In the multi-target attack missions of Unmanned Combat Aerial Vehicle (UCAV) air combat, task priority sequencing serves as the foundation for attack decision-making and constitutes a crucial component of aerial combat assisted decision systems. This paper introduces a method for prioritizing multi-target attacks in air combat based on time-varying lethality performance. First, the concepts of time-varying lethality performance and fire field are introduced to characterize the dynamic fire damage characteristics of the formation. A Bayesian network for assessing the priority of attack tasks is constructed, which comprehensively considers not only target value and threat level but also incorporates the fire information and its dynamic damage characteristics. Subsequently, the states of each node in the network are analyzed, and a network condition probability table is established to prioritize the multi-target attacks in air combat. Finally, simulation analysis and comparison demonstrate that using this model for prioritization yields more reasonable results, which can enhance the effectiveness of future multi-UCAV air combat.