In view of the contradiction between the agility of technical iteration and the lag of system renewal in the emergency system during the intelligent transformation of coal mines, an effective coordination mechanism was constructed to improve the efficiency of safety governance. Specifically, the “technology-system collaborative threshold model” was proposed, and on this basis, a multi-source data fusion framework and a “technology-system-insurance” ternary collaborative mechanism were constructed as follows. At the technology level, the federal learning was used to dynamically optimize the data weight and the OPC UA protocol unified semantics, which effectively solved the heterogeneous problem of downhole sensor protocol. At the system level, the dynamic adaptation rules of AI authority and responsibility were designed. At the insurance level, the linkage mechanism of algorithm filing and liability insurance was pioneered to systematically fill the gap of intelligent emergency system. Empirical studies demonstrate that at Yujialiang Mine, multi-source data fusion can increase operational efficiency by 80％ and reduce the false alarm rate for gas emissions to 4.3％. Meanwhile, at Hongqinghe Mine, introducing an optimization mechanism with an environmental priority coefficient, can achieve a 60％ reduction in ventilation response time. In summary, the constructed coordination threshold model and its implementation mechanisms provide a crucial quantitative tool for transforming coal mine safety governance from a passive-reactive approach to a proactive-defensive one. Furthermore, it establishes an innovative paradigm of “technology-institution-ecology” synergistic development, thereby advancing the construction of green mines.