Abstract: In pressurized water reactor safety analysis, the point kinetics model and one-dimensional thermal-hydraulic model are always used in system codes to simulate the transient parameter variation of core and pressure vessel. This approach neglects the time-dependent variations in core power and completely mixes the coolant from different loops in the downcomer and lower plenum. This study aims to develop and validate an implicitly coupled thermal-hydraulic and neutronics code based on system code NUSOL-SYS and to perform transient analysis of a main steam line break accident combining the coupled code and the three-dimensional computational capability of NUSOL-SYS. The coupling was achieved using Picard iteration in the form of Gauss-Seidel, which allowed the two codes to alternate solving at each time step until the convergence criterion was reached. The coupling code was validated against the NEACRP REA benchmark. Two postulated transient cases involving a single assembly at different time steps were conducted to compare the performance of the operator splitting method and Picard iteration. The results indicate that as the time step increases, transient power oscillations occur in simulations using the operator splitting method due to the lack of parameter convergence within the time step. The amplitude of these oscillations increases with the time step size. Under the same conditions, the relative error of transient peak power of the explicit coupling method is 30 times greater than Picard iteration. Additionally, a detailed three-dimensional model of the pressure vessel of CPR1000 was established. The comparative analysis with one-dimensional pressure vessel and sensitivity analysis on crossflow and whether coolant flow is lost during the accident were conducted. The results show the developed detailed model can correctly reflect the diffusion and mixing process of the coolant in the downcomer. Compared to the one-dimensional pressure vessel model, the second power peak and the local power peak calculated with the three-dimensional model are both higher. Sensitivity analysis during the accident indicates that the impact of crossflow on local parameters is minimal when coolant flow is not lost, which means the axial flow is dominant. However, when the coolant pump fails during the accident, the impact of crossflow becomes more pronounced and reduce the core local peak power by 17.49%.