Page 1 SIMULATIONS OF HYDROGEN RELEASES FROM STORAGE TANKS: DISPERSION AND CONSEQUENCES OF IGNITION By Benjamin Angers 1, Ahmed Hourri 1 and Pierre Bénard 1 Jérôme Perrin 2 and Pascal Tessier 3 1 Hydrogen Research Institute, UQTR 2 Air Liquide 3 Membrane LP, Membrane Separation Systems Air Liquide

Page 2 Introduction Overall objectives of the project: –To estimate the risks associated with the releases from hydrogen high pressure storage systems –Examine the consequences of ignition for late ignition –Validate standard simulation techniques relevant to this work Specific objectives of this work: –Validate dispersion simulation models used for this approach –Validate and adapt AutoReaGas (ARG) overpressure calculations for hydrogen jet releases

Page 3 The dispersion simulations were performed using FLUENT The ignitions and subsequent deflagration simulations were performed using a customized version of AutoReaGas (Century Dynamics & TNO) taking into account the initial velocity distribution of the gas mixture. –ARG does not simulate dispersion from an outlet CFD tools

Page 4 Methodology dispersion simulations The hydrogen concentration profile was obtained from CFD dispersion simulations performed by solving the steady-state Navier- Stokes equations in the presence of turbulence using FLUENT Turbulence was modelled using the RNG k-  model with standard parameters The choice of the modelling assumptions were validated by comparing with hydrogen horizontal jet experiments by M. Swain et al.

Page 5 Validation: dispersion simulations Experimental data from M. Swain RNG k-ε Simulation (this work) After 45 seconds

Page 6 Validation – Dispersion model (RNG k-ε) Sensor location Experimental concentration (% (vol)) Simulation results (45 sec)

Page 7 Explosion simulation The incident overpressure is calculated from the gas explosion solver AutoReaGas. The steady-state velocity and concentration profile obtained from the dispersion solver is imported for use in ARG. Turbulence is modelled in AutoReaGas using the standard k-epsilon approach. The laminar combustion process is based on a one step irreversible reaction. A turbulent reaction rate calculated from the Bray turbulent flame velocity is used for the turbulent combustion. Transition of laminar to turbulent flame occurs when the turbulent burning velocity exceeds the laminar burning velocity.

Page 8 Methodology The concentration and velocity profiles obtained from Fluent are averaged over the coarser grid used in ARG and imported into the latter. The simulations were performed using the default values for hydrogen in the solver (3.5 m/sec) and using the laminar burning velocity of hydrogen as an adjustable parameter. –ARG is calibrated for methane.

Page 9 Validation – Overpressure from hydrogen jet explosions Vertical hydrogen jets were compared to data obtained by H. Seifert and H. Giesbrecht from BASF. –Maximum overpressure observed: 80 Pascals at 2 meters from the point of ignition. The pressure waves immediately following the ignition of hydrogen were studied for propane and methane jets resulting from subsonic outflows: 140, 190 and 250 m/sec (10 mm diameter outlet).

Page m/s 190 m/s 250 m/s Validation studies (Seifert et al): ignition of a vertical jet Vertical jet: –outlet: 10 mm diameter, 1 m above the ground –Outflow velocities: 140, 190 and 250 m/sec

Page 11 Validation studies : ignition of a vertical jet Flow velocity (m/s)Extension of the hydrogen cloud along the x axis (m) perpendicular to the jet at concentrations of Extension of the hydrogen cloud along the y axis (m) perpendicular to the jet at concentrations of Extension of the hydrogen cloud along the z axis (m) parallel to the jet at concentrations of 2%4%15%2%4%15%2%4%15% Extent of flammable cloud

Page 12 Flow velocity (m/s) Laminar burning velocity (m/s) Overpressure at 2 m (pascal) Overpressure at 5 m (pascal) Overpressure at 10 m (pascal) Flow velocity (m/s) Overpressure at 2 m (pascal) Overpressure at 5 m (pascal) Overpressure at 10 m (pascal) ± 2219± 812 ± ± 2225 ± 716 ± ± 2535 ± 1022 ± 6 Validation : incident overpressure Experimental results Seifert et al Simulation results

