It is difficult for managers to make scientific and extensible evaluation on metro station safety based on traditional mathematical statistical method induced by dynamic variation of fire safety indices of metro stations and coupling effect. Hence, a safety evaluation model for metro station in case of fire based on COWA operator and BP neural network is proposed. Firstly, a fire safety evaluation index system for metro station is established from the aspect of different stages of fire and considering the factors of personnel evacuation. And then the COWA operator is used to regroup the decisionmaking data of experts to weaken the influence of individual extreme value on the weight and improve the scientificity of empowerment. Finally, [JP2]the continuous adaptive learning and feedback capabilities of BP neural network are used to simulate the complex coupling of indices. The model is applied to the evaluation of fire safety of Zijinshan Station on Line 2 of Zhengzhou Metro; and the results show that the fire safety level of the station is high.