ML22095A236

RILXX-XX, Predicting High Energy Arcing Fault Zones of Influence for Aluminum Using a Modified Arc Flash Model, Draft

ML22095A236
Person / Time
Issue date:	04/05/2022
From:	Lafleur C, Gabe Taylor NRC/RES/DRA, Sandia
To:
Taylor, Gabriel 301-415-0781
Shared Package
ML22095A236	List: ML22095A236
References
Download: ML22095A236 (53)

Text

RILXX-XX PREDICTING HIGH ENERGY ARCING FAULT ZONES OF INFLUENCE FOR ALUMINUM USING A MODIFIED ARC FLASH MODEL Evaluation of a modified model bias, uncertainty, parameter sensitivity and zone of influence estimation Draft for public comment Research Information Letter Office of Nuclear Regulatory Research Date Draft Published: April 2022 Prepared by:

G. Taylor1 C. LaFleur2 1 U.S. Nuclear Regulatory Commission 2 Sandia National Laboratories Mark Henry Salley, NRC Project Manager

Disclaimer Legally binding regulatory requirements are stated only in laws, NRC regulations, licenses, including technical specifications, or orders; not in Research Information Letters (RILs). A RIL is not regulatory guidance, although NRCs regulatory offices may consider the information in a RIL to determine whether any regulatory actions are warranted.

iii ABSTRACT 1

This report documents the evaluation and modification of an existing arc-flash hazard model to 2

calculate the incident energy and zone of influence from high energy arcing faults involving 3

aluminum. The NRC has identified the potential for (HEAFs) involving aluminum to increase the 4

damage zone beyond what is currently postulated in fire probabilistic risk assessment (PRA) 5 methodologies. To estimate the hazard from HEAFs involving aluminum an existing model was 6

evaluated. Differences between the base model and nuclear power plant (NPP) fire PRA 7

scenarios were identified. Modification of the base model established from existing literature and 8

test data was used to minimize these differences. The modified model was evaluated against 9

NRC datasets to understand the model prediction and relative uncertainties. Finally, a range of 10 fire PRA zone of influences (ZOI) were developed based on the modified model, target fragility 11 estimates and update HEAF PRA methodology. The results are expected to be used to inform 12 an update to ZOIs used in fire PRA.

13 14 15 Keywords 16 17 High Energy Arcing Fault, Arc Flash, Electrical Enclosure, Fire Probabilistic Risk Assessment, 18 Zone of Influence 19 20 21 22

v TABLE OF CONTENTS ABSTRACT............................................................................................................................... iii LIST OF TABLES..................................................................................................................... vii LIST OF FIGURES.................................................................................................................... ix EXECUTIVE

SUMMARY

........................................................................................................... xi 1 Introduction.......................................................................................................................1-1 1.1 Background..................................................................................................................1-1 1.2 Overview of HEAF Research........................................................................................1-2 1.3 Objective......................................................................................................................1-2 1.4 Scope...........................................................................................................................1-2 2 Overview of Model and identification of areas of dissimilarities...................................2-1 2.1 Introduction..................................................................................................................2-1 2.2 Base Model Overview...................................................................................................2-1 2.2.1 Calculation process........................................................................................2-1 2.2.2 Inputs 2-2 2.2.3 Model outputs.................................................................................................2-3 2.2.4 Assumptions...................................................................................................2-3 2.2.5 Limitations......................................................................................................2-4 3 Evaluation..........................................................................................................................3-1 3.1 Evaluation Process.......................................................................................................3-1 3.2 Overview of data..........................................................................................................3-1 3.3 Base model evaluation.................................................................................................3-3 3.4 Breach modification evaluation.....................................................................................3-6 3.4.1 Time to initial breach (tb).................................................................................3-8 3.4.2 Time to open from initial breach (to)................................................................3-9 3.5 Model breach and open time evaluation..................................................................... 3-11 3.6 Generator fed - energy conservation modification evaluation..................................... 3-12 3.7 Model Sensitivities...................................................................................................... 3-15 4 Zone of influence...............................................................................................................4-1 5 Summary............................................................................................................................5-1 6 References.........................................................................................................................6-1 APPENDIX A Arc Voltage......................................................................................................... 1 APPENDIX B ZOI TABLE MODELING INPUT........................................................................... 1

vii LIST OF TABLES Table 1. Summary of open box tests........................................................................................3-1 Table 2. Summary of Medium-voltage Switchgear Tests..........................................................3-3 Table 3. PRA scenario time to Model Time for Decrement Scenarios.................................... 3-14 Table 4. Constant duration decrement current modification based on an initial decay current of 32kA (IT = 32kA).............................................................................. 3-14 Table 5. Medium-voltage Switchgear [30kA, 6.9kV, 0.09in Steel Enclosure]............................4-2 Table 6. Medium-voltage Switchgear [30kA, 6.9kV, 0.125 Aluminum Enclosure].....................4-2 Table 7. NSBD Medium-voltage [30kA, 6.9kV, 0.09 Steel Enclosure]......................................4-3 Table 8. NSBD Medium-voltage [30kA, 6.9kV, 0.125 Aluminum Enclosure].............................4-3 Table 9. Low-voltage Switchgear [32kA, 0.6kV, 0.09 Steel Enclosure].....................................4-4 Table 10. Low-voltage Switchgear [32kA, 0.6kV, 0.125 Aluminum Enclosure].........................4-4 Table 11. Arc Voltage Measurements for Low-voltage Tests....................................................... 2 Table 12. Voltage Measurements for Medium-voltage Tests....................................................... 3 Table 13. CIGRE 602 estimates of arc voltage for 2018 tests..................................................... 5

ix LIST OF FIGURES Figure 1. Open box medium-voltage test configuration. Vertical sensor array (right),

Horizontal sensor array (bottom left).................................................................3-2 Figure 2. Medium-voltage Switchgear Test Configuration (Test device shown at left, Instrument Racks #2 and #3 used for evaluation).............................................3-2 Figure 3. Open Box - Base Model VCB (Bias = 0.68; SD_M = 0.44) Alum - blue, Cu -

Orange..............................................................................................................3-3 Figure 4. Switchgear MV - Base Model, short duration blue, long duration red (Bias =

2.33; SD_M = 0.74) [Blue : 2 sec; Red : 4 sec]..................................................3-4 Figure 5. Switchgear MV - Base Model by Duration. Left 2 seconds overprediction (Bias

= 3.56; SD_M=0.24) Right 4 seconds underprediction (Bias = 0.93; SD_M=0.19).....................................................................................................3-4 Figure 6. Photographs of 2 sec test enclosures (Left 2-19 25.8kA 39MJ / Right 2-22 32.0kA 51MJ)....................................................................................................3-5 Figure 7. Photographs of 4 sec test enclosure (Left 2-21 26.6kA 101MJ / Right 2-24 29.8kA 122MJ)..................................................................................................3-6 Figure 8. Illustration of Arc timeline (Not to Scale)....................................................................3-7 Figure 9. Modified burn-through steel enclosure (Bias = 0.90, SD_M = 0.18)...........................3-8 Figure 10. Steel enclosure mass loss versus electrical energy with aluminum bus bars (2018 arc energy estimated).............................................................................3-9 Figure 11. Steel enclosure mass loss versus arc energy (with aluminum or copper bus bars)............................................................................................................... 3-10 Figure 12. Switchgear MV - Breach Modification Model by Duration. Left 2 seconds overprediction (Bias = 2.29; SD_M=0.24) Right 4 seconds underprediction (Bias = 0.78; SD_M=0.17)..................................................... 3-11 Figure 13. Open Box MV - Breach Modification Model by Duration (note OBMV05 did not breach). (Bias = 0.79; SD_M=0.26)........................................................... 3-12 Figure 14. Reference plant decrement curve......................................................................... 3-12 Figure 15. Example of constant time approxiation showing current level and corresponding duration for select cases based on 32kA base Decay curve.... 3-14 Figure 16. Comparison of Decay Curve Discretizing Approaches.......................................... 3-15 Figure 17. Base Case comparison between HCB and VCB................................................... 3-16 Figure 18. Model Sensitivity Plots (Clockwise from upper left: Current, Conductor Spacing, Volume, and System Voltage........................................................... 3-17 Figure 19. Voltage Profile prior to (System Voltage) and during (Arcing Voltage) initial arcing duration (Voltage shown as Phase Voltage).............................................. 1 Figure 20. Interval Plot (left) and Box Plot (right) of LV Arc Voltage Data.................................... 3 Figure 21. Model evaluation for Low-voltage box tests (Left) and Low-voltage Switchgear voltage (Right) to the CIGRE 602 arc voltage model. All voltage shown

x are Line-to-Line. [Low-voltage box Bias = 0.91, SD_M=0.29] [Low-voltage switchgear Bias = 1.33, SD_M=0.18]................................................................... 4 Figure 22. Model Evaluation for Medium-voltage box tests to the CIGRE 602 arc voltage model [Medium-voltage Bias = 1.08, SD_M=0.17]............................................... 5 Figure 23. Bus Duct Specification (Aluminum bus and duct)....................................................... 1 Figure 24. Input Parameters for non-segregated bus duct.......................................................... 2 Figure 25. Medium-voltage Switchgear....................................................................................... 3 Figure 26. Input Parameters for Zone 1 and 2 MV Switchgear.................................................... 3 Figure 27. Low-voltage switchgear.............................................................................................. 4 Figure 28. Input Parameters for Zone 3 LV Switchgear............................................................... 5

xi EXECUTIVE

SUMMARY

1 2

PRIMARY AUDIENCE: Fire protection, electrical and probabilistic risk assessment engineers 3

conducting or reviewing fire risk assessments related to high energy arcing faults.

