Based on the complex network theory, this study constructs a causal chain identification model of coal mine gas explosion, and reveals the dynamic coupling mechanism between key causal nodes and propagation paths. By improving the PageRank algorithm, the node importance ranking method is proposed, and the key nodes such as local fan shutdown with an influence value of 0.89 are identified in the case of Shanxi high gas mine. A three-level early warning system including the path vulnerability index is constructed. The empirical results show that the red early warning response time ≤15 min can reduce the accident probability by 72%. The proposed intelligent power-off device and wind-water linkage explosion-proof facility improve the control efficiency of high-risk working surface by more than 40%, providing a new paradigm for coal mine safety prevention and control.