14. Results and discussion

Despite the limitation related to comprehensiveness of data, the present report relied on the HIAD and H2tools database to prepare this Part of the report. This is because, amongst publicly available data, this is the most reliable. The authors also acknowledge that for such incident databases, relatively more data exists in the early phases of the database creation as more information is sought at that stage. For both HIAD 2.0 and H2tools, a steady increase in the number of recorded accidents from 1970 to 2009 is observed, followed by a decrease in the next decade Figure 14.1. Nonetheless, one should bear in mind that there is also a steady increase in global hydrogen demand Figure 14.2 since 1975 (IEA, 2019[1]). To address both points, the normalised risk, defined by the number of accidents per million ton (Mt) of hydrogen demand, was calculated (orange curve, Figure 14.1) and plotted.

The lack of sufficient safety measures is likely to be the cause for the initial increase in normalised risk between 1970-2009, during which global hydrogen demand increased from 18.2 Mt to 62.4 Mt (IEA, 2019[1]) Figure 14.3. The later decrease in normalised risk indicates the impact of regulation implementation, codes and standards regarding the H2 industry, as well as increased learning regarding hydrogen safety from past accidents.

The overall trend in the number of hydrogen accidents is also reflected by the number of fatalities and injuries caused by these accidents, with these numbers reaching a maximum in the 2000’s. There is a noticeable decrease in the number of both fatalities and injuries after that, implying that, despite the increased global hydrogen demand and the new hydrogen applications and technologies, the number of serious hydrogen accidents is lower Figure 14.3. Part of the effect might also stem from a decrease in participation to the database, after the initial effort at their creation in the late 2000s would have subsided, although incidents resulting in personal injury and death, being notable, might be the less affected by such a trend. Normalising the number of fatalities and injuries caused by hydrogen accidents against the global hydrogen demand showed that the normalised annual fatality rate decreased from 0.064 fatalities per Mt hydrogen per year in the decade from 2000 to 2009, to 0.013 fatalities per Mt hydrogen per year in the following decade. Similarly, the normalised injury rate from hydrogen incidents also decreased, from 0.238 in the 2000s to 0.033. This observation (or trend) can again be attributed to increased learning about hydrogen safety, insights from lessons learned from past H2 incidents & accidents as well as implementation of H2 safety codes and standards.

Nevertheless, one should remember that since certain industrial sectors are bound to investigate and report accidents while others are not, the reported numbers may be inaccurate. However, the numbers reported still enabled us to observe a trend of improved safety in hydrogen technology over the years.

Out of the 695 accidents reported in the two databases, 131 accidents (19%) are related to Scenario 1 Table 14.1. In HIAD 2.0, Chemical & Petrochemical Production is classified as an application stage. We reasoned that hydrogen applications at this stage are related to Scenario 1 since water electrolysis, hydrogen compression, pipework transportation as well as storage are all covered at this application stage.

Hydrogen production is associated with a lower normalised fatality risk1 as compared to other conventional energy sources based on historical data Figure 14.4, confirming its potential as a fuel to replace oil and natural gas that are widely used today.

In addition, work by (Spada, Burgherr and Boutinard Rouelle, 2018[4]) based European data until 2012 yielded a fatality per Terawatt-hour at ca. 0.03.2 These two works suggest that the use of hydrogen does not present an elevated risk to the safety of people when compared to current energy sources.3

In order to study which component at water electrolysis facility is the major risk contributor, we summarised the number of accidents arising in 5 key components of an installation Table 14.2. These components are commonly used in all types of water electrolysis facilities. In agreement with scientific literature (Pan et al., 2016[5]), (Skjold et al., 2017[6]) and previous review on electrolysers (FCH 2 JU, 2020[7]), hydrogen compressors are the major risk contributor, while water electrolysers are considered to be a very safe technology. The hydrogen compressor is associated with the highest Death/Accident ratio, which is 3-5 times higher than the ratios for other components. On the contrary, there is only one accident recorded from a water electrolyser itself and a fatality did not result from that accident.

