Improvement of the Control-Oriented Modelling of the GEN-IV Lead-Cooled Fast Reactor: Development of Reduced Order Methods

The study of the system dynamics is usually carried out relying on lumped-parameter or one-dimensional modelling. Even if these approaches are well suited for control purposes since they provide fast-running simulations and are easy to linearize, they may not be sufficient to deeply investigate the complexity of the system, in particular where spatial phenomena have a remarkable impact on dynamics. Reduced Order Methods (ROMs) can offer the proper trade-off between computational cost and solution accuracy, combining the high-detail modelling usually adopted for design purposes with the requirements demanded for a control-oriented tool, firstly the computational efficiency. In this work, ROMs are used in order to improve a control-oriented plant simulator of a Lead-cooled Fast Reactor (LFR), i.e., substituting some components based on zero-dimensional modelling approach with ROM-based models. In particular, the attention is focused on the reactor core neutronics, and on the thermal-hydraulics of the reactor pool as well. The plant simulator is based on the object-oriented Modelica language and is implemented in the Dymola simulation environment. As for the neutronics, a spatial model for the reactor core has been developed aimed at substituting the classic point kinetics currently used in control-oriented tools. Different choices of spatial basis and test functions have been considered, i.e., Modal Method, Proper Orthogonal Decomposition (POD) and the newly developed Adjoint Proper Orthogonal Decomposition. The spatial neutronics approach has been tested in a simple 3D case, and implemented in the control-oriented simulator, proving the feasibility of employing ROM-based components. In addition, the full core model of the Advanced Lead Fast Reactor European Demonstrator (ALFRED) has been set up in order to evaluate the performance of the different modelling choices in reproducing the reactivity insertion following a temperature change or a control rod movement. As for the thermal-hydraulics, a spatial model of the reactor pool has been developed. This model is based on the POD-FV-ROM procedure, developed on purpose for extending the literature approach based on Finite Element to the Finite Volume (FV) approximation of the Navier-Stokes equations. Starting from the proposed procedure, a parametric ROM-based component of the coolant pool of the ALFRED reactor has been developed. In particular, the lead velocity at the steam generator outlet has been considered as parametrized boundary condition, being a possible control variable. The simulation results, both for neutronics and thermal-hydraulics examples, show that the ROM approach can provide a better physical description and a high modelling accuracy with respect to the classic 0D/1D modelling usually employed in control-oriented tools without increasing the computational burden.