Abstract: The complex environment underground in coal mines imposes strict requirements on wireless communication networks for high reliability, low latency, and differentiated services. This paper proposes a coal mine underground self-organizing network method based on multi priority dynamic access. Firstly, data frames are divided into different priority queues based on the level of business urgency, and a hierarchical M/G/1 queuing model is established to characterize the randomness of data arrival and differences in service time. Secondly, design a dynamic access mechanism based on threshold control, by calculating the lowest and highest priority thresholds and dynamically adjusting the intermediate priority threshold in combination with the business ratio, to achieve balanced optimization of network load and transmission success rate. A dynamic priority scheduling strategy is introduced to trigger preemptive transmission in the event of sudden high priority business（such as gas over limit alarms, collision avoidance instructions）. At the same time, a joint compensation mechanism of service counting and deadline is used to avoid long-term retention of low priority data. The experimental results show that this method can ensure a stable transmission success rate of over 99.4% for high priority data in various scenarios under high interference and multi hop self-organizing network environments in coal mines, with an average delay of less than 52 ms. In the event of an emergency situation of sudden gas exceeding the limit, it can achieve a 100% transmission success rate with an average delay as low as 37 ms. At the same time, this method effectively alleviates the phenomenon of multipath clustering in coal mines, improves channel utilization by 40%, and increases spectral efficiency by 0.6 bit/s/Hz.