Reactor power distribution detection and estimation via a stabilized gappy proper orthogonal decomposition method

Abstract

The proper orthogonal decomposition (POD) method has been applied in nuclear reactor physics to extract features of dominant flux or power. In particular, the gappy POD method is used to handle field reconstruction problems based on incomplete data due to limited observations. However, one important question is that the observation data is never noise-free in practice, which might result in blurred property of the gappy POD method, for example, the estimation error might no longer converge or even diverge as the data size increases. To address this issue, a stabilized gappy POD method is proposed within the data assimilation framework, which involves the preknowledge of the manifold structure of the physical states and the L-curve for setting regularization parameter. This method is applied to simulate the power distribution during the control rods movement process for the HPR1000 reactor core. In the numerical experiments, the stabilized gappy POD approach uses some dominant POD basis vectors and a small number of sensor signals, to obtain accurate reconstructed field with low computational cost. In addition, the stabilized gappy POD method has been integrated into the CORCA-3D code package to assimilate the observation data of nuclear reactors, which shows more stable result than the 3D-Var and CC methods used at present.