In this study a probabilistic approach for optimal sizing of cogeneration systems under long-term uncertainty in energy demand is proposed. A dynamic simulation framework for detailed modeling of the energy system is defined, consisting in both traditional and optimal operational strategies evaluation. A two-stage stochastic optimization algorithm is developed, adopting Monte Carlo method for the definition of a multi-objective optimization problem. An Italian hospital facility has been used as a case study and a gas internal combustion engine is considered for the cogeneration unit. The results reveal that the influence of uncertainties on both optimal size and annual total cost is significant. Optimal size obtained with the traditional deterministic approach are found to be sub-optimal (up to 30% larger) and the predicted annual cost saving is always lower when accounting for uncertainties. Pareto frontiers of different CHP configurations are presented and show the effectiveness of the proposed method as a useful tool for risk management and focused decision-making, as tradeoffs between system efficiency and system robustness.