CORE DESIGN AT PAKS NPP USING NEURAL NETWORKS
23rd Symposium of AER on VVER Reactor Physics and Reactor Safety (2013, Štrbské Pleso, Slovakia)
Fuel management issues
Abstract
The main objective of this research is to develop a new loading pattern designing method
using artificial neural networks and apply it to the Paks Nuclear Power Plant. We expect from
this technique to give us a tool which may allow, combined with the classical refueling
optimization method, a more economical functioning of the reactors of Paks NPP.
In the first part of this paper a short summary can be found about the main characteristics of
the core design at Paks, which is followed by a short introduction into the theory of neural
networks.
The main part of this paper contains the description of our neural network and the developed
computational apparatus. Neural networks are used to predict the value of two target functions
which characterize the safety and the economics of loading patterns. Using the appropriate
network type and structure we can check much more patterns than with the classical
optimization method. From these patterns we can choose the best ones based on the predicted
value of the target functions, the most economical one will be found after the exact
calculation of the function values. This paper contains only the first results of this new
research, we hope in the near future our method can be used as a standard tool for core design.