United States Nuclear Regulatory Commission - Protecting People and the Environment
Home > NRC Library > Document Collections > NUREG-Series Publications > Staff Reports > NUREG 0933 > Section 3. New Generic Issues- Issue 167: Hydrogen Storage Facility Separation (Rev. 2)

Resolution of Generic Safety Issues: Issue 167: Hydrogen Storage Facility Separation (Rev. 2) ( NUREG-0933, Main Report with Supplements 1–35 )


Historical Background

Issue 106 was resolved with the issuance of Generic Letter 93-06, "Research Results on Generic Safety Issue 106, ‘Piping and the Use of Highly Combustible Gases in Vital Areas,’" dated October 25, 1993,1547 which included evaluation of the risk from (1) the storage and distribution of hydrogen (H2) for the volume control tank in pressurized-water reactors (PWRs) and the main electric generator in boiling-water reactors (BWRs) and PWRs, (2) other sources of H2 such as battery rooms, the waste gas system in PWRs, and the offgas system in BWRs, and (3) small, portable bottles of combustible gases used in maintenance, testing, and calibration. However, the potential risk from large H2 storage facilities outside the reactor, auxiliary, and turbine buildings was not addressed. Studies performed during and subsequent to the resolution of Issue 106 raised concerns about the magnitude of the excluded risk.1534,1535 Thus, in December 1993, Issue 167 was identified1532 to address this excluded risk.

U.S. Nuclear Regulatory Commission (NRC) Information Notice 89-44, "Hydrogen Storage on the Roof of the Control Room,"1552 was issued in April 1989, and each NRC regional office was expected to determine whether the plants in its region had similar safety-related concerns. The information compiled by these offices was reviewed and issued in the preliminary report, SCIE-EGG-103-89, "Draft Technical Evaluation Report on U.S. Commercial Power Reactor Hydrogen Tank Farms and Their Compliance with Separation Distance Safety Criteria," in March 1990.1535 The storage of gaseous or liquid H2 at 119 power plants was then investigated, and possible accident scenarios resulting from a fireball, explosion, or presence of unburned H2 gas in ventilation air intakes were examined. Explosion was identified as the scenario posing the greatest risk potential. The analysis in SCIE-EGG-103-891535 focused on explosion, with all quantification performed relative to this accident only.

Safety Significance

The safety concern was whether or not there is adequate physical separation between H2 storage facilities and buildings or structures housing systems important to safety at nuclear power plants.

Possible Solutions

Possible solutions included relocation (or placement in pits) of storage facilities, buildings, and equipment and the construction of blast shields, or a combination of these. The resolution for this issue was assumed to be the construction of concrete walls enclosing the H2 storage facility. This structure would serve as a blast shield in the event of an explosion, essentially eliminating the risk.


The NRC staff assigned a LOW priority ranking to this issue in 1994. This section presents the NRC staff analysis for prioritizing this issue, which was published in 1995. This analysis, which includes frequency, consequence, and cost estimates and a value/impact assessment, has not been updated in the 2011 revision of this issue.

Hydrogen gas and cryogenic H2 storage tanks are designed, fabricated, tested, and stamped in accordance with Section VIII, Division 1, of the American Society of Mechanical Engineers (ASME) Boiler and Pressure Vessel Code (ASME Code) for unfired pressure vessels. The containers for gaseous H2 are seamless, single-walled containers. Liquid H2 is stored in vacuum-jacketed or double-walled vessels. The "Handbook of Chemical Hazard Analysis Procedures" (the Handbook) (published jointly by the Federal Emergency Management Agency, U.S. Department of Transportation, and U.S. Environmental Protection Agency) lists accident rates and percentages of volume released for use in analyzing potential accidents involving H2 storage containers. For single-walled containers, the accident rate suggested in this Handbook was 10-4/tank-year. For these containers, the Handbook suggested that 90 percent of spills are terminated, while 10 percent are instantaneous total release of contents. Thus, for single-walled containers, the frequency of release of 100 percent of container contents was estimated to be 10-5/tank-year. For double-walled containers, the accident rate suggested was 10-6/tank-year. In this case, the entire container contents are released instantaneously 100 percent of the time. A frequency of 10-5/tank-year for instantaneous release of 100 percent of vessel contents was assumed in this analysis.

