A6. Scenario A: High pressure releases

Overview
This generic scenario relates to failure of high-pressure hydrogen storage associated with a hydrogen or fuel cell system. High-pressure in this context refers to the maximum pressure associated with any system, which will most often be associated with the fuel storage and could be 45 bar, e.g. system generating hydrogen with on site compression, 200 bar, i.e. system using hydrogen cylinders, or potentially even higher, as in automotive applications where pressures of 350 and 700 bar arise. Phenomena related to high-pressure releases have been covered in this scenario, specifically jet fires, un-ignited jets and the delayed ignition of the flammable cloud formed by a release. The modelling work carried out within this scenario was intended to aid the understanding and hence evaluation of the risks associated with high-pressure releases.

Partners CEA, KI, NCSRD and UU contributed to the modelling in this section and the work carried out in within this scenario was closely linked to the experimental work performed by HSL and INERIS within WP5 and described in Appendix 5 and Section 3.2 of [1].  

 

A 6.4.2 Summary and conclusions of WP4 activities in scenario A
The high-pressure release of hydrogen is a potential scenario originating from the storage tank of a stationary fuel cell installation, and methods for assessing the consequences of un-ignited or ignited high-pressure releases are essential for risk evaluation of design and exploitation, and risk management of fuel cell installations.

Hydrogen safety engineering for high pressure hydrogen storage and its surroundings requires knowledge on the hazardous characteristics of un-ignited jets (size and shape of a flammable envelope or ‘detectable’ envelope of hydrogen within a predetermined level of concentration e.g. 50%, 25% or 10% of the lower flammability limit), as well as ignited jets (thermal effects from jet fires and pressure effects from explosions of delayed ignition of a forming flammable cloud).

Hydrogen safety engineers should be equipped with straightforward correlations, graphs and/or nomograms where it is possible. For example, as a result of this research project an engineering nomogram was developed, which requires input of only two parameters – storage tank pressure and physical leak size. Unfortunately, simple engineering correlations are not always available, especially for complex geometries, and more sophisticated predictive tools such as CFD should be used for hydrogen safety engineering. However, such contemporary tools should be verified against widely accepted theories and widely validated against experimental data.

Within scenario A the following activities were undertaken by partners to facilitate risk evaluation through the application of existing and development of new modelling capabilities

• Numerical simulation of an under-expanded jet structure and its comparison with experimental data from the University of Alabama on a micro-scale air jet proved to behave as “normal-scale” under-expanded jets (KI); KI’s validation exercise was an initial step before modeling of un-ignited impinging jets.
• Comparative analysis of existent notional nozzle approaches (CEA), use of existent (CEA and NCSRD) and original (UU) notional nozzle theories to model effective jet source for numerical simulation of hydrogen dispersion from highly under-expanded jets;
• Verification (CEA, against Chen and Rodi correlations) and validation (CEA against Takeno et al experiment [6] NCSRD and UU, against HSL test 7 [6]) of high pressure hydrogen release and distribution models and simulations (both RANS and LES);
• Development of a methodology for numerical simulation of blow-down phenomenon with a time-dependent mass flow rate and effective diameter (UU). This approach was applied for pre-test simulations of HSL experiments. Further research on blow-down phenomenon is needed taking into account findings of Sandia on the role of heat transfer;
• Validation of the LES high-pressure jet fire model and simulations (UU) against the largest known hydrogen jet fire experiment by Sandia National Laboratories. UU performed a validation study to assess the sensitivity of their modeling approach to changes in boundary specification (notional nozzle) and grid resolution. Experimental data published by Sandia was used as a validation case and predictions for flame length and shape were found to be within 10% of the experimental values. The length was slightly over-predicted which may be attributed to the fact that heat loss due to radiation was not accounted for at this stage, nonetheless this means that the results are slightly conservative which is acceptable from a safety point of view;
• Comparison of simulated pressure effects from delayed ignition of high-pressure releases with available experimental data of Takeno et al [6] (CEA). CEA used the results of dispersion simulations of the Takeno et al experiment to model delayed ignition.
• Joint modelling activities in conjunction with experimental data relevant to scenario A moved forward our understanding of high pressure releases, increased confidence in CFD as a contemporary tool for safe design of hydrogen and fuel cell installations and the environment, and contributed to the development of the principles of hydrogen safety engineering, in particular to the evaluation of safety distances.

