A review of hydrogen delivery technologies for energy system models

Transcription

1 A review of hydrogen delivery technologies for energy system models UKSHEC Working Paper No Paul E. Dodds and Will McDowall UCL Energy Institute, University College London Central House, 14 Upper Woburn Place London WC1H 0NN, UK p.dodds@ucl.ac.uk 1

2 Contents 1 Introduction UKSHEC hydrogen delivery analyses Methodology for producing energy system model data Calculating technology investment and O&M costs Converting costs for use in UK MARKAL Hydrogen delivery infrastructure Liquefaction Road tanker transport Ship transport Tube trailer transport Pipeline transport Hydrogen pipeline costs from the literature Hydrogen pipeline costs from UK natural gas experience Energy system costs for a hydrogen pipeline system for the UK Pipeline lifetime Modelling hydrogen transmission pipelines as lumpy investments Transporting hydrogen using natural gas infrastructure Fuelling stations Representing hydrogen fuelling stations in energy system models Technology data for energy system models Summary of technologies in current MARKAL models Recommendations for the future Liquid hydrogen delivery Tube trailer delivery Pipeline delivery Fuelling stations Annual fixed O&M costs Energy inputs and technology lifetimes Uncertainties in technology costing Learning curves Counterfactual technologies Electric vehicles Liquid hydrocarbon fuels Road tankers Fuelling stations Conclusions References

3 1 Introduction UK greenhouse gas (GHG) emissions predominantly result from the combustion of fossil fuels to provide energy services in all sectors of the economy. In the future, it will be necessary to either capture the emissions or to utilise zero-carbon sources of energy. Hydrogen has been identified as a potential zero-emission energy carrier for the future, primarily for the transport sector but also for energy storage and CHP applications. Although micro-scale hydrogen production systems are being developed for domestic use by some companies, it is likely that a network of hydrogen fuelling stations would be required to supply fuel to motorists or hydrogen-fuelled vehicles would suffer the same range limitations as electric vehicles. Small-scale hydrogen production from electrolysis or biomass would be possible at fuelling stations but large plants would otherwise be required for fossil fuel feedstocks so that the CO 2 by-product could be captured and stored using CCS. In the short-term, small-scale SMR could also be deployed at refuelling stations to facilitate the transition to a hydrogen economy but these would not be compatible with decarbonisation in the long-term. Hydrogen production technologies are examined in UKSHEC II Working Paper 6 (Dodds and McDowall, 2012). Hydrogen is relatively difficult to store and transport in comparison with petroleum fuels. Gaseous hydrogen (GH 2 ) has two principle drawbacks. Firstly, the unusually low volumetric energy density of gaseous hydrogen means that the gas must be compressed to extremely high pressure to be used as a transport fuel. Secondly, the tiny molecules have a higher propensity to leak than other gases and require particularly complex storage materials. One method of avoiding these two difficulties is to compress the hydrogen into a liquid (LH 2 ), but this is energetically-expensive and difficult to handle because LH 2 boils at around 253 C. Other forms of hydrogen, for example metal hydrides, are currently being researched but are at an early stage of development. Three methods are commonly used to deliver hydrogen to refuelling stations (Yang & Ogden 2006). GH 2 pipelines require large up-front capital investments but can transport large amounts of hydrogen very cheaply over short and sometimes long distances. The most cost-effective method of delivering small amounts of GH 2 over short distances is using gaseous tube trailers. Over longer distances, LH 2 delivery by road tanker becomes optimal. 1.1 UKSHEC hydrogen delivery analyses The UKSHEC I project identified several qualitative delivery pathways (Eames and McDowall, 2005) and simulated these using the UK MARKAL model (Balta-Ozkan et al., 2007). MARKAL is a partialequilibrium E4 optimisation model (energy-environment-economics-engineering) that is used to represent the entire UK energy system. The current hydrogen delivery technology characterisation is based on Yang & Ogden (2007), although a separate spatially-disaggregated analysis was performed under a DfT-funded project (Strachan et al., 2009). The UKSHEC II project is revisiting and extending these analyses. The first step is to update the delivery technology data, for which we have identified some shortcomings and inconsistencies with other fuel delivery pathways. The aim of this paper is to: (i) recommend consistent representations of hydrogen delivery technologies for the UK MARKAL and TIAM-UCL models (TIAM-UCL is a global E4 optimisation model that has been recently developed); (ii) ensure that the assumptions underlying the hydrogen technologies are also consistently applied to other competitor technologies; and, (iii) identify technologies for which the deployment costs are likely to substantially reduce with large-scale deployment. The third objective reflects the greater emphasis on innovation and technology learning in UKSHEC II. 3

