Safety

This work is part of the MULTHYFUEL E.U. research program [1] aiming at enabling the implementation of hydrogen dispersers in refuelling stations. One important challenge is the severity of accidents due to a leakage of hydrogen from a dispenser in the forecourt. The work presented in this paper deals with the quantification of the leakage scenarios in terms of frequencies and severities. The risk analysis exercise although performed by experts showed very large discrepancies between the frequencies of leakages of the same categories and even between the consequences. A large part of the disagreement comes from the failure databases chosen as shown in the paper. The mismatch between the components on which the databases have been settled and the actual hydrogen components may be responsible for this situation. However as it stands limited confidence can be laid on the outcome of the risk analysis.<br/>A new method is being developed to calculate the frequencies of the leakage and the flowrate based on an accurate description of each component and of each hazardous situation. For instance the possibility for a fitting to become untight due to pressure cycling is modelled based on the contact mechanics. Human errors can also be introduced by describing the tasks. In addition of the description of the method the application to a disperser is proposed with some comparison to experiments. One of the outcomes is that leakage cross sections can be much larger than expected.

Hydrogen Fuel Cell Electric Vehicles (HFC EVs) represent an alternative to replace current internal combustion engine vehicles. The use of these vehicles with storage of compressed gaseous hydrogen (CGH2) in confined spaces such as tunnels underground car parks etc. creates new challenges to ensure the protection of people and property and to keep the risk at an acceptable level. The HYTUNNEL-CS project sponsored by the FCH-JU was launched to develop validated hazard and risk assessment tools for the behavior of hydrogen leaks in tunnels. Among the experiments carried out in support of the validation tools the CEA has conducted tests on gas dispersion in a full-scale tunnel geometry. In the tests carried out hydrogen is replaced by helium under a pressure of 70 MPa in a 78 liter tank. The car is simulated by a flat plate called chassis and the discharges are made either downwards under the chassis or upwards to take into account a rollover of the car during the accident. Different thermally activated pressure relief device (TPRD) diameters are examined as well as different orientations of the discharge. Finally the mixing transient of helium with air is measured for distances between -50 and +50m from the release. Performing CFD simulations of such an under-expanded jet in an environment as large as a road tunnel demands a compressible flow solver and so a large computational cost. To optimize this cost a notional nozzle approach is generally used to replace the under-expanded jet by a subsonic jet that has the same concentration dilution behavior. The physics at the injection point is then not resolved and a model of these boundary conditions has to be implemented. This article first reviews the main experimental results. Then a model of boundary conditions is proposed to have a subsonic hydrogen jet that matches the dilution characteristics of an under-expanded jet. Furthermore this model is implemented in the TRUST LES computer code and in the Neptune-CFD RANS computer code in order to simulate some helium dispersion experiments. Finally results from the CFD simulations are compared to the experimental results and the effect of the exact shape of the tunnel is also assessed by comparing simulations with idealized flat walls and real scanned walls.

Natural ventilation is a well-known passive mitigation method to limit hydrogen build-up in confined spaces in case of accidental release [1-3]. In most cases a basic design of H2 infrastructure is adopted and vents installed for natural ventilation are adjusted according to safety targets and constraints of the considered structure. With the growing H2 mobility market the demand for H2 refueling infrastructure in our urban environment is on the rise. In order to meet both safety requirements and societal acceptance the design of such infrastructure is becoming more important. In this study a novel design concept is proposed for the hydrogen refueling station (HRS) by modifying physical structure while keeping safety consideration as the top priority of the concept. In this collaborative project between Air Liquide and the University of Delaware an extensive evaluation was performed on new structures of the processing container and dispenser of HRS by integrating safety protocols via passive means. Through a SWOT analysis combined with the most relevant approaches including analytical engineering models numerical simulations [4] and dedicated experimental trials an optimized design was obtained and its safety enhancement was fully evaluated. A small-scale processing container and an almost full-scale dispenser were built and tested to validate the design concepts by simulating accidental H2 release scenarios and assessing the associated consequences in terms of accumulation and potential flammable volumes formation. A conical dispenser and a V-shaped roof-top processing container which were easy to build and implement were designed and tested for this proof-of-concept study. This unique methodology from conception fundamental analysis investigation and validation through experimental design execution and evaluation is fully described in this study.