Figure 14.5 summarises the information in Table 14.2 in a pie chart using colours to label risk level (number of deaths per accident). The component that is most prone to failure is hydrogen pipework, which made up 44.6% of the total accidents that are related to Scenario 1. This is followed by storage (23.1%) and hydrogen compressor (18.5%).

Furthermore, the root causes of the reported accidents are analysed and presented in Figure 14.6. The majority of accidents were caused by equipment failure, followed by deficiency in procedure and unknown. Equipment failure relates to failures that are unexpected and cannot be eliminated by minor changes in procedure. An example of such accidents is HIAD ID 660: the failure of a 400 mm pipe at 1.7 MPa caused release and ignition took 10 seconds. It is therefore recommended to evaluate guidance on the expected lifespan of critical components.

The transport of hydrogen through high pressure pipelines is not yet widespread (most hydrogen pipelines that are currently operating are located in industrial sites), a fact that is reflected by the low number of reported incidents that are related to hydrogen pipelines. Only 9 such incidents have been recorded in the HIAD 2.0 database, while no incidents are recorded in the H2Tools database. Although these incidents are related to hydrogen pipelines, the operating pressure of these pipelines, as well as information about the pipe components and physical dimensions, is not reported to enable a more thorough analysis.

Globally, there are currently approximately 4 500 km of pipelines globally dedicated to hydrogen transport (Shell, 2017[8]) which is much less than the several million km of pipelines used for natural gas transport (Placek, 2021[9]). Nevertheless, a comparison between the normalised leakage incident rate from natural gas pipelines and hydrogen pipelines in recent years (2015-19) shows that the calculated normalised incident rate for hydrogen pipelines is 0.09 incidents per 1 000 km of pipeline per year, which is slightly lower than the leakage incident rate for natural gas pipelines, which is reported to range from 0.13 to 0.16 incidents per 1 000 km of pipeline per year Table 14.3. However, the hydrogen pipeline leakage incident rate will most likely change in the future, once the usage of pipelines for hydrogen transport becomes more widespread. Please refer to the section on Normalisation calculations: Scenario 2 - Pipeline transport: leakage from high pressure pipeline for more details on the normalised leakage incident rate calculations.

Further investigation into the 9 reported incidents found in the HIAD 2.0 database revealed that 2 of the incidents did not involve hydrogen ignition, while 5 incidents resulted in hydrogen fires. The remaining 2 incidents were found to have resulted in an explosion Figure 14.20.

The root causes behind the reported incidents were also studied. While for 3 incidents the root cause was unknown/unreported, for the other 6 incidents the causes were error in the design of the pipelines, human error, inadequate maintenance of the pipeline, as well as deficiency in procedure Figure 14.8. A characteristic example of an incident caused by human error is the leakage of hydrogen due to improper depressurisation of the pipeline during maintenance, as reported for event No. 345 in the HIAD 2.0 database. The incident that was caused by deficiency in procedure was the damage of a hydrogen pipeline by an excavator during excavation works, due to the fact that the company was not notified about the presence of the pipeline at that location.

These types of incidents are typical of those, which occur in any major hazard pipeline carrying hazardous substances such as flammable gas (methane) or liquid hydrocarbons. Therefore, the same causative risk profile can be assumed for hydrogen as well.

Normalisation calculations: Scenario 2 - Pipeline transport: leakage from high pressure pipeline

We normalised the number of incidents against the length of pipelines used for gas transportation per year for hydrogen and natural gas. Only incidents between 2015 and 2019 were considered, as they are more relevant to the current state of things than older incidents.

Data published by Shell (Shell, 2017[8])showed that the total length of hydrogen pipelines globally was 4 542 km as of 2017. Only two incidents involving hydrogen pipelines were recorded between 2015 and 2019, so these were normalised against the pipeline length.

