Deep learning driven inverse solving method for neutron diffusion equations and three-dimensional core power reconstruction technology

Online monitoring of nuclear reactor core plays a significant role in safe-operation and economic improvement of nuclear power plant. In the process of reactor online monitoring, limited amount of the timing measured data inside and outside the reactor will be used to solve the core power distribution. The traditional methods such as interpolation and harmonic-based methods still have room for improvement in power reconstruction accuracy and robustness. This paper introduces the basic principle of solving neutron diffusion equation and the general framework of power reconstruction driven by deep learning techniques. This method has good performances in online monitoring, even under the conditions of limited measurement data, missing boundary conditions, and partial detector failure. The key techniques of multi-source data fusion, inverse solution of diffusion equations, and detector failure correction with the actual boundary condition missing are proposed in the work. We conducted several standard benchmarks to confirm the accuracy of the solution to neutron diffusion equations based on deep learning method. Additionally, we validated the new technique for power reconstruction, demonstrating its accuracy and effectiveness through an engineering problem simulation. Hence, a new technical approach for reactor core power monitoring is explored in this work.