Numerically efficient determination of kinetic parameters of the VR-1 nuclear reactor based on experimental data and ODE-constrained optimization

Abstract

A method of adjusting nuclear reactor kinetic parameters to experimental data is proposed. The fractions of neutrons delayed via different precursor groups are of interest. Their values originally calculated by Monte Carlo simulations are modified to bring the power output of the reactor predicted by the point kinetics equations closer to the measured values. The measurements were performed on the VR-1 zero-power training reactor in the Czech Republic. Three reactivity patterns were investigated to account for the different reactor transients. The resulting ODE-constrained optimization problem is solved numerically, using the adjoint equations to obtain the gradient of the loss functional and applying a specifically tailored gradient descent technique. The performance of our approach is compared to other variants of gradient-based optimization. As a side result, a gradient descent step size adaptivity algorithm is proposed. Finally, discussion on the physical relevance of the obtained results is provided.