DOI: 10.55176/2414-1038-2019-2-180-188
Authors & Affiliations
Adeev V.A.
KOLA NPP, Polyarnye Zori, Russia
Adeev V.A. – Head of Laboratory, Cand. Sci. (Tech.), Kola NPP. Contacts: Polyarnyye Zori, Murmansk reg., Russia, 184230. Tel: +7(921)734-78-31; e-mail:
Abstract
The tightening of safety requirements leads to the introduction of new boundary and framework pa-rameters for the characteristics of the core. The selection has become much more complicated, because it is necessary to consider the fulfillment of criteria for the effectiveness of emergency protection, a separate organ and a working group of CPS, the temperature of repeated criticality, etc. The use of burnable absorber does not allow you to estimate the energy release peak factors only at the beginning of fuel cycle, analysis is required during the entire time of its burning, because the change of peak factors is non-monotonic. New limits on energy release have been introduced that require the performance of fuel cell calculations – linear load of fuel rods depending on their burn-up. The correct choice of reactor core determines the success of its operation during the whole year. The choice is complicated and subjective and depend on the skill of the designer.
The paper focuses on the implementation and application of evolutionary strategies and genetic algorithms (ES and GA) to optimize the core pattern of VVER reactors. A method of formalizing the core pattern when using ES and GA is described. A scheme for selecting options, performing calculations and displaying results is shown. Methods of graphical representation and results of core pattern search with maximum reactivity margin for different criteria are described. Methods for reducing the search space, ensuring the finding of a global optimum and reducing the number of calculations using heuristics and neural networks are proposed.
The author has developed a program for optimizing core pattern of VVER using ES and GA. The program is used in the design of fuel cycles and selection of core pattern at nuclear power plants. The program allows you to reduce labor costs, find a better core pattern, objectively assess the impact of boundary and framework safety parameters on the result of the choice.
Keywords
core design, optimization algoritms, genetical algoritm, neural network
Article Text (PDF, in Russian)
UDC 621.039