Abstract: Aiming at the difficult problems of ventilation regulation and control caused by the synergistic action of spontaneous combustion risk and hydrological disturbance in the mining of extra thick coal seams, an intelligent ventilation optimization method based on multi-source data fusion is proposed, and an intelligent system architecture consisting of dynamic perception network, seepage-ventilation coupling model, and layered regulation and control algorithm is constructed. The system collects parameters such as oxygen concentration, water pressure, temperature and others in real-time through heterogeneous sensor clusters, couples modeling the seepage inhibition effect in the oxidation reaction process, and utilizes a rolling optimization mechanism to achieve multi-goal dynamic balance of spontaneous combustion risk, energy consumption, and hydrological stability. In the 6-month test of the 5106 working face in Ningwu, Shanxi Province, compared with the traditional ventilation scheme, the system shortens the spontaneous combustion early warning response time from an average of 42 minutes to 8 minutes, the ventilation electricity consumption per ton of coal is reduced from 0.218 kW·h/t to 0.167 kW·h/t, and the number of overlimit times the sealed wall water pressure is reduced from 12 times/month to 3 times/month. The research results provide theoretical support and engineering path for safe and efficient ventilation of water-richness coal seams, breaks through the difficult problem of response lagging of "regulating air to promote combustion" under the seepage disturbance condition.