Unveiling genetic affinity between highly mature source rocks and crude oils: A machine learning approach driven by geochemical data
AAPG Bulletin · 30.06.2026
DOI: 10.1306/05012625009 ↗Kurz gefasst
Fachartikel in der Zeitschrift AAPG Bulletin, online erschienen am 30.06.2026. Verfasst von Tianjiao Liu und 7 weiteren Autor:innen.
Abstract (Originalsprache)
The Yongjin area is a significant hydrocarbon-producing region within the central Junggar Basin, surrounded by several hydrocarbon-generating sags. Understanding migration pathways in this setting requires accurate correlation between the oils found in the Upper Jurassic Qigu (J3q) Formation and the potential source rocks in adjacent sags. This task is complicated by the fact that many of source rocks have continued to be buried and continued to thermally mature since the time of oil expulsion. As source rocks reach high thermal maturity, common geochemical indicators such as biomarker ratios and absolute compound-specific carbon isotope values are significantly altered. These changes reduce their reliability for oil-source correlation. However, the shape or pattern of compound-specific carbon isotope curves can remain diagnostic of depositional environment, even as absolute values shift. In addition, inorganic geochemical parameters, including major, trace, and rare earth elements, are more resistant to thermal alteration and can retain useful environmental signatures. We hypothesize that combining the shape characteristics of organic isotope curves with stable inorganic geochemical ratios allows oils to be matched with highly mature source rocks. To test this, we applied a hybrid machine learning workflow using Constrained Matrix Factorization and Multi-Objective Particle Swarm Optimization to oil and source rock samples from the Yongjin area. The analysis identified the Middle Permian Wuerhe (P2w) Formation as the primary oil source. Crude oil samples from the northeastern part of the Yongjin area show a significant secondary contribution from Lower Permian Fengcheng (P1f). Oils sourced from the Jurassic and Middle Permian Xiazijie (P2x) are less significant. These findings support a two-phase charge model in the J3q Formation: an earlier episode derived mainly from the P1f, followed by a more widespread, later-stage charge from the P2w. This framework enhances our understanding of migration history and offers practical guidance for future exploration in superimposed basins with highly mature source rocks and mixed-source systems.
Zitieren
APA
Liu, T., Chang, X., Shi, B., Zhang, G., Wang, C., Xu, Y., Zhang, J., & Peng, Z. (2026). Unveiling genetic affinity between highly mature source rocks and crude oils: A machine learning approach driven by geochemical data. AAPG Bulletin. https://doi.org/10.1306/05012625009
Vancouver
Liu T, Chang X, Shi B, Zhang G, Wang C, Xu Y, et al. Unveiling genetic affinity between highly mature source rocks and crude oils: A machine learning approach driven by geochemical data. AAPG Bulletin. 2026. doi:10.1306/05012625009
BibTeX
@article{Liu2026,
author = {Liu, Tianjiao and Chang, Xiangchun and Shi, Bingbing and Zhang, Guanlong and Wang, Chenyu and Xu, Youde and Zhang, Junjian and Peng, Zhuanghao},
title = {Unveiling genetic affinity between highly mature source rocks and crude oils: A machine learning approach driven by geochemical data},
journal = {AAPG Bulletin},
year = {2026},
doi = {10.1306/05012625009}
}
EUVIMED