The paper investigates the use of non-linear model predictive control (NLMPC) as an advanced control strategy for the control of multi-stage flash (MSF) desalination plants, For this purpose, a comprehensive non-linear dynamic model was developed for the MSF plant and was validated with actual plant data. A control structure for the plant was then developed from which two control objectives were selected, One control objective is to maximize the distillate product and the other is to maximize the performance ratio while maintaining the manipulated variables in both cases under tight hard constraints. The NLMPC algorithm was then tested with these two objectives for both nominal and non-nominal model cases, In the non-nominal case the effects of parametric modeling error on the NLMPC performance were examined. Closed-loop simulations show that the advanced control strategy allows good control performance and substantial energy savings even when the plant is under the influence of unavoidable model uncertainties, The performances of the NLMPC algorithm were compared with a conventional proportional-integral (PI) controller.