Local Search Algorithm for Lattice Enrichment Optimization with Heuristics

Open Access
Borza, Gregory Michael
Graduate Program:
Nuclear Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
Committee Members:
  • Kostadin Nikolov Ivanov, Thesis Advisor
  • Nuclear
  • Lattice
  • Optimization
  • Local Search
  • Enrichment
  • Holtec
  • SMR
The objective of this study was to develop a lattice optimization tool to assist the loading pattern optimization of the HOLTEC SMR-160. Due to the use of cruciform control blades in the HOLTEC SMR-160 assembly design for controlling reactivity, a non-uniform pin enrichment distribution must be implemented. A non-uniform pin enrichment distribution helps mitigate the heterogeneity in the moderator distribution, allowing the lattice to have a flatter pin power shape. The optimization of the enrichment distribution within a lattice is a tedious yet important task to reduce high pin peak powers in full core simulations. The Standard Lattice Optimization Tool with Heuristics (SLOTH) was developed to automate the lattice enrichment optimization process while still allowing the user to have full control over the desired properties of the lattice (e.g. lattice average enrichment and burnable absorber distribution). It utilizes a local search algorithm in which the enrichment is incremented to minimize the pin peak power. Sensitivity studies indicate that the number of pins incremented at each iteration can be increased to five to ten pins without negatively impacting results. This allows SLOTH to optimize a lattice in approximately 90-150 minutes. Time is a key component in order for SLOTH to be an effective tool to assist in LP optimization. A heuristic biasing feature was included in SLOTH which allows the user to control the location of the pin peak power. This feature can be effective at gaining margin to thermal-hydraulic limits in instances where certain geometrical locations within the lattice are more likely to violate thermal-hydraulic limits than others. This feature was shown to be more effective in lattices with burnable absorbers, reducing pin peak power is specified locations by as much as 12.4% (using a 10% bias), compared to lattices without burnable absorbers.