Uncertainty and optimization analysis of advanced nuclear fuel cycles with generation iv reactors

Résumé

The introduction of new technologies and industrial processes, and of Generation IV reactors in particular, will translate into a more sustainable nuclear energy in terms of efficiency, nuclear waste management, safety and economic competitiveness due to the adoption of strategies based on Partitioning and Transmutation. However, as a consequence of the numerous existing designs as well as their different peculiarities, this introduction necessarily requires the evaluation of the impact of these new systems in the nuclear fuel cycle. In this manner, it will be possible to determine, based on the particular objectives of each country or region, the best way for transitioning from the current systems to these advanced technologies, finding and identifying in this process the possible limitations that may appear as a consequence of their implementation and use. Besides, given the complexity of these analyses, the study of nuclear fuel cycle scenarios is closely linked to the development of computer codes and tools for the simulation of these scenarios. It is therefore not surprising that these tools are currently being used for experts and policy makers around the world for studying and comprehending these new advanced fuel cycles. However, the conclusions that can be derived from these analyses will be as good as the quality of the results provided by the used tools, being this aspect of special importance in a world surrounded by uncertainties. In this context, this thesis has been focused on improving the reliability of the results provided by the nuclear fuel cycle simulators; to date the existing studies on this topic are scarce and present some limitations as a consequence of the strong hypotheses and assumptions they made. To that end, different methodologies that allow for taking into account the effect of the uncertainties appearing in this kind of studies in a generic way have been developed and implemented. All of this has been carried out through the study of several advanced electronuclear scenarios. More precisely, these scenarios have been inspired in realistic cases proposed at international level through different projects and collaborations given their interest from the point of view of transmutation and waste management, among others. The first step for meeting these objectives has been to upgrade TR_EVOL code (the nuclear fuel cycle simulator developed by CIEMAT) with different methodologies that allow for both uncertainty propagation and electronuclear scenario optimization. Meanwhile, a lot of effort has been made for improving this tool in such a way it becomes more versatile and faster. Once the necessary tool has been upgraded, the work of this thesis has been divided in two major parts. The first one is focused on the quantification of the effect produced by the different uncertainties, while the second part has been dedicated to the study of how the uncertainties may affect the optimization of electronuclear scenarios, and consequently, how they can affect decision making. The uncertainty quantification has been performed through the study of three scenarios, covering in each one of them one different family of uncertainties. Firstly, the uncertainties in the input parameters of the fuel cycle have been studied. To that end, a hybrid methodology making use of local and global methods has been used. In this way, the most relevant variables from the uncertainty propagation point of view can be identified in a quick and efficient manner, and in a second step, over this group of selected variables it is possible to study their detailed impact identifying non-linear effects or interactions. In conjunction with a surrogated model building based on the Polynomial Chaos expansion, this technique has allowed reducing very notably the computational demand of the global uncertainty propagation techniques. In the second scenario, the effect of the nuclear data uncertainties has been addressed. For this case, a methodology based on Monte Carlo methods and the generation of multiple perturbed nuclear data libraries has been developed and verified with the GODIVA integral experiment. After that, this methodology has been applied to an open fuel cycle in which the effect of the nuclear data has been compared with the effects produced by the most relevant parameters of the cycle in order to determine its relevance. In third place, the differences due to the fuel cycle simulation tools have been analyzed. Through a collaboration with the SCK·CEN research center, the same electronuclear scenario has been simulated with two different codes. The obtained differences which have been shown to be intrinsic to these tools, and therefore they will appear in any fuel cycle simulation whatever the quality or precision of the codes used, have been quantified through the comparison again with the effects due to the uncertainties in the input fuel cycle parameters. These analyses have shown that both the uncertainty in the nuclear data and the modeling and approximations made by the different fuel cycle codes, produce an effect in the simulation that is comparable to the uncertainty in the fuel cycle parameters. Additionally, these effects gain special relevance in advanced scenarios using Partitioning and Transmutation in which the multrecycling of the materials is pursued for its optimum use. On another note, the effect of the uncertainties in the fuel cycle parameters have shown to have a strong dependency with both the particular selected observable, and the electronuclear scenario under study. Finally, the thesis concludes by showing the relevance of the uncertainties in optimization studies. This has been evaluated through a multiobjective optimization problem which has been solved twice: the first one does not take into account the uncertainties while in the second one it has been assumed that some of the input parameters are unknown. The chosen problem has been based on an advanced scenario focused on reducing both the transuranic inventories and the fuel cycle cost with the use of advanced technologies. For the optimization, an evolutionary algorithm which iteratively evolves towards the best solution without making any assumption has been used. Although both solutions seem to produce similar results in the objective space being the case with uncertainties case contained in the case without uncertainties, the configurations leading to these solutions have shown to be different. In addition, the detailed study of the solutions without uncertainties has shown that, in their presence, they are unstable. In this manner, the uncertainties may not only lead to sub-optimal solutions but also can compromise the viability of the scenario if they are not considered during the optimization process.