Abstract: Addressing the challenges of multi-source data fusion and poor real-time early warning capabilities in traditional coal mine safety risk monitoring methods, which result in unsatisfactory performance in real-time risk warning and dynamic response, this paper proposes a coal mine safety risk management method based on knowledge graph. The construction of the coal mine safety risk management knowledge graph included key processes such as risk knowledge acquisition, dynamic hazard extraction, and dynamic risk management. Risk knowledge acquisition: Potential risks were identified through various standardized methods. A structured ontology model was built using languages such as OWL, and risk point instances and their attributes were entered into the enterprise risk knowledge graph to form a semantic network, laying a foundation for intelligent risk assessment and precise management. Dynamic hazard extraction: Multimodal data collected from different data sources were associated in real time with risk point instances in the knowledge graph, and the status of risk points was updated according to preset algorithms and rules. Dynamic risk management: For identified hazards, instant inference was realized through reasoning rules written in Semantic Web Rule Language (SWRL). Practical application results showed that this method could accurately and rapidly identify potential hazards in the production environment, effectively enhancing risk identification and early warning capabilities, and providing systematic support for coal mine safety management.