【Purpose/significance】 To study the endogenous relationship between the topological structure of Poisson network and the

evolution of public opinion, and use the strength of the connection as a controlling factor to investigate the influence on the two, which is helpful to better understand the internal mechanism of the evolution of public opinion.【Method/process】 Based on the opinion dy⁃namics Deffuant model, it is improved. By introducing the relationship strength and constructing a mathematical model at the micro level, the interaction between the evolution of public opinion and the network structure in the evolution process is discussed. Using the ABM research paradigm, a weighted Poisson network is constructed on the Repast platform based on Java to track the simulation re⁃sults under different parameter combinations and analyze the evolution process.【Result/conclusion】 The bounded trust range plays a decisive role in the formation of the number of opinion clusters. The dynamic changes of the network structure triggered by the relation⁃ship strength and opinion value will not have a subversive impact on the final result of the evolution of public opinion, but only change speed and rhythm of the evolution.【Innovation/limitation】 A dynamic Poisson network is constructed from the perspective of relation⁃ship strength, and the results of public opinion evolution are observed on this basis. The description of individuals and public opinion networks is not sufficient, and individual heterogeneity (psychology, memory, etc.), opinion leaders, community structure and other fac⁃tors can be integrated to improve the fitting degree with the real network.