Optimization of an ammonia-water absorption refrigeration system, minimizing operating and annualized capital costs, is described. A mixed-integer nonlinear programming (MINLP) model, with the inclusion of discontinuous functions for capital costs for the main components of the system, is presented. The minimum total annualized cost is achieved with the simultaneous optimization of seven design variables. Practical limits of the design variables are established for the purpose of generating design parameters in accordance with the type of system studied. The model is applied to two examples. The analysis considers two types of heat rejection media: cooling water and air. The results obtained indicate that the selection of cooling medium is dependent on the refrigeration level that is required. The optimum configuration and operating conditions do not correspond to the best process integration and the coefficients of performance, and irreversibilities are high. The design variables with the highest impact on the objective function are dependent on the heat rejection medium characteristics and the process requirements. A sensitivity analysis of the effects of the evaluation criteria on the minimization of operating and capital costs is presented.