Abstract: Based on the large-scale parallel computing platform, the numerical reactor uses advanced physical models, numerical simulation algorithms and fined modeling, so as to accurately simulate various physical phenomena of the reactor under normal operation and accident conditions. In order to solve the problems of low calculation resolution, low calculation accuracy and small application range caused by a series of theoretical approximations in the traditional reactor simulation technology, advanced simulation methods for numerical reactor were studied. In terms of neutronics resonance calculation, a resonance calculation method based on the globallocal coupling was proposed. The effects in resonance were divided into global effects and local effects. The Dancoff correction factor was used to deal with the global effects, and the pseudoresonantnuclide subgroup method was used to deal with the local effects. Based on the 2D/1D coupling method, an improved leakage splitting method was proposed to obtain the angular flux of each flat source region through the anisotropic angular reconstruction in the fuel rod, and a threelevel CMFD method was proposed to improve the computational efficiency of the traditional CMFD method. For the depletion calculations, a prediction/correction method based on reaction rate prediction was proposed to predict the change of nuclear reaction rate according to the first few burnup points and reduce the calculation time of subsequent burnup points. The Picard iterative method was used to calculate the full coupling of neutronics, thermalhydraulics and fuel performance. Based on the above methods, a numerical reactor code named NECPX was developed and verified by a set of internationally famous benchmark problems. It is proved that the code is of high calculation accuracy and efficiency. The code is then applied to the simulations of large commercial PWRs, research reactors and experimental reactors such as the second generation improved PWR M310, the third generation PWR AP1000, Xi’an Pulse Reactor, JRR-3M plate fuel reactor and Watts Bar reactor. For M310 core simulation, the critical boron concentration error of the code is within 30 ppm, and the maximum fuel effective temperature occurs at the beginning of the cycle and decreases with the burnup increasing. The change of the coolant temperature distribution is consistent with that of the fuel temperature distribution. Taking the solution from the MonteCarlo code as the reference, for the AP1000 simulation, the difference between NECPX and the reference in keff is 59 pcm. The difference in power for most assemblies is within 1.5%. For Xi’an Pulsed Reactor, the max difference in keff is 84 pcm, the max power difference is 5.2%. For JRR3M plate reactor, the max difference in keff is 97 pcm, the max power difference is 1.8%. For the multiphysics coupled simulation of Watts Bar reactor, the max fuel temperature calculated by the code is 1 709 K, the Mises stress of cladding changes little with time and space, the gap width is large at both ends of the fuel rod and small in the middle. The numerical results show that it can accurately predict key safety parameters of the reactor core, such as effective multiplication factors, power distributions, temperatures, stress, gap width, etc., and it is a reliable tool for the design and safety analysis of commercial largescale pressurized water reactors, research reactors and research reactors.