Refuelling infrastructure requirements for renewable hydrogen road fuel through the energy transition

.


Introduction
Let's look to the situation of road transport in just thirty years.We'll be in a world that will be both very different and very similar to today.Different, in that so much of the technology we know and use today will have vanished, replaced by newer systems that emit practically no greenhouse gasses.Similar, in that we will still need and want to move ourselves, and things, around.It seems inconceivable that we won't be using road transport in some form or another to do that.
At present, road transport is responsible for 11.9% of global greenhouse gas emissions (Ritchie and Roser, 2020).In order to reach the global targets for net zero greenhouse gas emissions these must be eliminated.
Until relatively recently, the default assumption among many has been that all future road transport in the UK will be electrified using batteries (Scottish Power, 2020).Furthermore, at the time of writing in 2022 a wide range of road vehicles is available as battery electric vehicles (BEV) in cars, light goods vehicles, and buses, and these are taking an increasing market share ( UK Government Department of Transport, 2019).However an alternative in the form of hydrogen is available, and awareness is increasing of this option.There have been numerous trials and developments of hydrogen as a fuel, along with expressions of support from governments.These include -but are by no means limited to -California (Clark and Rifkin, 2006), Japan (Chaube et al., 2020), South Korea (Choi et al., 2020), China (Ren et al., 2020), and also the EU (Cuevas et al., 2021).Specifically relevant to our study area, the Scottish (Scottish Government, 2020a) and UK (UK Government, 2021) governments have both recently released hydrogen strategic plans and policies.
The infrastructure requirement and associated cost has been examined by Robinius et al. identifying competitive relative costs for hydrogen infrastructure in Germany (Robinius et al., 2018).Greene et al. (2020) examine the challenges of deploying hydrogen refuelling infrastructure, concluding that hydrogen has the potential to supply a major share of the world's transportation energy demand.Whiston et al. (2022) elicit the views of experts in the field to consider various aspects of future hydrogen vehicle useamong other conclusions, they anticipate up to 5,000,000 hydrogen fuel cell electric vehicles (HFCEV) in China by 2040.
Here we assess the infrastructure requirements of using hydrogen for some or all road vehicle fuel.We examine needs right through the transition period to the time where essentially all road transport produces zero carbon emissions.
Geographically we limit ourselves to a case study of Scotland.However, the approach should be generally transferable to the rest of Europe and other areas of the world, provided local factors are taken into consideration.Local factors such as vehicle life, annual distance travelled per vehicle, and emissions from the production of hydrogen and generation of electricity, will all be significant.
We chose Scotland for the case study because It has an excellent track record to date of implementing emissions reduction methods in electricity generation (Scottish Government, 2020b).The high current planned and potential level of renewable electricity means that we can confidently expect that enough hydrogen from water electrolysis could be produced through carbon free means.
At present approximately 96% of Scotland's electricity is produced by carbon free means, predominantly wind or nuclear, having reduced electricity generation related greenhouse gas emissions by 89% since 2000 (Scottish Government, 2020b).There is now only one significant fossil fuel power station in use, the natural gas fuelled power station at Peterhead (Nedd et al., 2020).This implies that hydrogen generated from new electricity sources will be responsible for essentially zero carbon emissions when used as a fuel.If necessary, this could be backed up by locally available natural gas sources, which could be used to produce 'blue hydrogen' as an interim measure -albeit at less than complete, but still substantial, elimination of emissions, and potentially with significant short-term cost savings and availability benefits (Gonzalez et al., 2021).
Scotland has a defined and challenging set of emissions targets.The neighbouring jurisdiction in England has similar targets, although not identical (UK Government, 2018a); the potential for contamination by cross-border sales is limited.The Scottish government has adopted a series of relevant emissions related objectives (Scottish Government, 2017;Scottish Government, 2019a; UK Government News, 2020): • 2020 electricity generation to reach the equivalent of zero emissions for domestic use; • 2030 reduction in greenhouse gas emissions across all sectors of 75% from 1990 levels; • 2030 new hydrocarbon car and van sales are banned (UK government requirement); • 2045 reduction in greenhouse gas emissions across all sectors to net zero.
We do not aim to make a comprehensive study of safety issues in this paper.Refuelling stations and similar activities are well regulated in Scotland and the UK by the Health and Safety Executive (Health and Executive, 2022), and safe vehicle construction is regulated by the Driver and Vehicle Standards Agency (UK Government, 2022a).We assume for the purposes of this study that these organisations will ensure safe construction of hydrogen vehicles and fuelling facilities.
We aim to address the following questions: • For various scenarios of the extent of hydrogen fuel used within that transition, what quantity of hydrogen would be required over the energy transition period to 2050? • What production, distribution, compression, storage, and dispensing infrastructure will be required to deliver the hydrogen to end users?• What will be the capital cost?• Which scenario, or scenarios, will be most likely?
As a precursor, we will also have to consider: How will Scottish and UK government overall emissions objectives translate into road transport?And how fast should the transition to zero emission vehicles be to meet those emissions objectives?
We developed a Multi-Period model, incorporating Monte Carlo and Markov chain methods, to answer these questions; a similar approach has been taken in the forecasting of renewable electricity generation requirement (Hart and Jacobson, 2011).Here we present the results of the subsequent analysis We also present details of the construction of the model and the underlying assumptions.We conclude with recommendations of the optimum pathway to use hydrogen in road transport to meet the objectives, and appropriate steps to deliver it.It is important to note that we are not trying to predict what will happen; we are developing a range of options for what must happen in order to meet the objectives.
We incorporate the beneficial reuse of the existing widespread natural gas network as the most likely scenario in Scotland (Martin et al., 2021;Mouli-Castillo et al., 2021), allowing an understanding of a more integrated energy system than has traditionally been in place.We also consider the pace of change required to meet the Scottish government's targets, based on several scenarios.
We aim to contribute to the literature by assessing the size and cost of the necessary hydrogen infrastructure at a large scale in the long-term for vehicle refuelling purposes.We present a simple method for assessing the required infrastructure.This should inform policy makers not only for Scotland but also further afield, subject to incorporation of local factors.
This should also inform the debate around the question of battery electricity or hydrogen, although we hold the view that there are no winners and losers in that discussion.Just as petrol (gasoline) and diesel serve largely different needs at present, there will be a need for both future fuel types (and possibly others) as the demand grows exponentially.

