ML24304A829

PSAM17 Summary of Panel Session on Advanced Reactor PSA Needs

ML24304A829
Person / Time
Issue date:	10/10/2024
From:	Coyne K NRC/RES/DRA
To:
References
Download: ML24304A829 (5)

Text

Summary of PSAM 17 Panel Session on Probabilistic Safety Assessment (PSA) Needs for Advanced Reactors

Date: October 10, 2024; Sendai, Japan

Participants in Session: ~70 Panelists:

• Michelle (Shelby) Bensi (University of Maryland - UMD)

• Seung Jun Lee (Ulsan National Institute of Science and Technology, Republic of Korea -

UNIST)

• Eric Thornsbury (Electric Power Research Institute - EPRI)

Chair: Kevin Coyne (United States Nuclear Regulatory Commission - NRC)

Session Summary:

Kevin Coyne opened the session with a summary of the key points contained in the Panel Session extended abstract highlighting the motivation for the session, examples of several potential challenge areas for advanced reactor PSA, and a quick overview of the session format (which was intended to focus heavily on participant discussion and exchange of ideas). Each panelist then delivered opening remarks:

• Eric Thornsbury provided an overview of EPRI research on advanced reactors risk informed technology. He described the EPRI research roadmap which identified several gaps and opportunities for advanced reactor PSA (the roadmap is publicly available:

EPRI Report 3002026495). Eric also noted the need to leverage tools and technology to support the use of PSA. Some particular issues noted include data (including component performance but also performance under new material and environment conditions), the need for new technology neutral risk metrics, for streamlined risk approaches, and treatment of low frequency hazards (such as external events) which could benefit from improve treatment. EPRI has a taken a dynamic and flexible approach and incorporated new issues as needed, such as collocated hydrogen facilities, etc.

• Seung Jun Lee noted the evolution from fully digital instrumentation and control in modern light water reactors such as the APR1400 to new concept of operation that depart from traditional control room design and allocations of functions to operators such as that for NuScale small modular reactor design (SMR). Seung Jun noted that software errors are caused by human error. He also gave several examples of work done in Korea to evaluate every possible case for software inputs in order to generate statistical testing test cases. Seung Jun discussed work done for digital I&C reliability that identified a new human reliability assessment (HRA) performance shaping factor to better estimate software reliability and support collection of HRA data relevant to digital systems. He further noted the importance of self-diagnosis for software reliability and potential uses of artificial in telligence and machine learning for areas such as verification and validation or more direct support in safety systems. With regard to cyber security, Seung Jun noted use of reasonable design objectives such as a single compromise event should not lead to a significant consequence. Seung Jun identified several main

1 issues of concern, including: (1) software re liability; (2) cyber security (noting it may be difficult to develop a perfect strategy; therefore, risk information should be used to support cyber security priorities); and (3) HRA for SMRs.

• Shelby Bensi organized her opening remarks around five broad topical areas: new technology, new environments, regulatory needs, PSA methods and metrics, and workforce needs. Key points in each of these topics included the following: -

1. New technology: need to consider presumptions about safety and their impact on public perception(e.g., walk away safe); limited data and limited experience performing PSA for advanced reactors; advanced reactors are still largely conceptual - i.e., paper plants, and a lack of system knowledge.

2. New environments: Need to consider emerging threats, siting in new locations and near populations, climate change, new (old) hazards (some hazards not assessed in detail - individual and combined). Other considerations such as transportation of fueled modules.

3. Regulatory needs and guidance: non-stationary RIDM (e.g., considering a changing regulatory environment, guidance vs practice, regulatory uncertainty, guidance consistency, use of graded approach). More importantly, how can standards reflect state of practice when state of practice is being developed?

4. PSA methods & metrics: challenges include operational data scarcity, opportunities to apply innovative methods (which provide a chance for a fresh start and the potential to increase transparency of modeling); latent conditions; achieving transparency in analytical methods; and autonomous/automated operations.

5. Workforce needs: a talented workforce will be needed. to meet industry needs will need educational programs, which depend on workforce projections. These projections are better when information is shared. Also need to consider the competition that nuclear has with PSA trained students - other technical sectors need similar skills and knowledge (information technology, manufacturing, transportation, etc.).

Shelby also noted collaborative work being done between the University of Maryland and Virginia Tech and provided a link to a survey on PSA needs (https://go.umd.edu/AdvRxPSA ). This is a short, anonymous survey and responses to the survey will help develop advanced reactor research plans.

Discussion and Audience Questions:

The session had good engagement from participants with several audience members offering observations and questions for the panel. Key themes that emerged from the open discussion include the following:

• Broadly scoping research needs for advanced reactors - One comment/question from a session participant noted the need to more broadly consider risks across all areas of the fuel cycle. Further, leveraging knowledge from different techniques used for assessments at the

2 front-end (e.g., integrated safety assessment for fuel cycle facilities) and the back-end (e.g.,

waste packaging performance assessments) may be beneficial for reactor PSAs.

Furthermore, sharing lessons from the reactor community could benefit "front-end" and "back-end" applications. The panelists agreed with the need to consider broader scoping boundaries for the advanced reactor effort. With regard to fuel cycle risks, it was noted that we tend to keep these separate for the current operating fleet, but we should consider more integrated assessments for advanced reactors, particularly in light of potential changes to the fuel cycle including use of high assay lo w enrichment uranium, TRISO fuel forms, new storage considerations for spent fuel, etc. It was acknowledged that research activities would benefit from more voices, more perspective, and a broader community.

• Remote and autonomous operation - There are lessons to be learned from other infrastructure applications that rely on remote operations (including potential security challenges). We also need to look at the integrated facility - need to consider both likelihood and consequences of potentially challenging scenarios within the context of the whole facility. Implementation of simple and robust systems will be beneficial.