4 In this paper, we compare delivery technology data from the literature with existing hydrogen production data in the UK and US9R MARKAL models. We recommend revised technology data for the UKSHEC II project and consider the uncertainties in the data. The impacts of these uncertainties on model simulations will be considered in a separate paper. 2 Methodology for producing energy system model data Each hydrogen delivery pathway was broken into constituent technologies. For each technology, the E4 models require capital investment and operations and maintenance (O&M) costs, efficiency data, and fuel inputs and outputs. Few sources provided all of this data for each technology so we based our recommendations on a comparison of all available data rather than adopting a single source. This approach required a greater understanding of the factors underlying the costs and efficiencies of each technology which enabled us to consider data uncertainties. One potential drawback is that the recommended data might no longer represent a single plant design. We encountered two difficulties in particular. Firstly, the investment and O&M costs must be calculated for each technology using a consistent approach with other energy system technologies so that an unbiased comparison can be performed. Secondly, it is necessary to convert the costs into consistent monetary units. The following sections expand on these difficulties. 2.1 Calculating technology investment and O&M costs Direct investment costs refer to the actual cost of the installed equipment. All construction projects also have substantial indirect costs as well. System costs include design, site preparation, contingency costs and profits for the construction contractor (direct costs implicitly include profits for equipment manufacturers). The cost of land, licensing and permits vary between countries. Some plants have additional financial costs to cover financing for up-front fees. Estimates of technology costs always include direct costs but some indirect costs are often omitted. Since indirect costs are often more than 50% of the direct costs, the inclusion or omission of indirect costs can be one of the principle drivers of cost differences between studies. It is important that comparable costs are included for all technologies in the energy system models, both for hydrogen production and elsewhere, so that unbiased comparisons of technologies can be performed. There are presently no clear guidelines about which indirect costs are included in the UK MARKAL or TIAM- UCL models. In this study, we chose to include all indirect costs except for additional financial costs, which are separately accounted for by the model investment discount rate equations. O&M costs are also difficult to estimate because they include equipment outage costs, running costs, licensing costs and labour. Some are fixed each year while other variable O&M costs depend on the production rate. O&M costs can vary substantially for different hydrogen delivery technologies; for example, the predominant road tanker cost is driver wages while liquefaction costs are dominated by electricity consumption. Fuel costs are separately accounted for by energy system models so we concentrated on fixed O&M costs in this paper. Our approach calculated the annual O&M costs as a fraction of the capital investment costs for each type of technology. 2.2 Converting costs for use in UK MARKAL The UK MARKAL model uses British pounds in the year 2000 while the TIAM-UCL model uses US dollars in the year Investment costs in the literature are specified in both dollars and euros, for a range of years, so it is necessary to account for both currency conversion and inflation. The 4

5 Conversion factor to GB (2000) choices are (i) to adjust for inflation in the original currency and then apply the year 2000 currency conversion rate; or, (ii) to convert to /$ in the technology year then adjust for the UK inflationary difference to the year Exchange rates are more volatile than inflation rates so the first option produces a more consistent conversion factor (Figure 1). In this paper, in common with Lemus et al. (2010) and Bartels et al. (2010), the first approach is adopted. There is a 1.72 factor between the UK (2000) and US$(2005). However, there is no correct methodology as such and the conversion rates are a potential source of uncertainty when producing cost estimates of technologies. Figure 1: $: conversion factors for two conversion methods Conversion from US$ to GB (2000) Inflation adjustment then currency conversion Currency conversion then inflation adjustment US$ year of original cost data 3 Hydrogen delivery infrastructure We identified seven hydrogen delivery pathways and examined the constituent technologies: 1. National liquid hydrogen delivery (central liquefaction, road tanker delivery, LH 2 -GH 2 fuelling station). 2. International liquid hydrogen delivery (as national but with additional ship transport). 3. National high-volume gaseous hydrogen delivery (pipeline delivery from production plant to GH 2 -GH 2 fuelling station). 4. National low-volume gaseous hydrogen delivery (tube trailer delivery from production plant to GH 2 -GH 2 fuelling station). 5. On-site hydrogen production (at a large GH 2 -GH 2 fuelling station). 6. National gaseous hydrogen delivery to buildings as a replacement for methane (pipeline delivery from production plant to households). 7. Domestic micro-scale hydrogen production (no delivery technologies). 5

