Control of systems with strong nonlinearities is impossible or difficult by common linear systems methods. Sometimes the system linearization around the working point can be a solution, but it is not always applicable. Other, better possibility, is using online system identification during the whole process and based on it the controller parameters changing. In this case, the control is closely dependent on the process identification quality. In this paper, the quality of the control process using different identification methods and their settings is observed. A nonlinear process model based on the exothermic semi-batch reactor is used and the whole process is simulated using data of real process. Results are discussed.