International Energy Agency Hydrogen Implementing Agreement. Task 19 Hydrogen Safety. KNOWLEDGE GAPS IN HYDROGEN SAFETY A White Paper.

Transcription

1 International Energy Agency Hydrogen Implementing Agreement Task 19 Hydrogen Safety KNOWLEDGE GAPS IN HYDROGEN SAFETY A White Paper Prepared by Andrei V. Tchouvelev Subtask A Risk Management Leader This version R1: January 2008 For questions and comments please contact: A.V.Tchouvelev and Associates 6591 Spinnaker Circle Mississauga, ON Canada L5W 1R2 (416) (905) fax atchouvelev@tchouvelev.org Andrei V. Tchouvelev Page 1 of 54 R1_Jan-08

2 Table of Contents 1 Background Preamble Knowledge Gaps in Hydrogen Safety Knowledge Gaps Categories Knowledge Gaps in Codes & Standards Defining Hazardous Zones Safety Standards for Hydrogen FC Vehicles Safety Distances for Hydrogen Fuelling Stations Safety Standard for Hydrogen Detection Conclusion Knowledge Gaps in Risk Assessment Risk Criteria Ignition Probabilities Consistent Methodology for Site Risk Assessment Gaps in Fundamental Knowledge Auto Ignition Protective Barriers Consequence Modeling Wall Jets Discussion Next steps References...53 Andrei V. Tchouvelev Page 2 of 54 R1_Jan-08

3 Task 19 Hydrogen Safety Knowledge Gaps in Hydrogen Safety A White Paper 1 BACKGROUND The issue of knowledge gaps within Task 19 Hydrogen Safety was first raised during the expert meeting in Pisa in September At that time it was not really called knowledge gaps but rather what do we need to do to validate our models. Between Pisa and the meeting in Long Beach (March 2006), Dr. Pierre Benard from Hydrogen Research Institute (Canada) with contribution from Dr. Henri Paillere from CEA (France) prepared a draft list of experiments for Hydrogen Safety. At the Long Beach meeting, Dr. Andrei V. Tchouvelev from A.V.Tchouvelev & Associates and CTFCA (Canada) and the Leader of Subtask A Risk Management took the task to identify gaps in hydrogen safety knowledge and make recommendations for future testing and modeling programs. Drs. Tchouvelev and Benard reviewed the original list and expanded it to include various areas of hydrogen science and technology where they felt the gaps existed. The first draft of the Knowledge Gaps to Address via Experiments and Modeling document was released in early June The document was circulated within Task 19 experts and received good feedback. By the meeting in Vancouver (September 2006) we had circulated 4 th updated version. Considering the importance of the knowledge gaps for the whole Task 19 program, it was decided to dedicate a separate session within Subtask A agenda for the knowledge gaps discussion at the Vancouver meeting. Task 19 experts were asked to present the issues that are being considered as knowledge gaps in hydrogen safety in their countries. The goal of the Vancouver knowledge gaps session was to exchange opinions and reach a consensus on existing knowledge gaps to be addressed by future research, testing and modeling activities. The following experts made specific presentations on knowledge gaps issues: Andrei Tchouvelev (Canada) provided introduction to Knowledge Gaps working document and gave examples of knowledge gaps related to codes & standards development as well as fundamental knowledge Olav Roald Hansen (Norway) Validation status FLACS, knowledge gaps and needs for improvement Jay Keller (USA) - Research and Development for Hydrogen Safety, Codes and Standards - Sandia has extensive experience and knowledge on combustion processes, including hydrogen but certain aspects related to hydrogen impingement against protective barriers are not well understood. Pierre Benard (Canada) presented his perspective on knowledge gaps related to high pressure releases of hydrogen Andrei V. Tchouvelev Page 3 of 54 R1_Jan-08

4 Stuart Hawksworth gave a presentation on knowledge gaps identified by HSE/HSL and by the EU HySafe partners. Alessia Marangon gave a presentation on Knowledge Gaps concerning Risk Assessment Studies. Keiko Chitose gave a presentation on Research on Fundamental Properties of Hydrogen as performed in Japan. Jake DeVaal gave a presentation on knoeledge gaps related to Fuel Cell Vehicle Discharge Flammability & Potential Hydrogen Build-Up Testing. List of Experts Contributed to Knowledge Gaps Discussion Name Organization Country Pierre Benard Hydrogen Research Institute Canada Dag Bjerketvedt Telemark University College Norway Kieko Chitose Mitsubishi Heavy Industries Japan Jake Devaal Ballard Power Systems Canada Stuart Hawksworth HSL/HSE United Kingdom Olav Roald Hansen Gexcon Norway Jay Keller Sandia National Laboratories USA Chris Moen Sandia National Laboratories USA Jeff LaChance Sandia National Laboratories USA Henri Paillere CEA France Michael Swain University of Miami USA Andrei Tchouvelev AVT, CTFCA Canada At the end of the Vancouver meeting it was decided that Dr. Tchouvelev will prepare draft of A Knowledge Gaps White Paper that would address the safety related barriers to the widespread use of hydrogen. The initial focus of the White Paper could be the hydrogen infrastructure. 2 PREAMBLE Improved safety comes from understanding the outcomes and probabilities of undesirable events that may occur with new technologies, and by mitigating any unacceptable risks posed by these new technologies. In this regard, while a lot is known about hydrogen combustion and its safe handling, it is important to realize that hazards with new hydrogen technologies that are unrecognized or incompletely understood are difficult to mitigate against. As such, before appropriate mitigations can be developed, the underlying risks must be identified, quantified, and be well understood. Given this, the IEA Task 19 hydrogen experts have tried to name/identify knowledge gaps and barriers for selected applications and to indicate how it can be overcome. The intention of this activity is to focus limited resources on reducing the barriers in order to accelerate the use of hydrogen as a fuel globally. Andrei V. Tchouvelev Page 4 of 54 R1_Jan-08

5 This initial draft is based on the original knowledge gaps document released in June 2006, input from Task 19 experts received before the Vancouver meeting, knowledge gaps presentations and discussions at the Vancouver meeting as well as results from Canadian Hydrogen Safety Program projects, and papers and presentations developed in preparation for the ICHS 07 (September 2007). This paper also helps to define priorities for the next 3-year term of Task KNOWLEDGE GAPS IN HYDROGEN SAFETY 3.1 Knowledge Gaps Categories During Vancouver discussions three knowledge gaps categories were identified: Gaps in the existing codes and standards (C&S) as well as gaps needed to be filled for the ongoing C&S development; Gaps in the existing risk assessment methods and tools for their application to hydrogen systems; and Gaps to be filled in fundamental knowledge (that also directly relate to gaps in CFD modeling approaches and tools). It should be noted that the above categories are tightly related and the example cases provided below could easily be fitted into several categories. For example, a case related to defining hazardous zones could easily apply to gaps in risk assessment as consequence analysis would take the potential size of a flammable cloud into account. Case related to SAE J2578 development could be extended to gaps in fundamental knowledge as it tightly relates to better understanding flammability limits of hydrogen under various real life conditions. Issues related to risk criteria and ignition probabilities do not only relate to gaps in risk assessment but also to C&S, but also to fundamental hydrogen properties (spontaneous ignition). Problems raised within the fundamental knowledge gaps, are also related to CFD modeling and applicable to C&S development and risk assessment. The list may go on. It is, thus, important to stress that resolution of just one knowledge gap spans its effect across many areas related to hydrogen safety. The text below not only identifies certain knowledge gaps, but also specifies the needs and provides examples how those gaps could be overcome. 3.2 Knowledge Gaps in Codes & Standards Defining Hazardous Zones This gap relates closely to hydrogen properties and its behaviour in confined spaces and outdoors. It has been long suspected by hydrogen experts that, due to hydrogen high buoyancy and diffusivity, recommendations of IEC for determining sizes of hazardous zones are likely too conservative and result in inaccurate combustible volumes for hydrogen and higher than necessary hazardous zones. This, in turn, results in higher than necessary use of classified components that eventually increase the overall cost hydrogen systems. Andrei V. Tchouvelev Page 5 of 54 R1_Jan-08

6 The examples presented below contributed by Canada (Dr. Andrei V. Tchouvelev) as part of Canadian Hydrogen Safety Program illustrate this concern. In these examples, Computational Fluid Dynamics (CFD) software Phoenics was used to model potential releases of hydrogen under various conditions of a Hydrogen Station for both indoor and outdoor release scenarios. The obtained CFD numerical simulation results were compared with the requirements of the above standard Indoor Modeling Scenario Description This subsection discusses hydrogen release into the Generator Room of the Hydrogen Energy Station from a electrolytic hydrogen generator during self-purging start-up procedure [1]. At start-up, to ensure only high purity gas is directed for compression, hydrogen is being vented for 10 min. After 10 min a regulator driven by PLC logic re-directs hydrogen flow from vent to process. The point of potential release the vent pipe at the roof of the hydrogen generator. The outlet pipe size is 2 and the constant flow rate is Nm3/s. This is a low- pressure release (P = 1 psig). This worst case scenario assumes, conservatively, that at the start-up the generator vent pipe comes off and all hydrogen that is generated during these 10 min is vented inside the Generator Room, and hydrogen sensors fail to detect the leak until after 10 minutes. Thus, this potential hydrogen release has maximum duration of 10 min. Before Leak Simulation The existence of a louver and a exhaust fan in the Generator Room creates a steady-state airflow with 3-D fluid flow pattern. This airflow was simulated first, before trying to simulate the transient 3-D behaviour of hydrogen cloud introduced by a particular hydrogen release, since it provides initial fields of gas velocity and pressure for the release scenario considered below. Figure 1 shows ventilation velocities created by the louver and the exhaust fan. Figure 1. Ventilation velocities (X- and Y-planes) before leak. Andrei V. Tchouvelev Page 6 of 54 R1_Jan-08

