With the development of network services, social networking has become a new medium of information dissemination. Thus, many scholars pay close attention to the evolution of network public opinion which has great significance and practical value. In this paper, an opinion evolution model of social network based on information entropy is proposed, which is helpful to investigate the evolution of network public opinion. The model has two characteristics: firstly, it can reflect the psychological process when an individual makes a choice facing with two opposite views; secondly, it can reflect the influence of subjective and objective factors when the individual forms new ideas. In the simulation, some issues are discussed which includes the effects of the public opinion environment on the individual opinion evolution, the influences of the initial opinion and self-confidence on the individual opinion evolution, and the effects of leaders' opinions on the opinion evolution of populations. From the experimental results it follows that the model works well and can reflect the characteristics of individuals in real social networks. For example, the formation of the individual's opinion will be affected by public opinion environment, and an individual will not accept the opinions from others if he is very confident, and the opinion evolution of populations will be influenced when there exist the leader's opinions in network.