Parameter estimation in thermalhydraulic models using the multidirectional search method

Abstract

Some uncertainty is introduced in the simulation of the behavior of a nuclear power plant because some of the parameters of the models used for the simulation are undefined or unknown. The values of these parameters are usually determined by the analyst using their own judgement and thus a user effect is introduced into the simulation process. The uncertainty caused by user effects may be mitigated by the utilization of an automatic and systematic search and optimization method to estimate parameters for the simulation model of a particular plant. The multidirectional search algorithm is proposed as an optimization technique. Two different applications have been studied which successfully show the applicability of the optimization methodology. First, the methodology is applied using the RELAP5 code to determine the best boundary conditions to simulate an experiment performed at the facility of the Royal Institute of Technology. The second application consists of the development of a two loop reduced PWR reactor model and the determination of some of its effective parameters to adjust the reduced model behavior to a given reference associated with particular operational transients.