The evolution of industrial cluster Is a complex dynamic process. In the paper, a complex adaptive system (CAS) model with the virtual multi-agents of industrial duster If constructed. Based on the analysis of the mechanism of the Industrial cluster, the properties of the agents, and rules of the agents' moving, producing, cooperating and Interacting with the environments are described. Furtherer, the model is implemented on the Swarm platform and how the multi-agents behave to result In the emergence of the macro-economy behaviors id explained. Meanwhile, the simulation results show that the cooperative relation network has a series of characters of the scale-free network, so that the theory of scale-free network can be used to analyze the properties of Industrial cluster. This phenomenon is coincided with the real industrial cluster, so the model is effective to guide the decision-making of industrial cluster.