The paper presents a design method for robust multiva-riable control design and its application to retrofit a new additional control system in an old pressurized water reactor (PWR) nuclear power plant (NPP). The mathematical model of PWR NPP, used in the design process is briefly described. The 28th dimensional state equation model accounts for the main physical processes and plant subsystems. The optimal and suboptimal linear quadratic gain (LQG) additional control, which provides improved integration of the existing automatic feedback control loops is described and dynamical behaviour of the plant is simulated. Because of the process parameters variations, the design of a robust multivariable control system is considered and a new method which leads to improvement of system robustness is proposed. The method, based on LQG control improves the response of the existing control loops and increases the closed loop system robustness. The dynamical behaviour of NPP Krsko, Slovenia, for specified design plant perturbations is presented and the use of different control systems is compared.