Abstract
EDF R&D is seeking to access the potential benefits of applying Data Assimilation to a PWR's RCS (Reactor Coolant System) measurements, in order to improve the estimators for parameters of a reactor's operating setpoint, i.e. improving accuracy and reducing uncertainties and biases of measured RCS parameters. In this paper, we use balance (between primary and secondary systems) to improve the simplified semi-empirical 0D Model for RCS, using a "fitting" method for the bypass coefficient related to the part of the flow which is not in contact with the fuel assemblies in each quarter of a four-looped core. Thus, we get a model that can be used to generate state vectors containing most of primary parameters values. Then we describe how to use this model to define a Data Assimilation Approach, by generating random parameters and thus constructing a sample of random state vectors, from which the background vector and related error covariance matrix can be deduced. Finally, we apply our method with a focus on Normalized Integrated Neutron Powers, using twin experiments to evaluate its performances. Overall, calibrating the random parameter generator for Neutron Mapping on real data does improve the algorithm performance, though only moderately.
