Abstract : An accurate control of fluids flow is essential in any refrigeration system. As the conventional Proportional-Integral-Derivative (PID) control with invariable parameters can lead to unsatisfactory performance because of the variation of refrigeration unit parameters under disturbances, the aim of the work presented here is to develop a method for using Predictive Functional Control (PFC) to regulate the evaporator superheat, the condensing pressure and the cooling capacity on a variable-speed refrigeration system. Using a predictive controller requires to predic future change at the output of the process. This prediction is based on an internal model used as known model. By assuming that the behaviour of refrigerating machine heat exchangers can be represented using a first-order model, the implementation of PFC requires only three parameters: gain, time constant and time delay. In order to determine these parameters an original method has been developed which is based on the physical modelling of the machine. Physical models for different types of evaporator, condenser, compressor or expansion valve have been established to quantify heat transfer and refrigerant flow rate in these components. The control system created has been incorporated into an industrial programmable logic controller and used for experiments on two different refrigerating machines: the first one is composed of two shell and tube heat exchangers and a reciprocating compressor, whereas the second one is composed of a plate evaporator, a finned-tube condenser and a screw compressor. The tests performed show that PFC controller succeed in maintaining a precise chilled liquid temperature.