Unmanned aerial vehicle (UAV) cluster is a typical complex system. However, the reliability of the whole cluster is difficult to be directly evaluated and predicted for its loosely coupled structural characteristics and nonlinear correlation. For this issue, a Graphical Evaluation and Review Technique (GERT) network model of UAV cluster is constructed to describe the logic and transitive relation between the reliability state transition of UAV cluster and the reliability state transition of UAV system. Based on the topological structure of the UAV cluster, weight assignments are made according to the impact of each UAV system on the whole cluster in the GERT network. Combined with the technical achievements of Prognostics and Health Management (PHM), an application is realized where the reliability state prediction results of individual UAV systems are used to analyze the reliability state prediction of the UAV cluster. The simulation results indicate that the GERT network model can well reflect the loosely coupled structural characteristics of UAV cluster and achieve the association and conversion between the reliability state prediction of individual UAV systems and that of the UAV cluster.