The refrigeration sector has a fundamental and growing role in modern life. Its uses range from industrial processes and food preservation to applications of thermal comfort. Its economic weight size is reflected in the level of pollutant emissions of the segment. International agreements have been implemented with the objective of replacing refrigerant fluids that are harmful to the ozone layer or with a high global warming potential. These initiatives seek to mitigate the direct impact of the release of gases into the atmosphere. Other approaches aim to reduce the level of emissions by reducing the energy consumption of the operation of these systems.In this sense, this work investigates the implementation of a strategy based on predictive control for direct expansion systems. The objective was to verify if the control strategy based on this approach would be able to keep the internal temperature of an room within a defined range. The results show that the proposed control scheme was effective for the case of constant external temperatures, but ineffective in a dynamic context. Comparing the performance of this approach to the same system operating with a two velocities control scheme, the results were inferior. The predictive control strategy was able to keep the temperature within the desired range in 53 per cent of the simulation time versus 96 per cent of the two velocity aproach. In parallel, the application of the estimation method states in moving horizon was also studied. The objective was to discuss whether it was possible to infer parameters of the refrigeration cycle without actually measuring them. The results show that the refrigeration model proposed is observable under constant external conditions. In addition, a performance comparison between the system operating with different fluids was performed. The goal was to show the possibilities of this estimation method with respect to the calculation of performance metrics.