This paper presents new conceptions of the computer package for tuning optimization of power unit control loops.The proposed tuning consists of two main stages: the optimal model identification of a process and a controller parameters optimiaation. In turn, the suggested conceptions for the identification are based on the two-stage approach: the erst stage is the frequency response identification of a process, while the second one is the transfer function computation using frequency response so obtained. In the frame of the approach, optimization criteria are introduced and model accuracy is estimated by a correct controller tuning on the second stage. Additionally, this identification approach results in a calculation of process uncertainties. New optimal criteria for the controller parameters optimization are proposed considering two kinds of process non-linearities and passibility of the robust stability analysis.Based on the above conceptions the program package is developed in the Matlah environment and successfully applied for control loops tuning of power units at the Israel Electric Corporation.