The status of the 119 power plants was assessed1535 with respect to H2 tank farm separation guidelines in Electric Power Research Institute (EPRI) NP-5283-SR-A, "Guidelines for Permanent BWR Hydrogen Water Chemistry Installations." Sixteen percent were found not to meet the separation guidance with respect to explosion hazard. For the existing population of 110 plants, this translated into 18 light-water reactors not meeting the EPRI guidelines. (The permanently shut down Trojan PWR was excluded from this population. Thus, this analysis did not address H2 storage tanks located on top of the control room roof.) However, in NUREG-1364, "Regulatory Analysis for the Resolution of Generic Safety Issue 106: Piping and the Use of Highly Combustible Gases in Vital Areas," issued June 1993,1545 credit was assumed for an informal survey that showed mitigating factors to be insufficient at only three plants.

Frequency/Consequence Estimate

Of the types of accidents analyzed in risk assessments, H2 tank farm explosions seem most similar to some accidents classified as external events. Furthermore, because such explosions could cause large pressure forces to be exerted upon building walls or could cause possible impact by missiles, they would appear to be similar to tornadoes among the types of external events. However, unlike what is usually assumed for tornadoes, H2 tank farm explosions probably would not exert a uniformly high and destructive pressure on all buildings on site at one time. Thus, the consequences from H2 tank farm explosions were not expected to exceed those from tornadoes.

A review of available probabilistic risk assessments (PRAs) yielded the individual plant examination (IPE) for Oconee Nuclear Station (Oconee), Unit 3,1533 as most appropriate for this analysis. In this IPE, a fairly detailed assessment of tornado risk was performed, building on that from the earlier Oconee Unit 3 PRA.889 An added advantage to the selection of Oconee Unit 3 was that it provided a description of the site’s 48,000-scf H2 tank farm, which was deemed to be representative.1535 This farm consisted of six tanks that, using the suggested values in the Handbook, resulted in an accident frequency of (10-5/tank-year)(6 tanks/tank-farm)(1 tank-farm/reactor) = 6x10-5/reactor-year (RY) for an H2 accident that releases 100 percent of the contents of at least one tank.

To put the amount of H2 involved in perspective for the tank farm at Oconee, one tank contains 8,000 scf of H2. This was equivalent to 216.8 pounds (lbs) of trinitrotoluene (TNT), using the EPRI guidelines equivalence of 1,000 scf = 27.1 lbs of TNT for gaseous H2 storage. Six tanks in the same farm contain an equivalent of 1,300.8 lbs of TNT. In the terminology of the EPRI guidelines, both these amounts are considered to be in the "small equivalence" range (less than 4,000 lbs TNT). Terminology and equivalence notwithstanding, hypothesizing the detonation of one tank in a farm raises the question of subsequent damage to and detonation of adjacent tanks. The EPRI guidelines were based on the safety analysis of the failure of single vessels and did not address simultaneous failure of multiple storage vessels. There was factual support for using a basis of only one tank failure. The guidelines cited three events, two from reactor sites where H2 container explosions did not damage adjacent cylinders. Given release of the contents of one cylinder, it was assumed in this analysis that (1) detonation will occur, (2) possibly all of the tanks were involved, and (3) because the tank farm in question was assumed not to conform to EPRI guidelines, appropriately selected plant damage would ensue with an appropriately assigned conditional probability.

For the purpose of analyzing tornado-generated missiles, Duke Power Company (Duke Power) considered1533 two categories of tornado events: (1) tornadoes whose winds impact on Oconee Unit 3, and (2) tornadoes passing within 2,000 feet (ft) of Oconee Unit 3. The latter category was subsequently dismissed when analysis showed the probability of a core-melt due to tornado-generated missiles to be 100 to 1,000 times lower than that due to tornado wind loadings; therefore, only the first category was addressed. Duke Power assumed1533 that a tornado would render unavailable all offsite alternating current (ac) power sources except for one underground path. Tornadoes of intensity F-1 or less (i.e., with wind speeds less than 113 miles per hour (mph)) were assumed not to cause sufficient wind damage to generate a core melt. Oconee had been designed to withstand wind loadings of F-1 tornadoes.

The EPRI guidelines have been checked for responses for walls with static pressure capacities between 1.5 and 4.5 pounds per square inch (psi). Regulatory Guide 1.76, "Design Basis Tornado for Nuclear Power Plants,"42 indicated that a 1.5-psi pressure drop could be expected for a design-basis tornado with wind speed as low as 195 mph (the sum of rotational speed and minimum translational speed). This wind speed lies toward the upper end of the range for an F-3 tornado (158–206 mph). From Appendix B to the EPRI guidelines, for a small yield such as that (216.8 lbs TNT) from one tank at the tank farm in question, the separation distance can be fairly small (about 60 ft), even for a plant with moderate wall ductility (μ = 3) and low-end (1.5-psi) static design pressure. This lends some justification to the assumption made that the appropriate minimum blast force to use in the analysis corresponds to that from an F-3 tornado. Due to lack of knowledge of the exact number of tanks detonated (one to six) and other physical parameters involved, conditional probabilities of 1/3 were assigned to each of the resulting equivalent tornado forces F-3, F-4, and F-5. In other words, the 6x10-5/RY initiating event frequency derived from the Handbook was considered to be uniformly distributed among the assumed equally likely outcomes F-3, F-4, and F-5. This was believed to be conservative, at least with respect to F-4 and F-5.