Page 13 Volume of hydrogen as a function of flow velocity and fluid mechanics solver The grids of the fluid mechanics solver and the explosion module Autoreagas being different, an averaging program was used to map the data from the solver onto the Autoreagas (ARG) grid. Due to the averaging process performed to import the Fluent Data into the ARG mesh, there is a problem with conservation of mass during data transfert. Flow Velocity (m/s)Volume of H 2 in AutoReaGas (m 3 )Volume of H 2 in Fluent (m 3 )

Page 14 Horizontal jet, dispersion simulation: –Outlet: 6 mm diameter, 0.5 m above ground –Outflow velocities: 140, 190 and 250 m/sec –Ignition after sec for a 350 bars reservoir –Size of the CFD domain: 8 by 8 by 8 meters –Unstructured mesh: 279,026 cells Horizontal jets

Page 15 Horizontal jets : dispersion Flow velocity (m/s)Extension along the x axis (m) perpendicular to the axis of the jet Extension along the y axis (m) perpendicular to the axis of the jet Extension along the z axis (m) parallel to axis of the the jet 2%4%15%2%4%15%2%4%15% % 4%15%

Page 16 Overpressure generated by the ignition of the flammable cloud from a 250 m/sec outflow from a 6mm PRD device of a cylinder as a function of distance at seconds after ignition. Horizontal jets : overpressure

Page 17 Flow velocity (m/s) Laminar burning velocity (m/s) Overpressure at 2 m (pascal) Overpressure at 5 m (pascal) Overpressure at 10 m (pascal) *26.5*9.0* *44.62*14.4* *52.0*16.3* *Reflected peaks were larger Horizontal jets : overpressure

Page 18 Discussion For vertical jets, the use of the standard values in Autoreagas (burning velocity=3.5 m/sec) leads to larger overpressure peaks than experimentally observed by a factor of 4 to 8. Good agreement with the available experimental data from jets could be obtained by adjusting the burning velocity to about 1.5 m/sec. In the case of a 700 bars reservoir, this would correspond to an ignition delay of the order of a minute after release. Initial velocity profile of the release has an effect on the calculated overpressure.

Page 19 Conclusions Case considered represent essentially late ignition of a release from high pressure reservoir Issues: –Lack of experimental data for large, chocked,outflows expected when ignition occurs immediatly after release –Simulation issues Mesh: Importing dispersion data from Fluent to AutoReaGas (mass of hydrogen must be conserved when Fluent data is imported into ARG) Unavailability of outflow boundary condition in explosion solver may be a problem for early ignition of the release when large, sonic leaks occur. (Cannot predict steady state jet fire)

Page 20 Further work Obstacles The next stage of this project is to estimate the size and time dependant concentration profile from the release through a PRD of a fully filled hydrogen tank for late stage ignition, as well as the overpressure resulting from its ignition Investigate other explosion solvers. Effects of obstacles … Tackle early ignition

Page 21 References 1.Seifert H. and H. Giesbrecht, Safer design of inflammable gas vents, Loss prevention and safety promotion in the process industries: 5th International symposium of the European Federation of Chemical Engineering, Swain, M., Codes and standards analysis, 2004 annual program review meeting of the hydrogen, fuel cells & infrastructure program of the US Department of Energy, 2004.

Page 22 Acknowledgements Air Liquide

Page 23 Overpressure from a 250 m/sec outflow from a 10mm outlet from a vertical pipe as a function of distance.

Page 24 Overpressure as a function of cubic cell size for a 250m/s leak, using a laminar burning velocity of 1.35 m/s Cubic cell size (m) Overpressure at 2 m (pascal) Overpressure at 5 m (pascal) Overpressure at 10 m (pascal) Volume of H 2 in AutoReaGas (m 3 ) Volume of hydrogen in Fluent: 1.31 m 3