Data on natural gas leakage incidents from pipelines in the United States were obtained from the PHMSA incident database. Information about the length of natural gas pipelines in the United States over the years was obtained from Statista (Placek, 2021[9]).

Already normalised data on the number of natural gas leakage incidents per 1 000 km of pipeline per year for European pipelines was obtained from the 11th EGIG report (EGIG, 2020[11]).

For this scenario, incidents that involved vehicles transporting hydrogen, as well as incidents involving vehicles powered by hydrogen were considered. A total of 71 such incidents were reported, with 67 of the incidents being reported in HIAD 2.0, 2 incidents reported in H2tools and another 2 incidents reported in both databases Table 14.5.4 Of these incidents, 53 involved vehicles transporting hydrogen and 18 involved vehicles powered by hydrogen Figure 14.9.

In terms of the physical consequences of these events, most of them (26 in number, 37%) did not result in hydrogen release, 20 of them involved hydrogen release but no ignition, while 16 resulted in fire and 9 of them (13%) resulted in a hydrogen explosion Figure 14.10.

Table 14.6 shows a comparison between liquefied petroleum gas (LPG) vehicle incidents recorded in the Japanese High Pressure Gas Act Incident Database5 and hydrogen vehicle incident data.

Normalised incident rate suggests LPG vehicles are ca. one order of magnitude safer than hydrogen vehicles. LPG being a more mature technology certainly contributes to this observation. Nonetheless, most of the hydrogen accidents observed were traffic accidents and again, there were no novel hydrogen accidents when it comes to causation. We can therefore consider that hydrogen is not much more dangerous than LPG which is widely transported by vehicles and used as a fuel today.

For vehicles transporting hydrogen, the majority of accidents recorded were substance leaks due to collisions or traffic accidents and there were no novel hydrogen accidents when it comes to causation.

When the incidents are further divided into incidents that involve vehicles transporting hydrogen or incidents that involve vehicles powered by hydrogen, a difference in incident severity is observed.

When referring to vehicles transporting hydrogen, these are divided mainly into liquid hydrogen and compressed hydrogen tankers (involved in 26 of the incidents) and into trailers / trucks carrying hydrogen cylinders (involved in 26 of the incidents) Figure 14.11.

In terms of incident severity, 11 incidents resulted in no hydrogen release whatsoever and 18 incidents resulted in unignited hydrogen release, while 15 of the incidents involving hydrogen transport resulted in fires and 9 resulted in hydrogen explosions Figure 14.12. It should be noted that because of the relatively small number of incidents these ratios may change over time as the use of hydrogen increases.

In contrast, the 18 incidents which involved vehicles powered by hydrogen exclusively concerned fuel cell buses. These incidents occurred mainly during pilot projects such as the CHIC project (Müller, K. et al., 2017[12]), where the vehicle operation was more closely monitored and therefore minor incidents were reported that might not have been reported otherwise. This is reflected by the high number of reported incidents that resulted in no hydrogen release (15 incidents), with only two incidents resulting in unignited hydrogen release and one incident resulting in a hydrogen fire Figure 14.13.

The main root cause of all the reported incidents that are relevant to Scenario 3, was vehicles involved in a traffic accident, which accounted for 31 of the incidents (44%). Other significant incident causes were equipment failure, human error, as well as other external factors Figure 14.14.

Of the 11 incidents related to equipment failure, most involved flaws in the hydrogen tanks themselves, but there were also incidents that involved the failure of other components such as pressure relief bursting discs, hoses, valves and connections between hydrogen cylinders Figure 14.15.

We normalised the number of accidents against per registered vehicle per year for hydrogen and LPG. We considered only accidents between 2010 and 2021 as we believe they are more relevant to older accidents.

An article published by jupyter research (Jupyter Research, 2022[13]) estimated the number of hydrogen vehicles on road to be ca. 60 000 in 2022. We only considered registered accidents in the databases (HIAD 2.0 and H2tools) between 2010 and 2021 as we consider older accidents to be less relevant.