7 Leak Simulation: Release from hydrogen generator vent line As mentioned above, the leak scenario considers the case when, for whatever reason, during the CF hydrogen generator start-up self-purging procedure the hydrogen vent line on the roof of CF comes off, thus causing all hydrogen being produced during the self-purging procedure (10 min) to leak into the Generator Room. It is also assumed that all hydrogen sensors intended to shut down the CF during the self-purging procedure are disabled. Room ventilation is provided by the louver and the exhaust fan (1 m 3 /s). CFD predictions of 3-D hydrogen concentration distribution are shown in Figures 2 and 3. Figure 2 illustrates the H 2 (4% vol.) iso-surface at the end of the release (10 min) and Figure 3 shows the H 2 (2% vol.) iso-surface at the end of the release. It is seen that the sizes of these two clouds are very different. Figure 5. End of 10-min release from the CF vent line (4% hydrogen cloud). Andrei V. Tchouvelev Page 7 of 54 R1_Jan-08

8 Figure 3. End of 10-min release from the CF vent line (2% hydrogen cloud). Size of flammable mixture cloud (CFD approach) The size of the flammable cloud was calculated, using the programmability of the PHOENICS CFD software. Three global quantities, DOMV, P4H and P2H were defined as the volume (in m 3 ) of the whole domain (DOMV), the volume of the hydrogen cloud with more than 4% volume concentration (P4H) and the volume of the hydrogen cloud with more than 2% volume concentration (P2H) respectively. The printout from the global calculations file is shown below Global calculations: DOMV = P4H = E-02 P2H = It is seen that the 4% cloud volume (P4H), which is about m 3, is much smaller than the volume of cloud with 2% volume concentration (P2H), which is about m 3. Both clouds are much smaller in volume than the whole domain volume (DOMV), which is about 230 m 3. Andrei V. Tchouvelev Page 8 of 54 R1_Jan-08

9 Requirements of IEC IEC sets out the essential criteria against which the risk of ignition can be assessed, and gives the guidance on the design and control parameters that can be used in order to reduce such a risk. The important criteria are: the LFL of the gas, leak rate, concentration and grade of release the degree and availability of ventilation and if there are any obstacles Calculation to ascertain the degree of ventilation within Container Gas Compartment The following method of ventilation calculation was taken from IEC CHARACTERISTICS OF RELEASE Flammable material Hydrogen gas Source of release vent pipe Lower Explosive limit (LFL) 3.3 x 10-3 kg/m 3 (4% vol) Grade of release Secondary Safety factor, k (applied to LFL) 0.5 (secondary release) Release rate (dg/dt) = 12.6 Nm 3 /hr = (H 2 density = kg/m 3 at 20C, = density = at 273C) 12.6 Nm 3 /hr = m 3 /sec x 0.090kg/m 3 = kg/sec Gas concentration in release X o = 100% VENTILATION CHARACTERISTICS Generator room Volume = 7.3m x 7.5m x 4.2 m = 230 m 3 equipment volume = 185 m 3 Fan airflow = 1 Nm3/sec = 3600 Nm 3 /hour Number of air changes, C = 3600 m 3 /hr/185m 3 = 19.5/hr = /sec Quality factor, f = 2 (There are few obstacles to impede airflow) Ambient temperature, T = 35C (308K) Temperature coefficient, (T/293 K) = 1.05 MINIMUM VOLUMETRIC FLOW RATE OF FRESH AIR: 4 ( dg / dt) max T ( dv / dt) min = = = 0.175m 3 k LEL / sec Andrei V. Tchouvelev Page 9 of 54 R1_Jan-08

10 EVALUATION OF HYPOTHETICAL VOLUME V Z f ( dv / dt) C min V z = = = Conclusion 64.8m 3 The hypothetical volume V z of 65 m 3 is significant relative to the room volume of 230 m 3 and would likely result in a Zone 2 room classification for at least the upper third of the room. IEC predicts the size of 50% LFL cloud an order of magnitude larger vs CFD analysis (64.8 m 3 vs 6.2 m 3 ) Outdoor Modeling Scenario Description The scenario considered in this chapter addresses potential leaks from high-pressure piping between the hydrogen generator and high-pressure storage [2]. Assumed leak rate was substantial enough to cause a noticeable pressure loss from 400 bar storage and higher than a leak rate that might be anticipated from valves and fittings in smaller hydrogen systems where piping connections are welded or use compression fittings and diameters are typically 10 to 19 mm. From this, the hypothetical ignitable volumes for several scenarios were calculated using IEC expressions and these were also compared to the CFD modeling results for ignitable mixture volume and the extents of the mixture from the leak source. Two H 2 leak rates of 5 and 20 scfm ( and kg/sec) were used as a starting point and the necessary holes sizes to produce this leak rate were calculated for 400 bar storage pressure. Hydrogen gas leaks 5 and 20 scfm (2.37 l/sec and 9.47 l/sec) in downward, upward and horizontal orientations in a 0.5 m/sec wind were modeled. A low wind velocity was selected as it would allow the greatest accumulation of hydrogen and this velocity is what is considered by IEC to be the lowest outdoor ground velocity. 400 bar leaks from hydrogen piping vertical and horizontal outdoors Calculated orifice diameters: (a) 0.10 mm (0.004 in) (b) 0.20 mm (0.008 in) As representative examples only 20 CFM leaks images from 0.2 mm orifice are shown here. 20 CFM downward flow Extents of concentration envelopes: Concentration Horizontal (m) Vertical (m) 200% LFL (8% vol.) LFL (4% vol.) 3* (Touch the ground) 0.63* (Touch the ground) 50% LFL(2% vol.) 3*(Touch the ground) 3.31*(Touch the ground) Andrei V. Tchouvelev Page 10 of 54 R1_Jan-08

11 Hydrogen cloud volume: 0.412m 3 for LFL (4% vol.) and 3.735m 3 for 50% of LFL (2% vol.). Hydrogen volume according to IEC : 5.64m 3 for k=0.5, f=1. Figure 4. Small leaks from hydrogen piping 20 CFM, downward direction 20 CFM upward flow Extents of concentration envelopes: Concentration Horizontal (m) Vertical (m) 200% LFL (8% vol.) LFL (4% vol.) % LFL (2% vol.) Hydrogen cloud volume: 0.524m 3 for LFL (4% vol.) and 5.66m 3 for 50% of LFL (2% vol.). Hydrogen volume according to IEC : 5.64m 3 for k=0.5, f=1. Andrei V. Tchouvelev Page 11 of 54 R1_Jan-08

12 Figure 5. Small leaks from 40 Mpa hydrogen piping 20 CFM, upward direction 20 CFM horizontal flow Extents of concentration envelopes: Concentration Horizontal (m) Vertical (m) 200% LFL (8% vol.) LFL (4% vol.) % LFL(2% vol.) Hydrogen cloud volume: 0.106m 3 for LFL (4% vol.) and 1.399m 3 for 50% of LFL (2% vol.). Hydrogen volume according to IEC : 5.64m 3 for k=0.5, f=1. Andrei V. Tchouvelev Page 12 of 54 R1_Jan-08

13 Figure 6. Small leaks from 40 Mpa hydrogen piping 20 CFM, upward direction Considering the different release rates and release directions, there are 6 modeling scenarios, whose results are summarized in Table 1 below. Comparison between CFD Results and IEC Calculations for H 2 Leak Scenarios IEC is a standard referred to in the Canadian Electrical Code and NFPA 70 (National Electrical Code in USA) to determine ventilation requirements and hazardous locations. The standard uses a calculation to determine the hypothetical combustible volume caused by a fluid leak under specific conditions, and ventilation rates. Due to hydrogen s high diffusion rate and buoyancy, it is our opinion that the calculations in the standard are possibly too conservative and result in inaccurate combustible volumes for hydrogen. Table 1. Summary of the modeling scenarios and the numerical results. Flowrate (SCFM) Re # % vol. Hydrogen cloud volume (m 3 ) Horizontal cloud extension (m) Vertical cloud extension (m) IEC CFD 8 % vol. 4% vol. 2% vol. 8 % vol. 4% vol. 2% vol. 20 (down) * 3.31* 0.6 3* 3* 5 (down) * * 20 (up) (up) (horiz.) (horiz.) Notes: * These clouds touch the ground, which is 3 m below the leak orifice. Andrei V. Tchouvelev Page 13 of 54 R1_Jan-08