The turbine building was assumed1533 to be susceptible to wall damage from F-2 and stronger tornadoes. Wall damage could fail the 4,160-volt (V) (4-kilovolt (kV)) ac switchgear that powers safety equipment and/or the upper surge tank (UST), the prime suction source for the emergency feedwater pumps. The auxiliary building was assumed to be susceptible to wall damage from F-4 and stronger tornadoes (wind speeds greater than 206 mph), particularly the exterior walls of the west penetration room (WPR) and east penetration room (EPR). Damage to the WPR wall could fail piping and electrical penetrations, including those from the standby shutdown facility (SSF). This could lead to reactor coolant pump seal loss-of-coolant accidents, loss of the SSF backup for reactor coolant pump seal cooling, and loss of feedwater from the SSF. Damage to the EPR wall would cause similar failures, although the likelihood of piping failures there was judged to be about 10 times less due to tornado shielding by the reactor building. Other exterior components, such as the borated water storage tank, are also susceptible to failure from tornadoes. However, they did not appear in the listed cutsets for core damage and were not considered further.

Tornadoes falling within categories F-2 through F-5 resulted in accident sequences leading to various plant damage states. The total core damage frequency (CDF) (the sum of the accident sequence frequencies) was 9.74x10-6/RY. However, after eliminating the plant damage state not resulting in offsite releases, the frequency was calculated to be 8.2x10-6/RY.

There were 17 possible release categories associated with the plant damage state, and each was assigned a conditional probability of release.1533 When multiplied by the sequence frequency, each of these yielded the sequence frequency per release category. Associated with each category was a whole-body man-rem equivalent dose. The product of each release category frequency and its associated dose yielded a total risk of 9.11 man-rem/RY. These results for a tornado were then modified for an H2 tank-farm explosion.

Again, unlike tornadoes, H2 tank-farm explosions would not exert pressure on all site buildings at one time. Thus, multiple building wall failures were not expected as in the tornado accident sequences. To reflect this limitation, the cutsets of the tornado accident sequences were reviewed and it was found that nearly all contained conditional failure of the turbine building wall. Associated with this failure were failures of the UST and/or 4-kV ac switchgear. Failures of the walls of the WPR and/or EPR were contained in fewer of the cutsets of the tornado accident sequences. Therefore, an H2 tank-farm explosion was assumed to fail only the turbine building wall, resulting in failures of the UST and/or 4-kV ac switchgear; no failure of the WPR or EPR walls was assumed. This eliminated most of the T(F4) and T(F5) sequences. Using the remaining sequences, the Handbook derived an initiating event frequency of 6x10-5/RY, and the 1/3 conditional probability for each of the categories F-3, F-4, and F-5 resulted in a total CDF frequency from H2 tank explosion of 4x10-6/RY, which was less than that for tornadoes.

Using the same release category conditional probabilities and equivalent doses as for tornadoes, a total frequency over all release categories of 3x10-6/RY (less than the total CDF because not all accident sequences lead to offsite release) and a total risk of 2.9 man-rem/RY were obtained. Thus, the risk was also less than that for tornadoes. Assuming 18 affected plants with an average remaining lifetime of 23 years, the total risk reduction potential was 1,201 man-rem.

Cost Estimate

Industry Cost: It was reported1535 that the 48,000-scf Oconee tank-farm consisted of six tanks that covered an area 45 ft by 30 ft. It was surrounded by an exclusion fence and was always lighted. The proposed concrete enclosure was assumed to have the same dimensions as the six Oconee tanks combined. From the 1993 Means Building Construction Cost Data (51st edition), the costs were obtained for thick, smooth, gray, architectural precast concrete slabs 10 ft high and 6 inches thick. For a 20-ft length, the cost was $14.95 per square foot (ft2) (area) and, for a 30-ft length, the cost was $14.60/ft2. To form an enclosure at least 45 ft by 30 ft, two 20-ft sections and four 30-ft sections would be needed. A height of 10 ft was considered to be sufficient to protect the surroundings from horizontal blast effects; however, a thickness comparable to that of site building walls (about 18 inches) was necessary. Thus, the total number of precast concrete wall panels was as follows:

30-ft panels: (4/perimeter)(3 at 6-in thickness each) = 12
20-ft panels : (2/perimeter)(3 at 6-in thickness each) = 6
Total = 18

The enclosure would be 50 ft by 30 ft, yielding a total wall panel area of (2)(50 + 30)ft(10 ft)(3 panels) or 4,800 ft2. At $15/ft2, the cost of this enclosure would be ($15/ft2)(4,800 ft2) or $72,000.

