Abstract
After Fukushima Daiichi accident, more and more safety analyses of severe accident were done for Spent Fuel Pool (SFP) in Taiwan. The experience of Fukushima showed that a long term Station Blackout (SBO) may cause the rising of fuel temperature and hydrogen generation. Most of the SFPs do not have the equipment for hydrogen control, such as igniter, inert containment. For a case of severe accident, hydrogen may lead to deflagration or detonation. Thus, Hydrogen became one of the most important issues of SFP analysis. The purpose of this study is to calculate the hydrogen phenomena in severe accident of Chinshan nuclear power plant (NPP) SFP by MELCOR2.1/SNAP. MELCOR is a code developed by Sandia National Lab and it can calculate the severe accident phenomena such as core relocation, hydrogen generation, hydrogen burn and detonation, etc. The latest version MELCOR2.1 was used and combined with Symbolic Nuclear Analysis Package (SNAP). In this combination, MELCOR was used with a graphical user interface (GUI) that users can easily modify any detail of the model. An animation model of SNAP can also show the results of analysis easily. In this study, the Chinshan NPP SFP was set to be a conservative situation that full core was put inside the pool. There were three steps in this study. First, the MELCOR2.1/SNAP model was built and the results were compared to TRACE, CFD and MAAP. The results of Chinshan NPP SFP model were consistent with other codes. Second, two cases were calculated by this model. One was the case of long term SBO and the other was loss of coolant accident (LOCA). The heat and hydrogen generation were different in two cases because of the different between Zirconium-water reaction and Zirconium oxidation. The results were discussed in details in this study. The final step was the sensitivity study of the separation of MELCOR control volume (CVH). The calculation of hydrogen burn was done in each CVH component. So the sensitivity study separated the CVH component in this model to several smaller control volumes. The results showed that the hydrogen burned more locally in the new model and the cladding temperature was more conservative. But the CPU time was much longer than the old model in this case