14 Conclusion The hypothetical volumes predicted by IEC do not depend on the direction of the leak while CFD modeling predicts a substantial difference. Vertical leaks create a largest extent due to buoyancy, while wind and buoyancy tend to disperse horizontal leaks much faster. Also, as in the previous example, the IEC standard substantially overpredicts the size of hazardous zones. A recent paper prepared by HySafe WP 11 partners [3] comes to a similar conclusion and suggests to discuss the proper LFL criterion (LFL, ½ LFL or ¼ LFL) to be used for hydrogen s zone classification and zone extent between regulators and scientists. Based on present findings it seems that using ½ or ¼ LFL as basis for determination of the hazardous zone for hydrogen jets in open areas is an overly conservative approach. Key changes that need to be developed to appropriately update this key standard should include the development of a more realistic correlation between hydrogen concentrations and the sizes of corresponding volumes as well as take into account effects of congestion and geometry on the distribution of hydrogen concentrations Safety Standards for Hydrogen FC Vehicles The input to this section was provided by Dr. Jake Devaal, Ballard Power Systems, Canada [4]. Application: Transportation, Fuel Cell Vehicles (FCVs): Specifically, Hydrogen-fuelled passenger cars for widespread public use. Barrier: Public acceptance of FCVs fuelled with gaseous or liquid hydrogen; specifically, concerns that hydrogen may leak from the fuel storage system or from the operating vehicle and pose flammability/explosion issues, especially when the vehicle is parked inside a residential garage. Other (related) issues include concern that these vehicles may pose an undue hazard in the event of building fires. Acceptance that indoor parking is safe and these issues have been adequately addressed will be required by vehicle owners, local Authorities Having Jurisdiction (AHJs) and insurance companies before this barrier is completely overcome. Specific technical issues that must be addressed: FC Vehicles must be designed and built such that they cannot leak amounts of hydrogen that would pose a significant flammability concern when parked or operated indoors in residential garages, or, if this cannot be completely assured, provided with systems that would detect and eliminate or mitigate these releases. Also, it must be demonstrated that these vehicles do not behave worse than current vehicles when involved in building fires. Once these actions have been taken by the vehicle manufacturers and component suppliers, the public, AHJs and the insurance industry must also be shown that these potential issues have been addressed, thus providing assurance that these vehicles are safe to park and operate indoors. Knowledge gaps: Technically, a few; but likely none that cannot be overcome. In terms of gaps already addressed, the latest draft SAE J2578 standard identifies the hydrogen leak rates that vehicles must remain below to ensure safety when parked or when operated at idle indefinitely to prevent undesirable levels of hydrogen from building up, or posing local discharge flammability Andrei V. Tchouvelev Page 14 of 54 R1_Jan-08

15 concerns. The hazards posed by building and/or vehicle fires are also generally well understood, where the fuel supply systems are designed to safely vent in the event of a fire. The remaining technical challenges and knowledge gaps mainly reside in: 1) developing a good understanding of the probabilities of encountering component failures that could release hydrogen in unacceptable amounts, 2) the outcomes and probabilities of igniting these volumes in a residential garage, and 3) whether active mitigations of these releases are warranted to arrive at an acceptable risk level and to achieve public acceptance. As such, work with other hydrogen experts to better define and quantify these specific risks for evolving FCV designs is believed useful to achieve early and widespread acceptance of FCVs. The highlights of discussed issues are presented in the following three slides from Dr. Devaal s presentation. Andrei V. Tchouvelev Page 15 of 54 R1_Jan-08

16 3.2.3 Safety Distances for Hydrogen Fuelling Stations The input to this section was provided by Dr. Jeff LaChance, Sandia National Labs, Albuquerque, USA. Andrei V. Tchouvelev Page 16 of 54 R1_Jan-08

17 This section is based on the work conducted by the Sandia National Labs in the US [5] sponsored by the US Department of Energy. The text below uses the terminology accepted in the US where safety distances are often called separation distances but have the same principal meaning. Separation distances, however, may also include clearance requirements for inspection and maintenance purposes. The development of separation distances for the new generation of hydrogen facilities can be determined by SDOs and evaluated by facility designers in several ways. A conservative approach is to use the worst possible accidents in terms of consequences. Such accidents may be of very low frequency such that they would likely never occur. Although this approach bounds separation distances, the resulting distances are generally prohibitive. The current separation distances do not reflect this approach. An alternative deterministic approach that is often utilized by SDOs and allowed under some regulations is to select accident scenarios that are more probable but do not provide bounding consequences. In this approach, expert opinion is generally used to select the accidents used as the basis for the prescribed separation distances. Although anecdotal experience often forms the basis for the selection of the accidents, the frequency of accidents can also be used as a selection criterion. Figure 7 provides an example of deterministic separation distances based on one possible consequence of a hydrogen leakage event: the radiant heat flux from an ignited hydrogen jet. The figure shows the separation distances required to limit the exposure of a person to a radiant heat flux of 1.6 kw/m 2 which is generally accepted as a level that will not result in harm to an individual even for long exposures (this heat flux level is currently specified in the IFC [6] as a no harm criterion for designing hydrogen vent systems). The separation distances were calculated using a Sandia developed model for predicting the radiant heat fluxes and flammability envelopes from high pressure releases of hydrogen [7]. The calculated values are conservative since they assume free-forming jet fires that are not affected by the ground or structures and are orientated towards the target (jet fires that are orientated upwards result in separation distances that are roughly half the distance shown in Figure 7). As indicated in Figure 7, the required separation distances are significantly affected by the pressure of the hydrogen gas and the leak diameter. These insights suggests the gas storage pressure is an important parameter that should be considered when specifying separation distances and that the selection of a specified leak size and orientation is critical in determining the separation distances. The separation distances shown in Figure 7 are generally larger than those currently specified in the ICC and NFPA codes and standards even for low pressure systems and small leak diameters. One way to reduce consequence-based separation distances is to use a higher consequence level that introduces the potential for injuring the public or damaging structures. This is illustrated in Figure 8 which shows the separation distances that would be required for different radiant heat flux levels and hydrogen gas pressures. The separation distances were evaluated for a specified leak diameter of 2.38 mm and only consequences related to hydrogen jets are included in this plot (the potential consequences from other possible accidents such as vapor cloud explosions were not examined in this study but must also be considered when determining separation distances). Figure 8 also illustrates the separation distances required for hydrogen concentrations ranging from 2% to 8%. Andrei V. Tchouvelev Page 17 of 54 R1_Jan-08

18 Separation Distance (m) Pressure (MPa) Leak Diameter (mm) Figure 7. Separation distances required for an exposure to a radiant heat flux of 1.6 kw/m 2 generated by a jet fire. 45 Separation Distance (m) Consequence Parameter 1.6 kw/m2 4.7 kw/m2 25 kw/m2 Flame Length 2% Hydrogen 4% Hydrogen 6% Hydrogen 8% Hydrogen System Pressure (MPa) Figure 8. Separation distances required for a jet fire from a 2.38 mm diameter leak using different consequence parameters. An alternative approach for selecting the scenarios utilized to establish separation distances for a hydrogen facility involves the use of the estimated risk associated with the operation of the facility. Basically SNL suggestes using a modified EIGA conceptual framework [8]. This approach can provide a basis for eliminating large leakage events which have low frequencies and result in significant consequences that require large separation distances to protect the public, structures, and equipment from harm. The difference with EIGA concept is that SNL uses cumulative risk rather than frequency to identify risk criteria. This approach is illustrated in Figure 9 below. Andrei V. Tchouvelev Page 18 of 54 R1_Jan-08

19 1.0E-05 Increasing leak diameter Cummulative Risk (/yr) 1.0E E-07 Cummulative frequency of accidents requiring this separation distance Risk Criteria 1.0E-08 Separation Distance Separation Distance (m) Figure 9. Risk-informed approach for establishing safety distances Application Of Risk-Informed Approach To demonstrate the utility of the proposed risk-informed approach, a limited scope QRA was performed for an example hydrogen refueling station with an operating pressure that is not currently reflected by existing codes and standards. Specifically, the analysis was performed for a medium-size hydrogen refueling facility capable of refueling 100 cars/day at a pressure of 70 MPa. The scope of the analysis presented in this paper is limited to the gas storage area which is one of the focuses of the separation distances specified in most codes and standards. The facility was assumed to use 51 gas cylinders (250 liters in size) arranged in three cascades for a combined gas storage capacity of 500 kg (12.63 m 3 at 70 MPa). The facility design was assumed to meet codes and standard requirements including the presence of safety relief valves at important locations that release to elevated vent lines. The facility was also assumed to utilize properly located hydrogen and flame detectors to isolate the discharge pipes from the gas storage area upon detecting either a hydrogen leak or jet flame. However, this isolation capability does not terminate leakage from the gas cylinders and from other components upstream of the isolation valves. Gas flow check valves were assumed present to prevent back flow from one gas cascade to another to limit the duration of a leakage event. Intentional venting of hydrogen to reduce the duration of a detected leakage event was also not included in the current assessment. Only random component failures leading to hydrogen leakage were included in this example. Leakages initiated by human errors, or natural events, or by other mechanisms such as automobile accidents were not included. Leakage contributions from valves, piping, gas cylinders, connections, and instrument lines were included in the analysis. The assumed size of the piping and valves are Andrei V. Tchouvelev Page 19 of 54 R1_Jan-08

20 representative of a 70 MPa system and vary with location and function. The diameter of the discharge lines from the gas cylinders to the gas inlet and outlet manifolds were assigned a diameter of 6.35 mm and the manifolds for each cylinder cascade was assumed to have a diameter of 11.5 mm. Figure 10 illustrates the accident event tree that was used for evaluating the hydrogen release scenarios from the gas storage area. The accident sequence modeling is relatively simple since leaks in the gas storage area cannot be isolated. The accident event tree includes several phenomenological events that influence the accident sequence and resulting consequences. Included in these phenomenological events are the size of the leak, the potential for immediate ignition, and delayed ignition. The event tree illustrates the resulting consequences for each sequence which includes jet fires, flash fires following delayed ignition of hydrogen (the potential for vapor cloud explosion or hydrogen detonation was not included), and un-ignited gas releases. Gas Storage Cylinder Leak or Rupture Immediate Ignition of Hydrogen Jet Delayed Ignition of Hydrogen CYLINDER-L I-IGNITION D-IGNITION # END-STATE-NAMES 1 JET-FIRE 2 FLASH-FIRE 3 GAS-RELEASE cylinder leak - (New Event Tree) 2007/01/27 Page 0 Figure 10. Hydrogen gas storage leakage/rupture event tree. The QRA analysis also required the evaluation of the consequences for each hydrogen release scenario. As indicated previously, the consequences considered in this example QRA were limited to exposure to radiant heat fluxes and flash fires. A Sandia computer model developed by Houf and Scheffer [7] for predicting the behavior of jet flames and hydrogen concentrations was used to determine the resulting consequences for the hydrogen leakage events. The leak orientation was assumed to be directly at the target which results in the longest required separation distances. Examples of the results from this model were presented previously in Figures 7 and 8. Andrei V. Tchouvelev Page 20 of 54 R1_Jan-08

