An Adjoint Proper Orthogonal Decomposition method for a neutronics reduced order model. (April 2018)

• Record Type: Journal Article
• Authors: Lorenzi, Stefano
• Is Part Of: Annals of nuclear energy. Volume 114(2018)
• Page Start: 245
• Page End: 258
• Publication Date: 2018-04
• Subjects: Reduced order modelling -- Modal Method -- Proper Orthogonal Decomposition -- Spatial neutronics model -- Lead-cooled Fast Reactor; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 621.4805
• DOI: 10.1016/j.anucene.2017.12.029
• Languages: English
• ISSNs: 0306-4549

Abstract

Highlights: Reduced Order Methods (ROM) for neutronics applications. Different methods tested in order to assess the efficiency of the reduced order model. Adjoint Proper Orthogonal Decomposition (APOD) proposed for combining POD and the use of the adjoint flux. Analysis of the different methods in reproducing reactivity insertion due to temperature or CR change. APOD shows better results in terms of accuracy with a reasonable computational time. Abstract: This paper deals with the use of Reduced Order Methods for neutronics modelling. This approach is used whether both accuracy and computational efficiency are required. A very popular category of these methods is based on projection approaches which use spatial basis and test functions for the development of the reduced order model. The selection of the spatial basis and test functions used in the projection phase is a crucial issue since it has an impact on the accuracy and the computational cost. In this work, different methods for the creation of the spatial basis and the test functions are analysed. In particular, an Adjoint Proper Orthogonal Decomposition (APOD) method is proposed combining the properties of the Proper Orthogonal Decomposition and the use of the adjoint flux as test function in the neutronics framework. The different methods are applied to create a spatial neutronics model for the ALFRED reactor. The simulation results show that the APOD method gives better results compared to the other methods (Modal Method and standard Proper Orthogonal Decomposition) increasing the accuracy of the reduced order model or minimizing the computational cost.