4 SECONDARY AUDIENCE: Engineers, reviewers, utility managers, and other stakeholders who 5

conduct, review, or manage fire protection programs and need to understand the underlying 6

technical basis for the hazards associated with high energy arcing faults.

7 KEY RESEARCH QUESTION: How do you calculate the zone of influence for a high energy 8

arcing fault (HEAF) involving aluminum?

9 RESEARCH OVERVIEW 10 Operating experience has shown that HEAFs pose a hazard to the safe operation of nuclear 11 facilities. Current regulations and probabilistic risk assessment methods were developed using 12 limited information, and uncertainties require the use of safety margins to bound the hazard.

13 Testing aimed at providing additional data to improve realism identified a concern that HEAFs 14 involving aluminum may increase the hazard potential. Testing identified that the presence of 15 aluminum during a HEAF may increase the hazard potential.

16 Upon discovery of the potential hazard posed by aluminum, the NRC staff entered the issue into 17 the NRCs generic issues (GI) process and informed licensees of relevant operating 18 experienced in Information Notice 2017-004 [1]. The NRC GI process identified a need for 19 specific data of HEAF tests involving aluminum which were performed in 2018 on 20 medium-voltage switchgear and in 2019 on low-voltage switchgear and simplified box tests at 21 medium-and low-voltage levels. Planned testing on medium-voltage bus ducts and other 22 equipment was postponed due to the COIVD-19 pandemic. To make progress with available 23 test data and complete the risk assessment of the concern, an analytical effort was performed to 24 develop revised zone of influences (ZOI) based on available data, literature, and existing 25 models.

26 This report documents the results of evaluating an existing arc flash hazard calculation (base 27 model) to predict incident energy at various distances. The base model is modified due to 28 identified differences between model and fire PRA assumptions. This resulted in a modified 29 model that is evaluated against NRC data to understand model bias and relative uncertainty.

30 Using the modified model and target fragility estimates, ZOI estimates were developed for 31 specific HEAF PRA scenarios.

32 KEY FINDINGS 33 This research yields an simple empirically-based approach to estimate the zone of influence for 34 equipment containing aluminum components. Development and use of this approach identified 35 the following key findings:

36 The base model underpredicts the aluminum test data for open box test configurations 37 The base model underpredicts the incident energy for longer duration switchgear tests 38 (where the exothermic aluminum reaction was observed). The base model overpredicts 39

xii short duration tests, which is contributed to the lack of observed aluminum oxidation 1

reaction and differences in model assumptions and tests data.

2 Gap analysis identified a need to modify the model to account for enclosure breach 3

characteristics. The time to breach an electrical enclosure from a HEAF source can be 4

estimated based on existing models from gas insulated substation research, with 5

modification based on geometrical differences.

6 o The time to breach is dependent on enclosure material. Aluminum enclosures will 7

breach approximately 4 times faster than a steel enclosure of equivalent 8

thickness and fault current.

9 o Additional time is required for the initial breach opening to enlarge for sufficient 10 energy transfer to the targets. This can be estimated by analyzing existing data 11 sets.

12 Generator decay profiles for specific scenarios can be approximated by a constant 13 current (shorter duration) profile that conserves energy.

14 A sensitivity study found that the modified model is 15 o relatively insensitive to the conductor spacing, enclosure volume and system 16 voltage 17 o moderately sensitive to the arcing current 18 o has a linear relationship with the arc duration 19 o model configuration has an impact on the model output 20 Zone of influences (ZOIs) were developed based on the modified model and target 21 fragility estimates with the following results. These ZOI estimates do not include any 22 contribution from an ensuing fire.

23 o Non-segregated bus duct 24 The spherical ZOI increases from the current guidance (0.46 m [1.5 ft]) for 25 most scenarios. The largest ZOI involves aluminum bus duct with a 26 targeted aluminum bus duct which is estimated at 1.2 m (4 ft).

27 o Medium-voltage switchgear 28 Increase in ZOI estimates over current guidance was predicted for some 29 scenarios. The lower target fragility category (15 MJ/m2) had an upper 30 predicted ZOI of 1.6 m (5.4 ft), while the higher target fragility category 31 (30 MJ/m2) had a marginal increase for only a few of the longest fault 32 current scenarios 1.1 m (3.6 ft).

33 o Low-voltage switchgear 34 No scenarios involved the ZOI exceeding the current fire PRA guidance 35 of 0.91 m (3 ft).

36 37

xiii WHY THIS MATTERS 1

This report provides a numerical approach to estimate the zone of influence for HEAFs involving 2

aluminum. These estimates represent an informational input to support improvements to how 3

HEAFs are evaluated in NPP risk assessment to ensure the publics health and safety.

4 HOW TO APPLY RESULTS 5

Engineers performing fire probabilistic risk assessment method advancements involving 6

aluminum HEAFs should focus on Section 4 of this report.

7 LEARNING AND ENGAGEMENT OPPORTUNITIES 8

Users of this report may be interested in the following learning opportunities:

9 Nuclear Energy Agency (NEA) HEAF Project to conduct experiments in order to explore the 10 basic configurations, failure modes and effects of HEAF events. Primary objectives include 11 (1) development of a peer-reviewed guidance document that could be readily used to assist 12 regulators of participants, and (2) joint nuclear safety project report covering all testing and 13 data captured. More information on the project and opportunities to participate in the 14 program can be found online at https://www.oecd-nea.org/.

15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

1-1 1 INTRODUCTION 1

2 Events such as fires at a nuclear power plant can pose a significant risk to safe plant operations 3

when consequences of fires are not mitigated. Licensees combat this risk by having robust fire 4

protection programs designed to minimize the likelihood and consequences of fire. These 5

programs provide reasonable assurance of adequate protection from known fire hazards.

6 However, several hazards remain subject to a large degree of uncertainty, requiring significant 7

safety margins in plant analyses.

8 One such infrequent hazard comprises an electrical arcing fault involving electrical distribution 9

equipment and components comprised of aluminum. While the electrical faults and subsequent 10 fires are considered in existing fire protection programs, recent research [1] has indicated that 11 the presence of aluminum during the electrical fault can exacerbate the damage potential of the 12 event. The extended damage capacity could exceed the protection provided by existing fire 13 protection features for specific fire scenarios and increase plant risk estimated in fire 14 probabilistic risk assessments (PRAs).

15 The U.S. Nuclear Regulatory Commission (NRC) Office of Nuclear Regulatory Research (RES) 16 studies fire and explosion hazards to the safe operation of nuclear facilities. This includes 17 developing data, tools and methodologies to support risk and safety assessments.

18

1.1 Background

19 HEAFs are hazardous events in which an electrical arc leads to the rapid release of energy in 20 the form of heat, vaporized metal, and mechanical force. The guidance for modeling HEAF 21 events in fire probabilistic risk assessments (PRA) is documented in Appendix M of 22 NUREG/CR-6850, EPRI/NRC-RES Fire PRA Methodology for Nuclear Power Facilities [2].

23 This guidance postulates that HEAFs can occur in switchgear, load centers, and bus ducts with 24 a nominal voltage of 440V and above, and defines a zone of influence (ZOI) in which targets are 25 assumed to be damaged.

26 An OECD/NEA report [3], published in June of 2013, documented 48 HEAF events, accounting 27 for approximately 10% of the total fire events reports in the international fire records exchange 28 program database. These events were often accompanied by loss of essential power and 29 complicated shutdowns. To confirm the PRA methodology in NUREG/CR-6850, which was 30 formulated based on limited observational data, the NRC led an international experimental 31 campaign from 2014 to 2016 [4]. The results of these experiments uncovered an unexpected 32 hazard posed by aluminum components in or near electrical equipment and the potential for 33 unanalyzed equipment failures, which the current PRA modeling guidance does not address.

34 Upon discovery of the potential hazard posed by aluminum, the NRCs Office of Nuclear 35 Reactor Regulation conducted an immediate safety evaluation and concluded that no immediate 36 safety concern exists, but recommended that the NRCs Office of Nuclear Regulatory Research 37 (RES) begin the generic issues (GI) process. Additionally, RES staff conducted a review of 38 operating experience, and identified six events from the U.S. operating fleet where aluminum-39 related effects like those observed in testing were present. To inform licensees about the 40 findings of this review and results of testing, the NRC issued Information Notice 2017-004 [1].

41

1-2 NRC-RES staff proposed this potential safety concern as a GI in a letter dated May of 2016 [5].

1 The Generic Issue Review Panel (GIRP) completed its screening evaluation [6] for proposed 2

Generic Issue (GI) PRE-GI-018, HighEnergy Arc Faults (HEAFs) Involving Aluminum, and 3

concluded that the proposed issue met all seven screening criteria outlined in Management 4

Directive (MD) 6.4, Generic Issues Program. Therefore, the GIRP recommended that this 5

issue continue into the Assessment Stage of the GI program. The assessment plan, published 6

in August of 2018 [7] and revised in 2019 [8], requires the NRC to develop updated PRA tools 7

and methods for HEAFs to be used in pilot plant studies and risk evaluation. In August of 2021, 8

the NRC determined that the issue no longer met the timeliness criterion of the generic issue 9

program [9]. The NRC decided to continue to evaluate the issue by applying the NRC LIC-504, 10 Integrated Risk-Informed Decision-making Process for Emergent Issues [10].

