Valuable Lessons from Fuel Cycle Code Comparisons

Bo Feng

Argonne National Laboratory

Numerous nuclear fuel cycle system code benchmarks and comparisons have been performed over the last decade, coinciding with the rapid development of new codes due to improvements in computational capabilities, new software platforms, and the need for various institutions to provide technical feedback on potential fuel cycle strategies and policies. Many of these studies were performed within the framework of established international organizations or organized as an ad hoc study with voluntary contributions from participants. These studies achieved different levels of agreement depending on the scenarios analyzed and values compared, varying from excellent agreement in annual mass flows and inventories to general agreement in terms of trends. Since validation is challenging for these types of codes, these comparison studies helped develop confidence in the results from these forecasting codes. In addition, many of these codes were developed independently with limited feedback due to the lack of a widely-established user base. Therefore, such studies are also great opportunities for the developers and users to calibrate interpretations as well as modify/debug the codes themselves.

However, in almost all of these benchmarks and code comparisons, the same challenges and discrepancies keep re-appearing, and may continue to be “re-discovered” through future benchmarks. Therefore, this presentation is an effort to document such similarities and valuable lessons learned from these exercises so that future verification studies can focus on comparing more advanced capabilities and features rather than on these expected fundamental differences. The author presents some of the key lessons that he has observed through his participation in various code comparisons and benchmarks with the goal of stimulating discussion and encouraging the community to be aware of expected differences between codes. Some of the valuable lessons to be discussed include: 1) most of the differences in code results were not due to different code algorithms or calculation approaches, but due to different interpretations of the input specifications among the analysts, 2) the first specifications will almost never be the final version and will almost always require being iteratively updated with preliminary results from codes, 3) different codes often report or account for mass inventories at different times within a given year or even time-step that often lead to differences that are misleading, 4) for more complex scenarios, all codes may need to make approximations at one level or another (in the input or modeling of mass flows), and others.