Abstract: To address the nonlinearity and time variability of gas accumulation in underground coal mines and the control lag of traditional centralized monitoring, this paper proposed a design scheme for an intelligent active gas prevention and control system in coal mines based on edge computing architecture and fluid migration mechanism. The system adopted a master-slave collaborative star network topology and logically divided the prevention and control tasks into two layers, namely the sensing edge and the decision-making edge. The sensing edge completed oversampling and digital filtering of gas sensor signals at the source end and directly output high-confidence gas concentration values, thereby eliminating interference noise caused by long-distance transmission of analog signals at the physical layer. In coordination with the Time Division Multiple Access (TDMA) communication protocol, the sensing edge transmitted valid data only within the allocated time-slot window, which significantly reduced the probability of channel congestion and the overall system power consumption. Unlike traditional rigid threshold-based on-off control, the decision-making edge was embedded with a fuzzy PID algorithm based on the gas migration mechanism. When an abnormal deviation in gas concentration or an abnormal rate of change was detected, the system predicted the trend of gas accumulation and dynamically adjusted the ventilation fan speed through high-frequency pulse-width modulation signals, thereby avoiding the uncertainty caused by long-distance communication. The simulation and prototype test results showed that, under a harsh radio-frequency environment with a 20% packet error rate, the deterministic interaction delay of control instructions was minimized to 0.5 s. Under the conditions of ±20% model mismatch in flow-field parameters and sudden gas outburst, the algorithm showed very strong convergence robustness, and the dynamic overshoot was strictly limited to within 3.4%. The introduced safety-constraint priority mechanism shortened the full-link physical safety disposal time from 72.3 s under the conventional strategy to 41.7 s, while effectively avoiding mechanical fatigue damage to the ventilation fan and achieving rapid active defense against the risk of gas accumulation and smooth recovery.