A first approach to resilience of nuclear fuel cycle scenarios
Résumé
Nuclear fuel cycle scenario is a powerful decision-making tool, as it allows the study of the evolution of a nuclear reactor fleet and related fuel cycle facilities over a given period of time. The scenario studies make it possible to evaluate the impact of different decision options on the evolution of the nuclear fleet. However, the scenario parameters are usually contaminated with uncertainties which are due to many sources including the lack of knowledge or information about the future and can make the scenarios fragile and easy to fail. To solve this problem, one needs to enhance the capacity of the scenarios to perform well under uncertainty. This paper aims to make several contributions to the definition of resilience of nuclear fuel cycle scenarios and give out a method to make the considered scenarios resilient.First of all, different uncertainties in the nuclear fuel cycle scenario are identified and can be classified into six categories uncertainty of model error, uncertainty of nuclear data, uncertainty of scenario parameters for the past, uncertainty of scenario parameters for the future, uncertainty of scenario hypotheses and uncertainty of economic, societal and political context. The first three kinds of uncertainty can be tackled by validation. However, since the others involve the future, they are associated with the lack of knowledge or information and are thus trickier. In this paper, after defining the scenario resilience, we propose to deal with the uncertainties of scenario parameters for the future. In a resilient scenario, when the uncertainties appear and the scenario is predicted to fail, one can reorient the development of the scenario by using some pre-designed levers.Our methodological approach is based on the state-of-the-art SUR (Stepwise Uncertainty Reduction) inversion algorithm which can be used as an adaptive technique to point out how the levers should be adjusted for a resilient scenario. It is one of the most efficient methods to provide a fast estimator based on Kriging surrogate model to find out the scenarios of interest without calling the time-consuming scenario code. In this paper, an improved version of SUR algorithm, which allows considering the multi-constraint situation, is introduced. It is applied to an academic scenario exercise so as to demonstrate its potential application on the levers adjustment to reorient the perturbed scenarios to resilient ones.