6 This section briefly reviews each of these technologies with the aim of identifying suitable data for energy systems models. Our primary sources for current and future technology costs and energy efficiencies were the H2A (Steward et al., 2008) and TECHPOL (Krewitt and Schmid, 2004) technology databases. Further data were extracted from studies in the literature where available (E4tech, 2005; IEA, 2005; Syed et al., 1998; van der Zwaan et al., 2011; West, 2003; Yang and Ogden, 2007). Where possible, we compared these data with existing data from the current UK and US9R MARKAL models. Energy system optimisation models assume that technology costs are proportional to the energy throughput. This assumption is not always valid for hydrogen delivery technologies (and other transmission infrastructure) because the costs are partly determined by the spatial separation of production systems and consumers. This complication is most important for pipelines, tube trailers, road tankers and ships. Cost and efficiency data for six LH 2 and GH 2 delivery technologies are compared in Figures 2 and 3, respectively. All of these technologies, and tube trailers, are discussed in the following subsections. 3.1 Liquefaction Hydrogen undergoes liquefaction at a temperature of 20 K ( 253 C). Theoretically, only about 4 MJ kg -1 must be removed from the gas but the cooling process has a very low Carnot cycle efficiency (Fichtner, 2009) so even large plants require 30 MJ kg -1 to liquefy hydrogen. The principle disadvantage of LH 2 is the substantial amount of expensive electricity that is consumed. Larger plants have lower investment costs (Figure 2) and are more efficient (Figure 3) than smaller plants. There is some uncertainty over investment costs with estimates ranging from 3 GJ -1 y -1 to 16 GJ -1 y -1. The higher costs reflect actual plant costs during the 1990s (e.g. Syed et al., 1998). The H2A project (Steward et al., 2008) has similar costs but other studies forecast substantially lower investment costs that are consistent with the US Department of Energy target to dramatically reduce the cost of liquefaction (e.g. Mytelka, 2008). There is clearly widespread belief that the costs of large-scale production in the future will be substantially lower than the costs of existing small-scale plants, so it is reasonable to suppose that lower costs will be achieved in the future. The two MARKAL models reflect the costs of existing plant and these should be reduced for future technology vintages. The energy efficiency of liquefaction varies from 68% to 84%, with larger plants being more efficient (Figure 3). The two MARKAL models assume the efficiency of existing medium-sized plants. Since large, centralised liquefaction plants are most likely to be built in the future, using a higher liquefaction energy efficiency would be more appropriate. 6

7 Figure 2. Capital investment costs from the literature for hydrogen delivery technologies. 7

8 Figure 3. Energy efficiencies from the literature for hydrogen delivery technologies. 3.2 Road tanker transport Road tankers are currently used worldwide for fuel deliveries. Since the energy density by volume of liquid hydrogen is three times lower than that of gasoline, a much greater number of tankers would be required to deliver the same quantity of fuel. The relative fuel demand for hydrogen would be lower, however, as a result of the greater efficiency of hydrogen fuel cells compared to internal combustion engines. It is also possible that hydrogen production could be more decentralised than 8

9 current oil refinery production, reducing the distance between production plants and refuelling stations and so enabling road tankers to perform a greater number of sorties per year. Both the capital investment cost and the energy efficiency are determined by the average trip distance. The lower investment costs in Figure 2 assume an average round-trip of 200 km (240 deliveries per year), while the three higher costs assume a 1600 km round-trip (72 deliveries per year). The investment costs vary from 2.5 GJ -1 y -1 to 8 GJ -1 y -1 in 2000 and are forecast to reduce by 25% by For 200 km round-trips, the energy efficiency is virtually 100% (no losses). For 1600 km round-trips, the efficiency drops to 91%. The UK MARKAL model does not directly account for the road tanker investment cost. For transmission, it assumes an energy efficiency of 98%. The investment costs are simulated as a variable O&M distribution cost of 1.17 GJ -1, which is an output of the transmission model of Yang and Ogden (2007). The US9R MARKAL model has an optimistically-low investment cost ( 0.6 GJ -1 y -1 ) but pessimistically assumes an energy efficiency of 90%. 3.3 Ship transport Ships could be used to transfer hydrogen across the planet. Hydrogen ships are expected to be similar to the liquefied natural gas (LNG) ships that are widely used at present. Since ship transport suggests international energy transfers, the technology data is only relevant for the TIAM-UCL model. We only found ship data forecasts in the TECHPOL database (Krewitt and Schmid, 2004). Assuming that the ships make 33 round-trips each year, contemporary small ship investment costs would be 27 GJ -1 y -1 while larger high-capacity ships would cost only 3 GJ -1 y -1. Energy efficiency data were also only available from the TECHPOL database. All ships are forecast to have energy efficiencies of 95%, although this would depend on the transport distance. 3.4 Tube trailer transport Most existing hydrogen fuelling stations dispense fuel from compressed gas canisters that are delivered to the station. This is the most economic system for locations with low fuel demand, if the delivery distance is not too great (Yang and Ogden, 2007), and it might be the most suitable in the future in some locations. The principle advantage of tube trailer delivery is that it avoids the high liquefaction energy cost and high pipeline investment costs that affect other delivery systems. Tube trailer fuelling stations can be cheaper than other hydrogen fuelling stations because the hydrogen is dispensed directly from the tube trailer so little on-site storage is required. Fuelling stations tend to be small because a single tube trailer can store only kg and it is impractical to replace the trailer several times each day. High costs are the principle disadvantages of tube trailer delivery, particularly for long-distance deliveries. The lower costs in Figure 2 represent an average round-trip of 200 km ( deliveries per year) while the higher costs represent a round-trip of 1600 km ( deliveries per year). For long-distance delivery, investment costs are also much higher because a greater number of tube trailers are required to deliver the same amount of fuel. Investment cost reductions in the future will depend primarily on improvements to hydrogen storage technologies. 9

