Aimed at the problem of generator parameters dispersed under different disturbance scenarios, this article proposes a method to verify generator model parameters, which has taken the classification of disturbance scenes into account by combining probability and statistics theory. First, the obtained disturbance scenes are clustered and analyzed, and the target unit model parameters are checked via different disturbance scenarios under various disturbances. Then, based on the recognition dominant parameter and the gray distance measure, an interval for its value is given at a certain confidence level. Finally, because there are different disturbance types with their corresponding error intervals, different modes of generator parameters that provide a reasonable and sufficiently auxiliary basis for the selection and verification of simulation parameters are given. The method described in this article is verified by a large amount of measured information derived from random disturbances in the provincial power grid.