Uncertainty quantification for severe-accident reactor modelling: Set-up and first results of the Horizon-2020 project MUSA

The current Horizon-2020 project on “Management and Uncertainties of Severe Accidents (MUSA)” aims at applying Uncertainty Quantification (UQ) in the modelling of Severe Accidents (SA), particularly in predicting the radiological source term of mitigated and unmitigated accident reactor scenarios. A selected number of severe accident sequences of different nuclear power plant designs (e.g. PWR, VVER, and BWR) are addressed. The application of the Best Estimate Plus Uncertainty (BEPU) methodology to reactor accident scenarios requires a number of key steps: (i) the selection of severe accident sequences for each reactor design; (ii) the development of a reference input model for the specific design and SA-code; (iii) the definition of the figures of merit for the UQ-analysis; (iv) the selection of a list of uncertain model parameters to be investigated; (v) the choice of a statistical tool to propagate input deck uncertainties; (vi) the selection of a feasible approach (i.e., Monte Carlo versus order statistics) to address UQ by using a statistical software (i.e., UQ-tools DAKOTA, SUSA, URANIE, etc.); (vii) the running phase to achieve a high number of successful realizations with the SA codes; and, (viii) the statistical evaluation of the results (i.e., sensitivity analysis). This paper describes each of these steps such as settled in the reactor applications work package of the EU MUSA project and pays particular attention to the choices made by partners. It presents preliminary results also with an emphasis on the major challenges posed by BEPU application in the field of SA analysis