• Workforce considerations - Shelby highlighted the need for investment in education for the next generation workforce. Further, given the typical curriculum planning cycle for universities, strong communication is needed with the industry and regulator to help plan for future needs. Seung Jun noted limitations with the availability of experts to support certain application areas, resulting in the need to prioritize activities to make best use of available staff.

• Safety expectations for advanced reactors - The panel noted expectations for use of more reliable safety systems for advanced reactors and reactor designs that may minimize the release of radionuclides to the environment (e.g., low pressure reactor coolant, fission product retention within the reactor coolant, deployment of TRISO fuel, etc.). However, given the current global environment, it was noted that an accident anywhere is an accident everywhere (e.g., enterprise risk for the industry, reputational risk for regulators, public confidence, and other risks). There are also considerations associated with siting advanced reactors in areas with higher hazards levels or areas that may be impacted by climate changes. The scale of deployment of advanced technology should also be considered (e.g.,

potential numbers microreactors, SMRs or other designs). Lastly, the economic impact to investors and operators associated with accidents needs to be considered.

• Digital systems - Participants noted that digital technology is used with advanced reactors and across the fuel cycle (e.g., deployment in fuel facilities, waste management, etc.) Digital system failures arise from human failures - either in the software/hardware design, development of system requirements, or im plementation of requirements. Challenges include availability of data and use of digital technologies in potential new ways (such as supporting remote or autonomous operations). Also need to consider new concepts of operation associated with deployment of digital technology, including artificial intelligence, that may change the role of the human operator. A risk-informed strategy to address cyber-security should be developed that considers hardware and software reliability, physical security, and human reliability analysis. A potential path forward is to leverage the

3 knowledge gained in other industries, such as aerospace, to gain information on these technologies.

• Safety goal consideration and potential benefits of nuclear power - A participant questioned if a broader risk picture should be considered, for example, consider impact to society of not using nuclear energy in the future. The panel noted that in the US, safety goals compare the risks of nuclear power to other competing technologies. While this is a worthwhile measure to consider, it requires a more holistic consideration of societal risks. In other words, safety goals need to be within the context of society and the consideration of risk impacts of alternate power sources such as fossil fuels. Assessments should also consider the robustness of nuclear facilities and how quickly power generation sources can be brought back online after upset conditions.

• Other PSA methods and tools - Panelists noted that we should select the tools appropriate for the use. Dynamic PSA can provide insights that may have several uses such as validation of the static PSA or checking to see if there are missing sequences; however, we need to have explainable results. For exter nal hazard events, dynamics may matter for earthquake/high wind, combination of hazards. Hybrid tools will continue to be important, but we need to start thinking about simplicity and transparency for the tools. For example, can we use online risk management tools that go beyond stop light type of approach. We should strive to use the right tools for the right problem - and when possible, use the simpler tool. (One session participant noted that "simple" is in the mind of the beholder. What is "simple" to a generation trained to the use of current reactor PSA tools, including the familiar event tree/fault tree methodology, may not be simple - or even natural - to a generation of analysts trained to think in terms of more explicit/literal simulation modeling.) Also need to avoid becoming overly comfortable with our risk tools - what seems to the right now may not be in the future should do an integrated assessment, including consideration of other competing technologies. Changing from what is comfortable may allow us to develop simpler and more transparent risk models. Also noted a need to focus on development of methods where current methods are not adequate.

• PSA Challenges - Do simpler designs imply simpler PSAs? We need to consider types of uncertainty, as we get simpler and get simpler models need to get better with considering uncertainties. Its important to not repeat errors of the past - not doing PSA just puts us back in time, do not take away the systematic process. A key question is if the designs are truly simpler, or are we just changing where the complexity lies. For example, a passive system might be simpler from a hardware perspective, but require far more complex analysis and design than a more traditional active system. Public expectations for safety may be getting higher - completeness and fully evaluating the scope of uncertainty are important. In some cases, you dont know what you dont know for new designs. Need to think about risk and resilience. Although we can learn from other industries need to consider how far along nuclear PSA has come - for example, the chemical industry uses just mean value while nuclear uses mean plus uncertainty.

• Risk communication - Need better ways to communicate risk insights would be helpful. It was noted that the risk communication keynote speech for the conference was very lightly attended. A critical question is how does one go about communicating risk, how to develop

4 trust? Need to leverage and reach out to field of practitioners to increase the engagement of that community and improve communications (methods as well as practice). A broader community has the knowledge and communications experience needed by the PSA community. The panel also recognized that a nu mber (e.g., core damage frequency) is not a good way to communicate with the public - need to put risk in proper context to help people make informed decisions (e.g., can I live here?).

• Other considerations include development of new licensing and regulatory strategies for advanced reactors and potential transportation considerations. Consideration of a broad range of stakeholders and risks may be needed (e.g., enterprise risk, political, reputational, financial, public confidence, etc.). PSA should be part of the design process for advanced reactors Concluding Thoughts by Panel Members:

• Eric Thornsbury - The risk community has been gathering in conferences and other meetings for a long time. The technology was never perfect, but it has helped us improve safety and increase reliability. What else can we do with this technology to support advanced reactors?

• Shelby Bensi - A big insight from this discussion is that looking beyond just safety, we can step back a little bit to understand things a bit more holistically. We need to be ready to do things in new ways. And invest in workers - this includes education, internships, etc.

• Seung Jun Lee - Looking forward, need to consider higher levels of automation in advanced reactors and human reliability analysis. Korea has experience in successfully deploying a fully digital system in the APR1400, but the nuclear field has a high bar against new technologies; we need to understand we can derive benefit from these new technologies.

5