The identifiability problems of a simple power system can be solved by analytical approaches. But for large-scale power systems， it is very difficult. The relationship between the parameter identifiability and sensitivity is revealed， and the method is proposed， in which the parameter identifiability is judged by the phase angles of the sensitivities， and the key identifiable parameters are selected according to the magnitudes of the sensitivities. Finally the key identifiable parameters are optimized with ant colony algorithm. An example validates the feasibility of the above method.