Historically it has been a challenge to simulate the experimentally observed cellular structures and marginal behavior of multidimensional hydrogen-oxygen detonations in the presence of losses even with detailed chemistry models. Very recently a quasi-two-dimensional inviscid approach was pursued where losses due to viscous boundary layers were modeled by the inclusion of an equivalent mass divergence in the lateral direction using Fay’s source term formulation with Mirels’ compressible boundary layer solutions. The same approach was used for this study along with the inclusion of thermally perfect detailed chemistry in order to capture the correct ignition sensitivity of the gas to dynamic changes in the thermodynamic state behind the detonation front. In addition the strength of transverse waves and their impact on the detonation front was investigated. Here the detailed San Diego mechanism was applied and it has been found that the detonation cell sizes can be accurately predicted without the need to prescribe specific parameters for the combustion model. For marginal cases where the detonation waves approach their failure limit quasi-stable mode behavior was observed where the number of transverse waves monotonically decreased to a single strong wave over a long enough distance. The strong transverse waves were also found to be slightly weaker than the detonation front indicating that they are not overdriven in agreement with recent studies.

A growing share of renewable energy production in the energy supply systems is key to reaching the European political goal of zero CO2 emission in 2050 highlighted in the green deal. Linked to the irregular production of solar and wind energies which have the highest potential for development in Europe massive energy storage solutions are needed as energy buffers. The European project HyPSTER [1] (Hydrogen Pilot STorage for large Ecosystem Replication) granted by the Clean Hydrogen Partnership addresses this topic by demonstrating a cyclic test in an experimental salt cavern filled with hydrogen up to 3 tons using hydrogen that is produced onsite by a 1 MW electrolyser. One specific objective of the project is the assessment of the risks and environmental impacts of cyclic hydrogen storage in salt caverns and providing guidelines for safety regulations and standards. This paper highlights the first outcome of the task WP5.5 of the HyPSTER project addressing the regulatory and normative frameworks for the safety of hydrogen storage in salt caverns from some selected European Countries which is dedicated to defining recommendations for promoting the safe development of this industry within Europe.

Social risk is a comprehensive concept that considers not only internal/external physical risks but also risks (which are multiple varied and diverse) associated with social activity. It should be considered from diverse perspectives and requires a comprehensive evaluation framework that takes into account the synergistic impact of each element on others rather than evaluating each risk individually. Social risk assessment is an approach that is not limited to internal system risk from an engineering perspective but also considers the stakeholders development stage and societal readiness and resilience to change. This study aimed to introduce a social risk approach to assess the public safety of large-scale hydrogen systems. Guidelines for comprehensive social risk assessment were developed to conduct appropriate risk assessments for advanced science and technology activities with high uncertainties to predict major impacts on society before an accident occurs and to take measures to mitigate the damage and to ensure good governance are in place to facilitate emergency response and recovery in addition to preventive measures. In a case study this approach was applied to a hydrogen refueling station in Japan and risk-based multidisciplinary approaches were introduced. These approaches can be an effective supporting tool for social implementation with respect to large-scale hydrogen systems such as liquefied hydrogen storage tanks. The guidelines for social risk assessment of large-scale hydrogen systems are under the International Energy Agency Technology Collaboration Program Hydrogen Safety Task 43. This study presents potential case studies of social risk assessment for large-scale hydrogen systems for future.

Hydrogen is commonly used as feedstock in industrial processes and is regarded as a potential future energy carrier. However its reactivity and low density make it difficult to handle and store safely. Indoor hydrogen dispersion can cause a fire or explosion hazard if encountering an ignition source. Safety practices often use time expensive modelling techniques to estimate risk associated with hydrogen. A neural network based surrogate model could efficiently replace Computational Fluid Dynamics (CFD) modelling in safety studies. To lower the dimensionality of this surrogate model a dimensional analysis based on Buckingham’s Pi-theorem is proposed. The dimensional analysis examines stratified filling and highlights the functional parameters involved in the process. Stratified filling occurs for buoyancy dominated releases and is characterized by layers of decreasing concentration starting at the ceiling of the enclosure and developing towards the bottom. The study involves four dimensional cases that were simulated using Computational Fluid Dynamics (CFD) to demonstrate the usefulness of the proposed dimensionless time and dimensionless volume. The setup considered in this paper consists of a parallelepiped enclosure with standard atmospheric conditions a single release source and one pressure outlet to ensure constant pressure during the release. The results of the CFD simulations show a distinct pattern in the relation of hydrogen molar fraction and dimensionless time. The pattern depends on the dimensionless height of the measurement location. A five-parameter logistic (5PL) function is proposed to fit the data from the CFD models. Overall the paper provides insights into the functional parameters involved in the evolution of hydrogen mass fractions during stratified filling. It provides a nondimensional surrogate model to compute the evolution of the local concentrations of hydrogen during the development of stratification layers.