10 The tube trailer capacity varies between 250 kg and 500 kg in the literature. This has an important influence on the energy efficiency of delivery, particularly over long delivery distances. Figure 3 shows that tube trailers are particularly inefficient for hydrogen delivery over long distances when compared to road tankers, primarily because the cost of transporting a single load is very similar but the tube trailer load capacity is much lower than the typical road tanker capacity of kg. Tube trailers are not included in either the standard UK MARKAL or the US9R MARKAL models. They were included in the DfT version of UK MARKAL model but with zero investment costs. Instead, the model assumed an energy efficiency of 89% for transmission and simulated investment costs as a variable O&M distribution cost of GJ -1, which is an output of the transmission model of Yang and Ogden (2007). Tube trailers are competitive at short delivery distances, but if the demand is very high then pipelines are a more economic option. This leaves two principle roles for tube trailers in a hydrogen economy. Firstly, tube trailers can be the most economic technology during the transition to a low carbon economy when demand is low. Secondly, tube trailers can be the most economic technology for geographically-isolated fuelling stations in areas with low demand in a developed hydrogen economy. Tube trailers are omitted from many energy system studies because they occupy only a small niche in a developed economy that has little impact on model results. They should be considered, however, in a study of the transition to a hydrogen economy. 3.5 Pipeline transport Pipelines are the most efficient method of transporting large quantities of hydrogen, particularly over short distances. Almost 3000 km of hydrogen pipelines have been constructed since 1938 in Europe and North America (van der Zwaan et al., 2011). Transporting hydrogen through highpressure steel pipelines is more difficult than transporting methane because of hydrogen embrittlement, which makes strong steel pipes vulnerable to cracking, and because of hydrogen attack that allows reactions with the steel carbon atoms under certain operating conditions, again leading to cracks. Hydrogen has a lower energy density by volume than methane but a faster flow rate; this means that the total pipe capacity is around 20% lower for hydrogen than methane but the total hydrogen stored within the pipe is only a quarter of the total methane at the same pressure in energetic terms. Low-pressure hydrogen pipelines are not generally used for hydrogen except in a few niches, such as hospitals, but town gas, a mixture of hydrogen and carbon monoxide, was delivered at low pressure to buildings across the UK for 150 years. There is much more flexibility over the choice of pipeline material at low pressures. Pipeline investment costs can be split into four main categories: materials, labour, right-of-way fees and miscellaneous. Only the material costs are likely to differ from pipelines used for methane. One method is to estimate the cost as the equivalent methane pipeline cost + 20%, using the averaged data from Table 1 (although the H2A study (Steward et al., 2008) assumes a 10% cost increase). However, this approach does not account for the wide cost bandwidth for both types of pipeline in Table 1 that has been identified by van der Zwaan et al. (2011). Pipeline costs are affected by the diameter but also crucially by the topography, land use and labour costs as shown in Table 2; the most notable data in this table are the high cost of construction in the UK and in urban areas. These factors explain the high interannual pipeline construction cost variability over the last 30 years in the study of van der Zwaan et al. (2011). It is difficult to identify any long-term trends from their data but it seems likely that average construction costs have been stagnant or slightly rising over time. 10

