A model-driven design and optimization methodology for sizing the components of refrigeration cassettes for light commercial applications (i.e., cooling capacities ranging from 0.5 to 1.5 kW) is presented. Mathematical models were devised for each of the system components and their numerical results were compared with experimental data taken with different cassettes. It was found that the model predictions for the working pressures, power consumption, cooling capacity and coefficient of performance (COP) showed maximum deviations of +/- 10%. A genetic optimization algorithm was used to design the condenser and evaporator and also to select the compressor model based on an objective function which considers both the COP and cost. The optimization led to two improved cassette designs, which were assembled and tested. One of the optimized cassettes showed a COP/cost ratio approximately 50% higher than that of the baseline system. (C) 2011 Elsevier Ltd. All rights reserved.