Uncertainty Quantification in Advanced Reactors: The Coupling of SAS4A/SASSYS-1 with RAVEN and Dakota

摘要

Advancements in the knowledge state of nuclear reactor performance has led to an increased understanding of the value of uncertainty quantification (UQ) and sensitivity analyses (SA) in the advanced reactor domain. The role of UQ spans many facets in the nuclear industry, including design, optimization, licensing, and probabilistic risk assessment. As the result of independent increases in computing capabilities, a more robust and mechanistic treatment of uncertainties in reactor design and analysis efforts is possible such that comprehensive UQ assessments can now be performed. To address the need to perform SA and UQ for advanced reactors, SAS4A/SASSYS-1, a system-level safety analysis modeling tool for sodium cooled fast reactors (SFRs) developed and maintained by Argonne, is being coupled with two actively supported UQ/SA tools, RAVEN and Dakota. The RAVEN software, maintained by Idaho National Laboratory, was initially developed to implement control logic, generate dynamic event trees and enable high level uncertainty quantification analyses for RELAP-7. However, the code's capabilities have been expanded such that RAVEN now has the ability to interface with any software and provides the user access to a wide range of SA/UQ methodologies, surrogate models, and data visualization. The Dakota software, maintained by Sandia National Laboratories, is an UQ and optimization toolkit with over 20 years of supporting development. At a high level, both tools perform the same functionality and offer similar features, although both codes do not offer equivalent advanced capabilities. This paper explores the capabilities of RAVEN and Dakota, and provides demonstration applications of SAS4A/SASSYS-1 coupled with RA VEN with respect to UQ for transient analysis in an advanced reactor. Specifically, an unprotected loss of heat sink and transient overpower scenarios in the Advanced Burner Test Reactor are presented. In these cases, the uncertainty analyses largely focus on inherent reactivity feedback mechanisms, which are vital to ensuring reactor safety in an SFR. While this work quantifies the effects of uncertainties on a key safety component (inherent reactivity feedback) of an SFR at a high level, the primary objective is to demonstrate the coupling capabilities and flexibility of RAVEN, Dakota, and SAS4A/SASSYS-1.