11 Table 1. Pipeline construction costs and bandwidths for methane and hydrogen. Data from van der Zwaan et al. (2011). Gas Construction costs (kus$(2000) km -1 ) Construction cost bandwidth (kus$(2000) km -1 ) Methane Hydrogen Table 2. Location and terrain factors affecting methane pipeline construction costs. Data from Schoots et al. (2010), who reproduced it from IEA (2002). Country/region Location factor Terrain Terrain factor UK 1.2 High urbanisation Up to US$700,000 Europe 1.0 > 50% mountainous land 1.5 Australia/New Zealand 1.0 < 20% mountainous land 1.3 Japan 1.0 Cultivated land 1.1 US/Canada 1.0 Jungle 1.1 Equatorial Africa 0.9 Stony desert 1.1 Middle East 0.9 Wooded land 1.1 North America 0.8 Grassland 1.0 South America 0.8 South-East Asia 0.8 China/Central Asia 0.7 Indian subcontinent 0.7 Russia 0.7 South Africa 0.7 Two characteristics of pipelines complicate their representation in energy system models: (i) investment costs are proportional to the length of the pipeline; and, (ii) costs are dominated by high initial investment costs that are independent of the hydrogen throughput, so the utilisation rate is an important factor determining the economic viability if there is a long transition period to using hydrogen. These factors complicate pipeline simulation in energy system models because investment costs are normally assumed to be proportional to the energy throughput. A fixed lumpy investment can be used to correctly represent the deployment of a full pipeline but the size of the investment will depend on the spatial characteristics of the pipeline and must be calculated offmodel. This approach was taken by the UK MARKAL DfT project (Strachan et al., 2009). A further complication for representing pipelines is that the throughput is not fixed to the pipeline diameter. The throughput can be increased by increasing the pressure difference across the pipeline length but at a cost of greater energy consumption (Krewitt and Schmid, 2004). Table 3 shows how increasing the throughput in a pipeline reduces the energy efficiency but also substantially reduces the investment cost. This characteristic of pipelines, together with the wide range of pipeline diameters and lengths, explains the broad cost bandwidth of pipeline capital investment costs in Table 1. In practice, it is therefore necessary to optimise the pipeline diameter for the required throughput to balance investment costs and fuel consumption. This calculation is not trivial because hydrogen can traverse the pipeline as a laminar or turbulent fluid so the flow rate cannot be estimated as a simple function of the pipeline length and the pressure difference along the pipeline. 11

12 Table 3. Investment costs and energy efficiencies for 500 km pipelines of varying sizes and throughputs. Original data from Krewitt and Schmid (2004). Diameter (m) Throughput (kg d -1 ) Operations energy efficiency Total investment cost (EU( 2000) (kg d -1 ) -1 ) Total investment cost (EU m(2000) km -1 ) UK MARKAL investment cost (GB (2000) GJ -1 y -1 ) ,143 99% 53, , % 12, , % 14, , % 19, ,286 77% 1, ,286 95% 1, ,286 98% 1, ,000,000 93% Hydrogen pipeline costs from the literature Pipelines are one of the most difficult delivery technology to represent in energy system models because the capital investment costs and energy efficiencies depend on the pipeline length (and hence the spatial characteristics of the region), the diameter and the chosen throughput. For transmission pipelines, the UK MARKAL model currently makes broad assumptions about these factors based on Bossel et al. (2005). For distribution pipelines (within urban areas), all costs are represented using a variable O&M cost that is taken from the results of the transmission model of Yang and Ogden (2007). The DfT project calculated transmission pipeline costs using a spatiallydisaggregated approach but does not appear to have considered distribution pipeline costs (Strachan et al., 2008). One method of estimating the spatial characteristics of a UK hydrogen pipeline system is to assume that it would resemble the existing natural gas system. The national gas transmission network has a total length of 7600 km (National Grid, 2011a). Gas leaves the transmission network at 53 locations (National Grid, 2011a). Some large power generation and industrial consumers are supplied directly but most consumers receive low-pressure gas from one of thirteen regional mains distribution networks. These networks are much larger than the transmission network with a total length in 2010 of 280,000 km (ENA, 2010). This comprises approximately 12,000 km of high pressure pipes, 35,000 km of intermediate and medium pressure pipes and 233,000 km low pressure pipes (Transco, 1999). Service pipelines link smaller buildings to the mains distribution networks. They are the narrowest and shortest pipes in the system but since there are approximately 23 million of them across the country, they represent a substantial investment that is not borne by large gas consumers. It is necessary to make some simplifying assumptions to define and cost a pipeline system. Based on the existing natural gas pipeline system, we recommend that four types of pipeline are considered in the first instance: 1. Transmission large, high-pressure pipelines linking production plants to city boundaries. 2. High-pressure distribution regional pipes linked to the transmission network, supplying high-demand customers and the low-pressure network. 3. Low-pressure distribution smaller, low-pressure pipes that supply low-demand customers. 4. Service small pipelines that link customers to the low-pressure distribution system. 12

