Reactor optimization is central to increasing the efficiency of nuclear fuel cycles and critical for making meaningful comparisons between different design options. Optimization algorithms work by generating trial parameter sets which can be used as inputs to reactor physics models. Unfortunately, many reactor physics codes require substantial CPU time, making optimization of large parameter sets impractical. We have developed a method for finding optima within an N-dimensional parameter space using a fast, flexible reactor physics model that is capable of performing fuel burnup calculations on the order of once per second. Global optima found in this way can then be verified using a high fidelity reactor physics code. We demonstrate our approach by considering a simple fuel pin pitch optimization for a light-water reactor, and we find our code executes in 5 minutes. Repeating this approach using a high-fidelity Monte Carlo simulation requires approximately 15 days of runtime by contrast.

This content is only available via PDF.
You do not currently have access to this content.