11 1.2 Overview of HEAF Research 12 The objective of the NRCs HEAF research program is to develop tools and methods to assess 13 the risk posed by high energy arcing fault events based on experimental data, operating 14 experience, and engineering judgment. These tools and methods will account for the primary 15 factors that influence the occurrence and severity of HEAF events, including the presence of 16 aluminum and plant electrical configuration and protection schemes.

17 To leverage the expertise of collaborative partners, NRC-RES and the Electric Power Research 18 Institute (EPRI) formed a joint working group under the NRC-RES/EPRI memorandum of 19 understanding (MOU). This working group has developed a list of tasks needed to support 20 advancements to modeling HEAF in fire PRAs. One of these tasks includes development and 21 validation of a HEAF hazard model. The working group has collaboratively focused on using 22 computational fluid dynamic tools to estimate the HEAF hazard. This report provides an 23 alternative approach using a modified empirically derived approach. The work documented in 24 this report was prepared by NRC staff, with PRA scenario development support from Sandia 25 National Laboratories.

26 1.3 Objective 27 The objective of this report is to document how the arc flash model in IEEE 1584-2018, Base 28 model for Performing Arc-Flash Hazard Calculations [11], has been modified for the application 29 needs of the NRCs assessment of high energy arcing faults (HEAFs) involving aluminum. This 30 report also documents model bias and uncertainty, along with a parameter sensitivity evaluation 31 of the modified model.

32 1.4 Scope 33 The scope of this hazard modeling sub-task is to 34

1. present an overview of the IEEE arc flash model, 35

2. identify differences between the model and HEAF probabilistic risk assessment (PRA) 36 scenarios 37

3. provide basis for modification to model 38

4. evaluate the model against empirical data available to the NRC 39

5. document estimated HEAF ZOIs for an array of scenarios.

40

2-1 2

OVERVIEW OF MODEL AND IDENTIFICATION OF AREAS OF 1

DISSIMILARITIES 2

2.1 Introduction 3

Modeling the hazards associated with electrical arc flashes has been ongoing since at least the 4

early 1980s by the well-known work of Ralph Lee [12]. However, experience over time indicated 5

that Lees formulas didnt reconcile the greater thermal effect on persons positioned in front of 6

open doors or removed covers [11]. Work by Doughty, Neal and Floyd identified the contribution 7

of thermal energy from electrical enclosure interior surfaces. In 2002, the Institute of Electrical 8

and Electronic Engineers (IEEE) published its initial model using new, empirically derived 9

models based on statistical analysis and curve fitting of the overall test data. While the 2002 10 model has been used successfully, it was also recognized that not enough arc-flash incident 11 energy testing had been done from which to develop models that accurately represent all of the 12 real applications [13].

13 In 2018, the IEEE revises its model for performing arc-flash hazard calculations [11] (referred to 14 as the base model in this report). The revision to the base model was based on a multi-year, 15 multi-million-dollar research program organized by IEEE and the National Fire Protection 16 Association (NFPA) with support the industry vendors and organizations. The model 17 documented in the base model uses over 1,860 tests to develop an empirical model based on 18 statistical analysis and curve fitting. The base model expands the number of configurations from 19 two in the 2002 edition to five in the 2018 edition.

20 2.2 Base Model Overview 21 The 2018 version of the base model used to estimate incident energy at a specified distance 22 and the arc flash boundary for personal using electrical system characteristics as inputs. The 23 base model is applicable over a wide parameter range that covers the majority of the fire PRA 24 scenarios [11]. The base model is empirically derived based on statistical analysis and curve 25 fitting to the overall test data available with the understanding of the underlying electrical arc 26 physics. Based on the results of the IEEE/NFPA research, the base model incident energy 27 calculation is primarily dependent on arc current, duration, and distance. Bus gap has a smaller 28 influence on incident energy than these three parameters. The base model is linearly dependent 29 on arc duration, and an inverse exponential relationship with distance. The reader should review 30 IEEE 1584-2018 for a full description of the model and an understanding of its development and 31 use.

32 The subsequent sub-sections provide a description of the overall calculation process, inputs and 33 outputs, assumptions and limitations. Where differences between the base model and the 34 needs for the NRC assessment exist, they are identified. Evaluation and resolution of those 35 differences are contained in Section 3.

36 37 2.2.1 Calculation process 38 The process used to perform the calculation is presented in the base model [11]. It involves 39 collecting the model input parameters, followed by performing a number of iterative calculations 40 of arcing current and incident energy to arrive at the final incident energy and arc flash 41 boundary.

42

2-2 For the purposes of the NRC assessment, the calculation is the same, however, the process is 1

different in several ways based on the information available and differences in targets damage 2

levels. These differences are summarized below.

3 The arc flash boundary is not needed 4

o This is based on limits for sustaining injury to humans and not applicable to fire 5

PRA targets such as electrical cables and bus duct enclosures. As discussed 6

below the zone of influence (ZOI) will be determined using the HEAF PRA target 7

fragility thresholds [14]. The basis for these thresholds is presented in Target 8

Fragilities for Equipment Vulnerable to High Energy Arcing Faults (U.S. Nuclear 9

Regulatory Commission, Electric Power Reserach Institute 2022).

10 Model uses bolted fault current 11 o Updated HEAF PRA scenarios (EPRI / U.S. NRC 2021) present arcing fault 12 current and not bolted fault current. Bolted fault current will be iteratively solved 13 for convergence to the desired arcing fault current for the HEAF PRA scenario.

14 The distance to target is not known.

15 o The distance from the initiating source to the target is a desired output and not an 16 input. The calculation will be solved iteratively to find the distance where the 17 incident energy converges to the target fragility threshold. The distance where 18 the hazard incident energy equals the target fragility threshold is then the zone of 19 influence (ZOI) for the specific hazard scenario.

20 Additional modifications to the calculation are made to account for dissimilarities 21 between the base model assumptions and the PRA model assumptions, including 22 insights from operational experience. The specific modifications are discussed in Section 23 3.4 and include; 24 o Enclosure breach and opening delay 25 o Model bias adjustment 26 o Generator fed decay 27 28 2.2.2 Inputs 29 Several inputs are required, as described below.

30 Conductor Gap Spacing (G) 31 This is the gap distance between the electrodes. In most cases, acquiring this information may 32 be difficult. However, several resources online are available to provide representative distances.

33 For the evaluation of the base model, the test data sets included this measurement.

34 Open Circuit Voltage (Voc) 35 This is the pre-fault voltage. For this input the actual system nominal voltage or utilization 36 voltage can be used. Examples of nominal system voltage include 480V, 4.16kV, 6.9kV, 13.2kV, 37 etc.

38 Arc Duration (T) 39 This is the total duration of the arc. The base model does not identify an upper limit.

40

2-3 Distance (D) 1 The distance between the arc and the fire PRA target.

2 Bolted fault current 3

Determining the bolted fault current at the fault location requires detailed knowledge of the 4

electrical distribution system design and set points. This information is not readily available to 5

the staff. However, the fire PRA scenarios have identified or calculated the ranges of arcing fault 6

current that were experienced during actual HEAF events (i.e., operational experience). The arc 7

fault current is an output of the base model. Therefore, the application of the model has been 8

adapted by performing a numerical iteration of the bolted fault current input to converge the 9

IEEE calculated arcing current to the specified arcing fault current that matches the HEAF 10 scenario.

11 Electrode configuration 12 The base model provides five configurations to select. Two of the configurations are for open 13 air, a third is for an insulated configuration that is not expected to be used. Therefore, three 14 configurations are not considered applicable to the HEAF scenarios. The remaining two have 15 applicability for the intended purposes and selection of a configuration will be dependent on the 16 location of the arc and the orientation of the targets.

17 Enclosure Dimensions 18 The model requires the enclosure dimensions as input. This includes the width, height, and 19 depth of the enclosure relative to the target. The model validation is limited by the size of 20 compartments used in tests. If the volume of the enclosure exceeds the model limits in either 21 direction, the input parameters should be adjusted to most accurately represent the zone of 22 interest.

23 2.2.3 Model outputs 24 The model outputs include arcing fault current and incident energy.

25 Arc fault current 26 As discussed above, the information available to the staff has identified a range of arcing fault 27 currents and as such, the bolted fault input is to be iteratively adjusted to match the calculated 28 arc fault current with the expected arc fault current. This is easily performed in Microsoft Excel 29 using the goal seek function.

30 Incident Energy 31 The model calculates the incident energy at the specified location based on the inputs. The 32 incident energy is reported in units of MJ/m2.

33 Zone of Influence 34 The zone of influence (ZOI) can be estimated by numerical iteration of the model by iterating the 35 distance (D) where the incident energy matches the target fragility threshold. The NRC and 36 EPRI joint working group has identified 15MJ/m2 and 30MJ/m2 for the targets of interest to fire 37 PRA [14].

38 2.2.4 Assumptions 39 For the fire PRA application the HEAF zone of influence is needed. This is the distance from the 40 enclosure where targets are assumed to be damaged. Since the base model doesnt perform 41

2-4 fire thermal dynamics, any target ignition and flame propagation beyond the initiating source 1

(HEAF enclosure) is not included the application of the base model. Therefore, the ensuring fire 2

is not addressed in this report, but should be considered in the overall assessment of the HEAF 3

hazard.