13 These pipeline types are also consistent with several pipeline cost studies. Fuelling stations would predominantly be connected to the high-pressure distribution network. Table 4 shows the investment costs of type of pipeline from these studies for low-urbanisation regions. Urban costs will be higher than these costs because right-of-way fees are substantially higher in urban areas, where roads must be dug up and relaid. It is often difficult to assess whether urban right-of-way costs are included in pipeline cost estimates. The urban uplift figure in Table 2 ($700,000) is an upper limit and costs will vary substantially according to the location. In a cost analysis of USA gas pipeline construction, Parker (2004) finds right-of-way costs of up to $150,000 for 15 cm pipes and up to $550,000 for 90 cm pipes. Kelly (2007) also finds that urban pipeline costs have a substantial cost uplift. Table 4. Costs of hydrogen pipeline deployment from a range of studies. These figures are valid for mixed lowurbanisation landscapes and would be higher in highly-urbanised areas. All figures have units US$(2000) (cm diameter) -1 (m length) -1 unless otherwise stated. Transmission Service Typical diameter (cm) TECHPOL (Krewitt and Schmid, ) Leighty et al. (2006) 12 Schmid (2006) H2A (Steward et al., 2008) van der Zwaan et al. (2011) 28 Pipeline costs based on the literature data in Table 4 are presented in Table 5. The UK costs are calculated by multiplying the non-uk costs by 1.2 (the UK location factor in Table 2). The urban cost increases are roughly estimated assuming the fixed cost increases in Table 5; actual increases will vary on a case-by-case basis. Energy system costs can be calculated from the recommended costs using the typical diameters shown in Table 5. The maximum throughput in each pipeline is shown in Table 5 for pipeline efficiencies between 98% and 100%. Efficiencies would reduce if the maximum throughput were increased. Actual energy system investment costs are not presented here because they are very sensitive to the design and operation of the pipeline network. Table 5. Recommended costs for energy system model calculations. All figures have units UK (2000) (cm diameter) -1 (m length) -1 unless otherwise stated. * High-pressure pipelines are assumed to be built predominantly in rural areas so do have an urban uplift. Transmission Highpressure distribution Lowpressure distribution Highpressure distribution Lowpressure distribution Service Typical diameter (cm) Maximum throughput (kg d -1 ) 1,700, ,000 2, Pipe capital costs ( m(2000)/km) Pipe capital costs ( m(2010)/km) Urban cost increase (US$(2000) km -1 ) +$500,000 +$300,000 +$150,000 $0 Urban cost increase ( m(2000)/km) Recommended pipe capital costs ( m(2000)/mm) 0.79* 0.44*

14 3.5.2 Hydrogen pipeline costs from UK natural gas experience A transmission pipeline was recently built to transfer gas from a new LNG import terminal at Milford Haven in West Wales to the transmission network in England, a total length of 316 km, at a cost of 1bn (ICIS Heren, 2007). We can use these data to estimate the cost of building a national transmission network: where c is a constant that accounts for the larger size of this pipeline (1.2 m diameter) compared to the average. Assuming c has a value of 75% and taking the capacity of the system from the highest recent quarterly throughput (1239 PJ in the first quarter of 2010) gives a capital cost for transmission pipeline investment of 2.57m/km in 2010 or 1.96m/km in This is substantially higher than the figure in Table 5. Part of the increase is likely to reflect the rough terrain that this particular pipeline crosses, particularly the Brechin Beacons National Park in Wales. Also, hydrogen pipelines might be expected to be 20% more expensive on average than natural gas pipelines. A national programme is replacing many of the original iron distribution and service pipes with new plastic pipes for safety reasons (HSE, 2001). This programme slowed in the late 1990s but was then accelerated in 2002 following several fatal incidents, with the aim of replacing all iron pipes within 30 m of buildings by 2030 (CEPA, 2011). The cost/km of building mains distribution pipes depends primarily on the diameter of the pipe (Figure 3). Costs for all pipeline sizes are higher in London than elsewhere but it is a reasonable approximation to represent the whole country with a single representative region. The cost of building the whole network depends on the proportion of each pipeline size within the network. This information was not available to us but the total length of each pipeline size being replaced in the period to 2030 is available from Frontier Economics (2011). We estimate the total pipeline length at each diameter using the assumption that proportion of iron pipes across the network is independent of the pipeline size. This assumption is likely to underestimate the proportion of larger pipes because these are less likely to be located within 30 m of a building and are more likely to have been built since 1970 to connect the regional distribution networks to the national transmission network that was constructed at that time. Most of these larger pipelines are likely to be operational until 2050 so our assumption is reasonable as long as substantial additional capacity is not added to the network before Under these assumptions, the cost of lowpressure distribution pipes is 0.13m/km and service pipes 0.08m/km. The low-pressure distribution cost is lower than that derived from the studies above but that might reflect that the pipes are being replaced rather than deployed, so existing fittings can be reused and the new pipes are often inserted inside the old pipes to minimise cost and inconvenience to the public. The cost of service pipes is similar using both approaches. 14

