Proceedings of the Eighth International Conference on Probabilistic Safety Assessment & Management (PSAM)
162 Method of Level 2 PSA Uncertainty Study for NPP Paks (PSAM-0470)
Download citation file:
- Ris (Zotero)
- Reference Manager
A level 2 PSA has recently been performed for the VVER-440/213 type nuclear power plant of Paks, Hungary. Uncertainties in large radioactivity release frequencies were assessed in a follow-on analysis of the baseline study. Uncertainties were analysed and evaluated both qualitatively and quantitatively. The quantitative part of the analysis is in the focus of attention in this paper.
Uncertainties were propagated from the level 1 PSA model to the level 2 PSA in the first phase of quantitative uncertainty analysis Quantification was based on the use of the minimal cut sets for the different plant damage states (PDS's). The reliability parameters for the basic events in a minimal cut set were treated as random variables with the associated probability distributions. Monte Carlo simulation was applied to generate samples of basic event probabilities and these samples were used to determine PDS frequencies by means of propagating uncertainties through the PDS level minimal cut sets. Dedicated software was developed and used for this purpose.
The Monte Carlo approach was used to quantify uncertainties in accident progression from a plant damage state to the different containment states and the associated release categories. First the important severe accident phenomena were determined. For these phenomena the available model in the MAAP4/VVER severe accident code was reviewed and refined. In this phase a survey of internationally available experiments and their results as well as sensitivity studies helped to identify the importance of different phenomena. Then model parameters were selected for the purpose of uncertainty calculations. Although 4–6 parameters were used initially for each phenomenon in sensitivity analyses, the number of variables treated as uncertain for MAAP4/VVER simulation could be reduced to 40. In the uncertainty analysis not only the input values to the MAAP code but other parameters, e.g. the ignition of burnable mixture and containment fragility were taken into account. Finally 50 variables were chosen for the random sampling in total. The samples from the range and distribution of the selected model parameters were generated by the Latin hypercube sampling. Severe accident calculations were done for each branch of the Containment Event Tree (CET). A calculation included MAAP4/VVER runs and processing of the results to get probability samples for the branches of a CET. 200 calculations were performed for each branch of a CET.
The uncertainty distributions for the PDS frequencies and for the CET branches were sampled and then the frequencies of containment failure states were calculated on the basis of this sampling in accordance with the logic of the CET sequences. The total uncertainty for a containment state was determined by combining the PDS level results for the given containment state. Also, the results obtained for the different containment states were further aggregated to yield overall measures of uncertainty in the so-called consequence categories defined for the purpose of the Paks level 2 PSA. A dedicated spreadsheet based tool was developed and used to propagate uncertainties between plant damages states and containment states/release categories.
This paper gives a summary of the uncertainty analyses performed for the level 2 PSA of NPP Paks. The most important analysis steps and methods are highlighted with examples of the results obtained. Also, by discussing the necessity and number of hydrogen recombiners, an example is given on how to use the results from uncertainty analysis to support the development of severe accident management.