21 Results of Example Analysis The results for the two accident sequences modeled for the gas storage area are presented in Figures 11 and 12. The results for a sequence involving spontaneous ignition leading to a jet fire (i.e., sequence 1 in Figure 10) are shown in Figure 11. Each point on the curves represents the cumulative risk from leaks exceeding a specific diameter and the resulting separation distance that corresponds to the selected consequence parameter for that diameter. For these sequences, the consequence parameters are exposure to radiant heat fluxes or direct flame contact. The separation distances required for three different radiant heat fluxes are illustrated (the distance required for the 25 kw/m 2 heat flux also are representative of the visible flame length and thus provides an indication of where direct flame contact would occur). The risk measures (shown on the y-axis) for these three radiant heat fluxes represent a no harm risk (1.6 kw/m 2 ), the risk of injury (4.7 kw/m 2 ), and the risk of significant harm leading potentially to a fatality (25 kw/m 2 ). The unusual shape of the curve reflects the contribution to risk due to leakage from the different components in the gas storage area. The results for a delayed ignition leading to a flash fire (i.e., sequence 2 in Figure 10) are illustrated in Figure 12. The delayed ignition is assumed to result in a flash fire out to a distance corresponding to the LFL of 4% that would result in a risk of significant harm possibly leading to death to anyone within the resulting flame distance. The separation distances that would be required for a 2% hydrogen concentration is also shown in the figure to illustrate the distances associated with a no harm risk criteria. A comparison of the two figures indicates that both the frequencies and separation distances for flash fire sequences are greater than for the jet fire sequences and thus the consequences from flash fires should determine the required separation distances. The figures also indicate that the risk-informed separation distance can be large for gas storage areas with operating pressures of 70 MPa. Thus, some additional means of mitigating the consequences of leaks is required in order to justify shorter separation distances for this example facility. Sandia is currently evaluating one such mitigation feature the use of barriers. Andrei V. Tchouvelev Page 21 of 54 R1_Jan-08

22 1.0E E-03 No Harm Criteria Cummulative Risk (/yr) 1.0E E-05 Injury Criteria Harm Criteria Heat Flux 1.6 kw/m2 4.7 kw/m2 25 kw/m2 1.0E Separation Distance (m) Figure 11. Risk results for unisolated jet fires originating in gas storage area. 1.0E-02 Cummulative Risk (/yr) 1.0E E E-05 No Harm Criteria Harm Critera Hydrogen Concentration 2% H2 4% H2 1.0E Separation Distance (m) Figure 12. Risk results for unisolated flash fires originating in gas storage area. The results in Figures 11 and 12 can be used to illustrate the interaction between different risk measures and consequence parameters on the selected separation distances. Superimposed on both figures are a set of example risk criteria that are consistent with the risk data from gasoline stations: a criteria of 1E-5/yr for consequences resulting in significant harm to individuals that may lead to death (e.g., flash fire sequences and jet fires resulting in high radiant heat fluxes), a criteria of 1E-4/yr for consequences resulting in less serious injuries (e.g., a radiant heat flux level of 4.7 kw/m 2 ), and a criteria of 1E-3/yr for consequences that do not result in harm to an individual (e.g., 2% hydrogen concentration and 1.6 kw/m 2 ). For the jet fire sequences, the sequence frequencies are below the no harm criteria indicating that a short separation distance based on that criteria could be possible. The Andrei V. Tchouvelev Page 22 of 54 R1_Jan-08

23 use of the injury and harm criteria suggests risk-informed leak sizes that provide separation distances of approximately 20 m and 35 m, respectively for the example facility. The leak diameters that result in these separation distances are 5 mm and 10 mm, respectively. The selection of any distance less than those determined by the harm criteria would potentially introduce residual risk from larger leaks that would violate the harm criteria. Similar results are illustrated for the flash fire scenarios. The no harm separation distance is approximately 25 m for a leak diameter of 2 mm and the harm separation distance is approximately 75 m for a leak diameter of 10 mm. These results indicate that although it may be desirable to use no harm or injury consequence measures (e.g., 1.6 or 4.7 kw/m 2 ) to determine separation distances, their use with a relatively high no harm and injury risk criteria may result in a risk of significant harm that violates an accepted harm criteria. Thus, it may be required to us a consequence level and risk criteria that reflects significant injury or death from hydrogen leaks in the risk-informed determination of separation distances. In the example above, the risk-informed leak diameter, based on a harm risk criteria of 1E-5/yr, would be 10 mm and the separation distance 75 m. For comparison, a consequence-based separation distance was calculated using 20% of the gas cylinder outlet manifold (d = 11.5 mm) flow area. The equivalent leak diameter is 5.1 mm and the required separation distance is approximately 37 m which are less than the risk-informed values. The risk of harm for this separation distance taken from Figure 15 is approximately 1E-4/yr which is an order of magnitude greater than the risk criteria. Thus, use of the consequence-based separation distance would result in a level of risk substantially greater than the acceptance criteria Conclusion An example application of the risk-informed approach has been performed to illustrate its utility and to identify key parameters that can influence the resulting selection of separation distances. Important parameters that were identified include the selected consequence measures and risk criteria, facility operating parameters (e.g., pressure and volume), and the availability of mitigation features (e.g., automatic leak detection and isolation). The results also indicate the sensitivity of the results to key modeling assumptions (e.g., leak orientation and duration) and the component leakage rates used in the QRA models. These insights are being used to generate a set of separation distances for use by SDOs in a risk-informed process to establish or modify hydrogen related codes and standards Safety Standard for Hydrogen Detection The input to this section was provided by Dr. Andrei V. Tchouvelev, AVT and CTFCA, Canada. This section is based on the work conducted within Canadian Hydrogen Safety Program [9]. Early hydrogen detection is one of the key risk mitigation measures associated with potential hydrogen releases from hydrogen containing equipment and pipework. But how early should it be implemented to effectively improve safety and not become an operational nuisance at the same time? Hydrogen Detection at Fuelling Stations: The 100 ppm requirement Some people have a perception that if the hydrogen detection limit is as low as possible then safety will be improved (or risks reduced) at hydrogen fueling stations. This belief is particularly strong in Japan within the circle of experts that design, build and operate hydrogen stations. An informal survey of related Japanese industries on the minimum required detection limit of hydrogen concentrations Andrei V. Tchouvelev Page 23 of 54 R1_Jan-08

24 revealed that the average weighted response was around a few hundred ppm. Thus, to ensure that a few hundred ppm are reliably measured, the requirement for a 100 ppm detection limit is justified. Others think that such a low detection limit has nothing to do with safety or risk, is not scientifically based (on hydrogen properties and risk perspectives) and is purely emotional (perception-based). If we presume that a concentration of 4% hydrogen by volume (40,000 ppm) is flammable outdoors (which, we know, it is not) there is a huge disparity between the 40,000 ppm flammability level and the minimum detection limit 100 ppm. If safety is primary concern, a factor of 10 rather that 400 would be more than adequate. From risk prospective, there is absolutely no reason to believe that hydrogen detection at 100 ppm will reduce risk of incidents during fueling instead of detecting hydrogen at, say, 4,000 ppm (which incorporates a safety factor of 10). Moreover, premature detection and initiated inadequate response may have adverse effects (see below), while timely detection of reasonably significant concentrations will allow the user to concentrate on mitigating real hazards. As Confucius advised: Don t use cannon against mosquitoes. There is nothing wrong with a 100 ppm detection limit in principle; certainly there might be a need for this type of measurement. The point here is: it has nothing to do with improving fueling safety and it will likely be very impractical under the conditions of a fuelling station. It is ironic that in Japan, within the circle of experts discussing maximum allowable hydrogen concentrations that could be emitted from fuel cell vehicles tailpipes, emissions containing up to 4% hydrogen by volume are considered to be potentially permissible. In the US during the development of a recommended practice for general fuel cell vehicle safety, the same tailpipe emissions with 4% hydrogen have been discussed. This means that a regular hydrogen fuel cell vehicle may routinely generate short term emissions of up to 4% hydrogen by volume (or 40,000 ppm) from its tail pipe during shut down or start up. This could happen immediately before or after fueling. One could predict that every normal (i.e. safe) fueling operation will likely set off an alarm of a highly sensitive detection apparatus. This might in turn have very interesting effects ranging from solidifying a sense of fear or unease in the public to strong annoyance from station operators against hydrogen detection followed by permanent disabling of detection system. This would really have an effect on safety--a negative effect. In order to answer the question what is the optimum lower detection limit at hydrogen fuelling stations AVT has conducted CFD modeling analysis of potential tail pipe emission situations that might reasonably occur during refueling. As was noted above in section 3.2.2, Society of Automotive Engineers (SAE) has been developing a new revision of the Recommended Practice SAE J2578 for General FC Vehicles Safety. This document is planned to become a performance based standard, in other words, the maximum allowable concentrations of hydrogen emissions will not be specified only the condition of nonflammability. Recent experiments conducted by University of Miami (Dr. M. Swain) and Ballard Power Systems confirm that lean ignition limit for horizontal hydrogen jets is around 8% vol. This means that FC vehicles may routinely emit (for a short time though during start up and shut down) hydrogen clouds containing concentrations exceeding official LFL for hydrogen of 4% vol. or 40,000 ppm. Andrei V. Tchouvelev Page 24 of 54 R1_Jan-08

