Uncertainty quantification for a severe accident sequence in a SFP in the frame of the H-2020 project MUSA: First outcomes

The Management and Uncertainties of Severe Accidents (MUSA) project, funded in HORIZON 2020 and coordinated by CIEMAT (Spain), aims at consolidating a harmonized approach for the analysis of uncertainties and sensitivities associated with Severe Accidents (SAs) focusing on Source Term (ST). In this framework, the objectives of the Innovative Management of Spent Fuel Pool Accidents (IMSFP – WP6), led by IRSN (France), are to quantify and rank the uncertainties affecting accident analyses in a Spent Fuel Pool (SFP), to review existing and contemplated SA management measures and systems and to assess their possible benefits in terms of reduction of radiological consequences. To quantify the propagation of the uncertainties of the input parameters to the output uncertainties of severe accident codes (ASTEC, MELCOR, RELAP/SCDAP), a diverse set of uncertainty quantification (UQ) tools (DAKOTA, RAVEN, SUNSET, SUSA) are used. The statistical framework used by the different UQ-tools is similar e.g. pure random (Monte Carlo) and Latin hypercube sampling (LHS). Fourteen partners from three different world regions are involved in the WP6 activities. The target of this paper is to describe the achievements during the first three years of the project. In a first part, a description is given of the SFP accidental scenario, of the key target variables and radionuclides chosen as ST Figures of Merit (FoM) and of the identified uncertainty sources in models and input parameters. A key element when defining the SFP scenario has been the consideration (or not) of the reactor building, as it is expected to significantly affect analyses. In a second part, the first insights coming out from the calculation phase of the project are presented. The review of existing SA management measures is also exposed, as well as systems whose benefits will be assessed in the second phase of the project. Finally, challenges that arise from such an exercise are discussed, as well as major difficulties found when applying UQ methodologies to SFP scenarios and solutions adopted.