Research and empirical validation of the application modes of multimodal large language models for coal mines

Authors

• WANG Guofa, State Key Laboratory of Digital Intelligent Technology for Unmanned Coal Minings, Beijing 100013, ChinaFollow
• WANG Jingyi, Yunding Technology Co., Ltd., Ji’nan 250014, China
• LIU Shuang, Yunding Technology Co., Ltd., Ji’nan 250014, China
• XU Zhe, Yunding Technology Co., Ltd., Ji’nan 250014, China
• LIU Mingrui, Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093, China;School of Cyber Security, University of Chinese Academy of Sciences, Beijing 101408, China

Abstract

Objective Under the technological backdrop of large language models (LLMs), big data, and knowledge fusion, multimodal LLMs have found rapid application across traditional industries, including coal mining. However, the successful implementation and sustained operation of these models remain challenging. To address issues concerning the implementation and usage of multimodal LLMs in the construction of intelligent coal mines, this study conducted a multi-dimensional survey of multimodal LLM applications in coal mines, aiming to establish an LLM application paradigm that is suitable for the current management status and organizational structures for safe coal mining. Method Using content analysis, case analysis, and empirical survey, this study sorted out and analyzed the LLM projects of representative coal mining groups. By combining an on-site empirical study, this study highly generalized the large-scale application paradigms of multimodal LLMs in coal mining enterprises, especially those with a three-tier organizational structure consisting of a corporate group, subsidiaries, and production mines. Consequently, a specific semantic space was determined, under which a large-scale application paradigm based on heterogeneous complex networks was developed. Finally, the proposed paradigm was empirically validated. Results and Conclusion The results indicate that the large-scale application paradigm of multimodal LLMs for coal mines should consist of five major elements: scenarios (S), tools (T), actors (A), resources (R), and policies (P). These elements are interrelated and can be individually classified across multiple dimensions based on their own characteristics. The multidimensional nature of both these five elements and their relationships determines the multidimensional architecture of the STARP paradigm. Specifically, the STARP paradigm represents a high-dimensional space defined by S, constrained by P, and characterized by interactions among the remaining three elements—T, A, and R. The implementation and application of multimodal LLMs comprise a multi-element fusion system integrating technical, managerial, service & decision-making, and social attributes, while technological or engineering advantages alone are insufficient to ensure the successful operation of related projects. The proposed paradigm in this study can provide a reference for the implementation and application of multimodal LLMs for coal mines.

Keywords

intelligent coal mine, multimodal, large language model (LLM), artificial intelligence (AI) application, model paradigm

DOI

10.12363/issn.1001-1986.25.11.0864

Recommended Citation

WANG Guofa, WANG Jingyi, LIU Shuang, et al. (2026) "Research and empirical validation of the application modes of multimodal large language models for coal mines," Coal Geology & Exploration: Vol. 54: Iss. 6, Article 3.
DOI: 10.12363/issn.1001-1986.25.11.0864
Available at: https://cge.researchcommons.org/journal/vol54/iss6/3

Reference

[1] 国家矿山安全监察局，财政部. 关于印发《煤矿及重点非煤矿山重大灾害风险防控建设工作总体方案》的通知[EB/OL]. (2022-11-02) [2025-10-22]. https://www.chinamine–safety.gov.cn/zfxxgk/fdzdgknr/tzgg/202211/t20221102_425584.shtml.

[2] 国家矿山安全监察局，应急管理部，国家发展和改革委员会，等. 关于印发《关于深入推进矿山智能化建设促进矿山安全发展的指导意见》的通知[EB/OL]. (2024-04-24) [2025-10-22]. https://www.gov.cn/zhengce/zhengceku/202404/content_6948041.htm.

[3] 戴春雷. 我国矿山智能化研究进展及大模型应用前景[J]. 工矿自动化,2024,50(增刊2):1−11 DAI Chunlei. Progress of intelligent mine research and the prospects for mine large model applications in China[J]. Journal of Mine Automation,2024,50(Sup.2):1−11

[4] WIGGERS K，CORRALL C，STRINGER A. ChatGPT：Everything released from the AI–powered chatbot in 2022[EB/OL]. (2022-12-31) [2025-10-22]. https://techcrunch.com/2022/12/31/chatgpt–everything–released–from–the–ai–powered–chatbot–in–2022/.

