D6A.2 & D6A.3 Quantitative Risk Assessment of the distribution grid and built environment in the Netherlands: application and case studies

To make an initial assessment of the risks of hydrogen in the Dutch distribution network, a quantitative risk assessment (QRA) has been conducted. It compares the risk between the current natural gas distribution system and the future hydrogen distribution system. The total risk in the analysis consists of the risk arising from leaks in the distribution network and the risk arising from leaks in the house itself.

The model simplifies the built environment in the Netherlands in order to control the number of variables and associated calculations. For this initial version, detached houses were chosen for validating the model. In addition, semi-detached houses were also used in the case study of a representative sample neighborhood. The distribution network is simplified by using a limited number of materials, pressures, and diameters. An important input parameter of the model is the failure frequency of the different components. The failure frequencies for the distribution network were determined based on historical failure reports. However, no reliable dataset for the components behind the meter (meter assembly, internal piping, and end-use appliance) could be found for the Netherlands, so values from the UK were used instead. Given the aforementioned assumptions, the model provides an approximation of the location-specific risk resulting from fires or explosions. All calculated location-specific risks in this study for both natural gas and hydrogen remain well below 1x10^-6 per year, indicating a very limited risk.

Based on the aforementioned failure frequencies in the dwelling, it is found that the location-specific risk for natural gas due to leaks behind the meter aligns well with the (limited) historical data. The probability of a fatal accident in the Netherlands resulting from an explosion or fire per dwelling, based on historical data, is 0.06 x 10^-6, excluding cases involving intentional gas releases. Additionally, the probability of injuries is 1.1 x 10^-6. The model yields respective values of 0.02 x 10^-6 for fatal accidents and 0.4 x 10^-6 for injuries, indicating a similar order of magnitude to the historical data. The risk scales linearly with the failure frequency.

With the same set of parameters and without additional mitigating measures, it is found that the effect of explosions with hydrogen is more severe than with natural gas. The location-specific risk for hydrogen is 3.8 times greater, i.e., 0.18 x 10^-6, when the risk of carbon monoxide poisoning is not considered. When comparing the risk between hydrogen and natural gas in the house, the risk due to carbon monoxide poisoning should also be taken into account. The mortality risk due to carbon monoxide poisoning is 0.37 x 10^-6 per natural gas connection. When this risk is included in the comparison, it is found that there is a shift from reduced risk due to carbon monoxide poisoning to increased risk from explosions. The total location-specific risk with the chosen set of assumptions and without additional mitigating measures is lower for hydrogen than for natural gas.

The effect of ventilation on the accumulation of (dangerous) concentrations in the house was determined by analyzing ten identical houses with different ventilation rates. For hydrogen, halving the ventilation rate increases the risk by a factor of 1.8, while doubling the ventilation rate reduces the risk by a factor of 2.2. Ventilation has a stronger effect on the risk in the dwelling for hydrogen compared to natural gas.

The factor between the total risk from the distribution network without additional measures for hydrogen and natural gas is nearly 2.5 times. Based on the assumptions made in the model and averaged per connection (7.2 million), the risk from the distribution network in the dwelling is approximately 0.2 x 10^-6 for hydrogen, assuming no additional mitigating measures are taken.

To gain a better understanding of the relative effects of leaks behind the meter and from the distribution network, a case study of a representative neighborhood was analyzed. This neighborhood consists of 57 homes connected to a 100mbar main pipe through service lines. The 100mbar network is fed by an 8 bar steel pipe running through the neighborhood. The 100mbar network is modeled in several segments with different materials and diameters. The homes are modeled based on their surface area and include both detached and semi-detached houses. Additionally, the risk posed by leaks behind the meter has been determined for each of the homes.

The likelihood of leaks leading to the accumulation of gas inside a home is higher for leaks behind the meter. For hydrogen, it appears that the majority of the location specific risk is caused by leaks behind the meter (approximately 73%). The remaining portion is caused by the main pipe and service line connected to the home, as well as nearby sections of the mains. Generally, the contribution of the 100mbar pipe is greater than that of the 8 bar pipe, depending on the distance between the homes and these pipes. The risks are highest for semi-detached houses compared to detached houses. Similar to the aforementioned risks, the impact of explosions is greater for hydrogen than for natural gas. However, the overall risk per dwelling in the neighborhood is lower for hydrogen compared to the risk posed by natural gas when considering the contribution of carbon monoxide poisoning. It is important to note that even without additional measures, the total risk remains well below 1x10^-6 in both cases, indicating a relatively small risk.

In summary, it is concluded that the risks calculated in the model are relatively small. In perspective, the total number of fatal accidents in the Netherlands in 2021 was approximately 6,500. The majority of these accidents were caused by accidental falls (5,430, corresponding to a risk per resident of approximately 3x10^-4/year). The total number of fatal accidents caused by smoke, fire, and flames in 2017 was 43, corresponding to a risk of 2x10^-6. The share caused by natural gas in the built environment is a fraction of this.

The risk associated with hydrogen can be reduced by achieving lower failure frequencies. It is found that spontaneous leaks in frequently occurring parts of the network (100mbar) contribute most to the total risk. Damage from interference is detected earlier, resulting in less frequent accumulation of gas to dangerous concentrations in enclosed spaces. The greatest effect can be achieved through mitigating measures that reduce the frequency of spontaneous failures in pipelines or components, such as periodic leak detection or the replacement of couplings that often lead to leaks. An initial assessment has been made of the impact of excess flow valves and gas sensors with acoustic signals applied to the risk of hydrogen in the case study of a typical neighborhood. This is based on (yet) unpublished initial calculations applicable to the UK situation. The approximation shows that excess flow valves, depending on the assumptions in the model, can achieve a potential risk reduction for hydrogen of approximately 20%. For gas sensors, the estimated reduction is about 27%.

The results described in this report were obtained considering the set of assumptions as described. In this simplified model of the distribution network and built environment, several refinements of the model are possible. It is recommended to further improve the model by incorporating a greater variety of housing types. The model mainly used detached houses. The effect of explosions is calculated for nearby homes in the model, resulting in a higher risk for semi-detached houses. In a refined version of the model, a distribution of housing types (detached/semi-detached/row houses/small apartments, etc.) should be applied. One of the assumptions used, considered currently as a limitation of the model, is that PVC pipes have the same leak size distribution as PE pipes. In practice, this may be different. Hard PVC has a more brittle character and may potentially lead to more brittle fractures. This results in a different leak size distribution, which consequently affects the calculated risk. Further research on the leak size distribution is recommended. Lastly, it is recommended to model the effect of the excess flow valve and gas sensors. Based on initial outcomes from the UK, an initial estimation has been made. It is advisable to expand the model for the Netherlands by simulating the implementation of these measures.