The development of a prediction model for the Unmanned Aerial Vehicle (UAV) accident rate is regarded as a crucial measure to improve the safety of UAV operations. The operational complexity of UAV system is derived from a multitude of influencing factors and their intricate coupling relationships in the processes of management, maintenance and operation. In addition, dynamic feedback characteristics are identified among these influencing factors. For instance, the safety awareness of operators is affected by the organizational commitment on safety, and the reverse effect also exists. A prediction model for the UAV accident rate based system dynamics is established from the perspective of Man-Technology-Organization (MTO) in this study. In the model, the coupling relationships and dynamic feedback effects among the operational, maintenance and management factors are fully considered. The accident rate of the MQ-1 Predator UAV is predicted by applying the established model, and the consistency between the predicted results and actual accident rate data verifies the rationality of the proposed prediction model. It is verified that the system dynamics method is more applicable to the modeling of UAV system with complex coupling relationships and dynamic feedback characteristics.