Resonance calculation based on the deep learning method for treating the non-uniform fuel temperature distribution in PWRs

In the previous work, a global–local resonance method was developed for the PWR self-shielding calculations, but there are still some difficulties when dealing with the non-uniform temperature profile of the fuel rod. This work proposes a deep learning based resonance calculation scheme for treating the non-uniform fuel temperature distribution. In this scheme, the deep learning model is investigated to replace the ultra-fine group method for the local calculations of the global–local resonance method, because the ultra-fine group method is too slow for performing the whole core resonance calculations when considering the temperature profile. The generation of training data sets, deep learning models and some numerical results are introduced. The given tests demonstrate that the new method could give accurate self-shielding cross sections and significantly improve efficiency.