In this paper, nonlinear identification of the excitation system (EXS) in the gas unit #2 of Rajaee power plant in Iran is presented. Two methods of modeling, i.e., grey-box and black-box modeling are used and compared. In the grey-box (classical) approach, first a block-diagram for the EXS is suggested, then a test procedure for identification of its parameters is outlined. The input-output data corresponding to each block of the system is obtained through field tests. In this approach, the only nonlinearities considered in the block diagram are the limits. The other nonlinearities are reflected in the change of parameters in the linear transfer functions at different operating conditions. In the black-box approach, identification of the system is carried out using discrete wavelet transform. A variable structure wavelet was tried to cope with system changes at different operating conditions, but a wavelet with fixed number of scaling functions or wavelet functions proved to be quite adequate. The simulation results and their comparison, show the good accuracy of both derived models. Although the model obtained through the black-box approach shows a better fit when its output is compared with the measured variables, the model obtained through the grey-box approach reflects the physical properties of the system and may be more useful for power engineers.