Ameliorating Xe-Induced Power Oscillations in FHR Monte Carlo Depletion Simulations

This paper introduces a novel systematic approach to optimize use of computational resources in Monte Carlo (MC) depletion simulations. The MC method in principle allows obtaining accurate solutions to neutronics problems, but statistical uncertainties, which may be exacerbated by nonphysical (numerical) simulation effects, may render this advantage difficult to realize. In particular, MC depletion simulations are known to lead to large unphysical spatial flux and power oscillation. This is even more pronounced in reactors with thermal/epithermal spectra, where MC uncertainties in flux and associated xenon (Xe) distribution may induce or amplify these oscillations. This paper examines MC depletion simulations of a fluoride-salt high-temperature reactor (FHR) with TRISO-based fuel in stationary fuel elements. Using very short depletion steps and at the same time enforcing a tight flux distribution convergence at every step would resolve the problem, but the required computational effort is not acceptable for most practical depletion analyses. This work presents a broad range of investigation of the optimal apportioning of computational effort through simulation parameters to ameliorate the magnitude of xenon-induced oscillations while still maintaining the problem tractable considering available computing resources. As expected, oscillations in final results are reduced when the simulation effort is increased via replica simulations, simulation of additional histories, and/or increasing resolution of the time grid. All of these tend to scale simulation time linearly with the total number of histories, however, the resulting improvement in accuracy and oscillations is different. Somewhat unexpectedly, performance of replica simulations in place of a single high fidelity simulation does not show statistical improvement in the final result, beyond that associated with the total number of histories. Also, further refining the timestep size beyond an already reasonable value (~1000 MWd/tU) provides little improvement in oscillation behavior, and computational effort is more effectively allocated to increasing the number of simulation histories. It is found that averaging results from successive timesteps tends to “cancel out” oscillations more than what would be predicted based on the number of histories and potentially lead to more accurate results. This observation is explained by anti-correlation of Xe concentration and power density distributions. Future work will therefore investigate performing multiple transport simulations at timesteps and averaging results in place of refining the timestep grid. Finally, while the method and results are expected to be broadly valid and provide new insights for thermal/epithermal systems, the FHR is a large reactor, with specific features different than a light water reactor (LWR). It requires many more scatterings to thermalization, has larger mean free path and migration areas, and consequently may behave somewhat different with respect to Xe oscillations. To the best of our knowledge, this is the first study examining in detail Xe oscillations in MC simulations of large plate-type FHRs.