15 Pipeline cost ( /m) Figure 4. Cost of replacing pipelines in each UK mains distribution network as a function of the pipeline diameter (data from Frontier Economics, 2011) London South East Wales & South West East of England West Midlands North West North East & Yorkshire Scotland <3 4 to 5 6 to 7 8 to 9 10 to to to 24 over 24 Pipeline diameter (inches) Energy system costs for a hydrogen pipeline system for the UK The UK transmission network is designed to transport natural gas from the North Sea around the UK. Much of this natural gas is located to the north-east of Scotland and is brought on-shore at a large terminal at Peterhead then transported to the principle demand centres in England by four pipelines. Hydrogen, by contrast, would either be manufactured or imported close to the demand centres so a UK national hydrogen transmission network would be much shorter than the current gas transmission network; Strachan et al. (2008) estimate that all of the major UK demand centres could be linked with a total network length of 3467 km. We use this length for a complete hydrogen transmission network. For the distribution networks and service pipes, we calculate the capital costs using the current UK network statistics in Section We estimate the capacity of the systems using two methods: (i) estimating the single pipe throughput from Table 3, with an increase reflecting the assumed number of effectively independent networks in each case (6 for transmission and 24 for high-pressure distribution); and, (ii) by assuming a throughput of 80% of the maximum natural gas throughput from the last 10 years. One issue with the latter option is that gas throughput varies substantially between seasons so the annual capacity for current uses is substantially lower than the maximum capacity. In UK MARKAL, it is not possible to represent pipelines using timeslices so the gas network capacity is the annual average. If the hydrogen were being used for transport then the average would be close to the peak because demand is stable throughout the year; however, if the hydrogen were instead used for residential or service heat then the average would be much lower than the maximum and the data would need to be changed. This issue can be avoided by using a model such as TIMES that allows timeslicing for pipelines. We estimate the capital costs using the two approaches in Sections and This means we have two approaches to estimate the capital costs and two approaches to estimate the capacity for each type of pipeline. The UK MARKAL model requires the capital cost per unit of capacity and a range of credible values can be derived by combining these approaches. Table 6 shows the range and median of these costs for each type of pipeline. It is recommended that the median value be used in UK MARKAL and the range be used for sensitivity studies. 15

16 Table 6. Hydrogen pipeline capital costs for UK MARKAL in units (2000)/GJ/a. Capital costs for UK MARKAL Low Median High Transmission High-pressure distribution Low-pressure distribution Service Pipeline lifetime Hydrogen transmission pipelines in the UK MARKAL model currently have a 30-year lifetime. This is roughly consistent with investment accounting periods in the literature but does not necessarily represent the actual pipeline lifetime (for example, the first hydrogen pipeline in the Ruhr region of Germany, which was constructed in 1938, is still in operation). New PVC pipelines in the UK are expected to have a lifetime of 80 years (National Grid, 2011b). If pipelines were to have a lifetime of 80 years and were not deployed before 2030 in the UK (i.e. assuming pipelines are only used in a mature economy), then most of the benefits would be realised in the period Pipelines therefore have the capacity to lock-in hydrogen technologies for the twenty-first century Modelling hydrogen transmission pipelines as lumpy investments A hydrogen pipeline system cannot be built incrementally as demand increases; the pipeline length is fixed between the points of production and consumption and the capacity at commissioning is designed to satisfy the expected long-term demand so the utilisation rate is very low at first. In modelling terms, transmission pipelines represent a lumpy investment since the pipeline must all be built at once. In reality, the lumps have many different sizes because they depend on the spatial configuration of the supply and demand centres. In UK MARKAL, the lumps must all be the same size, although the model can build as many as it requires; a multi-region model is really required to improve the representation. The maximum capacity of the network in Section ranges from 760 PJ to 4000 PJ, and the supply and demand centres in the network are assumed to be located such that there are six effective pipeline routes. We therefore suggest a lumpy investment of 600 PJ for the transmission network, but note that there is a large uncertainty over this value and that a multiregional model is required to properly represent the flows of hydrogen around the country Transporting hydrogen using natural gas infrastructure Several studies have raised the possibility of distributing hydrogen using existing natural gas infrastructures to distribute hydrogen (Haeseldonckx and D'haeseleer, 2007; IEA, 2003; NaturalHy, 2010). Most studies have rejected the use of natural gas transmission pipelines for hydrogen as technically infeasible and have instead focused on low-pressure distribution networks. Two broad options are discussed in the literature: 1. Mixtures of hydrogen and natural gas. Mixtures with relatively low proportions of hydrogen (5 20% by volume) lower the overall carbon content of gas but do not require significant technical changes to existing installed technologies, including domestic appliances. Only minor carbon emissions reductions are achievable at such low blends. Gas separation technologies are expensive so separating pure hydrogen (e.g. for use in a fuel cell vehicle) from such a blend is not economic at present. 2. Moving entire local distribution zones onto hydrogen. Many existing local distribution zone (LDZ) pipeline networks are thought to be largely suitable for the distribution of pure 16

