Large volume liquid hydrogen releases: key results and outcome of modelling exercises
04 August 2021
Decarbonising shipping and transportation remains a significant challenge as storing electrical energy from renewables in significant capacity makes it of little use against the requirements of running a ship at sea where space and load requirements inhibit battery options. Liquid hydrogen (LH2), commonly used as fuel in the space industry, forms one potential solution where the required energy density/volume can feasibly be met for ships of all sizes.
Figure 1 – A representation of a hydrogen-electric ferry (Image: Shutterstock)
(Click here to view article in the digital edition)
However, its use in the maritime sector poses significant technical and safety challenges around the scale of operations, number of usage points, and its proximity to personnel. Potentially critical scenarios for LH2 on ships can be described through a variety of phenomena including outflow, dispersion, accumulation, cryogenic exposure, ignition potential, explosion, and fire.
In 2018, the NPRA gave the go ahead for a hydrogen-electric ferry to serve one of the routes along the coast of Norway. One concept for zero-emission ferry transportation involves LH2 as a concentrated form of hydrogen storage (Figure 1).
To identify and quantify safety related issues which may need to be resolved, DNV was commissioned by Norway’s Defence Research Establishment and the Public Roads Administration to conduct a series of experiments investigating the behaviour of large releases of LH2 into the atmosphere and in a closed but ventilated area.
The build and operation of the experimental facilities at DNV’s Research and Development Centre in Cumbria, UK, delivered a programme of research pertaining to the outdoor and confined leakage phenomena.
A total of 15 large releases (up to 50 kg.min-1) from a liquid hydrogen storage tanker were performed with variations in source tanker pressure, flow rate, orientation, ignition, and atmospheric conditions – both directly to atmosphere or within a 27 m3 steel confined space simulating a Tank Connection Space (TCS) complete with ventilation mast.
The outdoor releases were initially illustrated using predictions from DNV’s GasVLE, FROST and Phast software packages to analyse the expected behaviour for:
- Outflow – assessments of the mass flow versus pipeline and outlet conditions are made including assessments of the level of flashing (liquid mass fraction) within the pipe.
- Pooling – measurements of ground surface and subsurface temperature around the releases were made and interpreted for the presence of liquid hydrogen or condensed products of air.
- Dispersion – determination of the extents of the flammable limits near to ground level from each release.
- Thermal properties - thermal radiation measurements recorded around each ignited release are interpreted against simple models and existing correlations and guidance for hydrocarbons.
Bulk LH2 delivery was from a bulk tanker located in a protective mound at the test site, which included a tall camera tower equipped with wind instrumentation. A 30x30m concrete pad was constructed to contain the release arrangement. This incorporated a 40m long vacuum insulated pipe with instruments to analyse temperature and pressure conditions, the release point was located in the centre of the 900m2 concrete pad. Some obstacles in the form of ISO containers and smaller infrastructure were located on the test pad, as shown in Figure 2. The release of LH2 was directed both downwards and horizontally.
Figure 2 – Experimental arrangement for outdoor releases
Measurements in the outdoor releases focused on the temperature of the surface and subsurface of the concrete close to the release while dispersion measurements using oxygen sensors were taken further afield. The oxygen sensors measured the hydrogen gas concentration by examining the depletion of oxygen in the atmosphere. Two of the experiments were ignited after a delay to gather information on thermal radiation and overpressure.
Where releases were conducted into confined spaces, the TCS was located at the release point with the releases directed onto the floor of the chamber as illustrated in Figure 3. The chamber was equipped with ventilation openings, ventilation mast and an explosion relief vent to avoid significant damage to the chamber in each experiment. The chamber itself was instrumented to acquire measurements of temperature, gas accumulation and explosion overpressure in addition to the field measurements of gas dispersion and temperature laid out for the outdoor releases.
All experiments were conducted within a 250m exclusion zone and controlled remotely. This followed the method for purging and cooling the pipe that went from gaseous nitrogen (N2) to liquid nitrogen (LN2) to gaseous helium (He2) to cold gaseous hydrogen (H2) to LH2.
The following video shows the varying views of the outdoor, downward and horizontal releases and an ignited confined release.
Using the legacy British Gas model FROST and the DNV PHAST models, prediction of the vapour quality by mass and outflow were considered, with knowledge of the geometry of the release point orifice and the saturations conditions near to it.
The liquid/vapour fractions along the pipe during releases were calculated based on the pressure decay and assuming isenthalpic expansion. Releases in all experiments produced high liquid mass fractions (i.e. the proportion of the mass flowing in the pipe being in the liquid phase) with experiments being driven above saturation pressure in the tanker producing higher liquid mass fractions (i.e. little or no flashing in the pipe).
