SUMMARY

The continued growth of the world population and the push to rely on clean energy has placed nuclear power at the forefront of energy policy. Efforts to modernize the existing nuclear reactor fleet have culminated in the Generation IV International Forum and six subsequent designs. Of these, the molten salt reactor (MSR) has received notable attention on the basis of characteristics from thermodynamic efficiency to safety. Proposed designs include liquid- and solid-fueled variants, chloride- or fluoride-based salts, and so forth. This range of options necessitates significant research to provide comparison metrics among MSR types and to current power reactors. Furthermore, it requires effective modeling techniques to accurately represent system behavior. Traditionally, reactor design has been left as a problem for experts; however, the nuclear field has steadily increased its reliance on computers to perform this work. In particular, optimization algorithms and machine learning techniques have proven reliable in simplifying complex problems.
The proposed work seeks to further the application of optimization to advanced reactors, particularly MSRs. This will be done in a computational sequence developed to analyze the effect of different variables on reactor performance through parametric analysis, machine learning, and genetic algorithm optimization. Thus the first part of the work will train, test, and validate a machine learning model to predict outputs of the Monte Carlo code Serpent from input variables for use in a genetic algorithm sequence. The second half will compare resultant core designs among each other as well as to a light water reactor fuel cycle. While neutronics codes have demonstrated the feasibility of advanced systems, MSRs provide unique challenges due to delayed neutron precursor drift and online re-fueling. Of interest to fuel cycle analysis are material isotopic evolution and online re-fueling. This work uses a Serpent extension to account for this by adding fuel and removing fission products. Various metrics of the fuel cycle will be compared, including isotopic production, refueling rate, and off-gassing. The work will culminate in a recommendation for moving forward with molten salt reactors and a foundation for policymakers to do so.