4 2.2.5 Limitations 5

The base model specifies several limitations, these include 6

Dimensions 7

o Limits on maximum enclosure dimensions with limit on opening area.

8 o Minimum width based on conductor gap spacing 9

10 11

3-1 3

EVALUATION 1

The purpose of this section is to present an evaluation of the base model predictions along with 2

parameter sensitivity. The evaluations use data acquired during the NRC testing program that is 3

different from the data used to develop the base model.

4 3.1 Evaluation Process 5

The base model was evaluated against the NRC dataset followed the same process as used in 6

the NRC/EPRI fire model validation found in NUREG-1824 [16]. This approach involves a 7

scatter plot consisting of the experimental (measured) values represented by the horizontal axis 8

and the model (predicted) values represented by the vertical axis. If a particular prediction and 9

measurement are the same, the resulting point falls on the solid diagonal line. To better make 10 use of these results, two statistical parameters are calculated for each model and each 11 predicted quantity. The first parameter,, is the bias factor. It indicates the extent to which the 12 model, on average, under-or over-predicts the measurements of a given quantity. For example, 13 if the bias factor is 1.10, this indicates that on average the model overpredict by 10%. The 14 second parameter is the relative standard deviation of the model,. This indicates the 15 variability of the model. In addition, the relative standard deviation of the experimental 16 measurements is presented as. The degree of model uncertainty is indicated by the extent to 17 which the data scatter outside the experimental bounds. The calculation of bias and relative 18 standard deviations was performed in Microsoft Excel'.

19 3.2 Overview of data 20 Two sets of data are used for evaluation. A series of tests performed on medium-voltage 21 switchgear performed in 2018 [16], and a series of simple open box configurations performed 22 2019 at medium-voltages [17]. The open box data provides a convenient evaluation of the base 23 model as they are of a configuration like those used to develop the base model (vertical closed 24 box). The 2018 switchgear provides valuable reference as to the event evolution in full-scale 25 equipment. The data summary is presented in Table 1 and Table 2. Photos of experimental 26 configuration is shown in Figure 1 and Figure 2.

27 Table 1. Summary of open box tests 28 Test Electrode Material System Voltage (kV)

Current (kA)

Duration (s)

OBMV01 Aluminum 6.9 14.3 3.18 OBMV02 Aluminum 6.9 29.1 1.12 OBMV03 Aluminum 6.9 14.4 5.05 OBMV04 Copper 6.9 14.3 5.08 OBMV05 Copper 6.9 28.6 2.32 OBMV06 Aluminum 6.9 14.6 2.05 29

3-2 1

Figure 1. Open box medium-voltage test configuration. Vertical sensor array (right),

2 Horizontal sensor array (bottom left) 3 4

5 6

7 Figure 2. Medium-voltage Switchgear Test Configuration (Test device shown at left, 8

Instrument Racks #2 and #3 used for evaluation) 9 Vertical Array Horizontal Array Test Box Rack #2 (0.91m [3ft])

Rack #3 (1.83m [6ft])

3-3 Table 2. Summary of Medium-voltage Switchgear Tests 1

Test Electrode Material Voltage (kV) Current (kA) Duration (s) 2-19 Aluminum 6.9 25.8 2.0 2-21 Aluminum 6.9 26.6 4.1 2-22 Aluminum 7.0 32.0 2.1 2-24 Aluminum 7.0 29.8 4.2

• Voltage was only measured at Generator and is not the same as arc voltage.

2 3.3 Base model evaluation 3

The IEEE 1584-2018 base model is first evaluated against the NRC data without any 4

modification (referred to as base model). The two data sets were evaluated separately due to 5

the differences in the experimental configuration. The comparison uses incident energy in units 6

of MJ/m2. This is one of the outputs of the model and is used to characterize target fragility.

7 The vertical closed box configuration in the base model represents the open box test 8

configuration. The model comparison to the open box test data is presented in Figure 3 and 9

shows a model underprediction of 32% (Bias = 0.68).

10 11 Figure 3. Open Box - Base Model VCB (Bias = 0.68; SD_M = 0.44) 12 Alum - blue, Cu - Orange 13 14

3-4 Next the base model was used to evaluate the medium-voltage switchgear results. The 1

switchgear test comparison showed a 133% overprediction as shown in Figure 4. Since the 2

configuration was different than the model assumptions this overprediction was expected.

3 However, it was also noted that the data sets were segregated by test duration. Splitting the 4

data sets up by duration 2 second vs 4 seconds and re-evaluating the model to each data set 5

indicated a 256% overprediction for the short duration tests and a 7% underprediction for the 6

longer 4 second duration tests. These results are presented in Figure 5 with photographs of the 7

post-test equipment shown in Figure 6 and Figure 7.

8 9

10 Figure 4. Switchgear MV - Base Model, short duration blue, long duration red (Bias = 2.33; 11 SD_M = 0.74) [Blue : 2 sec; Red : 4 sec]

12 13 14 Figure 5. Switchgear MV - Base Model by Duration. Left 2 seconds overprediction (Bias =

15 3.56; SD_M=0.24) Right 4 seconds underprediction (Bias = 0.93; SD_M=0.19) 16

3-5 1

Figure 6. Photographs of 2 sec test enclosures (Left 2-19 25.8kA 39MJ / Right 2-22 32.0kA 2

51MJ) 3 4

3-6 1

Figure 7. Photographs of 4 sec test enclosure (Left 2-21 26.6kA 101MJ / Right 2-24 29.8kA 2

122MJ) 3 4

As can be seen by Figure 5 the short duration tests group differently from the longer 4 second 5

duration. Observations of the video data suggest that the difference can be due to the amount of 6

time the calorimeters are exposed to the arc energy. In the short duration tests a larger fraction 7

of the test duration involves the heating up, melting and yielding of the steel enclosure. Since 8

the base model assumes direct exposure to the arc energy, the time required to create an 9

opening in the steel enclosure needs to be estimated to arrive at a more accurate exposure time 10 for incident energy calculations. Subsequent sections present the modification to the model to 11 better align with fire PRA scenarios and assumptions.

12 13 3.4 Breach modification evaluation 14 One difference between the base model and the fire PRA scenarios (including test data) is that 15 the base model assumes no barrier between the source and the target. However, in U.S.

16 events, it is common for the event to occur with the enclosure under fault in the closed condition 17 (no open doors or panels). While exceptions do occur, this analysis assumes that the enclosure 18 is closed. This assumption results in the need for the energy from the HEAF to be absorbed by 19 the electrical enclosure to cause breach, resulting in an initial energy to be lost to the enclosure 20 and not transmitted to the external targets. This evolution effectively results in a shorter time 21 duration for the exposure of external targets. Therefore, if the time to breach and time to create 22

3-7 an enclosure opening sufficiently large enough to effectively transfer heat to targets can be 1

estimated, this time can be used to offset the time used in the base model. The base model is 2

linearly dependent on time, and therefore time can be modified as proposed below.

3 Arc Duration (tarc)

Breach (tb)

Open (to)

Exposure (te) 4 Figure 8. Illustration of Arc timeline (Not to Scale) 5 6

7

=

(1) 8 Where, te, exposure time 9

tarc, arc time [PRA scenario estimate]

10 tib, time to initial breach enclosure 11 to, time to open from initial breach 12 13 The exposure time represents the time that the external targets are exposed to the HEAF. This 14 is calculated by subtracting the time to the initial breach and the time from initial breach to 15 sufficient opening from the arc time. These two quantities are estimated as described next to 16 reach the final form shown below. Note that the basis for this equation is developed in 17 subsequent sections (3.4.1 and 3.4.2) 18

2 103

()

(2) 19

Where, 20 k(steel) = 2,434 k(aluminum) = 566 21

= 0.9 Iarc, arc current (kA) 22 h, enclosure thickness (mm) tb, breach time (s) 23 Varc(L-L), arc voltage Line-to-Line (kV)

, material parameter (1 steel, 0.25 aluminum) 24

, configuration factor (3.9 Switchgear, 1.2 bus duct) 25 26 27

3-8 3.4.1 Time to initial breach (tb) 1 The time to breach (tb) represents the time from the arc initiation (t=0) to the time when the 2

enclosure starts to breach. The literature was reviewed and found existing models to predict the 3

time to breach of an enclosure. Most of the literature found was related to gas insulated 4

substations (GIS) [19, 20, 21, 22, 23]. These models follow the following general form:

5

=

(3) 6 Where k is a material dependent parameter, h is the enclosure thickness, I is the current, and 7

and are model fitting parameters proposed by the authors. Evaluation of these models against 8

NRC data indicated a large underprediction (63%) in the time to breach, which would cause a 9

conservative estimate on the time to exposure. Further evaluation of these models identified 10 that the geometrical differences between the GIS system and a flat panel system such as HEAF 11 equipment may be causing the model bias. The GIS systems use a cylindrical housing with the 12 conductor in the center versus the switchgear or bus duct enclosures that are made of flat 13 panels. Modification of the material dependent parameter by (to account for geometric 14 differences) brought the model predictions into better alignment (10% underprediction) as 15 shown in Figure 9.

16 17 Figure 9. Modified burn-through steel enclosure (Bias = 0.90, SD_M = 0.18) 18 19 The equation presented above also suggests the difference between the burn through time is 20 approximately 4.3 times longer for a steel enclosure versus an aluminum enclosure of the same 21 thickness and arc current. Similar results were derived during the NRC fragility evaluation [14].

