Development of in-core fuel management tool for AHWR using artificial neural networks. (January 2021)

• Record Type: Journal Article
• Authors: Thakur, Amit; Sarkar, Debasmit; Bharti, Vishal; Kannan, Umasankari
• Is Part Of: Annals of nuclear energy. Volume 150(2021)
• Publication Date: 2021-01
• Subjects: Artificial Neural Networks (ANNs) -- Burnup -- On-power refueling -- AHWR -- Fuel management; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 621.4805
• DOI: 10.1016/j.anucene.2020.107869
• Languages: English
• ISSNs: 0306-4549

Abstract

Highlights: Refueling scheme for AHWR was developed successfully using ANNs. Target discharge burn-up was achieved keeping operational parameters in design limit. Computational cost reduced significantly by using ANNs to predict core parameters. Better refueling scheme generated in comparison to earlier used heuristic approach. Abstract: In-core fuel management methods based on evolutionary algorithms and machine learning tools are being developed worldwide to improve the overall efficiency of the operating cycle in all type of nuclear reactors. The fuel cycle of Advanced Heavy Water Reactor (AHWR) is unique and requires complex fuel management involving in-core refueling and reshuffling operations. Using the Artificial Neural Networks (ANN) approach, we have evolved an effective fueling strategy for AHWR. A computer code based on ANNs has been developed and predictions are done for k-effective and maximum channel power for all possible refueling inputs. The best candidates are chosen to perform 3D diffusion simulations. Unlike LWRs, the AHWR has on-power refueling feature which results in continuously changing core configuration and core burn-up profile. The paper highlights the use of ANNs for arriving at an optimized refueling strategy for AHWR and makes a comparison with earlier results where heuristic approach was used.