面向AI4S的多圈层油气富集知识底座建设方法与应用实践
宋东桓 , 常志军 , 张旺 , 喻志超 , 邱若原 , 汪文洋 , 钱力
地学前缘 ›› 2026, Vol. 33 ›› Issue (6) : 392 -406.
面向AI4S的多圈层油气富集知识底座建设方法与应用实践
Construction and application of the AI4S knowledge foundation for multi-sphere hydrocarbon accumulation
多圈层耦合作用下的油气富集机理研究,面临数据高维异构、因果关系非线性以及领域知识难以有效融入计算模型的挑战。为突破上述瓶颈,本研究提出一套面向AI4S的多圈层油气富集知识底座系统化构建方法。该方法特别强调从地球系统视角补充关键含油气系统要素,以总有机碳及其相关的有机地球化学、微量元素与同位素指标为切入点,通过建立TOC专题数据库,强化了对烃源岩有机质丰度、沉积环境与古气候条件的综合表达。整体架构上,设计了一个以领域知识本体为语义基石,耦合标准化数据库的逻辑架构,为融合异构数据与约束智能推理提供统一范式。在此基础上,研发一套贯通数据全生命周期的智能化工具体系,实现从多模态原始文献与数据中进行细粒度知识抽取、本体驱动的语义标准化与多层次自动化质量控制。成效验证表明,该方法论构建知识的效率相较于传统人工方式提升了约6倍,且对图件库、观点库等复杂语义知识的自动化抽取质量具有高可靠性。并以此为基础,成功构建了一个包含11.1万条TOC专题数据、1.2万幅图件及2 400余条核心观点的原型平台,实现了智能问答与证据溯源、视觉知识类比及多维数据遴选等应用功能。研究表明,该方法体系为解决油气地质领域知识的碎片化与应用困境提供有效路径,为实现知识与数据双重驱动的科研新范式奠定了坚实基础。
Research on hydrocarbon accumulation mechanisms under multiphase coupling effects faces challenges including high-dimensional heterogeneous data, nonlinear causality, and difficulties in effectively integrating domain knowledge into computational models. To overcome these bottlenecks, this study proposes a systematic method for constructing a multispheric hydrocarbon accumulation knowledge base for AI4S. The method emphasizes augmenting key petroleum system elements from an Earth system perspective, using Total Organic Carbon (TOC) along with related organic geochemical, trace element, and isotopic indicators as a starting point. By building a TOC-themed database, it enhances the comprehensive characterization of source rock organic matter abundance, sedimentary environment, and paleoclimate conditions. The overall system follows a logical architecture that employs a domain-knowledge ontology as its semantic cornerstone, coupled with a standardized database. This provides a unified paradigm for integrating heterogeneous data and constraining intelligent reasoning. Subsequently, an intelligent tool system spanning the entire data lifecycle was developed. This system enables fine-grained knowledge extraction from multi-modal literature and data, ontology-driven semantic standardization, and multi-level automated quality control. Validation results indicate not only an approximately sixfold improvement in knowledge construction efficiency compared to traditional manual methods but also high reliability in the automated extraction of complex semantic knowledge (e.g., for geological figures and viewpoints). Based on this, a prototype platform was successfully built, containing 111000 TOC-themed data records, 12000 geological figures, and over 2400 core viewpoints. The platform supports applications including intelligent Q&A with evidence tracing, visual knowledge analytics, and multi-dimensional data selection. This study demonstrates that the proposed methodological framework offers an effective solution to the fragmentation and application challenges of knowledge in petroleum geology. It lays a solid foundation for a new, dual-driven scientific research paradigm powered by both knowledge and data.
| [1] |
|
| [2] |
朱日祥, 张旺, 王华建, |
| [3] |
周圣荃, 李以科, 王永志, |
| [4] |
刘超, 赵汀, 王安建, |
| [5] |
刘泽垣, 王鹏江, 宋晓斌, |
| [6] |
人工智能全球治理行动计划[EB/OL]. [2025-08-10]. https://www.mfa.gov.cn/zyxw/202507/t20250726_11677803.shtml |
| [7] |
AMERICA’S AI ACTION PLAN[EB/OL]. [2025-08-10]. https://www.whitehouse.gov/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf |
| [8] |
The AI Continent Action Plan[EB/OL]. [2025-08-10]. https://ec.europa.eu/newsroom/dae/redirection/document/114523 |
| [9] |
钱力, 张智雄, 伍大勇, |
| [10] |
|
| [11] |
李白杨. AI4S 背景下的图书馆数字学术服务创新发展[J]. 农业图书情报学报, 2023, 35(10): 19-31. |
| [12] |
周力虹. 面向驱动AI4S的科学数据聚合: 需求、 挑战与实现路径[J]. 农业图书情报学报, 2023, 35(10): 13-15. |
| [13] |
What is AI-Ready Open Data?[EB/OL]. [2025-06-20]. https://www.star.nesdis.noaa.gov/star/documents/meetings/2020AI/presentations/202010/20201022_Christensen.pdf |
| [14] |
|
| [15] |
专家热议Al4S数据库与知识库: 构建创新基座促进建立开放共享数据生态[EB/OL]. [2025-04-23]. http://www.xinhuanet.com/tech/20230812/6aa67e63ec8d4f52ae7f14e3156bc3f7/c.html |
| [16] |
孙坦. 支撑AI4S的智能知识服务: 需求、 趋势与任务[J]. 农业图书情报学报, 2023, 35(10): 6-9. |
| [17] |
张智雄. 切实发挥科技文献在AI4S中的知识底座作用[J]. 农业图书情报学报, 2023, 35(10): 9-13. |
| [18] |
|
| [19] |
|
| [20] |
|
| [21] |
Earthchem[DB/OL]. [2025-10-10]. https://ecl.earthchem.org/search_results.php?offset=100&a1=&e5=&g7=&h8=&h9= |
| [22] |
GEOROC[DB/OL]. [2025-10-10]. https://data.goettingen-research-online.de/dataverse/pre_0686438?q=&types=datasets&sort=dateSort&order=desc&page=1 |
| [23] |
SGP[DB/OL]. [2025-10-10]. https://sgp.stanford.edu |
| [24] |
USGS[EB/OL]. [2025-10-10]. https://www.usgs.gov/products/data/all-data |
| [25] |
GEOSCIENCE AUSTRALIA[DB/OL]. [2025-10-10]. https://ecat.ga.gov.au/geonetwork/srv/eng/catalog.search#/search?from=1&to=20 |
| [26] |
PANGAEA[DB/OL]. [2025-10-10]. https://pangaea.de/?t=Lithosphere |
| [27] |
ISTAR[DB/OL]. [2025-08-10]. https://www.halliburton.com/en/products/istar-intelligent-drilling-and-logging-platform |
| [28] |
GEOVERSE[DB/OL]. [2025-08-10]. https://www.viridiengroup.com/sites/default/files/2021-06/21DB-GL-250-V1_GeoVerse_Overview_1.pdf |
| [29] |
GEOMARK[DB/OL]. [2025-08-10]. https://www.geomarkresearch.com/ |
| [30] |
|
| [31] |
丁勇, 林静文, 潘泉涌. 塔里木盆地塔北地区三叠系烃源岩特征及展布[J]. 地质科学, 2024, 59(5): 1198-1211. |
国家重点研发计划项目(2023YFF0725600)
国家社科基金项目(24BTQ043)
地球多圈层相互作用的油气富集理论课题(THEMSIE04010101)
/
| 〈 |
|
〉 |