The Means manual cited above stated that, "[i]f the work is to be subcontracted, add the general contractor’s markup, approximately 10%." In addition, the enclosure will have to be anchored in place and penetrated for piping and access. Combined with the general contractor’s markup, these factors were assumed to increase the cost of the enclosure by about 50 percent, bringing the total cost to (1.5)($72,000)/plant or $108,000/plant.

Any industry operation and maintenance activities associated with this resolution were assumed to be performed as part of activities already in place, such as standard inspection and reporting procedures. No industry cost was anticipated for operation and maintenance. Thus, the total cost for 18 plants was $1.94 million (M).

NRC Cost: The resolution was assumed to require $100,000 for development of a "typical uncomplicated TS [technical specification] change" and an implementation cost of $11,000/plant (1988 dollars) with a 4-percent inflation rate. For 18 affected plants, the total cost was estimated to be $341,000.

Total Cost: The total NRC and industry cost associated with the possible solution was estimated to be $(1.94 + 0.341)M or $2.28M.

Impact/Value Assessment

Based on a potential public risk reduction of 1,201 man-rem and an estimated cost of $2.28M, the impact/value ratio was given by the following:

Other Considerations

The following other considerations relate to this issue:


The number of affected plants was determined by identifying those that did not conform to the EPRI separation criteria that were based on an H2-to-TNT detonation equivalency. Had more stringent criteria been used, the number of affected plants might have been larger. EGG-SSRE-8747, "Technical Report: Improved Estimates of Separate Distances to Prevent Unacceptable Damage to Nuclear Power Plant Structures from Hydrogen Detonation for Gaseous Hydrogen Storage," issued November 1993,1534 concluded that "the hydrogen to TNT detonation equivalency used in previous calculations should no longer be used." The stated reason for this was that "the separation distances results from previous calculations [including those of the EPRI criteria] can be either overconservative or unconservative depending upon the set of hydrogen detonation parameters that are used." Nevertheless, this analysis was considered to be sufficiently conservative, particularly with respect to the assumption that all tanks are single-walled and assignment of conditional probabilities of 1/3 for F-4 and F-5 resultant forces.

An informal survey of all plants cited in NUREG-13641545 showed that, of those plants that did not meet the EPRI criteria for separation distances for safety-related air intakes or structures, mitigation features were insufficient at only three of the plants. Use of 3 affected plants instead of 18 would also result in a low impact/value score.

(2) Assuming a license renewal period of 20 years, with the 18 affected plants operational 75 percent of this time, the additional risk reduction would be (18)(2.9 man-rem/RY)(20)(0.75) or 783 man-rem. Because there would be no increase in cost, the impact/value ratio would be $1,149/man-rem.


Based on the impact/value ratio and the potential reduction in CDF and public risk, this issue was given a LOW priority ranking in 1994. Consideration of a license renewal period of 20 years would not change this ranking.

The NRC staff conducted a review of this issue in 2010 to determine whether any new information would necessitate reassessment of the original prioritization evaluation.1964 The staff determined that the existing regulations and guidance adequately address this issue, and the operating experience has not indicated a change in the significance of this issue. The following discussion demonstrates the application of the NRC regulatory framework to this issue.

Between the publication of this generic issue in 1995 and the year 2000, most licensees committed to National Fire Protection Association (NFPA) 50A, "Standard for Gaseous Hydrogen Systems at Consumer Sites," and NFPA 50B, "Standard for Liquefied Hydrogen Systems at Consumer Sites," as part of their licensing basis.1969 These codes provided separation distances for gaseous and liquefied hydrogen, providing a basis for the generic issue.

In 2000, with the implementation of the Reactor Oversight Process, the NRC issued Inspection Procedures 71111.05AQ, "Fire Protection Annual/Quarterly,"1970 and 71111.05T, "Fire Protection (Triennial)."1971 These inspection procedures have the following objectives:

  • Evaluate the adequacy and implementation of the licensees’ fire protection programs.

