| 数字盆地多模态大模型关键技术研究——以鄂尔多斯盆地为例 |
| 投稿时间：2025-12-25 修订日期：2026-04-20 |
| 关键词:Multimodal Large Model Digital Basin Ordos Basin Intelligent Geological Analysis |
| 基金项目:鄂尔多斯盆地区域地质调查(编号：DD202401021)；中国石油大学(北京)科研基金(编号：2462024YJRC013) |
|
| 摘要点击次数: 278 |
| 全文下载次数: 0 |
| 摘要: |
| Research on Key Technologies of Multimodal Large Models for Digital Basins: A Case Study of the Ordos Basin |
| To address the low utilization of multi-source heterogeneous data and the lack of intelligent analysis tools in the construction of the Ordos Digital Basin, this paper develops a domain-specific multi-modal large model. The model utilizes the Contrastive Language-Image Pre-training (CLIP) architecture to achieve cross-modal feature alignment among geological maps, well logs, and text reports. By integrating Retrieval-Augmented Generation (RAG) and Chain-of-Thought (CoT) reasoning into a geoscience-specific language foundation, the model establishes a "retrieval-reasoning-generation" workflow. This approach enhances the reliability and reasoning depth of professional geological knowledge. Based on this framework, the system implements three key applications: intelligent parsing of geological reports, joint lithofacies-provenance analysis, and automatic recommendation of geological modeling parameters. Validation results demonstrate that the model enables a leap from massive data retrieval to deep knowledge reasoning. It significantly improves the efficiency of comprehensive geological research and provides technical support for the transition of digital basins from "data integration" to "intelligent services". |
|
| 查看/发表评论 下载PDF阅读器 |