Initial assumptions
We ignore the cost of construction and operation of wind turbines and other generation equipmentwe use announced contracted costs of offshore wind electricity supply, or wholesale costs of network supplied electricity, as inputs to the model where required; this figure accounts for all such construction and operation costs (Thomas, 2020).
We assume that hydrogen is initially produced locally from electrolysis of water.This can come from grid supplied, or local dedicated, renewable electricity.We model that centrally produced green hydrogen will gradually become available over a period from a variable start (2026)(2027)(2028)(2029)(2030) to finish (2040-2045), supplied through the repurposed natural gas network.This repurposing of the network is expected over that timescale in any case, to replace the existing supply of natural gas with hydrogen (SGN, 2019).We assume that, over time, hydrogen supplied in this way will displace locally produced hydrogen using grid electricity where practical, and that local re-purification cost to remove, e.g., odourants from the hydrogen will be low or insignificant (Staffell et al., 2019).Green hydrogen supplied in this way is assumed to be limited to the 85% of households currently connected to gas network (UK Government statistics, 2020), although this value is allowed to vary in the Monte Carlo cases in the model as low as zero; the rest will stay locally produced.For this locally produced element, we model variable proportions of grid supplied or local dedicated electricity.See worksheet 19 in the model, supplied as Electronic Supplementary Information (ESI).
Initial unit costs of hydrogen fuelling stations are derived from methods presented by Tlili et al. (2020), along with contract information published by a large manufacturer of electrolysers (Nel, 2020), and sense checked in discussion with a commercial manufacturer and installer of hydrogen refuellers.We apply learning rates to the costs of hydrogen refuelling equipment (Ruffini and Wei, 2018), varied among the Monte Carlo cases in the model.
We assume that sufficient supplies of hydrogen can be made available; indeed, one key purpose of this paper is to identify how much will be required so that appropriate provision can be made.However, in the event of competition for inadequate supply, this would be reflected in the supply price.For the purposes of this work, this operating cost is insignificant.
We assume, as a starting position, that the number of vehicles in each class and mileage will remain static from levels at the time of writing.This is modelled by keeping the sales of new and scrapped vehicles equal to the number of vehicles in the class divided by the average age at disposal of the vehicle (calculated in worksheet 6 of the model, supplied as ESI).This gives a modelled sales figure lower than the actual recent sales figure, due to the size of vehicle fleet increasing in recent years for most classes of vehicle ( UK Government Department of Transport, 2019), so this anticipates some active management of vehicle demand.However, the Scottish Government also has a target to reduce car use by 20% by 2030 (Scottish Government, 2022).We model this as a sensitivity analysis of no change, 10% reduction, and 20% reduction in average annual distance driven by cars by 2030, and unchanging after that.In reality such a reduction might manifest as a smaller number of vehicles covering the same annual distance, or the same number covering a smaller distance, or another equivalent variation; in terms of the fuelling requirement, the focus of this paper, that would make no difference.
One possible flaw in this projection is the potential for people to keep existing hydrocarbon vehicles running for longer.This would have the effect of reducing the number of new Zero Emissions Vehicles (ZEV) sold while maintaining the number of vehicles on the road, and hence extending the time before the emissions targets are met.However, estimating the effect of this, or remedial measures, is outwith the scope of this analysis.
We use the standard vehicle classes used by the UK and Scottish governments, shown in Table 1 along with their numbers and selected characteristics: A list of other model input assumptions, with sources and background, is presented in Appendix A.

Preliminary assessment of sector emissions targets
As a preliminary step, we examine what emissions reduction will be required in the road transport sector in Scotland.As at 2019, road transport is responsible for 20% of Scotland's greenhouse gas emissions (UK Government, 2019).
The key Scottish Government all-sectors targets are (i) reduction from 1990 totals by 75% by 2030, and (ii) zero net emissions by 2045 (Scottish Government, 2017Government, , 2019a)).There is also a recent UK Government requirement to eliminate solely hydrocarbon fuelled vehicle sales by 2030, with hybrids eliminated by 2035 ( UK Government News, 2020).We assume that all sectors' emissions decrease at an equal proportionate rate from the present day levels to reach these overall targets, with three differences: 1. Negative emissions allocated to land use change stay at current levels; they have not changed significantly for several years (Scottish Government, 2019b).2. Emissions from electricity production will reach zero before 2030.
The target for this (affected by the Covid-19 pandemic) was 2020, but the exact date (before 2030) it is achieved does not affect this calculation (Scottish Government, 2017).3. Emissions due to air travel reduce by only 10% in each of these two stages.This figure is somewhat arbitrary -there are no currently available zero emissions commercial aircraft (Dincer and Acar, 2016), but the calculation is not sensitive to variations in this value as air travel is only a small contributor to the total (Scottish Government, 2019b) as can be seen in worksheet 2 of the model (provided as Electronic Supplementary Information).

Scenarios
We based our analysis on a series of transition scenarios.These combine three pace of transition options and three fuel choice options, as follows:  Government. 2018.Road transport energy consumption at regional and local authority level UK Government.Available from: https://www.gov.uk/government/statistical-data-sets/road-transport-energy-consumption-at-regional .UK Government, 2019.National Atmospheric Emissions Inventory.Available from: https://naei.beis.gov.uk/reports/reports?report_id=991.
a The vehicle classes in the UK Government statistics are termed "Goods" and "Light Goods" for commercial vehicles over and under 3,500 kg respectively (other than buses and coaches).In this paper we use the less ambiguous common terms Heavy Goods Vehicle (HGV) and Light Goods Vehicle (LGV) respectively.