The following conclusions can be formulated concerning high pressure releases based on a literature review, experimental data available at the time this report was prepared, and modelling and numerical studies performed by four partners (CEA, KI, NCSRD, and UU) in the framework of this project.

Un-ignited jets

• The structure of micro-scale under-expanded jets is similar to the structure of ‘normal’ size jets.
• Modelling of the development of an un-ignited jet can now be performed either by analytical model or by a CFD methodology, combining a notional nozzle model to describe the under-expanded jet close to the release and a weakly compressible Navier-Stokes model to describe the rest of the domain. In the case of complex geometry and interaction with the ground, the CFD methodology is currently preferable.
• Different notional nozzle concepts, used in this study as a boundary condition with different effective diameters of hydrogen jet and velocity, give reasonable results. However, it is expected that models based on Harstad and Bellan theory [7] could give higher hydrogen concentration at the same location on jet axis due to higher value of effective diameter as follows from Chen and Rodi correlation. Further research should be done including comparison with wider range of experimental data, and CFD of the under-expanded jet structure.
• Use of notional nozzle theories not accounting for non-ideal behaviour of hydrogen at pressures above 100 bar should be avoided. The ideal gas law was found to overestimate the total mass of hydrogen release to the atmosphere by up to 50% at pressures characteristic for hydrogen storages.
• Both RANS and LES models reproduce well-averaged hydrogen concentration during dispersion from high-pressure jets in unobstructed environment. However, the LES model demonstrates ability to reproduce non-uniform instantaneous concentration field that can be essential for realistic simulations of subsequent combustion.
• The engineering nomogram for straightforward calculation of axial distance to the edge of 1%, 2% and 4% of hydrogen envelope is developed for high momentum jets. The nomogram can also be used to estimate flame length – the nomogram is given in the ‘jet fires’ section of the conclusions.   
• During blow-down of high-pressure hydrogen tank, the farthest propagating envelope of particular concentration results from the initial period of the release, in the same time, the volume of potentially dangerous hydrogen–air mixture (e.g. hydrogen concentration 20% by volume) is largest during the initial moment of the release and quickly decreases as hydrogen storage pressure drops.
• The presence of walls or ground along the high momentum jet in general increases the length of flammable envelope or safety distance. Qualitatively the same behaviour was also observed in simulations of short blow-down releases with release duration of the order of 10 seconds, but the increase in propagation distance was found larger for larger hydrogen concentrations.

Jet fires

• LES simulations of jet fires should be performed on grids with sufficient resolution of areas of interest, in particular effective jet source when notional nozzle concept is applied.
• Following validation of the modelling approach, a number of simulations were performed for a range of equivalent diameters calculated by the UU notional nozzle theory. This data has been combined with available experimental data, including the INERIS jet fire experimental programme and HSL data, to produce an engineering nomogram for evaluating flame lengths for high momentum jet fires with subsequent calculation of thermal effect by techniques published by Sandia. This nomogram is presented in Figure A.1.

Figure A.1 Nomogram for calculation of flame length of high momentum jet fire by a physical size of leak and pressure in a storage

 

Pressure effects following delayed ignition of high pressure release

• The numerical study performed by CEA confirmed experimental observations of Takeno et al that the earlier the ignition time following leakage, the larger the maximum over-pressure tended to be. The suggested explanation given by Takeno is that the flammable envelope with high turbulent burning velocity formed during initial mixing has greater effect on explosiveness than the total amount of released hydrogen does.
• Due to planning constraints on the project, the latest data from HSL could not be fully analysed and modelled by CEA. The lower initial pressure and inventory chosen leads to a lower cloud extension and overpressures compared to those obtained by Takeno, but the results show the same dependence of the overpressure on the ignition time. A systematic analysis and modelling of these results should be performed after completion of the HYPER project.