To address the defect of only considering the node characteristics in previous comparisons of entire network based on set theory, we proposed a new method to compare metabolic networks by taking into account the differences of the properties of nodes in network topology. This method combinesd the centrality of seven topologies of the metabolic network, used principal component analysis to select principal component, and designed the sequence of the principal components method, and then used wavelet transform to study the metabolic network structure characteristic curve, and finally compared the fuzzy function metabolic networks of different species. Our method can analyze the similarity of metabolic network topology quantitatively. An analysis of the similarity data reveals the species -specificity of metabolic networks, and finds that the similarity of metabolic networks of species in the same evolution stage is much higher than that of species in different evolution stages, which provides a mathematical basis for the research of the evolution of metabolic network.