25 The question is then: what is the optimum lower detection limit at the refuelling stations so that detection does not become an operational nuisance causing false alarms and, thus, interfering with normal refuelling process? Input conditions for simulations It was reasonably presumed that FC vehicles will emit hydrogen-air mixtures ranging from 4 to 10% vol. from their tailpipes for 5 seconds during their shut down start up and, i.e. before and after refuelling. Emissions flow rate was selected as 185 slpm normal idling flow rate for exhausts (as communicated previously by Ballard Power Systems). The produced cloud would be dispersed within the refuelling station domain under no wind conditions. No wind condition was selected as conservative considering that some station designs have protective barriers around them that would block the wind. Clouds dispersion was modeled up to 30 seconds after the end of the release. Tables 3 and 4 below provide comparison of dispersed hydrogen clouds behaviour and concentrations after the end of the 5-sec release from the tail pipe depending on the original hydrogen concentration in the FC exhaust. Scale on the left lad side of each picture shows the maximum hydrogen concentration at that (elapsed) time after the end of the release. Table 3. Comparison of dispersion of 5-sec 4% and 6% vol. hydrogen-air mixture exhaust out of FC tailpipe Time 4% vol. 6% vol. 5 s 10 s Andrei V. Tchouvelev Page 25 of 54 R1_Jan-08

26 15 s 25 s As it can be seen from Table 3, hydrogen concentrations around ppm are obtained around the canopy within 25 seconds after the release has stopped. Table 4. Comparison of dispersion of 5-sec 8% and 10% vol. hydrogen-air mixture exhaust out of FC tailpipe Time 8% vol. 10% vol. 5 s 10 s Andrei V. Tchouvelev Page 26 of 54 R1_Jan-08

27 15 s 25 s In the case of 8% vol. hydrogen, since the initial hydrogen concentration is higher than those shown in Tables 3, hydrogen would reach the canopy faster likely within seconds or so at concentrations around ppm. In the last case, with the initial hydrogen concentration of 10% vol., hydrogen would reach the canopy even faster likely within 15 seconds or so at concentrations around 1,000 ppm. If we suggest that a few vehicles would refuel one after another, hydrogen dilution would be slower and the chance for higher concentrations accumulation within the reported times, i.e. during refuelling, would increase. Since, however, most of the FC vehicles will have their emissions in check and, thus, maintain them at worst below 6% vol. (reasonable assumption), it is unlikely that hydrogen concentrations at the canopy level would exceed 1,000 ppm due to tailpipe emissions. The only guaranteed way when it happens if the exhausts would contain hydrogen concentrations around 10% vol. and this needs to be detected. So, based on the above analysis, it was recommended to establish the lower detection limit for hydrogen refuelling at 1,000 ppm Conclusion This chapter reviewed several examples of knowledge gaps in the existing codes and standards as well as indicated the needs and suggested pathways to close the gaps in the development of hydrogen specific codes and standards requirements. The considered examples clearly demonstrated that hydrogen codes and standards to be truly reflective of unique hydrogen properties need to take those properties into account as well as hazards associated with the use of hydrogen. Selection of appropriate risk criteria is one of the key conditions of developing uniform and consistent codes and standards requirements (see more on risk criteria in the next section). Use of CFD analysis as well as probabilistic risk assessment might be necessary to help develop those requirements. Andrei V. Tchouvelev Page 27 of 54 R1_Jan-08

28 3.3 Knowledge Gaps in Risk Assessment Risk Criteria Text for this subsection is contributed by Jeff LaChance, Andrei V. Tchouvelev and Jim Ohi [10]. Establishment of risk criteria is a key element in risk management decision making. Since the primary concern is the potential for personnel injury, risk criteria can be established for all the people exposed to the consequences of facility-related accidents, which could include the public located outside the boundaries of the facility, users of the facility, and the facility workers. Societal or public risk is generally the main focus in risk assessments. In most QRA applications, the risk levels for the public are generally set one to two orders of magnitude less than the level for workers. Depending on the accident consequence, the selected risk criteria could reflect acceptance levels for either injuries or fatalities. Risk criteria can be specified with regard to individuals or the society at large. Individual risk reflects the frequency that an average person located permanently at a certain location is harmed. Characterization of the population surrounding a facility is thus not required to evaluate individual risk. Societal risk reflects the relationship between the frequency and the number of people harmed. Evaluation of societal risk requires determination of the population surrounding a facility. For the application of QRA to determine the separation distances specified in codes and standards, the use of individual risk measures may be the most appropriate since they are site independent. Risk acceptance criteria for both individual and societal risk, though de facto exist everywhere, are not always obvious. In some world jurisdictions, like in most Western European countries and Australia, they are incorporated into law. In the U.S. and Canada, to the contrary, as in many other jurisdictions around the world, they are not defined in any way and are, thus, subject to interpretation. Selection of individual risk criteria should be based on sound arguments and reflect the consensus of all stakeholders. Ideally, the risk associated with the widespread development of hydrogen refuelling stations should not substantially increase the injury or fatality risk of an individual. As mentioned in the previous section, this concept is not new and in fact has been utilized in the nuclear power industry. A critical question is what level of risk should be utilized in this concept? Several options are discussed here for consideration by codes and standards groups and other decision makers. The first is to specify that the risk from hydrogen accidents be some fraction of the total risk to individuals from all unintentional injuries. This approach has been adopted by the U.S. Nuclear Regulatory Commission (NRC) in their efforts to risk-inform the regulations for nuclear power plants. The NRC risk criteria is based on the principle that the risk to an average individual in the vicinity of a nuclear power plant should be a fraction (0.1%) of the sum of the fatality risk resulting from other accidents to which members of the public are generally exposed in everyday life (e.g., fatal automobile accidents). At the time the NRC established this policy in 1995, the individual fatality risk in the U.S. was approximately 5 x 10-4 /yr. Recent data [11] suggest that the individual fatality risk from unintentional injuries in the United States is on the order of 3.8 x 10-4 /yr. The individual injury frequency from unintentional accidents is approximately 0.09/yr. Andrei V. Tchouvelev Page 28 of 54 R1_Jan-08

29 For comparison, the individual fatality risk from various causes in the Netherlands [12] is reported to be 2 x 10-3 /yr from smoking, 4 x 10-3 /yr from traffic accidents, 1 x 10-2 /yr from all diseases, and 5 x 10-4 /yr from natural radiation. A variation of this first option is to utilize just the individual fatality and injury risk associated with only fires and explosions. Considering that these are the major concerns associated with hydrogen facility operation, this may be a better approach but it requires careful consideration of available fire and explosion statistics. The individual fatality risk due to fires in the United States is 1.2 x 10-5 /yr and the corresponding value for explosions and overpressure events is 6.0 x 10-7 /yr [11]. Further examination of this data indicates that the individual fatality risk from fires involving highly flammable materials such as hydrogen is approximately 2.0E-7/yr, the risk from structure fires is 9.5 x 10-6 /yr, the risk from fires outside of structures is 8E-8/yr, and the risk from unspecified fire sources is 1.1 x 10-6 /yr. Data on fire-related injuries have not been identified, but the individual injury risk from fires is expected to be approximately two orders of magnitude greater than the values cited above. A better option, as mentioned above, is to specify that the risk associated with hydrogen refueling stations be at par with the risk associated with gasoline or compressed natural gas (CNG) stations. Unfortunately, no published risk assessments for either gasoline or CNG refueling stations that could provide those risk estimates have been identified. However, there are some limited data on the frequency of fires in public gasoline stations [13] for the five-year period of (no published data for CNG stations were identified) that could be used to establish such a comparative criterion. This data indicate that the average frequency of a fire at a gasoline station is approximately 7.4 x 10-2 /yr. A majority of the reported fires were initiated by vehicle fires, and only a small fraction (~4%) was related to spills of gasoline leading to fires or explosions. When vehicle fires are eliminated, the fire frequency is approximately 2.8 x 10-2 /yr, and when only spills are considered, the average fire frequency is approximately 3 x 10-3 /yr. The reported fires resulted in, on average, 2 deaths/yr and 70 injuries/yr. Since there were approximately 100,000 public service stations in operation during this period, the average frequencies of a fatality or injury associated with the operation of a single gasoline station are approximately 2 x 10-5 /yr and 7 x 10-5 /yr, respectively. If vehicle fires are eliminated, the average fatality and injury frequencies associated with operation of an individual gasoline station are approximately 1 x 10-5 /yr and 3.3 x 10-4 /yr, respectively. The corresponding fatality and injury frequencies attributable to gasoline spills are approximately 5 x 10-6 /yr and 9 x 10-5 /yr. An alternative (and, probably, the best) way to determine the risk associated with the existing gasoline infrastructure would be to conduct QRAs of a variety of existing gasoline stations using best available QRA and modeling tools and data. The same could be repeated for other existing types of fuelling stations (e.g., compressed natural gas) as well as for those with co-located fuels. That, however, would require full cooperation of station owners, which is hard if not impossible to achieve. There is no interest from their side to potentially expose themselves to additional liability due to surprises that might be uncovered in the course of such a rigorous analysis, which otherwise stay hidden within the grandfathered permitting and approval process, which is based on non-risk-informed RC&S. Harm scale aversion will have to be taken into account when establishing societal risk criteria. Recent studies surveyed by HSE [14] earlier this year (2007) clearly suggest that society in general is more willing to accept small but frequent accidents rather than a single accident with large consequences, although the resulting number of casualties would be equal in both cases. In the Andrei V. Tchouvelev Page 29 of 54 R1_Jan-08