  • Review the procedures to incorporate and implement changes to the respective fire protection programs.

  • Determine the adequacy of the licensees’ systems for taking corrective action when warranted by quality assurance programs, generic deficiencies, or events.

With respect to this generic issue, these inspection procedures verify that a licensee’s fire protection program includes the control of combustible material, including the appropriate storage of bulk flammable gases and liquids like hydrogen. To that end, inspection procedures also verify that the licensee’s fire protection program consists of a fire hazard analysis, which includes analyses for postulated hydrogen explosions. The fire protection program also includes the facility’s technical specifications, which include the appropriate limiting condition for operations to prevent the postulated fire conditions.

In December 2002, the NRC reported the results of the inspections under Temporary Instruction 2515/146, "Hydrogen Storage Locations," Revision 1, dated April 18, 2002.1972 The report highlighted findings related to the adequate separation of hydrogen storage facilities from risk-significant tanks or SSCs and from ventilation intakes. The licensees of these plants committed to taking appropriate corrective actions.

With respect to enforcement, in December 2008, inspectors identified a Severity Level IV noncited violation of Title 10 of the Code of Federal Regulations (10 CFR) 50.59, "Changes, Tests and Experiments," for the licensee’s failure to perform a safety evaluation associated with installation of a bulk hydrogen storage facility located directly above buried circulating water system return lines.1973

Based on the review of the NRC’s regulations and guidance related to this issue, the staff concluded that existing regulations and guidance adequately address this issue. Therefore, the staff changed the status of Generic Issue 167 and DROPPED this issue from further pursuit.


0042.Regulatory Guide 1.76, "Design Basis Tornado for Nuclear Power Plants," U.S. Nuclear Regulatory Commission, April 1974. [7907100297]
0889.NSAC-60, "A Probabilistic Risk Assessment of Oconee Unit 3," Electric Power Research Institute, June 1984.
1532.Memorandum for C. Serpan from W. Minners, "Identification of New Generic Issue: Hydrogen Storage Facility Separation," December 16, 1993. [9312290134]
1533. Letter to the U.S. Nuclear Regulatory Commission from M. Tuckman (Duke Power Company), "Oconee Nuclear Station, Docket Nos. 50-269, 50-270, and 50-287, Generic Letter 88-20," November 30, 1990. [ML080520411]
1534.EGG-SSRE-9747, "Improved Estimates of Separation Distances to Prevent Unacceptable Damage to Nuclear Power Plant Structures from Hydrogen Detonation for Gaseous Hydrogen Storage," Idaho National Engineering Laboratory, (Draft) November 1993. [9502070287]
1535.SCIE-EGG-103-89, "Draft Technical Evaluation Report on U.S. Commercial Power Reactor Hydrogen Tank Farms and Their Compliance with Separation Distance Safety Criteria," Scientech, Inc., March 1990. [9502070289]
1545. NUREG-1364, "Regulatory Analysis for the Resolution of Generic Safety Issue 106: Piping and the Use of Highly Combustible Gases in Vital Areas," U.S. Nuclear Regulatory Commission, June 1993.
1547. Letter to All Holders of Operating Licenses or Construction Permits for Nuclear Power Reactors from U.S. Nuclear Regulatory Commission, "Research Results on Generic Safety Issue 106, "Piping and the Use of Highly Combustible Gases in Vital Areas," (Generic Letter 93-06)," October 25, 1993. [ML031200621]
1552. Information Notice 89-44, "Hydrogen Storage on the Roof of the Control Room," U.S. Nuclear Regulatory Commission, April 27, 1989. [ML031180843]
1964. Memorandum for B.W. Sheron from B.G. Beasley, "LOW Priority Generic Issues," March 17, 2011.[ML092520025]
1969.NFPA 50A is referenced in the SRP (post-1979 plants) or Appendix A to Branch Technical Position APCSB 9.5-1 (pre-1979 plants).
1970.Inspection Procedure 71111.05AQ, "Fire Protection Annual/Quarterly," U.S. Nuclear Regulatory Commission, Washington, DC.
1971.Inspection Procedure 71111.05T, "Fire Protection (Triennial)," U.S. Nuclear Regulatory Commission, Washington, DC.
1973. Letter from R. Daley to D. Wadley, "Prairie Island Nuclear Generating Plant, Units 1 and 2, Evaluation of Changes, Tests, or Experiments and Permanent Plant Modifications Baseline Inspection Report 05000282/2008007; 05000306/2008007(DRS)," December 23, 2008. [ML083590119]