Current superstructure size for energy system design is increasing with computing facility improvement. Despite such approach allows to link and to analyse a lot of elements, demonstrating the advantage of a global optimization, non-negligible uncertainties are introduced. They may be due to the level of detail of modelisation, predictive aspect of the model, or simply the stochastic nature of some parameters (like outdoor temperature by example). In the present paper, different approach for optimizing energy system under uncertainty will be discussed, taking into account ability of the system to adapt to uncertain variables variations.