30 Netherlands, for example, the probability of 10 fatalities in a single incident can be at most once in 100,000 years for installations that include hazardous companies such as large petrochemical plants, LPG filling stations and hazardous material storage facilities. An incident involving 100 fatalities may at most occur once in 10 million years [15]. As can be seen, the aversion slope in this example is -2, i.e. for every order of magnitude increase in number of fatalities, the probability must drop by two orders of magnitude. As was stated in the EIHP2 report [16], The slope of the FN curve is designed to reflect the society s aversion to single accidents with multiple fatalities as opposed to several accidents with few fatalities. It is important to note, however, that though in the Netherlands individual (location-based) risk criterion (set at 1x 10-6 /yr) acquired legal status in July 2004 and is part of the External Safety (Installation) Decree, the societal risk criteria quoted above are not legalized and still remain as guidance values [15]. This is probably an indication that the individual risk should likely be the main focus of risk-informed codes and standards, while the societal risk will remain the focus of site QRAs for the foreseeable future. IEA Task 19 within its subtask on risk management intends to perform detailed analysis of similar data available within IEA countries and recommend uniform risk acceptance criteria for hydrogen refuelling infrastructure for implementation globally Ignition Probabilities Important issues related to knowledge gaps in risk assessment were addressed by Alessia Marangon from University of Pisa, Italy. The key issue from the above listed bullets is the uncertainty with hydrogen ignition probabilities. Andrei V. Tchouvelev Page 30 of 54 R1_Jan-08

31 Immediate ignition of hydrogen releases leads to different consequences than delayed ignition. Immediate ignition will lead to jet fires for continuous leaks and fireballs for rupture, whereas delayed ignition of a continuous or instantaneous leak leads to a flash fire and / or deflagration. Hence it is essential in the risk study to separately control both the immediate and the delayed ignition probability, which ideally should be in line with historical ignition probability data. Ignition probabilities of conventional fuels are derived from vast historical data. Unfortunately, there is very little data available for hydrogen for the range of operating pressures, sizes of piping and release scenarios that would be common for the environment of a refuelling station. This certainly creates a great uncertainty in determining ignition probabilities for a risk assessment of a commercial (generally plug-and-play) hydrogen system. Up to now there is no consensus on method or approach to this issue. In 2005 DNV conducted a quantitative risk assessment of a hydrogen station in Hong Kong [17]. During the assessment there was a need to establish an approach and method to determine hydrogen ignition probabilities. During the intensive discussions with the customer consultant (Dr. Andrei V. Tchouvelev) the following approach and method were developed. The DNV database provides historical ignition probability data, shown in Table 5. These data show the total historical ignition probability, which varies with the type of material released and the release rate; the reported ratio of immediate to delayed ignition probability historically is 2 to 1. RELEASE RATE CATEGORY Table 5. Historical Ignition Probability Data (Cox, Lees & Ang) RELEASE GAS CRUDE CLASS I CLASS RATE LEAK II (kg/s) CLASS III Small < Large Massive > The above data may be of limited help in this study, as all leaks in this study concern hydrogen rather than hydrocarbons, while the maximum hydrogen leak rate in this study is also much less than 1 kg/s. Following the DNV guide line would imply that all hydrogen leaks would have to be modeled with an overall ignition probability of 1 percent. However, for a given leak rate, hydrogen would form an 8 x larger flammable cloud than methane, as the cloud size this is determined by the flow in mole per second, rather than flow in kg/s. (8 x, as both hydrogen and methane have a similar lower flammable limit). It is obvious that for delayed ignition the ignition probability increases with increased flammable cloud size. Hence an argument may be made to change the critical release rates shown in Table 5 for hydrogen by a factor 8; i.e. 1 kg/s for methane is equivalent to kg/s for hydrogen, etc. Andrei V. Tchouvelev Page 31 of 54 R1_Jan-08

32 The flammable range of hydrogen is 4 to 75 volume percent, which is a factor 7.3 higher than for methane (5.3 to 15 volume percent). One may be inclined to think that this would significantly increase the likelihood of delayed ignition of hydrogen when compared to methane. However, this is contradicted by consequence modeling dispersion results for equal size clouds (i.e. for similar mole/s leak flow rates). For both methane and hydrogen, the size of a cloud above 15 mole percent is approximately 16% of the total size of cloud above LFL. This is due to extremely rapid air entrainment and dilution near the point of release. Hence the larger flammable range of hydrogen does not materially affect the delayed ignition probability. Given the very low minimum ignition energy for hydrogen (0.02 mj, when close to stoicheometric mixture) as compared to methane (0.29 mj), a 1 percent overall ignition probability for hydrogen leaks < kg/s would seem to be too low, as hydrogen may be easily ignited by very weak ignition sources including static, which may be caused by line friction, build-up static on operator clothing, rotating machinery, or accidental uncoupling of a re-fuelling hose. This would justify increasing the hydrogen release ignition probability to be higher than the 1 percent suggested by Table 5. An opposing argument is that the hydrogen ignition probability should be regarded as similar to methane, as at LFL hydrogen would require similar ignition energy as methane. Table 6 presents the research data published by Dr Swain, University of Miami, May 24 th, 2004, which shows how the minimum ignition energy required varies with hydrogen in air concentration. Table 6. Minimum Ignition Energy Required Versus Hydrogen in Air Concentration Hydrogen Concentration Minimum Ignition Energy Required (mj) 29% (stochiometric) % % % % % 1.0 5% 3.0 4% 10.0 Despite significant research, DNV has not been able to locate definitive data on historical hydrogen release ignition probabilities. Hence, based on the above discussion DNV proposed the following as a reasonable, but conservative approach: Andrei V. Tchouvelev Page 32 of 54 R1_Jan-08

33 Reduce the Table 5 shown leak flow ranges by a factor 8 for hydrogen, allowing for differential molecular weight as compared to methane, which directly affects the size of flammable cloud. Increase the Table 5 Gas (= methane) ignition probabilities by 16 percent, allowing for the ratio of the flammable range of hydrogen compared to methane, and allowing that the 15 vol.% to 75 vol.% portion of any hydrogen cloud (due to pressurised releases) constitutes only 16 percent of the total cloud size above LFL. Treat the ignition probability of hydrogen as similar to methane, allowing that for most of the flammable cloud size is near the lower flammable range, where the minimum ignition energy required is similar to methane. This approach leads to the hydrogen release leak size ranges and ignition probabilities shown in Table 7. Table 7. Proposed Hydrogen Ignition Probabilities RELEASE RATE CATEGORY HYDROGEN RELEASE RATE (kg/s) HYDROGEN TOTAL IGNITION PROBABILITY HYDROGEN IMMEDIATE IGNITION PROBABILITY HYDROGEN DELAYED IGNITION PROBABILITY Small Leak < Large Leak Massive Leak > Flammable H2 gas mixture within closed systems Not Applicable 1-1 This approach and method can be recommended as temporary solution until more precise method has been developed and more operating data have been accumulated. An appealing upside of the above method is its clarity and simple (and common sense) logic. Its downside it does not take into account an effect of high pressure releases on increased probability of spontaneous ignition. This is discussed in the chapter 3.4 Gaps in fundamental knowledge Consistent Methodology for Site Risk Assessment Contribution to this subsection was mostly provided by Dr. Olav Nansen, GexCon, Norway. There is a significant focus on regulations, codes, and standards (RCS) e.g. safety distance rules for handling hydrogen risk at the present time. CFD is used only occasionally, and mainly for worst-case evaluations. On the other hand, the validity of simplified methods, like venting guidelines, for hydrogen is unknown. Therefore, a comprehensive procedure for risk assessment of hydrogen applications considering its most important objects is needed and experiences from the oil and gas industry must be used. This includes relevant methodologies to define risk acceptance criteria, scenario and hazard identification including mechanisms of failure, frequency estimation, and consequence assessment and risk estimation as well as prediction of the risk effects for mitigation measures. Various levels of treatment, ranging from analyzing the worst-case scenario to a comprehensive study including ventilation, dispersion and explosion, to evaluate the probability for Andrei V. Tchouvelev Page 33 of 54 R1_Jan-08

34 unacceptable events is considered. This is illustrated in Figure 13 that shows that the necessary precision level will vary with factors such as application, data/information availability and goal of individual risk assessment studies. The green line indicates worst-case screening, the blue line is the worst-case estimates with more accurate tools, and the pink line is the probabilistic approach. Figure 13. Illustration of levels of precision in risk assessment. In QRA, the worst-case calculations are often unrealistic and predict unacceptable potential consequences. There are many advantages of a CFD-based probabilistic QRA including a good description of physics and dependencies, possibility to quantify a realistic risk level (slightly conservative where necessary), and transparent assumptions, which can be modified when more information is available. More simplified methods only use parts of the information available. For a realistic situation they are sometimes not applicable (due to limitations). If applicable, they may often be either non-conservative or far too conservative, and will usually not be able to predict the effect of a change or mitigation. However, they can still be valuable if applied properly as they can provide quick risk estimates. The applicability may become better with future developments (in progress within the HySafe project). Further, even CFD methods need to be simplified in order to achieve results in a reasonable length of time. It should always be made sure that these simplifications are realistic (and generally conservative). The schematic of the probabilistic QRA approach developed by GexCon and used primarily in the oil and gas industry is presented in Figure 14: Andrei V. Tchouvelev Page 34 of 54 R1_Jan-08