Pace options
We project the annual change in zero emissions vehicles (ZEV) in each vehicle class using standard logistics functions creating typical 'S' curves.The logistic function takes the form (Verhulst, 1845) Equation 1 Where L is the target value (discussed below)k is the gradient function, typically in the range 0.5-1.0.x is the year (or other period) under consideration xo is the mid-point year of the time series under consideration .
The first logistic function used represents new ZEV sales rising to meet the existing level of sales of all vehicles.The constant L here represents the number of sales of all vehicles.Where a class of vehicle has a mandatory date in legislation for ending hydrocarbon vehicle sales, the number of sales is forced to the full value of new vehicle sales (L) by the end of that year rather than allowing the function to produce a natural taper.
The second logistics function represents the future scrapping of ZEVs based on the typical average lifespan of vehicles in each class.Again it rises to meet the existing level of sales (required to keep total numbers constant).In this case, L represents the total number of vehicles scrapped in an average year.Clearly, for the case where the total number of vehicles remains constant, the two values of L will be equal.
As total numbers of vehicles and total annual sales are held constant, the annual change in number of ZEVs and hydrocarbon vehicle (HCV) sales is then in a matter of simple arithmetic.
In an iterative process, the constants k and xo, which control the two logistic functions, are varied (the same constants are used for both functions, within each scenario) until the emissions in use meet the objectives (see section 2.5).
The pace options used are - • Equal Paceall vehicle classes transition at an equal pace.That is, the logistic functions for all classes have the same initial midpoint and gradient.

Fuel choice options
These represent the share of zero carbon fuels between hydrogen and other fuels in the future.
• Large Vehicles Onlyall vehicles in the classes Buses & Coaches and HGVs are HFCEV.Other vehicles use other means of decarbonisation.
• Like for Like -There is a view that hydrogen is more suitable for longer range and larger vehicles, esepocially with restricted maximum weight, due to the high volume, light weight, and fast refuelling times (Cooper, 2018).Similarly, large and long distance vehicles favour diesel at present, albeit for other reasons.So this scenario has current diesel fuel vehicles replaced with HFCEVs, and current petrol vehicles replaced by BEVs.This means that all Buses & Coaches and HGVs, and 41% of cars and 97% of LGVs, will be hydrogen fuelled (UK Government Department of Transport, 2019).• 100% hydrogen.This replaces all hydrocarbon fuelled vehicles with hydrogen fuel cell vehicles.
These Pace and Fuel Choice options combine to give nine Transition Scenarios for evaluation, within which we examine the vehicle classes shown in Table 1.

General approach for analysis
For each Transition Scenario, we took the following steps, analysed over the period 2021 to 2050.The overall investigation is based on a Multi-Period model which we constructed using Microsoft Excel.The model incorporates a Monte Carlo analysis of 1000 randomly generated cases to test the effects of varying unknown and forecast quantities, and a Markov Chain projection of future hydrocarbon vehicle emissions.
We use the model to: 1. Identify annual change in numbers, and total numbers, of ZEVs required to meet the emissions targets.This means that in all scenarios, the targets relevant to that transition scenario will be met or bettered, with a +5% allowed variance.2. Identify demand for hydrogen as a fuel to supply these ZEVs to the relevant proportion.3. Identify the fuelling and related infrastructure required to deliver that quantity of hydrogen.Infrastructure here refers to equipment specifically needed for storing and dispensing fuel, upgraded or new distribution systems, and local or central production.4. Identify capital costs of providing the infrastructure.
These steps are also shown in Fig. 1 (section 2.8).

Approach for emissions calculations
First we examined the constraints on future HC vehicle emissions, based on current EU and UK legislation, and existing vehicle fleet emissions as set out in Table 1.We used a Markov chain assessment (meaning that each value generated in a series is affected by the previous value generated) to create future emissions.Over several time steps, we generated a random level of emissions for new vehicles.This random level was constrained to (i) not exceed the anticipated legislative constraints and (ii) not to exceed the previous time step value.This method was carried out for each vehicle class independently, and different random values were generated for each of the Monte Carlo cases described in section 2.4.
Next, we took the number of new hydrocarbon vehicles in each class for each year of the model and multiplied by the new vehicle emissions.
Then we removed the number of scrapped HC vehicles in each class, multiplied by the previous year's average emissions in class.
This allowed us to arrive at a total figure for each year's class emissions from HC vehicles, and the new class average per vehicle emissions.
In an iterative process, the constants in the logistic functions generating the numbers of new ZEVs were manipulated until the total emissions for the scenario reached the targets.This was done such that no random Monte Carlo case produced emissions exceeding the target by more than 5%.This process allows us to generate the required number of ZEVs of each class in use each year.
Due to the large amount of renewable electricity available now or imminently, emissions from ZEVs in use are taken as zero.Emissions in vehicle manufacturing are outwith the scope of this study.
We also limit this study to greenhouse gas emissions from fuel use.Particulate emissions from fuel use and non-fuel sources are also outwith the scope of the study.

Computation of hydrogen refuelling station numbers
We split the Monte Carlo cases into three sections of 330 for each of the three fuel choice options.These defined the proportions of ZEV in each vehicle class using hydrogen for each year, which then produces the actual numbers of vehicles when combined with the total number of ZEVs found in section 2.5.
The number of hydrogen fuelled vehicles is then used to calculate the total demand for hydrogen fuel each year.
Three Hydrogen Refuelling Station (HRS) sizes were defined, based on the sizes of typical existing hydrocarbon stations in use today.The proportion of fuel supplied by each size of HRS was taken as fixed at the proportion supplied at present by the corresponding sizes of petrol/ diesel station (see worksheet 18 in the model, supplied as ESI).
The proportion of each station's capacity that would be actually used on an annual average was given a central value of 71%, after Robinius et al. (2018).This was allowed to vary randomly in each case, constrained between 61% and 81%.
By combining these with the gross hydrogen demand for each year, the required gross capacity of each HRS size could be readily calculated.Dividing this by the capacity of the HRS, and rounding up to the next integer, gives the aggregate numbers of HRS of each size required in each year.Finding the numbers of new HRS required each year is then a matter of simple arithmetic.

