scale sequence stratigraphy analysis. The fine delineation of sublayers within sand groups has yet to receive sufficient attention. We propose a multitask encoder-decoder architecture based on SegNet to enhance sublayer interpretation. To improve the accuracy of stratigraphic division and correlation in the depth domain, the multitask SegNet model incorporates the mean squared error to formulate a joint loss function, which we use as the objective function. We first augment the training data using lithology-sensitive logs and a wavelet transform. Subsequently, we identify the sublayers of target wells using the trained neural network. Finally, we retrain the multitask SegNet model to conduct spatial stratigraphic interpretation constrained by previous artificial intelligence results. The principal distinction between well sublayer and stratigraphic surface interpretation is that the latter utilizes a seismic impedance model instead of well logs to achieve the SegNet results. We validate the effectiveness of our method by conducting a case study in the Shengli Oilfield, China. The results demonstrate that our method is significantly more efficient than the existing approach and facilitates stratigraphic subsurface characterization.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Dai, Y., Huang, X., Liu, H., Yang, H., & Song, H. (2026). From Automated Well Stratigraphic Interpretation to Its Seismic-Constrained Spatial Mapping Using Multitask Neural Network: Implications for Reservoir Property Modeling. Interpretation. https://doi.org/10.1190/int-2024-0165
Vancouver
Dai Y, Huang X, Liu H, Yang H, Song H. From Automated Well Stratigraphic Interpretation to Its Seismic-Constrained Spatial Mapping Using Multitask Neural Network: Implications for Reservoir Property Modeling. Interpretation. 2026. doi:10.1190/int-2024-0165
BibTeX
@article{Dai2026,
author = {Dai, Yue and Huang, Xuri and Liu, Haojie and Yang, Hongwei and Song, Haibo},
title = {From Automated Well Stratigraphic Interpretation to Its Seismic-Constrained Spatial Mapping Using Multitask Neural Network: Implications for Reservoir Property Modeling},
journal = {Interpretation},
year = {2026},
doi = {10.1190/int-2024-0165}
}
EUVIMED