Optimisation of used nuclear fuel canister loading using a neural network and genetic algorithm

• DOI: 10.1007/s00521-021-06258-2
• Publication Date: 2021-07-04T11:02:10Z
• Authors: Virginie Solans; Dimitri Rochman; Christian Brazell; Alexander Vasiliev; Hakim Ferroukhi; Andreas Pautz
• Keywords: neural network; Energy Engineering; 02 engineering and technology; assemblies; 7. Clean energy; Neural network; high-level nuclear waste; Energiteknik; Genetic algorithm; genetic algorithm; 0202 electrical engineering, electronic engineering, information engineering; High-level nuclear waste

Abstract This paper presents an approach for the optimisation of geological disposal canister loadings, combining high resolution simulations used nuclear fuel characteristics with articial neural network and a genetic algorithm. The fuels (produced in open cycle without reprocessing) considered this work come from Swiss Pressurised Water Reactor, taking into account their realistic lifetime reactor core cooling periods, up to final repository. case 212 representative assemblies is analysed, assuming loading 4 per canister, optimizing two safety parameters: decay heat (DH) effective neutron multiplication factor k $$_{\mathrm{eff}}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:msub> <mml:mrow /> <mml:mi>eff</mml:mi> </mml:msub> </mml:math> . In present approach, trained as surrogate model evaluate value substitute time-consuming-code Monte Carlo transport &amp; depletion SERPENT specific calculations. A algorithm then developed optimise simultaneously DH values. computed during using previously artificial network. allows (1) minimize number canisters, given assumed limits both quantities (2) differences among canisters. study represents proof-of-principle capabilities, will be applied future larger cases.