The paper presents the results of research on the restoration of signals of failed self-powered neutron detector (SPND) based on the application of neural network technologies. The general information on the development, improvement and importance of determining the linear energy release of the fuel rod in the VVER in-core monitoring system is presented. L-BFGS algorithm and the modified Levenberg - Marquardt method were used for the training of the neural network. These algorithms were used to study the neural network based on the in-core monitoring system data from the power units of ZNPP-5, KhNPP-1 and KhNPP-2. Self-powered neutron detectors with different degree of burn-up and in different core positions were chosen for simulation. The analysis of simulation data for selected SPNDs has shown that failed detector signal recovery is possible even at a short training set of the neural network with an error not more than 2 %.
Keywords: fuel rod linear energy release, self-powered neutron detector, neutron flux measuring channel, incore monitoring system, neural network.
1. Gorohov A. K. Method for determining the operational limit of linear energy release in the improved core of VVER-1000 and its ensuring under conditions of xenon oscillations. The abstract of the thesis for a degree of Cand. Tech. Sci. / A. K. Gorohov. – Podolsk, 2008. - P. 20. (Rus)
2. In-core monitoring systems for VVER type NPP / V. A. Bragin, I. V. Batenin, M. N. Golovanov, A. S. Kuzhil. – Moskva : Energoatomizdat, 1987. - P.128. (Rus)
3. Borysenko V. I. Model of formation of in-core neutron detector signal / V. I. Borysenko, Yu. F. Piontkovskyi, V. V. Goranchuk // Yaderna fizyka ta energetyka. - 2016. - Vol. 17, № 4. - P. 364 - 373. (Rus)
4. Experience in operating neutron and temperature measurement channels (NTMCs) at Tianwan NPP / M. G. Mitelman, V. M. Trotsenko, O. V. Titov, S. A. Kushmanov // Proceedings. The 6th International Scientific and Technical Conference «Safety Assurance of NPP with VVER». OKB «Gidropress», Podolsk, Russia, 26 - 29 May 2009. (Rus)
5. Ignatyev D. N. Experimental study of the behavior of VVER fuel elements in an accident with a large leak from the first loop of the reactor plant. The abstract of the thesis for a degree of Cand. Tech. Sci. / D. N. Ignatyev. - Podolsk, 2009. - P. 25 (Rus)
6. Khalimonchuk V. A. Top-Level Software for VVER-1000 In-core Monitoring System under Implementation of Expanded Nuclear Fuel Diversification Program in Ukraine / V. A. Khalimonchuk // Yaderna ta radiatsiyna bezpeka. - 2015. – Iss. 1. - P. 7 – 12. (Rus)
7. Turkcan E., Ciftcioglu Ö. Neural Network Benchmark for SMORN-VII – SMORN VII, Avignon, June 1995. P. 504 – 519.
8. Fantoni F. P., Mazzola A. Transient and Steady State Signal Validation in Nuclear Power Plants using Autoassociative Neural Networks and Pattern Recognition – SMORN VII, Avignon, June 1995. P. 423 – 434.
9. Haykin S. Neural networks. A Comprehensive Foundation. Second Edition. – Moskva : LLC "Viliams", 2006. - P. 1104. (Rus)
10. Haykin S. Neural networks and learning machines. Third Edition. McMaster University. - Hamilton, Ontario, Canada. - 2009. - P. 936.
11. ALGLIB - C++/C# numerical analysis library [Electronic resource] : [website]. – digital data. – Nizhny Novgorod 1999 - 2018 – Access mode: www.alglib.net (date 14.02.2018) – Title from the screen.