hydrogen release from a metal hydride tank

The event occurred at a laboratory, when testing a hydrogen prototype system. Its main components were a compressed hydrogen cylinder, an experimental metal hydride storage system and an air-cooled fuel cell.
The metal hydride storage system was being refilled from the compressed hydrogen source
After 3 min 30 sec from the start of the filling, the pressure relief device installed on the tank activated at 35 bar and hydrogen was released into the lab.

The hydrogen supply from the compressed gas cylinder was cut immediately and the window was opened. The operator was on site at the time of the incident. The amount of hydrogen released was estimate less than 1,000 normal liters (less than 900 g of hydrogen). The overpressure device closed after some minutes, indicating that the pressure had fallen below 35 bar.

The INITIATING CAUSE was the over-pressurisation of the metal hydride storage system and the consequent correct activation of the pressure relief device.

The post-incident investigation revealed the that the pressure reducer installed on the hydrogen cylinder was made for 200 bar input pressure and 0-200 bar output pressure.
With a 0 to 200 bar measuring range, the measure of the pressure on the secondary stage was imprecise at a charging pressure set at 20 bar corresponding to 10% of the full scala. A check with a different, electronic pressure gauge revealed a pressure of roughly 32 bar, very near the maximal operative pressure set by the activation of the pressure relief device.

On top of that, the "Supply Pressure Effect" occurred during the formation of the hydride. This effect induces in the pressure reducer an increase of secondary pressure when the primary pressure reduces. This led to a further increase of the secondary pressure, to the limit of the pressure relief device. Finally, an additional contribution to the increase of the pressure in the storage tank could have been provided by the increased temperature taking place in the first phase of hydride formation.
The ROOT CAUSE could be attributed to inadequate attention to the experimental set-up design.

pressure relief valve,
pressure reducer,
metal hydride tank

DESCRIPTION OF THE METAL HYDROGEN STORAGE SYSTEM
The tank was able to store 400 g of hydrogen in metal hydride at a pressure less than 15 bar.
For refuelling, the secondary pressure on the compressed hydrogen supply container was set to 20 bar and the absorption of the hydride was started. The hydrogen flow was not regulated. In the first filling phase, a sharp increase in tank temperature occurred, during the onset of the exothermic metal hydride formation. In a second phase, this temperature increase lead to a higher formation pressure, which automatically reduced the hydrogen uptake and the energy released by the chemical reaction. This provided a self-regulating temperature control under steady state pressure-temperature hydride formation. Therefore, the tank did was not equipped with an active cooling system.
The system had a pressure relief valve able to open gradually at 35 bar to guarantee the tank integrity and avoid overpressure.
The hydrogen was supplied to the tank by a compressed hydrogen cylinder with 200 bar nominal pressure. A pressure reducer was used to reduce the pressure at the inlet of the tank.

The operator was on site at the time of the incident. The hydrogen supply from the compressed gas cylinder was isolaed and the window opened.

This is event is a minor incident, and the safety system in place (the pressure relief valve) worked as designed as mitigating measure and avoided further consequences. It opened when the internal tank pressure exceeded its set value, to protect the tank from failure due to overpressure, and re-closed when the internal pressure fell below that value, minimising the amount of hydrogen released.

Nevertheless, the event provided the following reflections.
(1) the value of 35 bar chosen for overpressure protection was very low. H2TOOLS reports that, “based on the tank design, at least 60 bar would be acceptable”. A too high safety margin on one single parameter is not always the best recipe for ensuring safety. In this case, it is possible that by choosing a higher pressure value for the activation of the relief valve would have avoided the hazard of a hydrogen release in the laboratory. However, metal hydride tanks are complex system, due to the chemical reaction taking place during hydrogenation and the temperature-pressure evolution associated with the reaction. The H2TOOLS report does not indicate the type of metal hydride and the temperature range characterising the reaction(s). A correct safety design must consider the phase diagram of the metal-hydrogen system and the evolution to pressure-temperature inside the metal and in the gas phase of the tank, together with the integrity of the tank over the whole possible pressure-temperature range.
(2) The relief valve discharged hydrogen into the laboratory. Despite the small hydrogen mass released (less than 800 g), under certain condition it could have ignited and created an overpressure in the confined space of the laboratory. Any safety device designed to release hydrogen shall have a discharge point outside the laboratory.
(3) Gauges used for measurement of parameters which are critical to safety must be fit-for-purpose, maintained and properly calibrated (or verified, if calibration not possible). A pressure gauge with a scala up to 200 bar would not be indicated for the accurate measure and regulation of a pressure value around 20 bar. Especially mechanical pressure gauges are inaccurate if used in a narrow portion of the full scale, in this case only 10% of full scale. Moreover, pressure reducers tend to be used ‘as received’ and for a long period of time without inspection and checks.

To properly measure and set the hydrogen pressure from the hydrogen supply, the inaccurate mechanical gauge was replaced by a digital pressure transducer.
Moreover, the tank charging pressure was limited to 15 bar to slow down the hydride formation and avoid excessive temperature-pressure transients.