We were able to find detailed Japanese LPG data in their High Pressure Gas Act Database. A report by World LPG Association (World LPG Association, 2019[14]) reported the number of registered LPG vehicles in Japan to be 182 000 in 2018.

We identified 25 accidents that are related to Scenario 5: of which 9 accidents were reported in HIAD 2.0 and an additional 16 reported in H2tools. In contrast with most hydrogen-related accidents, the majority of Scenario 5 related accidents (15) are caused by the mal-function of the compressor or dispenser, and result in no hydrogen leak (Figure 14.16 and Figure 14.17). This is in agreement with an earlier study (Sakamoto et al, 2016) focused on Scenario 5 related accidents in Japan and the United States,6 which are not yet covered by either HIAD 2.0 or H2tools.7

The accident rate is normalised to be 1.19 x 10-7 accidents per refuelling (Appendix 5.2.3). As a comparison, we also calculated the (normalised) accident rate for LPG stations, which is at 2.52 x 10-7 accidents per refuelling. The numbers suggest that hydrogen refuelling stations, in their current states, can be considered slightly safer than LPG stations (see Table 14.9).

As expected, the majority of the accidents are caused by equipment failure Figure 14.17, which is dominated by dispenser and compressor failure. Other equipment failures are related to storage facilities and sealings Figure 14.18.

In specific, dispenser-related accidents are dominated by flexible hose failures (4), and the majority of human error was caused by fuel cell vehicles (FCV) users. The above-mentioned Japanese study (Sakamoto et al., 2016[15]) highlights that no user-induced accidents were reported in Japan due to the regulations prohibiting self-serviced hydrogen fuelling stations.8 Nonetheless, even self-serviced petrol stations were once prohibited in many countries based on state fire codes (National Association of Convenience Stores (NACS), 2022[16]). Therefore it is still necessary to provide safety measures for the prevention of FCV-user induced accidents as like petrol stations, we expect FCV-users would eventually be able to perform self-fuelling, even in Japan. The same Japanese study also mentions that the majority of leakages in Japan are caused by screw joints. Since joints are mainly welded in the United States, there is a reduced proportion of joints-related leakages from 81% (Japan) to 45% (United States). This example suggests a relatively small change in design can in some cases significantly reduce the risk associated with a certain component.9

We normalised the number of accidents against per refuelling for hydrogen and LPG.

A presentation by the National Renewable Energy Laboratory reported a number of refuelling per station per hour at 3.1. Based on this, we estimated a number of refuelling per day per station at 49.6 assuming hydrogen stations operate between 7:00-23:00. For accidents registered in the databases, we only considered those after 2004 as there is only one recorded accident before 2004 (1991) and it may be less relevant.

For LPG vehicles we once again relied on the Japanese High Pressure Gas Act Database. In addition, the (World LPG Association, 2019[14]) reported an average estimated number of registered LPG cars at around 2.2 x 105 and an average LPG consumption at 1.28 x 106 metric tons for the period 2004-2018. Since LPG tank sizes range between 20 and 140 litres, we used the median (80.25 L) to estimate the number of refuelling. A conversion factor of 1.96L/kg is used to convert this volume (80.25L) to weight (42.09 kg).

Note that there were no recorded LPG station accidents in Japan after 2012. If we consider only the period between 2004 and 2011, then the number of accidents per fuelling per year would be 5.75 x 10-8.

Work by (Park et al., 2006[17]) provided the total number of LPG accidents between 1992 and 2003. In addition, Korea Energy Economics Institute10 published LPG consumption data.

Besides the incidents that have been analysed in previous sections, which are related to specific scenarios, a number of incidents were identified which involve the storage. Since hydrogen storage can be related to scenario 1, 3 and 5, these incidents are analysed here separately.

