Improvement of a deterministic fuel management code using artificial neural network for liquid-fueled molten salt reactor

•A cross-section online updating surrogate model based on artificial neural network is proposed and developed.•The deterministic fuel management code was improved by integrating with the surrogate model.•The surrogate model and ThorNEMFM-ANN code were validated by the MSBR and MSFR benchmarks.•The applicability of the code is further validated with the fuel cycle simulation of the MCSFR. The fuel management code of liquid-fueled molten salt reactor (MSR) relies on the coupling of the neutronics code with depletion code. The deterministic method can improve the efficiency of neutronics calculation. However, updating the microscopic cross-sections in depletion calculation is still time-consuming using Monte Carlo code. A surrogate model based on artificial neural network (ANN) is developed to update the cross-sections online. Then a new fuel management code named ThorNEMFM-ANN was developed by integrating the surrogate model into the deterministic fuel management code. The surrogate model and ThorNEMFM-ANN code were validated with the molten salt breeder reactor (MSBR) and molten salt fast reactor (MSFR) benchmarks. Furthermore, the ThorNEMFM-ANN was applied to the fuel cycle analysis of a molten chloride salt fast reactor (MCSFR) for the demonstration of the applicability. The numerical results indicate that the surrogate model is feasible and effective in the fuel cycle simulation of liquid-fueled MSR.