With the steady increase of combined cycle units， their influence on the electric power system stability is being given increasingly great attention. It is of important practical significance to develop a mathematical model in conformity with their complicated nonlinear characteristics and easily accessible through the test of the parameters. Based on the classical nonlinear model of combined cycle units， this paper presents a new nonlinear model for testing and identifying the combined cycle units. The improved parameter identifying method is proposed based on the genetic algorithm and applied to a 400 MW combined cycle unit. The feasibility and effectiveness of the model and the method proposed are proved.