The invention discloses a method for predicting mixture critical properties on the basis of a group contribution theory. The method comprises the following steps that: through a pure-quality molecularstructure, dividing a first horizontal group library and a second horizontal group library; utilizing the experiment data of the pure-quality critical properties, carrying out fitting to obtain the group contribution value of each group for the critical properties; putting forward an association model between the mixture critical properties and the group contribution value, searching the criticalproperty experiment data of a binary mixture, and utilizing the experiment data of each category of system critical property to carry out fitting to obtain a corresponding mixture model parameter; and after the group contribution value and the mixture model parameter are obtained, predicting multicomponent mixture critical properties. The method has the advantages that pure-quality critical volumetric data is not required to serve as a basis, and in addition, the mixture critical properties can be accurately predicted under a situation that the mixture critical properties are unknown. When the critical temperature and the critical pressure are predicted, the group contribution values of the critical temperature and the critical pressure are independently used, the group contribution valueof critical volume is not required, and geometric isomers can be distinguished.