22 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0

0.5 1

1.5 2

2.5 3

Model [Predicted] (s)

Experimental [Measured] (s)

3-9 Therefore, the time to initial breach (tb) is calculated as follows, 1

=

2 (4) 2 k(steel) = 2,434 3

k(aluminum) = 566 4

= 0.9 5

I, arc current (kA) 6 h, enclosure thickness (mm) 7 tb, breach time (ms) 8 3.4.2 Time to open from initial breach (to) 9 Evidence from high speed videography indicates that the opening starts small and expands as 10 more energy is used to melt or yield the metal enclosure. The mass loss of the enclosure can be 11 estimated based on pre-and post-test measurements. However, the limited amount of mass 12 loss compared to the total mass of the enclosure excluded the ability to take a time resolved 13 mass loss measurement during test.

14 The total enclosure mass loss versus total estimated energy for the 2018 medium-voltage 15 switchgear experiments [16] is presented in Figure 10. Similar linearity with energy is found from 16 the open box test data [17] shown in Figure 11, including a notable difference between steel 17 mass loss dependent on electrode material (copper or aluminum). It is also noted that the full 18 scale 2018 data shows a lower mass loss per electrical energy input. This is likely due to 19 configuration differences (larger volume space in 2018 full scale). As such, the full-scale results 20 are used to estimate an opening area for switchgear and the smaller scale box tests used to 21 estimate the opening area for bus ducts.

22 23 Figure 10. Steel enclosure mass loss versus electrical energy with aluminum bus bars (2018 24 arc energy estimated) 25 26 0

20 40 60 80 100 120 140 160 Energy (MJ) 0 2

4 6

8 10 12 14 16 Mass Loss (Kg)

ML = 0.1471 (En) - 3.3458 R 2 =0.9788 Test 2-19 Test 2-22 Test 2-21 Test 2-24 Test Data Fit

3-10 1

Figure 11. Steel enclosure mass loss versus arc energy (with aluminum or copper bus bars) 2 3

Based on a review of the medium-voltage switchgear data and high speed video, the size of the 4

opening between initial breach and efficient energy transfer to targets external to the enclosure 5

was approximated at 550 cm2 (85 in2), which correlates to a mass loss of 0.99 kg and 6.75 MJ 6

of energy for the 2018 tests. Therefore, the time to open from breach is estimated as, 7

=

[]

(5) 8 For a PRA scenario, the arc power needs to be calculated using the following form:

9 For Medium-voltage:

10 Arc power = 3*(VL-L)*I = 3*(Varc(L-L))*(Iarc)*(1E-6) [MJ]

(6) 11 For Low-voltage:

12 Arc power = 3*(VL-L)*I = 3*( Varc(L-L))*(Iarc)*(1E-6) [MJ]

(7) 13 Therefore, 14 0 =

3.9

()106 [, ]

(8) 15 Where is 1 for Steel and 0.25 for aluminum enclosures and Varc(L-L) is 650VL-L for Medium-16 voltage and 350VL-L for low-voltage. See Appendix A for selection of arc voltage.

17

3-11 The analysis above was done using the MV switchgear data that is applicable for larger volume 1

enclosures. For smaller non-segregated phase bus ducts, the medium-voltage box tests appear 2

to be more appliable for the geometry. For the medium-voltage box tests, the same analysis of 3

the data and high-speed video was performed. The corresponding mass loss 0.77kg or 2.14 MJ 4

of energy, which results in a breach to open time (t0) of, 5

0 =

1.2

()106 [, ]

(9) 6 Where is 1 for Steel and 0.25 for aluminum enclosures and Varc(L-L) is 650VL-L for Medium-7 voltage and 375VL-L for low-voltage.

8 9

3.5 Model breach and open time evaluation 10 Similar to what was done in Section 3.3 the model predictions will be evaluated against 11 experimental data using the exposure time rather than the arc time. These results are presented 12 below.

13 14 15 Figure 12. Switchgear MV - Breach Modification Model by Duration. Left 2 seconds 16 overprediction (Bias = 2.29; SD_M=0.24) Right 4 seconds underprediction (Bias =

17 0.78; SD_M=0.17) 18 19

3-12 1

Figure 13. Open Box MV - Breach Modification Model by Duration (note OBMV05 did not 2

breach). (Bias = 0.79; SD_M=0.26) 3 4

3.6 Generator fed - energy conservation modification evaluation 5

Generator fed faults on a unit connected design (no generator circuit breaker) have been 6

identified as a leading contributor to long duration HEAFs [24]. Based on this, the NRC-RES 7

contracted with CESI to develop a decrement energy curve based on a typical nuclear power 8

plant design and reference data from an event occurring at a U.S. facility. This effort resulted in 9

developing a power decay curve shown in Figure 14.

10 11 Figure 14. Reference plant decrement curve 12 13 From this work, the arcing current is defined as, 14

=

(10) 15

3-13 Where, Tref = 4.75 1

It, initial arcing fault current (kA) 2 t, time (seconds) 3 it, decaying arcing fault current (kA) 4 The development of this curve is based on plant operational event data provided to the NRC-5 RES via a collaborative research agreement (Memorandum of Understanding). A subsequent 6

event that occurred in December 2020 has provided additional insights and suggests that the 7

current decay time constant (Tref) may be larger. The time constant is dependent on the plant 8

equipment characteristics and as such all events will have a slightly different time constant. For 9

the evaluation presented below, the CESI value is used and as such, may result in non-10 conservative estimates, if the final agreed upon time constant is larger than 4.75.

11 Arcing power can be determined as, 12

= 3 (11) 13 where, Pt, decaying arc power (MW) 14 VarcLL, arc voltage (kV) 15 it, decaying arcing fault current (kA) 16 17 and arc energy, 18

=

0

= 3

0

= 3

1

(12) 19 Since the base model was not built to input a decaying current profile, two approach for 20 conserving energy of the decrement curve but discretizing a constant power (i.e., current) for 21 the base model are proposed. The first approach is to adjust the time parameter in the base 22 model while maintaining the initial fault current (IT) constant. Since the model output (incident 23 energy) has a linear dependency with time this approach appears reasonable. Therefore, the 24 model energy is equated to the arc energy and solved for the model duration (tm) using the 25 scenario duration (ts) as shown below, with results of the constant current approach presented 26 in Table 3.

27

=

1

(13) 28 29

3-14 Table 3. PRA scenario time to Model Time for Decrement Scenarios 1

Scenario Time (t)

[seconds]

Model Time (tm)

[seconds]

Scenario Time (t)

[seconds]

Model Time (tm)

[seconds]

1 0.9 9

4.0 2

1.6 10 4.2 3

2.2 11 4.3 4

2.7 12 4.4 5

3.1 13 4.4 6

3.4 14 4.5 7

3.7 15 4.5 8

3.9 2

The second approach is to maintain the same duration but adjust the fault current such that the 3

total energy is conserved. This is illustrated in Figure 15, with results are presented in Table 4 4

using a 32kA base case (i.e., IT = 32kA).

5 6

Figure 15. Example of constant time approxiation showing current level and corresponding 7

duration for select cases based on 32kA base Decay curve 8

9 Table 4. Constant duration decrement current modification based on an initial decay current 10 of 32kA (IT = 32kA) 11 Duration

[seconds]

Current

[seconds]

Duration

[seconds]

Current

[seconds]

4 21.6 10 13.3 5

19.8 11 12.5 6

18.2 12 11.7 7

16.7 13 10.9 8

15.5 14 10.3 9

14.3 15 9.7

3-15 1

These two approaches were compared using the modified model prediction of incident energy 2

at three feet. The results are presented in Figure 16. The results indicate that the constant 3

current approach using the base scenario case current (IT) shows a slightly higher prediction.

4 Without data to baseline the modification approach, the constant current approach using the 5

initial fault current and a modified duration is suggested to reduce the likelihood of 6

underprediction.

7 8

Figure 16. Comparison of Decay Curve Discretizing Approaches 9

10 3.7 Model Sensitivities 11 The modified model requires several inputs to perform its calculation. This section evaluates the 12 sensitivity of the modified model to each input, holding other inputs constant. The base scenario 13 is as follows:

14 Parameter Value Parameter Value Voltage 6.9 kV Configuration HCB Arc Current 28.5 kA Duration Variable (4 - 15s)

Gap Spacing 5.75 in Enclosure 36 in x 36 in x 36 in (WxHxD)

Target Crit. IE 15 MJ/m2 Enclosure Thickness 0.09 in Enclosure Material Steel Enclosure Zone MV Switchgear 15 The results are presented as plots of zone of influence (ZOI) versus duration. Since it is known 16 that duration is linearly proportional to the incident energy calculated, this was chosen to be the 17 dependent variable plotted on the horizontal x-axis. The results are presented in Figure 18. The 18 range of parameter variation was based on information from operating experience, judgement, 19 or model limitations. The ranges for each are described below.

20

3-16 Current: A low value of 23kA and a high value of 32kA was based on operational experience 1

review of actual HEAF events.

2 Voltage: Medium-voltage was evaluated by selecting a common lower and higher medium-3 voltage level (4.16kV and 13.8kV).

4 Gap Spacing: Equipment manufactures, and other organizations have published typical bus bar 5

spacing ranges for voltage level and equipment types. Based on this literature, a lower spacing 6

of 10.2 cm (4 in) was based on the upper range of 5kV equipment, the difference between the 7

upper range of 5kV and 15kV equipment or 4.5 cm (1.75 in) above the base case resulted in the 8

upper value of 19 cm (7.5 in).