[5] KAPLAN J，MCCANDLISH S，HENIGHAN T，et al. Scaling laws for neural language models[EB/OL]. 2020：arXiv：2001. 08361. https://arxiv.org/abs/2001.08361.

[6] DEEPSEEK–AI，GUO Daya，YANG Dejian，et al. DeepSeek–R1：Incentivizing reasoning capability in LLMs via reinforcement learning[EB/OL]. 2025：arXiv：2501. 12948. https://arxiv.org/abs/2501.12948.

[7] DEEPSEEK–AI，LIU Aixin，FENG Bei，et al. DeepSeek–V3 technical report[EB/OL]. 2024：arXiv：2412. 19437. https://arxiv.org/abs/2412.19437.

[8] TEAM K，BAI Yifan，BAO Yiping，et al. Kimi K2：Open agentic intelligence[EB/OL]. 2025：arXiv：2507. 20534. https://arxiv.org/abs/2507.20534.

[9] TEAM G，ANIL R，BORGEAUD S，et al. Gemini：A family of highly capable multimodal models[EB/OL]. 2023：arXiv：2312. 11805. https://arxiv.org/abs/2312.11805.

[10] TEAM G，GEORGIEV P，LEI V I，et al. Gemini 1. 5：Unlocking multimodal understanding across millions of tokens of context[EB/OL]. 2024：arXiv：2403. 05530. https://arxiv.org/abs/2403.05530.

[11] TOUVRON H，LAVRIL T，IZACARD G，et al. LLaMA：Open and efficient foundation language models[EB/OL]. 2023：arXiv：2302. 13971. https：//arxiv. org/abs/2302. 13971.

[12] 王海军. 矿山行业大模型建设路径探索与应用展望[J]. 煤炭科学技术,2024,52(11):45−59 WANG Haijun. Construction exploration and application prospect of the large model in mining industry[J]. Coal Science and Technology,2024,52(11):45−59

[13] 王国法,任怀伟,赵国瑞,等. 智能化煤矿数据模型及复杂巨系统耦合技术体系[J]. 煤炭学报,2022,47(1):61−74 WANG Guofa,REN Huaiwei,ZHAO Guorui,et al. Digital model and giant system coupling technology system of smart coal mine[J]. Journal of China Coal Society,2022,47(1):61−74

[14] 新华网. 赋能煤矿人工智能建设“太阳石矿山大模型”正式发布[EB/OL]. (2024-04-08) [2024-08-12]. http://www3.xinhuanet.com/energy/20240408/a1c6d90938594471aeb51be73150421a/c.html.

[15] 讯飞工业大模型(矿山版)震撼煤炭界[EB/OL]. (2023-09-28) [2024-08-12]. https://mp.zgkyb.com/m/news/88341.

[16] 中国电信. 开源星辰AI 大模型：央企中首个完成LLM 研发和开源的选手诞生[EB/OL]. (2024-01-17) [2024-08-12]. https://cloud.tencent.com/developer/article/2380117.

[17] 新华网. 驶向“AI星辰大海”中国电信发布行业大模型[EB/OL]. (2023-12-14) [2025-10-22]. https://www.news.cn/info/20231214/d0ba9927dbe14c728b3856026e1985fb/c.html.

[18] 李远. 中科慧拓发布愚公YUKON矿山大模型[N]. 中国能源报，2024-06-10(16).

[19] SUN Yizhou,HAN Jiawei,YAN Xifeng,et al. PathSim:Meta path–based top–K similarity search in heterogeneous information networks[J]. Proceedings of the VLDB Endowment,2011,4(11):992−1003.

[20] SHI Chuan,LI Yitong,ZHANG Jiawei,et al. A survey of heterogeneous information network analysis[J]. IEEE Transactions on Knowledge and Data Engineering,2017,29(1):17−37.

[21] ZHOU Dengyong，HUANG Jiayuan，SCHÖLKOPF B. Learning with hypergraphs：Clustering，classification，and embedding[M]//SCHÖLKOPF B，PLATT J，HOFFMAN T. Advances in neural information processing systems 19. Cambridge：The MIT Press，2007：1601–1608.

[22] GAO Yue,ZHANG Zizhao,LIN Haojie,et al. Hypergraph learning:Methods and practices[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2022,44(5):2548−2566.