Calculation of costs
For the calculation of costs, we identify two types of HRS: Those which produce hydrogen in-situ with renewable electricity, and those which use hydrogen taken from the re-purposed natural gas network.The in-situ producing HRS are subdivided into those which use electricity from the electricity grid, and those which have a direct, dedicated connection to a local source of renewable electricity.
The Monte Carlo cases have random allocations of the following variables (introduced in section 2.1) related to the type of HRS: -The start and finish dates of the conversion of the natural gas network to hydrogen.-The maximum proportion of network supplied HRS, reached at the end date of the network conversion through typical logistic function ('S' curve).-The proportions of in-situ producing HRS which use a dedicated connection to a renewable electricity source or a connection to the electricity grid.For each case, the numbers of HRS of each type are computed.The capital cost associated with each type is found by combining the initial cost and the learning rate, described in section 2.1.The capital cost of the electrolyser capacity required to produce the hydrogen distributed through the network is assessed separately, in a similar way.

Modelling tool
The model is outlined as follows: The model input values, assumptions and sources are presented in Appendix A.
The calculations used are detailed in Appendix B.
A reviewable version of the model as used, including all worksheets referred to in Fig. 1 and elsewhere in this paper, is presented as Electronic Supplementary Information.

Preliminary assessment of emissions targets
The key Scottish Government all-sectors targets are a reduction in emissions from 1990 totals to 75% by 2030, and net zero by 2045 (Scottish Government, 2017Government, , 2019a)).There is also a UK government requirement to eliminate solely hydrocarbon fuelled vehicle sales by 2030.
Projecting the emissions data gives target residual road transport emissions of 5586 and 1185 kt/yr CO 2 equivalent by 2030 and 2045 respectively, or a required reduction from 2018 levels of approximately 45% by 2030 and 88% by 2045.These are the emissions targets that we work to in this project (model worksheet 2).
Fig. 2 shows these targets in the context of the recent emissions records and the primary energy related emissions sources.

Pace of transition
The Pace of Transition question depends on how quickly end-of-life hydrocarbon vehicles can be replaced with ZEVs rather than new HCVs.This is independent of whether the ZEVs are HFCEVs or something else (e.g.BEV).Figs. 3 and 4 illustrate the rate of increase of new ZEV sales required to meet the targets, and their effect on greenhouse gas emissions.
We see that the Accelerate Bus & Truck options require a smaller number of total ZEVs than Equal Pace in the early years, while meeting the same targets.This means that the Accelerate Bus & Truck scenarios might be easier to implement, due to reduced demand on manufacturers.Also, given that large vehicles are typically owned in fleets, fewer decision makers will need to be influenced.However, smaller vehicle transition will still be required at a good pace and cannot be ignored.The Laid Back scenario allows a significantly slower transition in all vehicles.This is illustrated in Fig. 3.
Fig. 4 shows the range of emission profiles arising from these pace options.This shows that meeting the demanding 2030 objectives should to lead to a considerable overshoot of the 2045 objectivethis could create some headroom in other harder to decarbonise sectors.Conversely, the Laid Back scenario emissions shows that even if it proves impossible to meet the interim targets for road transport, meeting the ultimate 2045 target should be much more achievable.This would still not prevent the overall objectives for 2030 from being met, if sufficient early gains could be made in other sectors.
For each scenario, it can be seen that the variation in emissions between Monte Carlo cases is smallthis means that variation in the forecast future level of per-vehicle emissions is much less significant than the pace of removing hydrocarbon fuelled vehicles altogether.

Quantity of hydrogen required
By forecasting the energy demand for the numbers of HFCEVs identified, we modelled the quantity of hydrogen fuel required for each of the transition scenarios.This is found from the energy provided by liquid fuels to the hydrocarbon vehicles removed from the road, adjusted to account for the different levels of efficiency.The results of this assessment are presented in Fig. 5. shows the Laid Back option.As can be seen, the Laid Back pace results in meeting the same ultimate demand, but not until around 2050; the same holds for all other fuel choice options (not illustrated).
Fig. 5 also shows the quantity of hydrogen required in terms of weight as kilotonnes per year; this is only illustrated for Like for Like/ Equal Pace options.

infrastructure requirement for refuelling
Existing fuelling stations hydrocarbon were categorised as "company owned", "dealer owned", and "hypermarket" in a 2012 study carried out for the UK Government (Deloitte, 2012).By extrapolating the numbers from the 2012 survey in proportion to the overall UK numbers today (based on correspondence with the UK Petroleum Retailers' Association, unpublished but available on request from the authors), along with UK government statistics on fuel sales (UK Government, 2018b), we arrived at the numbers of stations and average fuel volumes shown in Table 2.
We then calculate equivalent quantity of hydrogen to provide the same useful energy as these three filling station sizes.This takes account of the improved efficiency of the fuel cell over internal combustion engines, and uses the lower heating value of hydrogen (Mazloomi and Gomes, 2012).These values are then rounded to provide useful sizes for small, medium and large hydrogen fuelling stations.These results are also shown in Table 2. See worksheet 18 in the model for calculation.
We assume that the numbers of each size of fuelling station will be in the same proportion as the numbers of each size of hydrocarbon filling station at present.
Carrying out the process described in section 2.6 yields 330 possible out-turns for each of our transition scenarios.Fig. 6 shows the maximum, minimum and mean total HRS numbers for the pace options Equal Pace and Accelerate Bus & Truck, with the fuel options of Like For Like and Large Vehicles Only.Other scenario options are presented in worksheet 25 of the model.
We can see from Fig. 6 that there is only a little variation between Accelerate Bus & Truck and Equal Pace in the Like For Like option, with a more pronounced variation in the early stages of the Large Vehicles Only option.This is perhaps unsurprising.
It's also clear from Fig. 6 that numbers of HRS required in the Like For LGV sales by that date, forcing the curve to the maximum value.Without this, the curves would naturally meet the maximum around 2035.This chart applies to all fuel choice options.Like options doesn't exceed the ultimate likely minimum of the Large Vehicles Only options until around 2030.This has implications about the confidence of investing in, or supporting, the early stage development of HRS.We explore this further in section 5.

