Methodologies for sensitivity/uncertainty analysis using reactor core simulators with application to pressurized water reactors

• Autores: Emilio Castro González
• Directores de la Tesis: Nuria García Herranz (dir. tes.)
• Lectura: En la Universidad Politécnica de Madrid ( España ) en 2018
• Idioma: español
• Tribunal Calificador de la Tesis: Carolina Ahnert Iglesias (presid.), Oscar Cabellos (secret.), Rafael Miró Herrero (voc.), Marianna Abrámova (voc.), Rafael Federico Mendizábal Sanz (voc.)
• Programa de doctorado: Programa de Doctorado en Energía Sostenible Nuclear y Renovable por la Universidad Politécnica de Madrid
• Materias:
  • Física
    • Física atómica y nuclear
      • Energía nuclear
      • Reactores nucleares
  • Ciencias tecnológicas
    • Tecnología nuclear
      • Reactores de fisión nuclear
    • Tecnología del espacio
• Enlaces
  • Tesis en acceso abierto en: Archivo Digital UPM
• Resumen

  • Traditionally, safety analyses of nuclear power plants were carried out using pessimistic methods and hypotheses, forcing the results to be conservative. In recent years, and thanks to the increased accuracy in computational tools, realistic simulations are being considered for application in safety analysis. However, the realistic or best-estimate predictions must be complemented with a rigorous quantification of the impact of the input parameters uncertainties on the responses of interest. These approaches are named best-estimate plus uncertainty (BEPU) methods.

    Aware of the importance of the role of uncertainties, the OECD/NEA Nuclear Science Committee launched the “OECD Benchmark for Uncertainty Analysis in Best Estimate Modeling for Design, Operation and Safety Analysis of LWRs”. This is an international exercise with the aim of developing uncertainty quantification methods for light water reactor simulations.

    This thesis has been done in the framework of the mentioned UAM Benchmark, with the main objective of implementing sensitivity analysis and uncertainty quantification techniques in the core simulators developed at Universidad Politécnica de Madrid (COBAYA4 and SEANAP), converting those best-estimate approaches in BEPU methodologies. The source of the uncertainties can be any input parameter to the codes: nuclear data as well as technological, thermalhydraulics and thermomechanics parameters; and the response of interest will be any parameter calculated by the codes: reactivities, power distributions, among others.

    The first system, SEANAP, is a complete set of tools for the simulation of PWR, covering all steps from refueling to core simulations, which has demonstrated a good agreement between calculated and measured values in several Spanish nuclear power plants. The uncertainty analysis methodology implemented in the system is based on random sampling of the input parameters. In addition, and thanks to a collaboration with AREVA GmbH, a data assimilation technique has been also implemented to include measurements of past cycles in order to improve the predictions of the subsequent cycles. The application of this system to the design of a cycle of a PWR core is included.

    The second system is based on COBAYA4, evolution of the core simulator included in SEANAP. The most relevant features of COBAYA4 are: i) its ability to perform multigroup and multiscale (nodal and pin-by-pin) simulations; ii) its flexibility to work with different lattice physics codes; iii) its capability to deal with multiphysics problems through its coupling to thermalhydraulics codes, allowing different resolutions in the neutronics and thermalhydraulics domains. SCALE/NEWT is used as lattice code, and COBRATF as thermalhydraulics solver, so the whole system is referred to as SCALE6.2/COBAYA4/COBRATF.

    In this second system one sensitivity analysis and two uncertainty quantification techniques have been implemented. The sensitivity analysis uses an adjoint sensitivity procedure to calculate the sensitivity coefficients of the response k-eff to the few-group macroscopic cross sections. The first uncertainty quantification technique is deterministic and uses those sensitivity coefficients in the Sandwich Rule together with the covariance matrix of the few-group homogenized constants to obtain the uncertainty in the response ($k-eff). That covariance matrix can be generated thanks to the uncertainty quantification capabilities of SCALE6.2, chosen as upstream code for COBAYA4, so that the few-group cross-sections can be provided with their covariance matrices. The second uncertainty analysis methodology is based on random sampling of the input parameters: nuclear data using the SAMPLER sequence from SCALE; technological, thermalhydraulics, and thermomechanics parameters using other sampling tools like DAKOTA.

    Three applications of this second system are included: i) propagation of nuclear data uncertainties in standalone neutronics calculations to full core results, using SCALE and COBAYA4; ii) propagation of technological, thermalhydraulics and thermomechanics uncertainties in coupled calculations using COBAYA4 and COBRATF; iii) propagation of nuclear data uncertainties on the few-group homogenized cross sections along burnup using SCALE.

    All in all, this thesis provides an uncertainty quantification framework for the BEPU analysis of pressurized water reactors using core simulators.