35 Figure 14. Schematic overview of GexCon probabilistic QRA-approach (consistent with NORSOK Z- 013 Annex G guidelines) It is possible to use this approach as the basis for developing QRA methodology for hydrogen applications. The building blocks of such an approach can be represented as: 1. Exact representation of geometry 2. Well-validated and efficient CFD model for a. Ventilation b. Dispersion c. Explosion d. Fire 3. Simulation of many different scenarios with probabilistic representation of a. Ventilation conditions b. Leak parameters (location, direction, rate) 4. Time-dependent probabilistic ignition modeling with contributions from: a. Spontaneous ignition b. Intermittent ignition sources c. Constant ignition sources 5. Use of equivalent stoicheometric gas cloud size: a. Limiting the number of scenarios b. Non-homogeneous clouds Smaller stoicheometric clouds 6. Probability of exceedance curves for pressure (and impulse) 7. Acceptance criteria checked against frequency-load curve It is hypothesized that if the method is constantly evaluated and improved, weak assumptions will lead to experiments and further studies, and gradually be improved with more experience and knowledge. The methods should be flexible allowing adjustments according to the requirements and needs of the different areas. Harmonization of the different partners methodologies within IEA Task 19 will be a part of this development. Andrei V. Tchouvelev Page 35 of 54 R1_Jan-08

36 As a subsection of this task, there is a need to address a gaping need in reliable failure frequency data. Due to lack of failure frequency statistics, other approaches like Bayesian analysis to establishing hydrogen leakage frequencies, phenomenological event probabilities and mitigating components failure probabilities need to be seriously considered. This suggestion was made by Jeff LaChance from SNL, Albuquerque. 3.4 Gaps in Fundamental Knowledge Auto Ignition As was pointed out in the presentation by Dr. Stuart Hawksworth from HSL, UK, there are considerable knowledge gaps related to hydrogen ignition. Andrei V. Tchouvelev Page 36 of 54 R1_Jan-08

37 A slide below illustrating the work by Dr. Dryer was provided by Dr. M. Radulesku from University of Ottawa. Andrei V. Tchouvelev Page 37 of 54 R1_Jan-08

38 3.4.2 Protective Barriers The study of effects of protective barriers as means to reduce clearance or safety distances for compressed hydrogen releases is important for the development of installation codes and risk mitigation requirements. So far there is no guidance when and how to use protective barriers and, what s even more critical, no clear understanding of conditions when protective barriers could lead to escalation of a hazardous situation. This issue is being currently addressed in detail by Sandia national Labs in Livermore, USA. Below are the slides from the presentation provided by Dr. Jay Keller. Andrei V. Tchouvelev Page 38 of 54 R1_Jan-08

39 Andrei V. Tchouvelev Page 39 of 54 R1_Jan-08

40 Andrei V. Tchouvelev Page 40 of 54 R1_Jan-08

41 Andrei V. Tchouvelev Page 41 of 54 R1_Jan-08

42 3.4.3 Consequence Modeling Development of comprehensive consequence models is required to support a risk-informed permitting process, including the development of risk-informed codes and standards. The QRA tools should be proactively used for AHJ education and training so that regulators can learn about potential consequences, frequencies/probabilities, and risk of selected failure scenarios related to likely station designs and associated technologies. The tool may also be used to help design engineers in understanding compliance with codes and standards, risks of typical component failures, and for developing adequate mitigation strategies. In summary, the consequence modeling tool should include simple but realistic consequence models for all possible accident types (i.e., jet fires, vapor cloud explosions, flash fires, BLEVEs, pool fires, etc.). Alternatively, it may be possible to limit the modeling to address just the dominant accident types. The following modeling tools flow chart prepared by Pierre Benard from HRI, Canada, will be analyzed for accuracy and completeness within the next 3-year term of Task 19, and an updated chart summarizing appropriate modeling tools will be developed. Andrei V. Tchouvelev Page 42 of 54 R1_Jan-08

43 Andrei V. Tchouvelev Page 43 of 54 R1_Jan-08

44 3.4.4 Wall Jets Contribution to this subsection was provided by Dr. Pierre Benard and Dr. Andrei V. Tchouvelev [18]. The properties of a high pressure jet originating from either a pressure relief valve or a small crack in the piping of a storage vessel depends on the leak location and size, the release conditions such as pressure and temperature and the physical properties of the gas such as velocity of sound, specific heat and molecular mass. High pressure jets will also be influenced by the presence of obstacles in the immediate surroundings, either impinging surfaces or turbulence inducing structures. From hydrogen safety considerations, interest lays in the determination of the extents of the flammable clouds which are very important parameters in the establishment separation distances and sizes of hazardous zones in the hydrogen codes and standards. Birch et al [19] proposed a methodology to evaluate the decay of the mean concentration field along the centreline of a supercritical jet. The distance taken for the mean volume fraction concentration to decay to a given value in such flows is proportional to the diameter of the source and inversely proportional to the square root of the density of the jet fluid. In their analysis they showed that the concentration field behaves as if it were produced by a larger source than the actual nozzle source diameter; this is referred to as the pseudo-source. Later in 1987, Birch et al [20] reformulated their effective diameter definition based on the conservation of both mass and momentum. In a recent study, Houf et al [7] reused the Birch method to determine the concentration decay of unignited hydrogen jets. In their implementation, Houf et al reformulated the effective diameter of the pseudo-source by replacing the velocity at the end of the expansion region by an effective velocity originally suggested by Hess et al [21] for under-expanded gas jets. They also removed the discharge coefficient in the effective diameter definition. Joint study by HRI and AVT obtained CFD simulations results of both free (i.e. unbound by a surface) and so-called wall (i.e. bound by a surface) horizontal and vertical jets using commercial software FLACS Hydrogen and Phoenics. Particular attention was given to the effects of proximity to the surface for horizontal and vertical hydrogen jet releases, which will impact the concentration decay. The results were compared to methane jets numerical simulations and to the predictions of the Birch correlations for the size of the flammable cloud Birch Effective Diameter Approach And The Mean Concentration Decay In the original implementation presented by Birch et al [19], the effective diameter d eff is given by an expression of the form d d C V ρ 2 2 eff = d, (1) V3ρ 3 where d is the jet exit diameter, C d is the discharge coefficient, V 2 and V 3 are the velocities of the gas at the exit of the reservoir and at the end of the expansion region respectively. Similarly ρ 2 and ρ 3 are the density of the gas at the exit and at the end of the expansion region. Birch et al made the hypothesis that ρ 3 =ρ g, where ρ g is the density of the gas at ambient conditions. The velocity V 2 and the densityρ 2 are obtained from isentropic relations. The effective velocity at the end of the expansion region V 3 is given by an expression of the form Andrei V. Tchouvelev Page 44 of 54 R1_Jan-08

45 γt R =, (2) 3 V3 m mol where γ is the specific heat ratio, T 3 is the temperature at the end of the expansion region, R is the universal gas constant and m mol is the molecular weight. The decay of the mean mole fraction η along the axis of a constant vertical jet can be expressed by η Kd eff a =, (3) x + x 0 ρ ρ g where K is the axial decay constant, d eff is the effective diameter of the pseudo-source, x is the position along the centerline of the jet, x 0 is the virtual origin displacement of the jet, ρ a is the density of air. The value of x 0 x 0 is generally neglected since x>>x 0. The distance x 0 can also be approximated to be equal to 10% of d eff. In their 1983 paper, Birch et al used a value of K = 4.9 based on experimental results obtained with natural gas, and a discharge coefficient value of C d = 0.85 for the calculation of the effective diameter. In the implementation by Houf et al, the following expression is used for the effective diameter d eff V ρ V ρ 2 2 d eff = d, (4) 3 3 where it is assumed that ρ 3 =ρ g. The velocity V 2 and the density ρ 2 are calculated using isentropic conditions. The effective velocity V 3 at the end of the expansion region is given by an expression of the form, V V P P = 2 +, (5) ρ2v2 as originally suggested by Hess et al [21]. P 2 and P 3 are the pressure of the gas at the exit and at the end of the expansion region respectively. P 2 is calculated using isentropic relations. P 3 is assumed to be equal to the ambient pressure. Houf et al used the axial decay constant value K = 5.4 for hydrogen in the expression for the decay of the mean mole fraction in conformity with the correlation obtained by Birch et al Dispersion Simulation Results Using Flacs A constant flow rate from an 8.48 mm diameter orifice of a bar storage unit was studied numerically for both hydrogen and methane. The scenarios simulated here are: a) Horizontal and vertical free jets of hydrogen and methane b) Wall jets: In this case horizontal and vertical jets of hydrogen and methane are simulated in the presence of a surface. In all cases, the jets are located m away from the surface. Andrei V. Tchouvelev Page 45 of 54 R1_Jan-08

