Licentiate seminar: Core Design and Fuel Design Optimization for SMR

Opponent: Jan Dufek, KTH

Examiner: Stephan Pomp

Abstract:

Nuclear energy remains a central component of global decarbonization strategies, offering a mature, dispatchable, and low-carbon energy source of electricity. Recent interest in Small Modular Reactors (SMRs) has renewed the need for advanced analytical tools capable of supporting economical and safe reactor‑core design. Among these SMR concepts, AP300, a 300 MWe SMR from Westinghouse, is characterized by a distinct design choice, featuring a substantially reduced power density while retaining a relatively large core size. These characteristics offer potential benefits in neutron economy and fuel utilization. However, the design choice introduces new challenges in the context of material performance, as meeting current discharge burnup levels requires substantially increased in-core fuel residence times. In addition, for this design, an extended cycle length (36‑month cycles) is envisioned, which, from a core‑design perspective, may challenge established design criteria adopted in large Pressurized Water Reactors (PWRs). Demonstrating the feasibility of these novel design choices, and understanding their economic implications under stringent time and computational constraints, becomes crucial for project success, particularly when large teams of experts are required to assess these innovations.

In-core fuel management, an inherently combinatorial and multi-objective optimization problem in which the objective is the determination of optimal fuel composition and fuel-loading patterns, is one of the tasks most affected by these issues. For a long time, such problems were addressed using heuristic techniques such as expert judgment, but with increasing fuel costs it has become crucial to develop formal optimization frameworks capable of supporting extensive design-space exploration. Among in-core fuel-management activities, equilibrium-cycle optimization has remained a less popular field due to the increased complexity of the problem and the substantial computational requirements needed to evaluate equilibrium cycle performance. In these activities – conducted for long-term cost projections and fuel-procurement planning – the goal is the determination of an economically optimized, repeatable fuel-management scheme that guarantees adequate safety margins. With improvements in optimization algorithms and the development of modern surrogate modeling techniques, such as artificial-neural-network-based surrogate models, new research opportunities have emerged that substantially improve computational efficiency for equilibrium-cycle optimization.

In this work, the AP300 core layout has been selected for equilibrium-cycle optimization, which has been conducted in two stages. First, using a Linear Reactivity Model coupled with the Westinghouse Integrated Levelized Cost of Electricity methodology, the effects of core reload fraction and cycle length on economic performance have been investigated, showing that 36-month cycles provide a favorable zone in terms of fuel utilization and sensitivity to uncertainty in economic input data. Following additional analysis of the role of reload fraction in economic performance, a specific reload fraction has been selected for subsequent in-core fuel-management optimization. In the second stage, an optimization algorithm based on Evolutionary Strategies has been developed and employed to perform full in-core fuel-management optimization. In this context, a data-driven surrogate model has been developed to further enhance computational efficiency, exploiting a graph-based representation of equilibrium core-refueling schemes to evaluate configuration safety margins. These activities demonstrate that equilibrium-cycle optimization, although characterized by slow convergence, can be effectively studied within formal optimization frameworks, yielding substantial improvements in computational performance.

This thesis illustrates the equilibrium-cycle modeling techniques, the optimization algorithm developed, and the architecture of the surrogate model, with particular emphasis on comparing scenarios with and without explicit axial fuel-assembly composition optimization. The results highlight the positive role of an explicit axial description on surrogate-model predictive capabilities, particularly for axially dependent quantities, at the cost of increased optimization complexity.