Treatment of modelling uncertainties for deterministic safety analysis

Treatment of modelling uncertainties is a generic issue related to quantifying margins from a safety analysis and obtaining confidence in safety conclusions drawn from the safety analysis. This issue is particularly important for analysis of design-basis accidents (DBAs). It is believed that the limit-of-operating envelope (LOE) methodology (i.e., setting the key operating parameters at their limiting values) is conservative so that the effects of modelling uncertainties are compensated by the LOE approach. This belief has been confirmed by an integrated best-estimate and uncertainty (BEAU) analysis for some transients such as the first 5-s power pulse transient in a large-break loss-of-coolant accident (LOCA), during which a large number of operating parameters play a dominant role in the simulation of the transient. However, this belief cannot be generalized and may not be true for other transients, especially for those (e.g., a loss-of-fueling-machine cooling accident) where there are only one or two operating parameters and the calculation results are predominantly affected by many modelling parameters. The presentation will discuss the relevant requirements stipulated in CNSC document REGDOC-2.4.1, Deterministic Safety Analysis, particularly the need for an explicit treatment of modelling uncertainties, even when the LOE methodology is used in a safety analysis for DBA. The presentation will also review the CANDU Owners' group report Guidelines for Application of the LOE/ROE Methodologies to Deterministic Safety Analysis (COG-11-9023 R01) and substantiate its applications. (author)