Agent-based combat modelling and simulation is an important method for UAV swarm surveillance effectiveness evaluation. The agent behaviour mechanisms and model structures in previous agent-based combat modelling and simulation studies are difficult to describe self-organized surveillance behaviours in complex mission scenarios. In this paper, taking a counter-piracy surveillance mission as an example, a combat modelling and simulation method oriented for swarm self-organized surveillance is proposed. By the hierarchical composite behaviour mechanism and the composable modular model structure, complex behaviours are described by the composition of simple behaviours, and the self-organized path planning and flight control are realized by the interaction of the pheromone map module and the path management module. Simulation experiments are conducted to demonstrate the feasibility of the proposed method, and compare the surveillance effectiveness with different path planning methods. The results show that, self-organized path planning is more adaptive to dynamic events interrupting surveillance. In large-area, long-duration and low-target-density area coverage surveillance missions, merely decreasing revisit intervals by improving path planning methods cannot significantly increase the probability of finding targets.