As part of executing 25 hydrogen-based Power to X (PtX) projects our team of Safety consultants has completed safety and risk assessments for a number of hydrogen production developments. Drawing on this experience we will present the importance of making comparisons between hydrogen specific data sources such as HyRAM and conventional oil and gas data sets and calculation methods to ensure that project design is carried out to the most appropriate data and provides a robust solution to demonstrate risks are managed. This presentation will be based on case studies where Fire and Explosion Risk Assessments (FERA) and Quantitative Risk Assessments (QRA) were conducted. The frequency calculations for these assessments used the release frequencies and ignition probabilities provided in HyRAM. However it is noted that the HyRAM ignition probabilities are derived from a correlation from oil and gas assessments in the 1990s. The oil and gas approach has moved on from this data source and now derives ignition probabilities based on the type of facility and fluid characteristics. To address this evolution a comparison was made between the leak frequencies for equipment in hydrogen service and established oil and gas release frequencies from IOGP. In addition a comparison between the HyRAM recommended ignition probabilities and the correlations used for oil and gas (from OEUK formerly UKOOA) was conducted. By taking this approach it was confirmed that the UKOOA data was more conservative and sensitivity calculations were carried out. It was also noted that as hydrogen technologies are emerging there is a level of uncertainty around the data and comparisons must be regularly made to ensure the most appropriate basis for calculations is used.

A series of unconfined hydrogen detonation bench-mark experiments are analyzed with respect to CFD code validation and safety measures development. 1-Dimensional in-house code COM1D was applied for validation against experimental data for unconfined detonation of a hemispherical envelope of about 3- and 5-m radius with hydrogen-air mixtures from 20 to 30% hydrogen in air. The code demonstrates a very good agreement with experimental data and allows an adequate simulation of the unconfined hydrogen detonation. All calculated data were scaled in Sachs coordinates to compare with experimental data and to approximate the data for practical evaluation of safety distances. Numerical experiments with different hydrogen inventories from 50 g to 50 kg and different sizes of the cloud from 1 to 2 m radius of the same amount of hydrogen 50g were carried out to clarify the problem of energy of gaseous explosion responsible for the strength of blast wave. Additionally a comparison of hydrogen-air explosion pressure with blast wave properties from the hypothetical cloud of hot compressed combustion products (P=Picc; T=Ticc) and simply a hot air of the same initial pressure and temperature as combustion products showed very good agreement of shock wave strength at far distances beyond the cloud. This confirms the governing role of energy of combustion on blast wave propagation and its ability to scale the strength of blast waves. The dynamics of the explosion process and combustion product expansion were also analyzed experimentally and numerically to evaluate the dimension of the heat radiation zone and heat flux from combustion products. To demonstrate the capability of tested COM1D code the modeling and analysis of high-pressure hydrogen tanks rupture at 350 and 700 bar were conducted to investigate blast wave strength and evaluate the safety distances.

Hydrogen flames are much thinner than hydrocarbon flames. They have a higher propensity to wrinkle and are subject to thermo-diffusive instabilities in lean conditions. The large scale experiments of Sherman under partially vented conditions have shown that the transition to detonation is possible with only modest flame acceleration to approximately 200 m/s which is much lower than the commonly accepted limits corresponding to choked flames. At present the reason for this transition is not known. Vented H2-air explosions have also demonstrated the role played by expansion/flame interactions in deforming the flame. The state of the art on flame burning rate enhancement by expansion waves will be provided along with the recent experimental and numerical results of head on interaction of flames with an expansion wave conducted in our group. We show that the expansion wave interaction can generate local burning rate increases by more than an order of magnitude. The role of thermo-diffusive instability is also assessed. The mechanism of flame deformation is via the vorticity generation by the misaligned pressure gradient controlled by the expansion wave and the density gradient of the flame. Expansion waves originating from the unburned gas severely elongate the cells until the flame folds burn out. Expansion waves originating from the burned gas side first invert the flames then elongate them by the same mechanism. The rate of elongation is controlled by the volumetric expansion of the gas and the curvature-enhanced growth.