annual investment
Following on from that, we can estimate the annual investment required, as described in section 2.7.This is illustrated for the Equal Pace option only in Fig. 7, for the three fuel choice options.The Accelerate Bus & Truck pace option is essentially identical to this.
The Laid Back cases, not displayed, require a similar total expenditure, but skewed significantly to the later years as might be expected.Fig. 7 also shows the estimated annual value of capital refurbishment and upgrade for hydrocarbon filling stations.Clearly this is approximated in a very wide range, but we can see that it is comparable in magnitude to the costs of establishing new HRS, which would all be required by 2045.The main curves in Fig. 7 relate only to the costs of establishing new HRS, not to the ongoing expenditure associated with maintaining or renewing them as they agethat would of course continue indefinitely beyond 2045/50.The potential range of investment required suggested by the Monte Carlo analysis in the model is also shown here.A significant source of variability, especially in the later years, is the question of what proportion of fuelling stations are supplied from the gas grid, compared to using local generationgrid connected stations being considerably cheaper since they don't require their own electrolyser.The other main source of variability is the usage rate of the fuelling stations; as would be expected, if fuelling stations are used more intensively, fewer are required.

Sensitivity analysis
We carried out this analysis on the basis that vehicle use would stay the same.However, the Scottish Government has a further target to reduce car use by 20% by 2030.We carried out a sensitivity analysis, considering the effect of a 10% or 20% reduction in car use by 2030 with use remaining constant beyond that date.We modelled this by keeping the vehicle numbers constant, but reducing the average annual distance covered.
This meant that for each of the Pace options, the constants in the logistic were revised so that the same emissions targets were met.
In the Equal Pace options, this also reduces the pace of conversion to ZEV of other vehicle classes; with all vehicles converting at the same pace, all see the benefit of the reduced car emissions.Thus, even for the Large Vehicles Only fuel choice options, there is a short term reduction in the numbers of fuelling stations required.However, the Accelerate Bus and Truck pace options have buses and HGVs on a different path from smaller vehicles.Here, motorcycles and LGVs share a reduced pace of decarbonisation, but buses and HGVs are unchanged.This way the Large Vehicles Only fuel choice options show no change in infrastructure requirement from the base case.The Laid Back option was not considered in this sensitivity analysis.
The outputs from this analysis are combined in Fig. 8.The maximum numbers of hydrogen refuelling stations are found from the unreduced car use, in the Accelerate Bus & Truck pace option in both Like for Like and Large Vehicles Only fuel choice options.The minimum numbers are found from the reduced car use and Equal Pace option for both fuel choice options.This is as expected.For comparison, Fig. 8 also shows the mean number of HRS required for both unreduced and 20% reduced car use.

Table 2
Estimated current (2020) numbers of various types of refuelling station in use, with current fuel sales (shaded boxes), along with equivalent capacities of hydrogen refuelling stations (white boxes).The hydrogen station sizes take account of the different efficiency of the fuel and engine types, to supply the same useful energy.The effect of a 10% reduction lies between the 20% reduction and no change, as expected.However, as the 20% reduction has only a limited impact on our conclusions and recommendations, we have not considered further the effect of a 10% reduction.

In the future
We can envisage the future zero emissions road transport system.In this future transport world, all vehicles will be zero greenhouse emissions at the point-of-use.The technology required to do this has existed from the early transition, and all new vehicle sales will have been zero emissions from 2030 for smaller vehicles and approximately 2035 for larger vehicles (UK Government News, 2020).The technology exists today to deliver a zero carbon road transport system for Scotland, at an achievable financial cost, and hydrogen should play a substantial role in that.
There will always be a need for a choice of fuel types; different fuels serve different purposes.Much as the pre-transition road fuel system uses both petrol and diesel for different, but overlapping, purposes, so we can expect that both battery electricity and hydrogen will be used for different purposes.Quite possibly other fuels that are not available in the early stages of the transition will become viable as well.As a fuel, the important characteristics of hydrogen for users are: lower weight (similar to existing hydrocarbon systems), fast refuelling time (a few minutes), long range readily achievable, and fuel cost competitive with electricity when made at scale (worksheet 4 in the model, supplied as ESI).Conversely, internal space is more compromised by the large volume the fuel tank requires, and for some users overnight recharging could be convenient (Cunanan et al., 2021;Turoń, 2020).All of this means that, just like diesel in the present day, hydrogen will lend itself predominantly to larger and longer distance vehicles.No one type of fuel is likely to become universal.
So what proportion of vehicles will be fuelled by hydrogen?It seems likely that essentially all buses and coaches and HGVs will be hydrogen fuelled (this paper's Large Vehicles Only option).It also seems likely that some car and van owners will choose hydrogen fuelled vehiclesit is quite conceivable that all pre-transition diesel vehicles will be replaced with hydrogen (this paper's Like For Like option).
This future fleet of hydrogen vehicles will be supported by a network of around 300 (Large Vehicles Only) to 820 (Like For Like) hydrogen refuelling stations of various sizes (see Fig. 6).The majority of these HRS will be supplied with hydrogen generated from offshore wind electricity.This forms part of the Scottish government's programme of developing renewable hydrogen production, for use domestically and for export, from North Sea wind, announced in 2020 (Scottish Government, 2020a).This hydrogen will be supplied through the national gas network, which will be converted over a period from the late 2020s to the early 2040s, from its original purpose of distributing natural gas (SGN, 2019).At the upper end of this scale, the number of HRS would be similar to the number of petrol & diesel fuelling stations currently in service (861 -model worksheet 18); this suggests that the ultimate spacing between them could be similar to that at present, with provision available in more remote areas.Of course, the HRS fuelling industry might develop into a smaller number of larger stations, or vice versawhich could have an impact on the service provision in rural areas.
Those hydrogen refuelling stations which couldn't be sited with a connection to the hydrogen gas network will probably use their own dedicated renewable electricity supply -most likely wind turbines -to produce their own hydrogen in-situ.This approach will be widespread in the early days of the transition before the gas grid is fully converted to hydrogen.The production cost of the fuel is likely to be broadly similar (model worksheet 4), so there won't be a significant commercial disadvantage in one supply method or the otherother than the risk of the local electricity source not producing for some time.So, just like the early hydrogen refuelling stations, these future non-grid hydrogen filling stations will require a backup supply in the event that their own electricity supply is inactive for too long.This is likely to be in one of three forms: a connection to the electricity grid, but this tends to be very expensive in actual use, adding around 80% to the wholesale fuel cost; a larger on-site storage facility than would normally be required; or a supply delivered by road using a tube trailer (a specialised tanker).
An alternative scenario could be around the use of several hydrogen hubs as production and distribution centres.This is considered in the Scottish Government's recent consultation on hydrogen (Scottish Government, 2021), but analysis of this option is outwith the scope of this paper.

