Fast and accurate GPU PWR depletion calculation. (July 2018)

• Record Type: Journal Article
• Authors: Heimlich, A.; Silva, F.C.; Martinez, A.S.
• Is Part Of: Annals of nuclear energy. Volume 117(2018)
• Page Start: 165
• Page End: 174
• Publication Date: 2018-07
• Subjects: GPU -- Parallel processing -- Nuclear fuel burn-up -- Exponential matrix; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 621.4805
• DOI: 10.1016/j.anucene.2018.02.045
• Languages: English
• ISSNs: 0306-4549

Abstract

Highlights: Two GPU-based Runge Kutta Methods used to evaluate the burn-up in a PWR reactor. Exhibit speed improvement exceeding 100 times over the sequential. Exhibit precision and accuracy lower than 0.0001%. Abstract: Fast and accurate simulations of isotopic inventory are of fundamental importance to the conceptual design, refueling, and disposal of fuel from nuclear power plants (NPP). The determination of the fuel's isotopic composition requires a high computational effort arising from the complexity of solving the large system of coupled ordinary differential equations (ODE's). The system of ODE's is related to the physical mesh, fuel and coolant temperatures, time history of power, previous isotopic concentrations, and radioactive decay chains under analysis. This study surveys two methods used to simulate fuel burn-up on pressurized water reactors (PWR) implemented in Graphics Processor Unit (GPU). The accuracy of methods was also studied, by comparing the inventory simulation of one cycle burn-up looking for the benchmark obtained from Chebyshev Rational Approximation Method (CRAM), and shows the Square-Root-Mean-Error (SRME) less than 0.0001%. A performance comparison of sequential version and GPU methods exhibit a speed improvement exceeding one hundred times.