In total, 28 incidents related to hydrogen storage were reported, with 22 of the incidents being reported in HIAD 2.0, 4 incidents reported in H2tools and another 2 incidents being reported in both databases Table 14.14. Notably most of the incidents took place before 2010, with only 4 of the incidents taking place after 2010 Figure 14.19. This could be an indication of the success of the stricter regulations and safety requirements regarding the storage of hydrogen.

Incidents that involve hydrogen storage have the potential to be severe due to the high pressure of the stored hydrogen and its large mass. In terms of incident severity, 15 of the reported incidents resulted in hydrogen explosions, 2 resulted in fires, while 10 incidents resulted in unignited hydrogen release and 1 incident involved no hydrogen release whatsoever Figure 14.20.

Since most of the incidents occurred before 2000, in many cases the available information is limited and the causes of the incidents were unclear or unreported. Overall, the main causes of accidents involving hydrogen storage are the failure of the storage equipment, errors in the design of the hydrogen storage or deficiencies in the hydrogen handling procedure Figure 14.21. An example of deficiency in procedure observed in a recent incident is the release and ignition of hydrogen at the gas storage station of a nuclear power plant in France. The accident took place because, when a pallet of empty hydrogen cylinders was being replaced, the pallet was not disconnected from the gas supply line. The hose connecting it to the pressure relief system was then accidentally torn off during the pallet removal by a forklift. This incident revealed that the safety procedures had not been properly adapted to the specific storage conditions. Issues were discovered which could be contributing causes to the incident, such as uncontrolled access to the storage area, lack of respect of the ATEX11 distance for welding work and the abnormally high frequency by which the gas pallets were being replaced.

In terms of component failure, only 5 incidents that were caused by failed components were identified. The components involved in these incidents were pipes, a compressor, a high-pressure valve and a pressure controller Figure 14.22.


[3] Brook, B. et al. (2014), “Why nuclear energy is sustainable and has to be part of the energy mix”, Sustainable Materials and Technologies, Vol. 1-2, pp. 8-16, https://doi.org/10.1016/J.SUSMAT.2014.11.001.

[11] EGIG (2020), 11th Report of the European Gas Pipeline Incident Data Group (period 1970 – 2019), https://www.egig.eu/reports (accessed on 6 May 2023).

[7] FCH 2 JU (2020), Minutes of the FCH 2 JU workshop of electrolysis on 18 November 2020, Fuel cells and Hydrogen 2 joint undertaking.

[2] IEA (2022), Global demand for pure hydrogen, 1975-2018, https://www.iea.org/data-and-statistics/charts/global-demand-for-pure-hydrogen-1975-2018 (accessed on 7 June 2023).

[1] IEA (2019), The Future of Hydrogen: Seizing today’s opportunities, OECD Publishing, https://doi.org/10.1787/1e0514c4-en.

[13] Jupyter Research (2022), Hydrogen Vehicles: Market Forecasts, Key Trends & Segment Analysis, https://www.juniperresearch.com/researchstore/key-vertical-markets/hydrogen-vehicles (accessed on 6 May 2023).

[12] Müller, K. et al. (2017), Clean Hydrogen in European Cities: Final Report including Part A - Final Publishable Summary Report, Part B - Final Restricted Full Report. Daimler Buses.

[16] National Association of Convenience Stores (NACS) (2022), The History of Self-Fueling, https://www.convenience.org/Topics/Fuels/The-History-of-Self-Fueling.

[5] Pan, X. et al. (2016), “Safety study of a wind-solar hybrid renewable hydrogen refuelling station in China”, International Journal of Hydrogen Energy, Vol. 41/30, pp. 13315-13321, https://doi.org/10.1016/J.IJHYDENE.2016.05.180.

[17] Park, K. et al. (2006), “Incident analysis of Bucheon LPG filling station pool fire and BLEVE”, Journal of Hazardous Materials, Vol. 137/1, pp. 62-67, https://doi.org/10.1016/J.JHAZMAT.2006.01.070.

