Abstract: Accurate coal mine disaster risk identification is essential for underground personnel safety and mine safe production. Addressing low early warning data utilization and low fusion identification accuracy of gas, roof, water inrush, fire and rock burst in domestic coal mines, an intelligent multi-disaster risk fusion early warning technology is proposed based on large model and GIS spatial analysis coupled with big data, cloud computing and artificial intelligence. A standardized multi-source heterogeneous disaster data governance scheme and supporting disaster data middle platform are developed to realize orderly data management and on-demand scenario-based sharing. A mechanism-mathematics dual driven risk identification index system is established with multimodal fusion and deep learning, matching supporting risk prediction small models and cause analysis decision-making large models. A risk-embedded 3D geological digital twin is built to couple monitoring data and simulation results and visualize disaster risk dynamic evolution. The integrated intelligent early warning system classifies risks into four grades with accurate cause positioning and targeted control measures. Field tests in Shanxi and Shaanxi coal mines show that the system realizes rapid multi-disaster risk identification within seconds and reduces hidden danger-induced disasters. In high gas and strong ground pressure coupled scenarios, the large model optimizes coal-rock fracture strength indicators to predict gas emission, with abnormal gas risk identification accuracy up to 90.56%. The technology strengthens complex-condition multi-disaster early warning capacity and provides solid support for coal mine intrinsic safety production.