In practical situations, decision makers often need to handle severe uncertainty. Such severe uncertainty can be caused by a variety of reasons, including lack of data, limited expert opinion, computational constraints on the model, and a lack of understanding of how to mathematically best represent certain aspects of the real system. First, I will discuss one particularly interesting way of representing and quantifying uncertainty, based on a behavioural concept called desirability, I will argue that this approach is more suitable for dealing with severe uncertainty than standard probability [1,3,4]. Next, I will explain how this approach can be related to possibility measures, belief functions, and robust Bayesian analysis . Finally, I will present a practical application of this approach in renewable energy .