Nicholas Rollins
Advised by Dr. Jason Hou
11:00am – 2:00pm
LMP 444
Abstract
Power uprates are a vital part of the global effort to support and sustain the longevity of our aging nuclear reactor fleet. These efforts most often take the form of the narrow optimization of a specific geometric variable or design element for a specific power plant and are therefore not widely applicable. New, modern tools are urgently required to improve and increase the speed, efficiency, and reliability of the engineering process for designing and optimizing nuclear reactors. This dissertation presents novel functionalities that have been developed for existing optimization capabilities to enhance the investigation of power uprates in Pressurized Water Reactors (PWRs) by developing novel ML optimization tools capable of taking consideration for the complex multiphysics, economic, and safety-related impacts of the power uprate as an integrated part of the optimization process to create a generally-applicable modular optimization framework. This included the development of fully-realized equilibrium cycle shuffling schemes that represent an optimization of the loading pattern and fuel inventory of a PWR equilibrium cycle, the development of multiphysics coupling tools capable of informing the optimization process through the analysis of transient scenarios, and the development of the ability to approximate the economic impact of each proposed solution through fuel cycle cost calculations. Accomplishing the aims presented in this work necessitated an overhaul of the MIDAS optimization framework to enhance its optimization capabilities and provide a more accessible, data-driven, multiphysics tool for modeling and optimization of complex engineering challenges.
A major component of this work is the capability to optimize for a fully-realized equilibrium cycle shuffling scheme. Equilibrium cycle optimization is often the ultimate objective of reactor core design, as it enables consistent, repeatable gains in reactor performance across continued plant operation. Traditional optimization tools typically focus on optimizing the loading pattern of a single cycle. This significantly decreases the scope of the solution-space and complexity of implementation but is limited in its real-world applicability, requiring additional user time and effort to complete the design process. In contrast, optimizing for an equilibrium cycle ensures that performance improvements persist across the continued operation of the nuclear reactor, mitigating the risk of overfitting solutions to a single cycle and enhancing the long-term operational efficiency of the reactor.
This work has successfully implemented an expansion of the optimization tool functionality to incorporate internal multiphysics coupling using the nuclear codes PARCS and TRACE, to enable feedback from both neutronic and system thermal-hydraulic performance into the optimization process. The design and operation of a nuclear reactor is highly multidisciplinary due to the complex and coupled nature of the physics governing its operations. The design of a nuclear reactor core requires considerations to be made for all of the relevant interactions. Optimization tools that only partially address these considerations require additional engineering validation and iteration, leading to inefficiencies and potential rejection of proposed designs. By integrating multiphysics coupling within the optimization workflow, crucial performance data becomes readily accessible to the ML algorithm, allowing for more informed and robust solutions without complicating the existing workflow. Such capabilities can expedite the licensing process for operating reactor fuel cycles by significantly decreasing the engineering time and effort required.
Beyond equilibrium cycle and multiphysics considerations, this work introduces enhancements to existing tools by incorporating fuel lattice optimization, which improves the performance of reactor core shuffling schemes by optimizing the internal design of fuel assemblies. The performance of even the most robust reactor core optimization tools must necessarily be limited by user-supplied assembly designs and by optimizing the spatial arrangement of fuel rods and integrated absorbers within assemblies, further improvements in fuel utilization, power distribution, and reactivity management can be achieved. The iterative process developed through this work involves initial equilibrium cycle optimization of the shuffling scheme followed by targeted fuel lattice optimization to reduce local power peaking, ultimately feeding back into the equilibrium cycle calculations for further refinement. Combining equilibrium cycle optimization, multiphysics feedback, and fuel lattice enhancements through this work has resulted in a comprehensive toolset for designing and optimizing fuel assembly-based reactors, addressing both the current and future needs of nuclear reactor technology.
