Abstract This paper focuses on the modelling and predictive control of a real medium temperature solar cooling system made up of a solar collector, an absorption chiller, a gas heater, hot and cold water storage tanks, connecting pipes, pumps, and a number of valves, allowing to modify the plant configuration in order to optimize its performance in terms of electric power and gas consumption. A detailed model of the plant is first obtained as a mix of first principle equations, validated with data collected on the plant, and identification techniques. Then, the model is simplified and used to formulate a hybrid Model Predictive Control (MPC) problem computing the optimal plant configuration over a prediction horizon of six hours, or more, and based on the available forecast of the solar radiation and user demand. The developed MPC algorithm has been used to control the plant; some of the collected results are reported here to witness the potentialities of the proposed approach.