9 Volume: The lower and upper limits were based on symmetry around the base value and the 10 model limitations. All volumes were cubical in geometry. Thus, 61 cm (24 in) and 122 cm (48 in) 11 square dimensions were used for the sensitivities.

12 As expected, the results indicated that the model is most sensitive to the duration and current 13 inputs. The current variations indicted a 9-13% increase or decrease in the zone of influence 14 over the base case. The conductor gap spacing showed minimal variation with different gap 15 spacings with maximum variations of 4-6%. System voltage was near identical for 4.16kV and 16 6.9kV, however the 13.8kV did show a minimal increase over the former two levels by 17 approximately 6-7%. Volume also showed minimal change between the base case and a larger 18 enclosure (1-3%) but did show a slight increase in the smaller volume enclosure of 6-9%.

19 As a final comparison, the VCB configuration was run and compared to the HCB configuration 20 using the base case parameters. These results are shown in Figure 17 and show that the VCB 21 ZOI differs from the HCB by approximately 30%.

22 23 Figure 17. Base Case comparison between HCB and VCB 24 25

3-17 Current Gap Spacing Voltage Volume Figure 18. Model Sensitivity Plots (Clockwise from upper left: Current, Conductor Spacing, Volume, and System Voltage.

4-1 4

ZONE OF INFLUENCE 1

This section presents zone of influence (ZOI) estimates for an array of possible HEAF scenarios 2

using the modified model with bias corrections as described in Section 3.4. The ZOIs are listed 3

in tables that are based on the scenario configurations developed by a joint U.S. Nuclear 4

Regulatory Commission Office of Nuclear Regulatory Research (RES) / Electric Power 5

Research Institute (EPRI) HEAF PRA working group. Presenting the results in this way allows 6

for comparison to modeling results develop by the working group using computational fluid 7

dynamic models such as Fire Dynamics Simulator. Tables are presented as follows:

8 Switchgear, medium-voltage, steel enclosed (Table 5) 9 Switchgear, medium-voltage, aluminum enclosed (Table 6) 10 Non-segregated bus ducts, medium-voltage, steel enclosed (Table 7) 11 Non-segregated bus ducts, medium-voltage, aluminum enclosed (Table 8) 12 Switchgear, low-voltage, steel enclosed (Table 9) 13 Switchgear, low-volage, aluminum enclosed (Table 10) 14 To minimize the breadth of the ZOI tables, the scenarios presented in this section use single 15 point estimates for the arc fault currents, which the NRC-RES/EPRI working group agreed on.

16 While the actual event fault current will be dependent on system configuration and fault 17 conditions, the values used here are expected to be representative and, in some cases, bound 18 plant fault condition observed from operational experience reviews. The durations used in the 19 estimates are based on either a constant current (stiff supply), a decrement current (decaying 20 supply) or a combination of both a constant and decrement current. For the decrement current, 21 the joint working group agreed to a 130MJ decrement curve which resulted in a decrement 22 equivalent arc time of 3.85 seconds.

23 The zone of influence (ZOI) was developed by iterating the model to identify the distance where 24 the incident energy matches the target vulnerability threshold. For cables this is 15MJ/m2 for 25 thermoplastic jacketed cable and 30MJ/m2 for thermoset jacketed cable. For aluminum enclosed 26 bus ducts with this is 15MJ/m2 and 30MJ/m2 for bus ducts with steel enclosed. These critical 27 values for incident energy are based on work performed by the Joint NRC/RES - EPRI HEAF 28 Working Group [14].

29 The modified model does not evaluate the ensuing fire or thermal heat transfer from the 30 thermally hot enclosure. This is beyond the scope of this report but is being addressed by the 31 joint NRC/RES-EPRI HEAF working group and is expected to be published in the future.

32 33

4-2 Table 5. Medium-voltage Switchgear [30kA, 6.9kV, 0.09in Steel Enclosure]

1 Fault Duration Stiff (s)

Fault Duration Decrement Equivalent (s)

Equivalent Time (s) 15 MJ/m2 (m [ft])

30 MJ/m2 (m [ft])

1 0

1 N/A N/A 2

0 2

0.55 [1.8]

0.36 [1.2]

3 0

3 0.76 [2.5]

0.51 [1.7]

4 0

4 0.94 [3.1]

0.62 [2.0]

5 0

5 1.10 [3.6]

0.73 [2.4]

0 3.85 3.85 0.91 [3.0]

0.61 [2.0]

1 3.85 4.85 1.07 [3.5]

0.71 [2.3]

2 3.85 5.85 1.22 [4.0]

0.81 [2.7]

3 3.85 6.85 1.35 [4.4]

0.90 [2.9]

4 3.85 7.85 1.48 [4.8]

0.98 [3.2]

5 3.85 8.85 1.60 [5.2]

1.06 [3.5]

Note: N/A indicates results are not applicable due to ZOI less than 0.3m [1 ft].

2 3

Table 6. Medium-voltage Switchgear [30kA, 6.9kV, 0.125 Aluminum Enclosure]

4 Fault Duration Stiff (s)

Fault Duration Decrement Equivalent (s)

Equivalent Time (s) 15 MJ/m2 (m [ft])

30 MJ/m2 (m [ft])

1 0

1 0.39 [1.3]

N/A 2

0 2

0.64 [2.1]

0.43 [1.4]

3 0

3 0.84 [2.7]

0.56 [1.8]

4 0

4 1.01 [3.3]

0.67 [2.2]

5 0

5 1.16 [3.8]

0.77 [2.5]

0 3.85 3.85 0.98 [3.2]

0.65 [2.1]

1 3.85 4.85 1.13 [3.7]

0.75 [2.5]

2 3.85 5.85 1.27 [4.2]

0.85 [2.8]

3 3.85 6.85 1.40 [4.6]

0.93 [3.1]

4 3.85 7.85 1.53 [5.0]

1.01 [3.3]

5 3.85 8.85 1.64 [5.4]

1.09 [3.6]

Note: N/A indicates results are not applicable due to ZOI less than 0.3m [1 ft].

5 6

4-3 Table 7. NSBD Medium-voltage [30kA, 6.9kV, 0.09 Steel Enclosure]

1 Fault Duration Stiff (s)

Fault Duration Decrement Equivalent (s)

Equivalent Time (s) 15 MJ/m2 (m [ft])

30 MJ/m2 (m [ft])

1 0

1 N/A N/A 2

0 2

0.39 [1.3]

N/A 3

0 3

0.54 [1.8]

0.35 [1.1]

4 0

4 0.67 [2.2]

0.43 [1.4]

5 0

5 0.79 [2.6]

0.51 [1.7]

0 3.85 3.85 0.65 [2.1]

0.42 [1.4]

1 3.85 4.85 0.77 [2.5]

0.50 [1.6]

2 3.85 5.85 0.88 [2.9]

0.57 [1.9]

3 3.85 6.85 0.99 [3.2]

0.63 [2.1]

4 3.85 7.85 1.08 [3.6]

0.70 [2.3]

5 3.85 8.85 1.18 [3.9]

0.76 [2.5]

Note: N/A indicates results are not applicable due to ZOI less than 0.3m [1 ft].

2 3

Table 8. NSBD Medium-voltage [30kA, 6.9kV, 0.125 Aluminum Enclosure]

4 Fault Duration Stiff (s)

Fault Duration Decrement Equivalent (s)

Equivalent Time (s) 15 MJ/m2 (m [ft])

30 MJ/m2 (m [ft])

1 0

1 N/A N/A 2

0 2

0.44 [1.4]

N/A 3

0 3

0.58 [1.9]

0.38 [1.2]

4 0

4 0.71 [2.3]

0.46 [1.5]

5 0

5 0.83 [2.7]

0.53 [1.7]

0 3.85 3.85 0.69 [2.3]

0.45 [1.5]

1 3.85 4.85 0.81 [2.7]

0.52 [1.7]

2 3.85 5.85 0.92 [3.0]

0.59 [1.9]

3 3.85 6.85 1.02 [3.3]

0.65 [2.1]

4 3.85 7.85 1.11 [3.7]

0.72 [2.3]

5 3.85 8.85 1.20 [4.0]

0.77 [2.5]

Note: N/A indicates results are not applicable due to ZOI less than 0.3m [1 ft].

5 6

4-4 Table 9. Low-voltage Switchgear [32kA, 0.6kV, 0.09 Steel Enclosure]

1 Current (kA)

Fault Duration Decrement Equivalent (s) 15 MJ/m2 (m [ft])

30 MJ/m2 (m [ft])

32.0 1

N/A N/A 32.0 2

0.36 [1.2]

N/A 32.0 3

0.48 [1.6]

0.34 [1.1]

32.0 4

0.58 [1.9]

0.41 [1.4]

32.0 5

0.66 [2.2]

0.47 [1.5]

32.0 6

0.74 [2.4]

0.52 [1.7]

3.4 40 0.57 [1.9]

0.41 [1.3]

Note: N/A indicates results are not applicable due to ZOI less than 0.3m [1 ft].

2 3

Table 10. Low-voltage Switchgear [32kA, 0.6kV, 0.125 Aluminum Enclosure]

4 Current (kA)

Fault Duration Decrement Equivalent (s) 15 MJ/m2 (m [ft])

30 MJ/m2 (m [ft])

32.0 1

N/A N/A 32.0 2

0.44 [1.4]

0.31 [1.0]

32.0 3

0.54 [1.8]

0.39 [1.3]

32.0 4

0.63 [2.1]

0.45 [1.5]

32.0 5

0.71 [2.3]

0.50 [1.6]

32.0 6

0.77 [2.5]

0.55 [1.8]

3.4 40 0.62 [2.0]

0.44 [1.4]

Note: N/A indicates results are not applicable due to ZOI less than 0.3m [1 ft].