Present day
Let's now consider what has to be done in the short term in order to permit that future to unfold.Primarily, enough hydrogen refuelling stations of sufficient capacity need to be constructed to allow the use of hydrogen fuel vehicles up and down the country.The problem being that we don't know how many to build, or where.Key questions about the future are presently unanswerable with accuracy: Will all vehicle classes transition at the same rate (this papers Equal Pace option)?Or might it be possible to encourage larger vehicles to transition to hydrogen faster, giving more carbon dioxide reduction per vehicle replaced (this paper's Accelerate Bus and Truck option)?Or even, might a more relaxed transition take place, missing out the Scottish government 2030 75% emissions reduction target (Laid Back option)?For planning purposes we should assume that the 2030 target will be met.This also means that the 2045 transport target compatible with overall net zero will almost certainly be comfortably exceeded.
For the early stage activities, it would be ideal if we could identify a way which minimises the risk of over-construction, while offering the maximum appropriate support for hydrogen fuelled vehicles.Fortunately there is enough overlap in the requirements to facilitate this -the ultimate number of hydrogen refuelling stations required for the smallest out-turn of the Large Vehicles Only option will still provide enough capacity for the first 8-10 years of the Like For Like option, as in Fig. 6.Conversely, following the highest predicted demand for the Like For Like scenario up to 2025 would provide enough capacity for any likely out-turn in that time, and would not be wasted if the Large Vehicles Only option transpired.
In section 4, we considered the implication of a reduction in car use of 20% by 2030, a policy goal of the Scottish Government.If this came about, the above balance of requirements would only change to the extent that the minimum ultimate requirement for Large Vehicles Only would provide enough for the Like for Like fuel option for 8-12 years; planning for such provision then would still not be wasted.
Putting actual numbers to this, then, shows that a sensible initial program to 2025 should consist of 9 large fuelling stations, 11 medium, and 36 small ones to service a total demand of 71,500 kg/day (26 kT/ year) by 2025.These will most likely use hydrogen produced locally.The best option for local production would be to bypass the national electricity grid and use dedicated wind turbines, or a specific offtake agreement with nearby wind farms.This would influence the location of the filling stations.
If our Accelerate Bus And Truck -Like For Like scenario holds good, then this capacity could be required by 2025.If the reduced demand, and initially slower, scenario Equal Pace -Large Vehicles Only is the outturn, then this capacity would be required by 2028.A reduction in car use of 20% by 2030 would extend this later date by less than one year.This initial construction programme can be expected to cost in the region of £140M.This compares very favourably to Scottish Power's forecast costs for electrification of road transport, even excluding the substantial electricity grid reinforcement costs (Scottish Power, 2020).Alongside this, though, the existing program of EV charger rollout should continue for some time -for the future envisaged here, there will be a need for a large capacity to charge battery electric vehicles as well.
There is always a risk associated with setting up this type of new infrastructure.Without extensive existing users, an operator may not be confident of being able to sell enough hydrogen to cover their costs.But without enough infrastructure, people and companies are not likely to buy new hydrogen vehicles to create the demand.To resolve this chicken-and-egg situation fast enough to meet the emissions targets will likely require some market stimulation or support to enable initial progress.
However, the annual costs involved are comparable in magnitude to the current expenditure on hydrocarbon refuelling renewables.Hydrogen is very important to the fuelling industry due to the substantially different requirement of electric charging, and the way hydrogen refuelling is carried out in a similar manner to liquid hydrocarbons at presenttaking just a few minutes at a pumpwhich means that the future business model will be similar to the present day.This means that the a substantial part of the costs might be reasonably borne by the industry; some underwriting of risk may be all that is necessary in terms of government support.This view was shared by a director of a large UK refuelling company in an informal discussion.
Availability, as in manufacturing capacity, of vehicles and infrastructure will be critical.ZEV HGVs have only recently been introduced as BEVs (Murray, 2020).HFCEV HGVs are expected by 2023 (Hyundai, 2021) and are already available in some markets, e.g.Switzerland, where a partnership between a vehicle manufacturer and a refuelling operator has helped to pave the way (Hyundai Motor Company, 2021).HFCEV cars are slowly becoming available (Benz, 2020;Hyundai, 2020;Toyota, 2020;Honda, 2020) and can reasonably be expected to become more popular once the fuelling infrastructure is in place to make their use practical.

