Through the extensive research on severe accidents, knowledge on severe accident phenomenology has constantly increased. Based upon such advance, probabilistic risk studies have been performed for some domestic plants to identify plant-specific vulnerabilities to severe accidents. Severe accident management is a program devised to cover such vulnerabilities, and leads to possible resolution of severe accident issues.
This study aims at establishing severe accident management framework for domestic nuclear power plants where severe accident management program is not yet established. Emphasis is given to in-vessel and ex-vessel accident management strategies and instrumentation availability for severe accident management. Among the various strategies investigated, primary system depressurization is found to be the most effective means to prevent high pressure core melt scenarios. During low pressure core melt sequences, cooling of in-vessel molten corium through reactor cavity flooding is found to be effective. To prevent containment failure, containment filtered venting is found to be an effective measure to cope with long-term and gradual overpressurization, together with appropriate hydrogen control measure.
Investigation of the availability of Yonggwang 3 & 4 instruments shows that most of instruments essential to severe accident management lose their desired functions during the early phase of severe accident progression, primarily due to the environmental condition exceeded ranges ofinstruments. To prevent instrument failure, a wider range of instruments are recommended to be used for some severe accident management strategies such as reactor cavity flooding.
Severe accidents are generally known to accompany a number of complex phenomena and, therefore, it is very beneficial when severe accident management personnel is aided by appropriately designed supporting systems. In this study, a support system for severe accident management personnel is developed using containment event trees and associated resources, which are resulted from probabilistic safety assessment. Complementary method is provided using safety objective trees. Simulation results shows that this supporting system provides appropriate and timely strategies and associated information as intended. Spatiotemporal neural network is introduced to predict the occurrence timing for essential events for severe accident management. Application predicting the timing for reactor pressure vessel failure shows that the failure timings are within tolerable errors.
