Cooling load calculation is the basis for the design of building cooling systems. The current design Methods are usually based on deterministic cooling loads, which are obtained by using design parameters or information. However, these parameters contain uncertainties and they will be different from that used in the design calculation when the cooling system is put in use. The actual cooling load profile will deviate from that predicted in design. By considering uncertainties of these parameters, the sizing and configuration of the cooling system can be improved. In this study, a design optimization method is proposed by considering uncertainties related to the cooling load calculation. Impacts caused by the uncertainties of nine factors are considered, including the outdoor weather conditions, internal heat sources and indoor set-points. The cooling load distribution is analyzed. The oversize problem is explained from the viewpoint of uncertainties. By presenting the probability distribution of the cooling load and the potential capital cost, the proposed method can determine the cooling system capacity with quantified confidence. Comparison between the cooling systems with different configurations is also conducted. With the distributions of their energy consumption, decision makers can select the optimal configuration based on quantified confidence. (C) 2015 Elsevier B.V. All rights reserved.