Hydrogen may be released by Fuel and Cladding interaction with the steam at very high temperatures into the containment during a severe accident at the nuclear power plant (NPP). Locally, high hydrogen concentration may be achieved that might probably result in detonation or fast deflagration and challenges the reliability of the containment. The mixing and distribution of hydrogen is a serious safety issue for scientists to preserve the structural reliability of the containment. Although the Three Miles Island (TMI) accident in 1979 was the initiator to study the production and buildup of hydrogen in the containment. Subsequently, the hydrogen explosion in the Fukushima Dai-ichi NPP accident (2011), modeling of the gas behavior turn out to be a significant topic in the nuclear safety analysis. Computational fluid dynamics (CFD) codes can be used to investigate the hydrogen distribution in the containment during accidental scenarios and predicts the local hydrogen concentration in different regions of the containment vessel. In such a way, the associated risk of hydrogen safety can be determined, and safety assessment and procedures can be measured. The current paper presents the results of systematic work done by using the HYDRAGON code, developed by Department of Engineering Physics, Tsinghua University, to study (I) Mesh sensitivity (II) Buoyancy-driven flows analysis, for various turbulence models i.e., a standard k-epsilon (SKE) model, a renormalization k-epsilon (RNG) model and a realizable k-epsilon (RLZ) model and, to demonstrate the HYDRAGON code thermal hydraulic simulation capability during a severe accident at the NPP. The HYDRAGON code simulation results were compared to the published data performed by Jordan et al., and it has been observed that the simulation results obtained by refined mesh have given generally satisfactory results. However, the global effect of the buoyancy term in the turbulence equations on the HYDRAGON code simulated results is very small.