Abstract: Individual risks in evacuation under toxic gas diffusion conditions were quantified on the basis of the human vulnerability model. The alert communication and people transition laws were used to determine individual risks at different deployment ratios of alert receiving devices. A bi-objective mathematical model of deployment for alert receiving devices around natural gas wells in operation was built by the graph theory and multi-objective optimization method,and based on the greedy algorithm a practical algorithm was introduced to solve a better non-inferior solution of the model. Finally,taking the emergency warning project of a montanic gas well as an example,the border of deployment sub-regions and the corresponding deployment proportion were achieved by program computation. The results show that,for the same risk mitigation goals,this method has a better economy than the average deployment method in reducing deployment costs and relieving road congestion.