Abstract: Based on evolutionary games theory on complex networks, from the perspective of clusters' microscopic heterogeneous agents, considering the characteristics of agents' expected learning behavior and adaptive behavior, a traditional industrial low-carbon evolution model by using the random game, learning model and multi-agent system simulation methods are established. The scenario simulations about low-carbon strategy competition,emergence and evolution are carried out. The results show that structure of complex network, heterogeneous expectation and decision behavior of agents are significant to low-carbon strategy's emission. The key factors to spread the new strategy are heterogeneity agents' low-carbon preference, network' external effect, cost and technological compatibility of strategy.

Key words: traditional industrial cluster, complex networks, evolutionary game, low carbon evolution