The project focuses on the intelligent transformation of coal mining, aiming to solve key problems in the safe production and mining continuity of coal production enterprises. By integrating large language models, knowledge graphs, and the Internet of Things, a "Coal Science Guardian Large Model" is created. A three-tier system of large language models in the coal vertical field, business scenario agents, and intelligent operation and maintenance platforms is built. The project focuses on solving industry pain points such as equipment operation and maintenance relying on manual labor, lagging fault warnings, insufficient identification of safety hazards, and inefficient production planning and spare parts allocation. Practical application shows that the system reduces equipment information query time by 30%, improves operation and maintenance response efficiency by 40%, and increases fault warning accuracy by 30%, effectively improving safety production and operation and maintenance management levels.