5 6

5-1 5

SUMMARY

1 The report documents the evaluation, modification and application of an approach to predict the 2

arcing hazard from high energy arcing faults (HEAFs) involving aluminum. Gaps between the 3

underlying assumptions and intended application of a base model were identified and 4

addressed to improve the prediction capabilities. These modifications were based on existing 5

literature and available test data. The modified model was then exercised in conjunction with 6

target fragility thresholds to develop zone of influence estimates to support refinements in fire 7

PRA methods. The results from this work are expected to be used in conjunction with other 8

efforts under development by the U.S. Nuclear Regulatory Commission (NRC) Office of Nuclear 9

Regulatory Research (RES) and Electric Power Research Institute (EPRI) joint working group 10 on HEAF. The end user is reminded that the information presented in this report is not 11 regulatory guidance, nor does it represent regulatory requirements.

12 13 14

6-1 6

REFERENCES0F1

1.

Information Notice 2017-04, High Energy Arcing Faults in Electrical Equipment Containing Aluminum Components, ADAMS Accession No. ML17058A343 U.S.

Nuclear Regulatory Commission, Washington, DC 20555-0001, August 2017.

2.

NUREG/CR-6850, EPRI 1011989, EPRI/NRC-RES Fire PRA Methodology for Nuclear Power Facilities: Volume 2: Detailed Methodology. Electric Power Research Institute (EPRI), Palo Alto, CA, and U.S. Nuclear Regulatory Commission, Washington DC 20555-0001, 2005.

3.

NEA OECD Fire Project - Topical Report No. 1, Analysis of High Energy Arcing Fault (HEAF) Fire Events¸ Nuclear Energy Agency, Committee on The Safety of Nuclear Installations, Organization for Economic Cooperation and Development, June 2013.

4.

NEA HEAF Project - TOPICAL REPORT No. 1, Experimental Results from the International High Energy Arcing Fault (HEAF) Research Program - Phase 1 Testing 2014 to 2016, Nuclear Energy Agency, Committee on The Safety of Nuclear Installations, Organization for Economic Cooperation and Development, May 2017.

5.

Memorandum, Submittal of Possible Generic Issue Concerning the Damage Caused by High Energy Arc Faults in Electrical Equipment containing Aluminum Components, U.S.

Nuclear Regulatory Commission, Washington, DC 20555-0001, (ADAMS Accession No. ML16126A096), May 2016.

6.

Memorandum, Results of Generic Issue Review Panel Screening Evaluation for Proposed Generic Issue Pre-GI-018, High Energy Arc Faults involving Aluminum (ADAMS Accession No. ML16349A027). Washington, DC 20555-0001, July 2017

7.

Assessment Plan for Pre-GI-018, Proposed Generic Issue on High Energy Arc Faults Involving Aluminum (ADAMS Accession No. ML18172A185). Washington, DC, August 2018.

8.

Update to the Assessment Plan for Pre-GI-018, Proposed Generic Issue on High Energy Arc Faults Involving Aluminum (ADAMS Accession No. ML19127A202). Washington, DC, July 2019.

9.

Memorandum from R. Furstenau to A.D. Veil regarding Closure of Proposed Generic Issue PRE-GE-018 High Energy Arc Faults Involving Aluminum, (ADAMS Accession No. ML21237A360), U.S. Nuclear Regulatory Commission, Washington DC, April 2021.

10.

Integrated Risk-Informed Decision-making Process for Emergent Issues, (ADAMS Accession No. ML19253D401), U.S. Nuclear Regulatory Commission, Washington DC, March 2020.

11.

IEEE 1584-2018, IEEE Guide for Performing Arc-Flash Hazard Calculations. New York:

The Institute of Electrical and Electronic Engineers, Inc., New York, NY, 2018.

12.

Lee, Ralph H. "The Other Electrical Hazard: Electric Arc Blast Burns." IEEE Transactions on Industry Applications 246-251, 1982.

13.

IEEE 1584-2002, IEEE Guide for Performing Arc-Flash Hazard Calculations. New York:

Institute of Electrical and Electronic Engineers, Inc, New York, NY, 2002.

14.

Draft report for comment, Target Fragilities for Equipment Vulnerable to High Energy Arcing Faults, (ADAMS accession No. ML21326A010), U.S. Nuclear Regulatory Commission, Washington, DC 20555-0001 and Electric Power Research Institute, Palo Alto, CA, November 2021.

1 Publicly available NRC published documents are available electronically through the NRC Library on the NRCs public Web site at http://www.nrc.gov/reading-rm/doc-collections/. The documents can also be viewed online or printed for a fee in the NRCs Public Document Room (PDR) at 11555 Rockville Pike, Rockville, MD; the mailing address is USNRC PDR, Washington, DC 20555; telephone (301) 415-4737 or (800) 397-4209; fax (301) 415-3548; and e-mail pdr.resource@nrc.gov.

6-2

15.

NUREG-1824, Verification and Validation of Selected Fire Models for Nuclear Power Plant Applications (Supplement 1). NUREG, Washington DC / Palo Alto, CA: U.S.

Nuclear Regulatory Commission / Electric Power Research Institute, 2016.

16.

RIL 2021-10, Report on High Energy Arcing Fault Experiments, Experimental Results from Medium Voltage Electrical Enclosures, U.S. Nuclear Regulatory Commission, Washington, DC 20555-0001, NIST TN 2188, National Institute of Standards and Technology, Gaithersburg, MD, SAND2021-12049R, Sandia National Laboratories, Albuquerque, NM, 2021.

17.

RIL 2021-18, Report on High Energy Arcing Fault Experiments, Experimental Results from Open Box Enclosures, U.S. Nuclear Regulatory Commission, Washington, DC 20555-0001, NIST TN 2198, National Institute of Standards and Technology, Gaithersburg, MD, SAND2021-16075 R, Sandia National Laboratories, Albuquerque, NM, 2021.

18.

RIL 2021-17, Report on High Energy Arcing Fault Experiments, Experimental Results from Low Voltage Electrical Enclosures, U.S. Nuclear Regulatory Commission, Washington, DC 20555-0001, NIST TN 2197, National Institute of Standards and Technology, Gaithersburg, MD, 2021.

19.

Babusci, G., et. al., Assessment of the Behavior of gas-insulted Electrical Components in the Presence of Arc, 1998.

20.

Bernard, G., Electrical faults mastery in high voltage SF6 insulated substations, EDF, 1982.

21.

Trinh, N.G., Evaluation of the Risk of Burn through due to Internal Arc of Gas-Insulated Equipment, CEA Trans, March 1989.

22.

Boeck, W.A., and Kruger, K., Arc motion and burn through in GIS, IEEE Transactions on Power Delivery, Vol. 7, p 254-261, 1992.

23.

Chu, F., et. al., Estimation of Burn-Through Probability in SF6 insulated substations, Power Apparatus and Systems, IEEE Transactions, PAS-101, 1391-1399, 1982.

24.

EPRI 3002015992, Nuclear Station Electrical Distribution Systems and High-Energy Arcing Faults Events, Electric Power Research Institute, Palo Alto, CA, July 2019.

25.

CIGRE 602, Tools for the simulation of the effects of the internal arc in transmission and distribution switchgear. ISBN: 978-2-85873-303-3, CIGRE, 2014.

A-1 APPENDIX A 1

ARC VOLTAGE 2

Arc voltage is an important parameter for determining arc energy and for the time to breach and 3

time to open model modification. The arc voltage is the voltage drop of an arc between the arc 4

roots. Arc voltage varies because of arc looping, change in electrode configuration, 5

compartment size, and pressure. Arc voltage is different from system voltage as shown in 6

Figure 19. The arc is purely resistive and as such has a dependency on resistivity and arc 7

current. While arc voltage is not an input to the model, it is used to calculate the energy of 8

specific HEAF PRA scenarios and for the breach adjustment approach. Thus, the arc voltage 9

needs to be known or estimated. The following provides a review of the evidence available to 10 estimate the arc voltage for different voltage classes of equipment.

11 12 13 Figure 19. Voltage Profile prior to (System Voltage) and during (Arcing Voltage) initial arcing 14 duration (Voltage shown as Phase Voltage) 15 16 17

-0.1

-0.05 0

0.05 0.1 0.15 0.2 time (seconds)

-6

-4

-2 0

2 4

6 Phase-Neutral (kV)

A-N B-N C-N

A-2 A.1 Test data 1

2 The test data is limited due to the varying locations where the voltage measurements were 3

made. Arc voltages were acquired for the low-voltage switchgear [18] and low-voltage box 4

tested in 2019 [17]. The switchgear bus spacing on center is 5-inches and the box bus spacing 5

on center is 3.5-inches. These results are presented in Table 11 and Figure 20. The low-voltage 6

switchgear shows a slightly higher arc voltage than the box enclosure but comparison between 7

the two test configurations show a 95% confidence interval for the mean to only vary 10-20 volts 8

between the two. Thus, these similarities suggest combining the two data sets, which results in 9

a mean of 350VL-L and an 95% confidence interval for the mean of 320 VL-L to 375VL-L. Based on 10 this a mean of 350VL-L for low-voltage equipment is suggested. If conservatism in the estimate is 11 needed the upper value of 375VL-L could be used for low-voltage equipment.