The transition phase
We can turn our thoughts to what will happen between this initial phase of investment and the longer-term.We can expect the number of hydrogen vehicles to increase enormously over that period 2025 to 2045 in the case of our Like For Like scenarios; so the number of hydrogen refuelling stations required would increase as well.
The ultimate number of hydrogen refuelling stations of the sizes we have considered could be around 820, somewhat fewer than the 860 petrol and diesel refuelling stations in service today, reflecting the fact that a substantial part of the vehicle refuelling load would be taken by charging of battery vehicles.The potential reduction in car use by 20% would reduce this ultimate number to around 760.The effects of this, though, would be seen more clearly after some years, by which time market forces and other effects should be better understood.
If the out-turn were Large Vehicles Only -and the position between these two scenarios would be subject to market forces driven both by fuel costs and user preference -there would be a need for around 300 hydrogen refuelling stations (possibly fewer but larger, since almost all of the relevant demand would come from larger vehicles).Over this period we can also expect the existing natural gas network to become fully converted to supply hydrogen produced at a centralised location.We therefore anticipate that new hydrogen refuelling stations would use this as a source of hydrogen, where a network connection can practically be made.
In the earlier part of this period, we can anticipate that the cost of installing and operating the infrastructure, along with vehicles, should reduce enough to permit any market stimulation to be withdrawn.We also expect that the extent of renewable electricity available offshore will increase dramatically in line with, or exceeding, the Scottish government announced targets (Scottish Government, 2020a).However, the total renewable electricity requirement for hydrogen generation (at around 73% of road transport energy requirement) could ultimately be provided by a windfarm/s supplying around 18 TWh per year (calculated in model, worksheet 19).
For context, Scotland's 2020 renewable energy generation was around 39 TWh (of which onshore wind electricity 19.5 TWh, and offshore wind 3.5 TWh with the balance being solar, hydro-electricity, heat and biofuels), compared with an all sector energy demand of 155 TWh (Scottish Government, 2020b).Consented further offshore wind generation represents around 22 TWh per year (Offshore Wind Scotland, 2020).A contribution might also be drawn from curtailed wind generation; in 2019, 1.9 TWh of potential wind powered electricity generation was curtailed in Scotland (Renewable Energy Foundation, 2020).In addition, the January 2022 Scottish offshore wind electricity leasing round indicates a further 25 GW capacity is expected to be delivered in due course (Government, 2022), which should provide over 100 TWh per year of additional renewable electricity.
In December 2020 the Scottish Government announced its hydrogen strategy (Scottish Government, 2020a), which includes the production of 5 GW equivalent of hydrogen by 2030 and 25 GW by 2045 for a range of uses including export.These are well in excess of the requirements we forecast for road transport, which are equivalent to approximately 1 GW in 2030 and around 2.5 GW in 2045 (based on a Like for Like fuel demand option).
At current fuel and energy prices and tax rates, hydrogen at a large scale should be cheaper to the consumer than petrol or diesel however it is produced; if grid electricity is avoided, it should be also be cheaper than electricity for batteries (Model worksheet 4; although note that because of the volatility in energy and fuel prices at the time of writing, this indicates relative position better than actual values).The question of how to equitably charge customers supplied by different routesthat is, network and non-network gas, and local dedicated or grid supplied electricitywill have to be addressed as a policy decision.This may be tied into taxationin the long run, governments will undoubtedly seek to replace some of the lost fuel duty currently paid though hydrocarbon fuel sales.
Overall, the key constraints are more likely to be the availability of vehicles, fuelling equipment and hydrogen generation.We see the practical delivery of these in the in the required timescale as a bigger hurdle than the cost or the development of new technology.
We think the emissions targets are achievable, but they are also extremely challenging.This a very big undertaking, and time is of the essence.