[10] PHMSA (2022), PHMSA pipeline incident database, https://portal.phmsa.dot.gov/analytics/saw.dll?Portalpages.

[9] Placek, M. (2021), US pipeline system mileage 2004-2020, Statista, https://www.statista.com/statistics/197932/us-pipeline-system-mileage-since-2004/ (accessed on 6 May 2023).

[15] Sakamoto, J. et al. (2016), “Leakage-type-based analysis of accidents involving hydrogen fueling stations in Japan and USA”, International Journal of Hydrogen Energy, Vol. 41/46, pp. 21564-21570.

[8] Shell (2017), “Energy of the Future?”, https://epub.sub.uni-hamburg.de/epub/volltexte/2017/68185/pdf/shell_h2_study_new.pdf.

[6] Skjold, T. et al. (2017), “3D risk management for hydrogen installations”, International Journal of Hydrogen Energy, Vol. 42/11, pp. 7721-7730, https://doi.org/10.1016/J.IJHYDENE.2016.07.006.

[4] Spada, M., P. Burgherr and P. Boutinard Rouelle (2018), “Comparative risk assessment with focus on hydrogen and selected fuel cells: Application to Europe”, International Journal of Hydrogen Energy, Vol. 43/19, pp. 9470-9481, https://doi.org/10.1016/J.IJHYDENE.2018.04.004.

[14] World LPG Association (2019), Autogas Incentive Policies: A country-by-country analysis of why and how governments encourage Autogas and what works, https://www.wlpga.org/wp-content/uploads/2019/09/Autogas-Incentive-Policies-2019-1.pdf.


← 1. Normalised risk: A measure of risk created by mathematically adjusting a value in order to permit comparisons.

← 2. A similar figure plotted by (Spada, Burgherr and Boutinard Rouelle, 2018[4]) is presented in Output 2.

← 3. In addition to these figures, comparative data with normalised number of road accidents and / or fatality rates among hydrogen-powered vehicles and other types of vehicles have been requested by the Japanese authorities.

← 4. An additional accident involving liquid hydrogen release from a rail tanker was reported in HIAD 2.0, however as rail transport of hydrogen was determined to be beyond the scope of this review, this was not included in this report.

← 5. High Pressure Gas Act Incident Database, 2021, the High Pressure Gas Safety Institute of Japan (KHK).

← 6. Recorded in the Japanese High Pressure Gas Safety Act Database (in Japanese) for period 2005-2014 and US HIRD (hydrogen incident reporting database) database for period 2004-2012.

← 7. While H2tools and HIAD 2.0 cover a broad range of hydrogen accidents from across the world, they rely on the gradual collation of information from smaller hydrogen databases and user reports, so such gaps in coverage are not unexpected.

← 8. The Self-Serviced Hydrogen Station Guidelines (JPEC-TD 0004, in Japanese) released by the Japan Petroleum Energy Center (JPuEC) in 2018 allows driver-performed hydrogen fuelling provided they have gone through required safety training.

← 9. Nonetheless, the complex layouts of hydrogen refuelling stations can make welding operations difficult and there was limited data on material strength of welded parts in high-pressure hydrogen environments when the study was published (2016).

← 10. http://www.keei.re.kr/main.nsf/index_en.html?open&p=%2Fweb_keei%2Fen_Issues01.nsf%2F0%2FFBCEC343E68337DF49256E2900483FB5&s=%3FOpenDocument, accessed 20/07/2022.

← 11. ATEX is the name given to European Directives 99/92/EC and 2014/34/EU which define the minimum requirements for improving the health and safety protection of workers potentially at risk from explosive atmospheres, as well as directing the laws of Members States concerning equipment and protective systems used in potentially explosive atmospheres.

Metadata, Legal and Rights

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area. Extracts from publications may be subject to additional disclaimers, which are set out in the complete version of the publication, available at the link provided.

© OECD 2023

The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at https://www.oecd.org/termsandconditions.