UC Berkeley

Interest in pebble bed reactors (PBRs) has surged in the past decade, given distinct unique designs and potential for inherently safe operation. Nevertheless, accurately modeling and simulating the behavior of PBRs, especially determining the flux spectrum distribution and pebble trajectories, remains an intricate task due to methodological challenges and extensive computational demands. In contrast with methods using spectral zones that aggregate pebbles over large core regions, this research proposes a novel approach called hyper-fidelity depletion (HxF), in which each pebble's history is individually tracked.

The dedicated HxF tool, a Python-based wrapper code, follows this approach. It uses the Cerberus/Serpent 2 interface at its core, which allows to finely control the simulation of a PBR operation in a time-dependent iterative manner. Running the HxF tool for an extended period, the equilibrium state of a core can be determined. Serpent 2 embeds numerous features that make the geometry modeling simple and the neutron transport processes accurate and efficient. In addition, the integration of domain decomposition ensures the feasibility of individual pebble depletion without prohibitive memory requirements. The time for obtaining a full-scale equilibrium core from scratch is on the order of magnitude of days on an average cluster, which, while not considered fast for design exploration or optimization, makes it acceptably short.

The motion and fuel handling components are synchronized with depletion within the HxF tool. Successive positions are determined either through a streamlined discrete motion sequence or by integrating granular data from a discrete element method (DEM) simulation. DEM uses a spring-dashpot model to represent mechanical interactions between pebbles and is coupled to an innovative looping method, drastically saving computational time. These positions are then communicated to Serpent 2. Pebbles undergo discharge, discard, re-insertion, and fresh fuel introduction throughout depletion. Comprehensive pebble-wise data such as position, nuclide concentrations, detector (power, flux), burnup, decay heat, activity, or passes can be tracked and saved. The discrete motion approach offers efficiency in the simulation process and faster computation times. Though effective, it has limitations in addressing complex geometries and varying pebble velocities. The DEM instead provides more precise pebble trajectories, which is especially important for more geometries with fueling and defueling chutes with the price of increased computational demand.

A preliminary exploration of the thermal-hydraulics is also provided by incorporating thermal-hydraulics coupling into HxF using the GeN-Foam solver through the Serpent 2/OpenFOAM multiphysics interface. A porous media solver and a pioneering double-heterogeneous power model enabled a better understanding of the interplay between temperature, density distributions, and neutronics within the PBR core. In particular, accounting for thermal distributions in determining accurate core behavior is essential. As temperature varies, so does the core reactivity, thus emphasizing the need for this coupling in the HxF tool.

These capabilities were applied to the generic FHR (gFHR), a benchmark model made by Kairos Power. First, a comparison is performed between the equilibrium state obtained with HxF using discrete motion and the one with the company's house code KPACS. While both tools presented satisfactory results for the equilibrium core, HxF provided a more acute understanding of parameter distributions within the core and their temporal variations. Then, using a DEM with the gFHR and a more realistic geometry, HxF displayed enhanced accuracy in depicting pebble trajectories and velocity profiles, especially when confronted with complex geometries like fueling and defueling chutes.

The HxF method's adaptable design showcases remarkable versatility, making it suitable for various applications. It can explore slow transients, diverse fueling strategies, waste management insights, fuel performance analyses, and non-proliferation evaluations. Furthermore, the rich data generated by HxF is a valuable resource for training machine learning algorithms, enabling accurate predictions and operational assessments.Critical areas for future development include adjusting the carbon-to-heavy metal (C/HM) ratio, enhancing the discard threshold, and improving depletion through dedicated, dynamic predictor/corrector methods. Additionally, enhancements in thermal-hydraulics modeling and addressing compatibility issues with Serpent 2 and Cerberus are essential to unlock HxF's full potential.

By better simulating PBRs operations, the community can better understand their behavior, optimize their design, explore new possibilities and ideas, and predict future operation conditions.