In this paper, by considering the collaborative scientific behaviors, a model for evolving hyper-network is proposed based on the hypergraph theory. We study analytically the dynamics of authors’ hyper-degree using mean-field approach. Results of both theoretical and numerical studies indicate that such a hyper-degree distribution follows a power-law decay, with the exponent γ=1+L/M, where L/M suggests the author growth rate of an evolving scientific network. In addition, empirical studies on ten-year journal data of Science China and Acta Physica Sinica from 2003 to 2012, shows that the present model can find good agreement with the empirical data. The proposed model may shed some light on the in-depth understanding of the structure and scientific collaboration networks.