Neutronic design and optimizing the mixed core of VVER-1000 nuclear reactor via ANN-GA approach as machine learning method

Abstract

The core of an operational power plant can include hundreds of fuel assemblies. In cases where more than one single fuel assembly design is present in a core—whether through a change in fuel manufacturer (vendor), offering a better design, or other reasons—the core is described as a mixed core. Ensuring that different types of fuel assemblies do not interact in a harmful manner that causes problems in the reactor core is essential to ensuring nuclear safety. This article thoroughly explores the neutronic modeling of the core of a VVER-1000 reactor using TVS-2 M fuels and the investigation of the mixed core created by loading a UTVS fuel assembly with the optimal fuel composition and location in the core of this reactor. The research is conducted thoroughly to ensure the desired UTVS assembly is placed in the reactor's core with minimal tension regarding neutronic parameters while maintaining the core's performance. The core of the VVER-1000 reactor was initially modeled using DRAGON and PARCS codes. Then, the new UTVS fuel assembly was loaded into each fuel assembly station with varying UTVS fuel enrichments. Neutronic parameters were obtained for each of the created states using computational codes. These results were used to create an artificial neural network in MATLAB. By connecting this neural network to a genetic algorithm, optimization was performed to determine the optimal fuel location and enrichment for loading the UTVS fuel assembly in the desired reactor core. © 2025 Elsevier B.V.