12 13 Table 11. Arc Voltage Measurements for Low-voltage Tests 14 Test ID LV SwGr Current (kA)

System Voltage (VL-L)

Arc Voltage (VL-L)

Test ID LV Box Current (kA)

System Voltage (VL-L)

Arc Voltage (VL-L) 2-13A 9.800 480 389 OB01a 1.052 1,000 347 2-13B 9.973 600 420 OB01b 1.030 1,000 308 2-13C 11.650 600 298 OB02 14.016 1,000 271 2-13D 9.266 600 426 OB03 13.804 1,000 314 2-13E (Cu) 10.388 600 305 OB04 27.786 1,000 276 2-13F 9.733 480 302 OB05 1.018 1,000 359 2-13G 10.707 600 330 OB06 11.959 1,000 424 2-18A 19.146 480 290 OB07 12.952 1,000 431 2-18B 19.349 600 415 OB08*

24.870 1,000 537*

OB09 4.794 1,000 297 OB10 4.869 1,000 381 Average 353 Average 341*

Maximum 426 Maximum 431*

• Rod ejection resulted in abnormal arc voltage during Test OB08 and removed 15

A-3 1

Figure 20. Interval Plot (left) and Box Plot (right) of LV Arc Voltage Data 2

3 The medium-voltage tests did not include an arc voltage measurement for the 2018 switchgear 4

tests. The only voltage measurement was made at the generator and voltage losses between 5

the generator and the test device cannot be estimated. The medium-voltage open box tests did 6

measure voltage at the test device [17]. The available medium-voltage data is presented below.

7 8

Table 12. Voltage Measurements for Medium-voltage Tests 9

10 Test ID MV SwGr Current (kA)

System Voltage (kVL-L)

Gen Voltage (VL-L)

Test ID MV Box Current (kA)

System Voltage (kVL-L)

Gen Voltage (VL-L)

Arc Voltage (VL-L) 2-19 (AL) 25.8 6.9 767 OBMV01 (AL) 14.3 6.9 623 543 2-21 (AL) 26.6 6.9 769 OBMV02 (AL) 29.1 6.9 830 468 2-22 (AL) 32.0 7.0 869 OBMV03 (AL) 14.4 6.9 606 475 2-24 (AL) 29.8 7.0 876 OBMV04 (CU) 14.3 6.9 597 458 OBMV05 (CU) 28.6 6.9 775 406 OBMV06 (AL) 14.6 6.9 671 493 Average 820 Average 683 475 Maximum 876 Maximum 830 543 11

A-4 A.2 Arc voltage model 1

There are several models available for estimating arc voltage. The predominant models are 2

presented below with reference to the NRC test data of known arc voltage measurements.

3 4

CIGRE 602, Tools for the Simulation of the Effects of the Internal Arc in Transmission and 5

Distribution Switchgear [25] presents a model for estimating arc voltage. For MV metal 6

enclosed switchgear in air with copper electrodes, the arc voltage can be estimated as, 7

8

= 30

+ 1 2

40 9

10 where, Uarc, phase-to-phase (LL) arc voltage (volts) 11 d, distance between electrode centers (cm) 12 Irms, the effective short circuit current (kA) 13 14 Using the arc voltage measurements available, the model is compared for low-voltage and 15 medium-voltage as shown in Figure 21.

16 17 18 Figure 21. Model evaluation for Low-voltage box tests (Left) and Low-voltage Switchgear 19 voltage (Right) to the CIGRE 602 arc voltage model. All voltage shown are Line-to-20 Line. [Low-voltage box Bias = 0.91, SD_M=0.29] [Low-voltage switchgear Bias =

21 1.33, SD_M=0.18]

22 23

A-5 1

Figure 22. Model Evaluation for Medium-voltage box tests to the CIGRE 602 arc voltage 2

model [Medium-voltage Bias = 1.08, SD_M=0.17]

3 4

The box tests bus bar spacing for medium-voltage equipment was smaller than the 2018 MV 5

switchgear tests. Since measurements of arc voltage were not made during the 2018 test 6

series, the arc voltage is estimated for those tests as shown below using the CIGRE model and 7

correction factor (1.08). Using this information to correct the CIGRE 602 model prediction, the 8

arc voltage was estimated for each of the 2018 tests with the results presented in Table 13.

9 From this a representative value to use for medium-voltage is 650VLL.

10 11 Table 13. CIGRE 602 estimates of arc voltage for 2018 tests 12 Test ID Irms (kA)

Ugen (VLL) measured Uarc (VLL)

Uarc (VLL) corrected 2-19 25.8 767 681 631 2-21 26.6 769 687 636 2-22 32.0 869 730 676 2-24 29.8 876 713 660 Average 650 VL-L [375 VL-N]

Maximum 676 VL-L [390 VL-N]

3rd Quartile 665 VL-L [384 VL-N]

13 14 15

B-1 APPENDIX B 1

ZOI TABLE MODELING INPUT 2

3 This appendix documents the input parameters that were used to calculate the zone of influence 4

(ZOI) presented in Section 4.

5 6

B.1 Non-segregated phase bus duct 7

Based on equipment procured for testing as specified by NRC-RES/EPRI Working Group as 8

being garden variety bus duct, shown in Figure 23. Enclosure thickness is 11Ga (3.2mm [1/8 9

in]) steel or (2.3mm [0.091 in]) aluminum. Model parameters are shown in Figure 24. Note that 10 the time and critical incident energy values were varied to the specific scenario. Critical incident 11 energy values were either 15 MJ/m2 or 30 MJ/m2 dependents on the target. The enclosure type 12 steel or aluminum was also evaluated. VCB configuration was used as this would postulate 13 damage to targets above or below the bus duct.

14 15 16 17 Figure 23. Bus Duct Specification (Aluminum bus and duct) 18 19

B-2 1

Figure 24. Input Parameters for non-segregated bus duct.

2 3

4 B.2 Zone 1 and 2 MV Switchgear 5

Based on equipment tested during the 2018 MV Switchgear tests, shown in Figure 25.

6 Enclosure exterior panel thickness is 13 Ga (2.3mm [0.09 in]) steel. Model parameters are 7

shown in Figure 26. Note that the time and critical incident energy values were varied to the 8

specific scenario. Critical incident energy values were either 15 MJ/m2 or 30 MJ/m2 dependents 9

on the target. The enclosure type steel was the only one evaluated. Configuration selected was 10 dependent on the equipment type location. For the supply vertical section, the VCB 11 configuration was used. For the Load and Main Bus Vert Section, the HCB was selected since 12 this representative of the load configuration and bounding for the main bus.

13 14 INPUT PARAMETERS Conductor Gap Spacing G

3.63 in Arc Fault Current Iarc 30 kA Configuration VCB Open Circuit Voltage Voc 6.9 kV Duration T

Varied seconds Enclosure Width W

33 in Enclosure Height H

36 in Enclosure Depth D

14.25 in Critial Incident Energy IE 15 or 30 MJ/m2 Enclosure Type Typical Bolted Fault Current Ibf 34.18 kA Enclosure Thickness 0.09 in Enclosure Material Steel Enclosure Zone LV Switchgear

B-3 1

Figure 25. Medium-voltage Switchgear 2

3 4

Figure 26. Input Parameters for Zone 1 and 2 MV Switchgear 5

6 7

INPUT PARAMETERS Conductor Gap Spacing G

6.02 in Arc Fault Current Iarc 30 kA Configuration HCB Open Circuit Voltage Voc 6.9 kV Duration T

Varied seconds Enclosure Width W

36 in Enclosure Height H

36 in Enclosure Depth D

36 in Critial Incident Energy IE 15 or 30 MJ/m2 Enclosure Type Typical Bolted Fault Current Ibf 34.51 kA Enclosure Thickness 0.09 in Enclosure Material Steel Enclosure Zone MV Switchgear Load Supply Main Bus

B-4 B.3 Zone 3 LV Switchgear 1

Based on equipment tested during the 2019 LV Switchgear tests, shown in Figure 25. Enclosure 2

exterior panel thickness is 13 Ga (2.3mm [0.09 in]) steel. Model parameters are shown in Figure 3

28. Note that the time and critical incident energy values were varied to the specific scenario.

4 Critical incident energy values were either 15 MJ/m2 or 30 MJ/m2 dependents on the target. The 5

enclosure type steel was the only one evaluated. Configuration selected was dependent on the 6

equipment type location. For the supply vertical section, the VCB configuration was used. For 7

the Load and Main Bus Vert Section, the HCB was selected since this representative of the load 8

configuration and bounding for the main bus.

9 10 11 12 Figure 27. Low-voltage switchgear 13 14 Supply Main Bus Load

B-5 1

Figure 28. Input Parameters for Zone 3 LV Switchgear.

2 3

INPUT PARAMETERS Conductor Gap Spacing G

1.26 in Arc Fault Current Iarc 32 kA Configuration HCB Open Circuit Voltage Voc 0.6 kV Duration T

Varied seconds Enclosure Width W

24 in Enclosure Height H

24 in Enclosure Depth D

14 in Critial Incident Energy IE 15 or 30 MJ/m2 Enclosure Type Typical Bolted Fault Current Ibf 45.06 kA Enclosure Thickness 0.09 in Enclosure Material Steel Enclosure Zone LV Switchgear