Conclusion and policy implications
Our aim in this analysis was to estimate what extent of fuelling infrastructure, and associated costs, would be required to support the use of hydrogen as a fuel for road.We did this primarily using a Multi-Period model, incorporating Monte Carlo and Markov Chain components, which we constructed for the purpose using Microsoft Excel.
Our key findings were that (i) the most probable scenario meeting the targets is Equal Pace -Like for Like, and (ii) that providing an initial seed network for this scenario would still be well within the needs of the likely minimum out-turn, the Large Vehicles Only option.
This most likely scenario means anticipating that (i) all vehicles will transition to zero emissions vehicles at a similar pace, and (ii) existing diesel vehicles will be replaced with HFCEVs, and petrol vehicles will be replaced with BEVs or other technology.Because larger vehicles use proportionately more energy, this would mean ultimately around 73% of fuel energy being transported as hydrogen.
The infrastructure to supply this in the first 5 years would be around 9 large scale (5700 kg/day) hydrogen fuelling stations, 11 medium sized (2850 kg/day) ones, and around 36 smaller ones (1000 kg/day).However, if the Large Vehicles Only option transpired, as an effective minimum likely out-turn, the same infrastructure would still be required within 7-8 years.This infrastructure would cost around £140 million, which compares very favourably to the cost of electrification for battery electric vehicles.
Ultimate numbers for the Like For Like option could reach around 820 hydrogen fuelling points of various sizes; significantly fewer would be required under Large Vehicles Only at about 300.These would cost around £740M and £320M respectively, expressed as NPV to 2050 at 6%, or £2.1bn & £670M as a simple aggregate.Annual expenditure would peak at around £100M/year between 2028 and 2038 for Like for Like, or around £40M for Large Vehicles Only.The cost of providing for alternative zero emissions fuels to fill the gap between these scenarios would be likely to be substantially higher than the apparent saving in hydrogen infrastructure, based on the costs set out by Scottish Power (Scottish Power, 2020).
It may become possible to accelerate the deployment of larger vehicles, that is buses and HGVs, as in our Accelerated Bus & Truck pace option.This would have advantages in reducing the number of zero emissions vehicles required in the short term; the early years' expenditure on refuelling infrastructure would be slightly higher (up to 5% extra).However, this may not be achievable over the next few years due to the different development stages of the vehicle types.
We propose that this initial programme of hydrogen refuelling locations and charging points should be pushed forward as a seed and development network, as a matter of some urgency.This should, however, add to rather than replace the ongoing programme of expanding the network of battery electric vehicle chargers.It also seems likely that the vehicle fuelling industry should be able to fund a large part of the hydrogen infrastructure within the existing pace of commercial investment.
Shorter term market stimulation for hydrogen fuelling systems, vehicle sales and/or fuel costs might be required, until commercial risks and costs reduce to a level similar to current fuel systems.Vehicle and fuel costs are anticipated to reach this level by around 2025 (Top Gear magazine, 2020;Reed et al., 2020).A policy of ongoing support might be needed in areas obliged to use more expensive technology.We also recommend developing a partnering strategy, involving government, academia, vehicle manufacturers, energy and fuelling companies, and others, as soon as possible to push this forward as efficiently and as quickly as possible, if these challenging targets are to be met.This could be developed into a permanent centre of excellence, supporting the development of Scotland's related industry to take advantage of the currently under-developed global supply chain.the previous timestep.This creates a Markov Chain value generation.We generate values for the key dates of 2020, 2025, 2030, and 2040.• %Hclass = Future zero carbon fuel market share of hydrogen, using the three input constraints from the fuel choice options of Large Vehicles Only, Like for Like, and 100% Hydrogen.• %EllUse = Electrolyser usage rates as a % of capacity.This is allowed to vary randomly between constraints, set at 81% and 61%, based on a central value of 71% (Robinius et al., 2018).• LocalElec = The proportion of in-situ produced hydrogen using local dedicated electricity generation, rather than grid supplied electricity.This is allowed to vary randomly between 0 and 100%.
For each modelled case, calculation steps 2-8 below were applied.
Step 2 This generates the future vehicle numbers, which control all the subsequent steps.
The numbers of ZEVs in each class for each year is generated, using a pair of standard logistics functions, where one represents new sales and the second represnts vehciles scrapped at end of life.The two are offset by the average age on scrapping of the class.
The values of the constants creating this distribution were adjusted manually for each vehicle class, until the emissions in use (calculated in step 3) met the relevant 2030 and 2045 targets within 5% for all modelled cases.The constants were varied for the different classes of vehicle as appropriate.The values of the constants developed are presented in the model in worksheet 12.
From there, the total numbers of ZEVs, and new sales, number scrapped, and total registered hydrocarbon vehicles for each year are then calculated using simple arithmetic.The sales figures are forced to 100% of cars and vans at the end of 2030, reflecting the UK Government's ban on hydrocarbon sales from that date.
Inputs: NewVehclass is the total annual sales of new vehicles of any fuel type for each class, taken as constant through the transition.
k is an arbitrary constant affecting the gradient of the curve x is the year in question xo is the mid-point of the time distribution, that is the point at which NZEclass = NewVehclass/2.VL is approximately average vehicle life, found from total fleet size/annual sales (this assumes a constant fleet size).EndHCdate = year after which hydrocarbon vehicle sales are to be stopped.TotVehclass = Total numbers of vehicles in class, assumed to be constant at current levels.
Outputs: TotEmclass is calculated for each vehicle class, and summed across all classes for each year.
Step 4 Calculates the required quantity of hydrogen (produced locally and centrally), based on the numbers of ZEVs of each class (Step 2) and the proportion of them using hydrogen (step 1).
The average demand for hydrogen is found from the average current demand for hydrocarbon fuels for each vehicle class, adjusted by an efficiency factor relating the energy required from petrol or diesel to that from hydrogen.
The annual total and proportionate demand for hydrogen is calculated:

Fig. 1 .
Fig. 1.Schematic of model created to analyse required fuel infrastructure and emissions to meet emissions targets.N.B.Worksheet numbers refer to the worksheets within the model.Worksheets 1-10 set out input information and carry out preliminary calculations; 13 is an input data summary page; un-numbered pages beyond 2 5 hold combined outputs from running multi-scenario macros.

Fig. 5
Fig. 2. Scottish emissions targets, showing overall government targets, road transport estimated target for 2030, and energy related emissions history and required trajectories to achieve ultimate net zero target.Non-energy emissions not shown but accounted for in the net-zero calculated end points.Historic data from Scottish Government Climate Statistics (Scottish Government, 2019b).

Fig. 3 .
Fig. 3. New sales of zero emission vehicles, projected for pace of transition scenarios.The step at 2030/31 reflects the ban on hydrocarbon car &LGV sales by that date, forcing the curve to the maximum value.Without this, the curves would naturally meet the maximum around 2035.This chart applies to all fuel choice options.

Fig. 4 .
Fig. 4. Emissions reductions achieved from the three Pace of Transition scenario options, using the vehicle sales numbers shown in Fig. 3.Each trace represents one result of the Monte Carlo simulation used to generate the emissions.This chart applies to all fuel choice scenario options.

Fig. 5 .
Fig. 5. Fuel requirement for hydrogen and hydrocarbon fuels.Only the Equal Pace pace option is shown (indistinguishable from Accelerate Bus & Truck in this graph), except for the Large Vehicles Only fuel option which also illustrates the Laid Back pace option for comparison.

Fig. 6 .
Fig. 6.Numbers of hydrogen refuelling stations for selected scenarios: Like for Like and Large Vehicles Only fuel options, with Equal Pace and Accelerate Bus & Truck pace options.Maximum and minimum cost cases are presented, along with the case derived from the core estimated model inputs.

Fig. 7 .
Fig. 7. Capital expenditure required for hydrogen, showing all scenarios with the Equal Pace option.Solid lines represent the outputs from the core values used in the Monte Carlo inout cases.Vertical bubbled lines represent the range of outouts from the Monte Carlo analysis.Approximate range of annual expenditure on capital upgrades to hydrocarbon infrastructure renewals is also shown as the grey lines.

Fig. 8 .
Fig. 8. Hydrogen fuelling stations required by year.This combined chart shows the extremes found from the base case and 20% car use reduction, Equal Pace and Accelerate Bus and Truck pace options, for the Like for Like and Large Vehicles Only fuel choice options.It also shows the mean numbers for both base case and 20% car use reduction, in the Equal Pace options.

Table 1
Vehicle classes, number of vehicles in service in Scotland, fuel types, fuel consumption, Carbon Dioxide emissions for fleet and single vehicles.Where more than one fuel type is in use, the split is based on the proportions for the whole UK.Petrol includes hybrid and plug-in hybrid.