In response to the strong suddenness of underground coal mine disasters and the difficulty of traditional static disaster avoidance path planning methods in adapting to dynamic disaster environments, this study proposes a real-time generation technology for dynamic disaster avoidance paths in coal mines based on multi-source data fusion, using the intelligent ventilation system of Sandaogou coal mine as the engineering background. By constructing a topology model of the mine ventilation network and integrating multi-source heterogeneous data such as gas concentration, wind speed, temperature, and personnel positioning information, a dynamically updated underground disaster evolution model is established; Based on the improved Dijkstra algorithm, the method of calculating the influence coefficient and equivalent length of roadway traffic is introduced to achieve dynamic optimization of disaster avoidance paths; The development of a real-time calculation module linked with the intelligent ventilation system ensures that the disaster avoidance path is dynamically updated as the disaster evolves. Engineering applications have shown that this technology can shorten the response time of path planning to within 30 seconds; In typical disaster scenarios such as gas exceeding limits, the accuracy of generating the optimal disaster avoidance path reached 98.7%, significantly improving the timeliness and reliability of emergency disaster avoidance in mines. This provides an innovative dynamic disaster avoidance solution for the construction of intelligent coal mines and has important engineering application value.