Consideration of Major Uncertainties in the Risk Study of a Nuclear Power Plant

Probabilistic risk assessment (PRA) is an important tool in assessing the safety of a nuclear power plant. Probabilistic approach allows the selection of plant modifications which efficiently decrease the plant total risk. Optimal decision making requires that the plant is modelled as accurately as possible. A single risk factor should not be underestimated or exaggerated. This thesis presents general uncertainties related to a risk study of a nuclear power plant. Specifically, challenges associated with the PRA model of Loviisa nuclear power plant are identified. The risk significance of primary circulating pump seal leaks is substantial, and the modelling of the accidents has involved several conservative assumptions. In this thesis, the risk significance of seal leaks is re-evaluated. The thesis studies two essential accident chains that might lead to primary circulating pump shaft seal failures resulting in a loss of coolant accident (LOCA). If a LOCA occurs, various safety systems are required to ensure sufficient reactor core cooling. The risk related to loss of pump relief magnetization and reverse rotation is proved insignificant. This is based on updated studies on the endurance of modern bearings and on braking tests which indicate that seal damage is highly improbable. Another accident chain analyzed is related to loss of seal cooling. The resulting leak is re-evaluated by identifying the most essential factors affecting the leak size. A discrete probability distribution is constructed for each factor by using expert judgement. Opinions of seven pump experts are inquired on the basis of Delphi method. Individual estimates are then aggregated with a criterion based on Kullback-information. Finally, a probability distribution for leak size is constructed with the help of scenario analysis and flow calculations. The distribution indicates that the leak size is significantly smaller than previously suggested, although several uncertainties remain in the modelling. More accurate evaluation would require tests for rubber ring seals in high temperatures, modelling of two-phase flow and re-evaluation of LOCA limits in the risk model. The influence of the results on plant total risk is evaluated by considering Fussell-Vesely importances and reductions in LOCA frequencies. An approximate estimate indicates that the risk decreases roughly 5%, thus implying a core damage frequency of 5,5·10 -5 a -1 . Number of pages 97