One of the key considerations when assessing the risk of releases is how much of the two-phase flow ends up on the floor, forming a pool, perhaps, or raining out from the jet. Taking the measurements of the surface and subsurface temperature of the concrete and in the experiments allowed DNV to conclude there was no evidence for any LH2 beyond 0.5m from the releases in the downward releases conducted onto concrete in the programme and there was zero evidence of any rainout in the horizontal releases.
Having set or determined the source conditions by looking at the outflow, the dispersion and ultimately, the lower flammable limits (LFL) of the dispersing cloud was analysed. The trends in the experimental results were also represented in the model results:
- Reducing concentration with distance
- Reducing concentration with radial departure from the downwind direction of about 90o
- Peak gas concentration versus the peak dip in temperature, giving a nominally linear relationship, worsening as the concentration gets higher
Figure 3 – Experimental arrangement for confined, ventilated releases
- Higher concentrations at higher positions above ground level
- LFL was not exceeded in downward releases beyond 30m from the release point and 50m in the horizontal releases.
The relationship between the peak concentration and the peak temperature drop across several experiments was further compared to the predictions from GasVLE, a fluid properties calculation tool from DNV, the predictions are slightly below that of the actual observations. This may be because the GasVLE model does not include heat transfer from the ground, possibly evident in the measurements.
At the point of ignition there is an explosion effect creating an initial expanding fireball which can give rise to local dynamic overpressures with the potential to cause harm. At all measurement locations, in all ignited experiments, the peak pressure did not exceed 30mBar in the horizontal jet and around about 15mBar in the downwards jet.
Results of the confined, ventilated releases and explosion experiments are still being analysed. In most cases, the releases were so large that the atmosphere within the chamber quickly became very cold and non-flammable (i.e. air was either condensed to liquid or displaced by the high-volume release). It was noticeable that no evidence of any flammable concentrations of hydrogen were found at any of the ground level measurement points down-wind of the ventilation mast.
Two of the confined experiments were ignited at the top of the ventilation stack after isolation of the release resulting in two very different severities of explosion; high overpressure was observed when the ventilation in the chamber was higher, with low overpressure observed when the ventilation levels were low. Considering the large array of variables influencing the severity of explosions in these geometries, particularly with cryogenic atmospheres, more experimentation is required to obtain greater understanding of large scale, cryogenic hydrogen explosions.
The releases conducted as part of this programme provide a valuable, extensive and unique dataset in relation to the phenomena contributing to the severity of events following a loss of containment of liquid hydrogen. The outdoor release data has been compared to predictions from existing software models and found to be of a high quality and suitable for further model development / validation.
Hydrogen is one of the most suitable solutions to contribute to the replacement hydrocarbons in the future and its consumption is expected to grow significantly over the next three decades. Despite being a clean fuel, concern surrounds its chemical and physical properties and the gaps in engineering techniques, tools and implementation experience and the potential impact on safety. Further investigation and model analysis is recommended.
About the authors:
Daniel Allason is a Chartered Physicist with over twelve years of experience in major hazard research at DNV’s Spadeadam Research and Testing facility. He specialises in large-scale major hazard studies designed to enhance industry knowledge and understanding through the conduct of experiments as close as possible to full-scale. He is currently leading experimental research programmes related to safety aspects of hydrogen as an energy carrier.
Clara Huéscar Medina is a chartered senior project engineer at DNV’s Spadeadam Research and Testing facility. She joined DNV in 2014. Clara has carried out a wide range of test programmes related to gas safety, pipeline integrity and oil and gas product testing. Clara holds a Master of Science in Energy and Environment and a PhD in Fire and Explosions Engineering from University of Leeds.
Anne Halford is a mathematician with over 25 years’ experience in the gas industry. She has developed mathematical models for dense gas dispersion, methodology development and performed consultancy work. Implementing and maintaining the models in onshore and offshore risk assessment packages. She has performed consequence calculations and quantified risk assessments for LNG import and export terminals, oil processing sites and gas reception terminals including sites handling sour hydrocarbons.
Jan Stene works with numerical simulations and consequence modelling as a Principal Mathematical Modeller at DNV. His experience ranges from developing methods for complex fully three-dimensional two-phase flows to more than 10 years of developing faster phenomenological consequence models for the world-leading software packages Phast and Safeti used in the Process Industry. He has written numerous academic papers and has strong programming skills. Jan holds a Master of Engineering in Applied Physics and Mathematics from the Norwegian University of Science and Technology (NTNU) and a Doctor of Philosophy in Applied Mathematics from the National University of Singapore (NUS).