Hydrates in porous media are in a variety of microoccurrence types, which affect the flow of fluid. In this study, multiphase digital cores containing particles, pores, and hydrates were reconstructed based on CT scan data of hydrate-containing cores. The hydrate network was extracted by the watershed method and the maximum ball method. Then, a parameter set was established to characterize the topological characteristics of hydrates. The self-organizing mapping (SOM) neural network was used to classify the hydrate types in the cores: the pore-floating hydrate accounted for 22.8%, the bridging hydrate accounted for 49.8%, and the cementing hydrate accounted for 27.4%. Furthermore, the influence of different hydrate micro-occurrence types on porosity, flow rate, and permeability was studied by flow simulation. It revealed the fluid flow characteristics in hydrate-containing cores: the existence of hydrates caused the large pores to be divided into small pores or blocked, which resulted in the reduction of the porosity, flow volume, and permeability. Bridging hydrates have the most obvious effect on reducing porosity and permeability, followed by pore-floating and cementing hydrates.