46 The jet outlet conditions, i.e. the leak rate, temperature, effective leak area, velocity and the turbulence parameters (turbulence intensity and turbulent length scale) for the flow, are calculated using an imbedded jet program in FLACS. FLACS can also calculate the time dependent leak and turbulences parameters data for continuous jet releases in the case of high pressure vessel depressurization. The program is based on isentropic conditions and avoids simulating the supersonic region immediately downstream of the leak source by using a pseudo source approach. The procedure largely follows the Birch method in reference [20]. However, the reference condition in FLACS is modified to account for high velocities and air entrainment, while Birch et al used the stagnation reservoir condition. FLACS also includes the enthalpy equation and thereby makes the assumption of recovered temperature at the equivalent turbulent jet origin unnecessary. The conservation equation for mass, momentum, and enthalpy in addition to conservation equations for concentration, are solved on a structured grid using a finite volume method. The SIMPLE pressure-velocity correction method is used and extended for compressible flows with source terms for the compression work in the enthalpy equation. FLACS uses the k-ε turbulent model and the ideal gas equation of state. For all the scenarios studied, the simulations were run as a function of time until steady-state was achieved, using a constant mass flow rate of about 1 kg/s for hydrogen and 2.7 kg/s for methane. For a horizontal wall hydrogen jet, the maximum horizontal extent obtained was 52.5 meters at 13.7 sec. This maximum was achieved within the transient part of the release, while the steady state value obtained was 44.9 m, which was reached at 23.5 seconds after the beginning of the release. For a free jet, the maximum extent obtained was 35 m and only a very small transient maximum was observed. For a vertical wall hydrogen jet, a maximum transient length of m was obtained at 48.8 seconds which eventually stabilized at 80 m after 85 seconds from the onset of the release. The extent of the free vertical jet was 42.4 m, in close agreement with the Birch/Houf et al result of m. Figures 15 and 16 below show the contours of hydrogen at 4% molar fraction in air at steady state for the horizontal hydrogen free and wall jets. Figure 15. Contour of constant concentration (4% volume) of hydrogen in air at steady state for the horizontal wall jet (side view: longitudinal cut along the X-Z plane at Y=0.1 M) Andrei V. Tchouvelev Page 46 of 54 R1_Jan-08

47 Figure 16. Contour of constant concentration (4% volume) of hydrogen in air at steady state for the horizontal free jet (side view: longitudinal cut along the X-Z plane at Y=0 m) In case of methane, the maximum extent of the flammable cloud at 5% (vol.) is reached at steadystate for all the scenarios studied except for the vertical wall jet. Similarly to hydrogen, the presence of a surface in the proximity of the jet has an effect on the extent of the lower flammability contour. In this case, Flacs results show a larger extent than that obtained for a free jets. Figures 17 and 18 show the contours of methane at 5% molar fraction in air at steady state for the horizontal methane free and wall jets. Figure 17. Contours of constant concentration (5% volume) of methane in air at steady state for the horizontal wall jet (Side view: longitudinal cut along X-Z plane at Y=0 m) Figure 18. Contours of constant concentration (5% volume) of methane in air at steady state for the horizontal free jet (Side view: longitudinal cut along X-Z plane at Y=0 m) Andrei V. Tchouvelev Page 47 of 54 R1_Jan-08

48 Dispersion Simulation Results Using Phoenics CFD modeling of Birch experiments with natural gas Birch s experiments on natural gas jets [19] were simulated with the commercial CFD software Phoenics using the properties of methane gas. Different symmetric domain sizes were used for the simulations. Figure 19 shows the volumetric concentrations obtained by the k-e RNG and LVEL turbulence models for 3.5 bars. The LVEL turbulence model yields the simulation results substantially deviating from the experiment data for high gas concentrations but shows good agreement with the experimental results around LFL concentrations. The k-e RNG model yields the results within 10% difference from the experimental for the whole range of concentrations. Figure 19. Comparison of CFD results with the experimental data from Birch et al [19] for a pressure of 3.5 bars. Simulations were further performed for higher pressures (up to 170 bars) using the RNG k-ε model. The natural gas turbulent diffusivity is assumed to be 0.7 of turbulent viscosity, which is calculated by the kinetic energy and turbulent energy dissipation rate (k-e). Figure 20 shows the simulation data versus the experimental correlations for the pressures from 3.5 to 170 bars. Andrei V. Tchouvelev Page 48 of 54 R1_Jan-08

49 Figure 20. Comparison of CFD results with the experimental mean concentration for high pressure natural gas results showing collapse of the data in terms of z / d P. Pressure range: 3.5 bars to 170 bars. It can be seen that the numerical simulations reproduce the experimental data for various pressures with acceptable errors (within 20% for a wide range of pressures). It is interesting to note that simulations for pressures beyond the Birch range, i.e. >71 bars, produce greater errors. In general, the above trials confirm the validity of Birch correlations based on the experimental results for free vertical jets. CFD Modeling of Hydrogen and Methane Wall Jets As specified above, both hydrogen and methane jet releases from storage tanks were simulated using a leak orifice of 8.48 mm ID and the stagnation pressure in the tank of bars at 1 m above ground. It is estimated that the choked release lasts for 80 seconds for hydrogen and 240 seconds for methane. A symmetric domain of 100m long 8 m wide 25 m high was used to save computational resources and an optimal grid size of was used to achieve accurate results with good convergence. Figures 21 and 22 show the transient hydrogen and methane cloud extents from the leak orifice for 60 and 90 seconds respectively. The buoyancy forces substantially shorten the LFL hydrogen cloud extent along the centreline extents in comparison with the maximum horizontal extent. This effect is not observed for methane. Also, at approximately seconds from the onset of the release, both gases experience a puff resulting in disconnect of a part of the cloud and its abrupt reduction in length. Figures 21 and 22 also show that the maximum extents of flammable clouds for both gases after the puff are close to each other (with the hydrogen one being a bit shorter). This is quite remarkable considering that the ratio of extents of hydrogen and methane free jets is approximately 3.5 to 1. This indicates that there is a significant effect of a surface on methane jet extent while it is quite weak for hydrogen. Andrei V. Tchouvelev Page 49 of 54 R1_Jan-08

50 Figure 21. Hydrogen maximum and centerline cloud extents with time. Figure 22. Methane maximum and centerline cloud extents with time. Comparison of Real and Effective Orifice Modeling (under investigation) A comparison was made between the real and effective orifice modeling approaches for a hydrogen release under appr. 430 bars. The results of this comparison showed that the effective orifice produced about 25-30% longer extents than the real orifice. The main reason for this difference is that the real orifice approach uses real gas hydrogen properties and, thus, takes its real compressibility into account, which deviates substantially from ideal gas law at high pressures. To the contrary, the artificial orifice approach is based on ideal gas law, which leads to higher deference with the real orifice results the higher is the tank pressure. Another contributing factor is the input velocity of sound. The real orifice approach uses the velocity of sound at critical temperature (around 243 K), which equals to 1,189 m/s, while the effective orifice calculations use the velocity of sound at room temperature (around 293 K), which equals to 1,305 m/s. Andrei V. Tchouvelev Page 50 of 54 R1_Jan-08

51 Conclusions The presence of a surface for horizontal and vertical hydrogen and methane jets has a major impact on the flammable cloud extent at steady state. The presence of a surface affected the maximal extent of horizontal hydrogen jets, to a lesser extent (30% extent increase) than horizontal methane jets (125% extent increase) due to buoyancy. For vertical jets, the presence of a wall affected the maximal extent for both gases in practically the same way (113% increase for methane and 126% increase for hydrogen). Unlike hydrogen, the centerline extent of horizontal methane jets is smaller than the full extent because of a downward bending of the flammable envelope, which is a result of the contribution of the smaller buoyancy of methane compared to hydrogen (which reduces the rise of the jet as a function of distance with respect to hydrogen) combined with the reflection of the methane jet on the surface. If real orifice results are taken into account, the deviation from Birch predictions for a free horizontal hydrogen jet becomes even greater. It may be generally concluded that the higher is the pressure the greater would be the deviation. The above findings stress the importance of conducting further investigations of wall jets behaviour under the wide range of pressures and proximity to surfaces. 3.5 Discussion Next steps IEA Task 19 partners in the next 3-year term will strive to close some of the critical gaps within its Subtasks A and B. For the work plan period through September 30, 2010, Subtask A will focus on four principal activities: A1: Develop uniform risk acceptance criteria and establish link with risk-informed codes & standards. Activity leaders: Jeff LaChance, SNL, USA and Angunn Engebo, DNV, Norway A2: Develop a list of appropriate engineering models and modeling tools. Develop simple but realistic physical effects models for all typical accident phenomena (i.e., jet fires, vapor cloud explosions, flash fires, BLEVEs, pool fires, etc.) for education and training, design evaluation and simplified quantitative risk analysis purposes. Activity leaders: Pierre Benard, HRI, Canada and Jay Keller, SNL, USA A3: Develop methodology for consistent site risk assessment based on HyQRA approach. Activity leaders: Olav Hansen, GexCon, Norway, Koos Ham, TNO, Netherlands and Alessia Marangon, UNIPI, Italy A4: Release updates (at least once) to all original Subtask A products: Risk assessment methodology survey, Knowledge gaps white paper and Review and comparison of risk assessment studies. Activity leader: Andrei V. Tchouvelev, AVT, Canada A relationship between the first three Activities and their contribution to quantitative risk assessment and risk-informed RCS (regulation, codes and standards) process is illustrated by the diagram below. Andrei V. Tchouvelev Page 51 of 54 R1_Jan-08

52 The above activities will result in internal products reports, the contents of which will be disseminated as per the policy adopted by the Task 19 participants. It is anticipated that the key results will be published at various hydrogen safety-related events like International Conference on Hydrogen Safety, through appropriate scientific journal publications and other national and international events. Partners will continue sharing information on their testing programs within the Subtask B, particularly on the topics identified herein as knowledge gaps, namely auto ignition, protective barriers and wall jets. Andrei V. Tchouvelev Page 52 of 54 R1_Jan-08