17 hydrogen, though technical uncertainties remain. If these are resolved, it would be possible to switch entire LDZs over to hydrogen, in much the same way as the transition from town gas to natural gas required the switching over of LDZs one by one. This would require wholesale replacement of end-user technologies (boilers, gas fires and cookers), as most of these appliances would not be able to run safely on pure hydrogen. Gas distributed through such networks would need to be odorized for safety reasons. This would mean that it would not be suitable for use in PEM fuel cells without purification or separation technology. As a result, it seems likely that the synergies of such an option with hydrogen transportation are not complete, i.e. even if there was hydrogen in the natural gas grid, this hydrogen would not be directly available for use as a transport fuel, as it would need to be cleaned and compressed for use in vehicles. The case for pursuing either of these two broad options rests on arguments about: (i) the relative costs of electricity and gas distribution infrastructure; and, (ii) the scope for interseasonal storage of hydrogen for heating, enabling reductions in peak energy supply capacity, against an all-electric scenario. If either of the above options were realised then markets for hydrogen could develop in both the residential and transport sectors, facilitating the establishment of a market for transporting hydrogen through the use of shared infrastructure (e.g. large pipelines) that could reduce overall costs in both sectors. The use of hydrogen in natural gas pipelines is not currently represented in UK MARKAL, but blends of hydrogen and methane are implemented within TIAM-UCL. Representing these options in energy system models is not straightforward. The benefits of hydrogen in natural gas grids are in part driven by temporal differences of energy supply and demand patterns, which MARKAL is unable to represent with any degree of sophistication. Much of the literature in this area focuses on the feasibility and possible implications of these options, rather than the likely costs and efficiency. This makes it difficult to identify data that can be used in models. Mixtures of natural gas and hydrogen could be introduced into UK MARKAL to represent the possibility of introducing small mixtures of hydrogen into the gas grid. One approach would be to define a new H 2 CH 4 energy carrier that uses existing infrastructure and becomes available at a time in the future. Another would be to add hydrogen as an input fuel to the distribution grid but with a constraint that it cannot exceed 10% of the gas supply. We recommend that these options are not added permanently until a sensitivity study has been performed in the model. Since MARKAL models are unable to simulate interseasonal hydrogen storage, one of the principle benefits of this option cannot be included in the model analysis. The TIMES model is able to represent interseasonal storage so this option would be more attractive in that model; for this reason, we recommend that the option continues to be included in TIAM-UCL. Transitioning the entire gas network to hydrogen would mean that existing residual capacity could be used by hydrogen. It would also mean that future pipeline investment beyond the residual life of current natural gas infrastructure should come at a similar cost to the natural gas pipeline costs in the model. This option would require the addition of hydrogen combustion end-use technologies, which would be essentially the same as natural gas combustion technologies but with a small cost increment (perhaps 5%). If the hydrogen were odorised throughout the distribution grid then it would be necessary to introduce a gas clean-up technology at refuelling stations to produce hydrogen of the required purity. We recommend that this option be tested in the energy system models in the short term; if it is frequently selected as the most economic technology then it will be necessary to consult with expert gas engineers on the technical feasibility of using hydrogen within current infrastructure before permanently implementing it in the models. 17

18 3.6 Fuelling stations Hydrogen is delivered to customers at fuelling stations. Stations can be split into two broad categories, according to whether the hydrogen is stored in liquid or gaseous form. Typical schematics of the two categories of fuelling station are shown in Figures 4 and 5. LH 2 is substantially cheaper to store than GH 2 because it has a much greater volumetric energy density and can be stored in a single tank, but fuel losses are higher due to boil-off (at temperatures above 253 C). Hydrogen liquefaction is expensive so LH 2 storage would only be economically-viable if the hydrogen were being delivered by LH 2 tanker. For GH 2 delivery or on-site generation, gaseous storage is cheaper because customers require high-pressure GH 2 for fuel cell vehicles. Figure 4: Liquid hydrogen refuelling station schematic. Image from Weinert (2005). Figure 5: Gaseous hydrogen refuelling station schematic. Image from Weinert (2005). LH 2 and compressed GH 2 are three and six times less dense than petrol, respectively, so both types of station will require more physical space and more equipment than is deployed at current petrol stations to supply the equivalent amount of fuel. This difference will be partially reduced by the higher efficiency of fuel cell vehicles relative to ICEs. Little storage is required for continuous GH 2 pipeline deliveries but bi-weekly LH 2 tanker deliveries would require a large LH 2 storage tank. Space limitations are likely to be most acute for stations with on-site hydrogen production. Current petrol fuelling stations are not suitable for hydrogen delivery and would have to be completely rebuilt. 18

19 Both types of fuelling station would require substantial capital investments. Figure 2 shows a wide range of investment costs for both LH 2 and GH 2 stations (small stations in these graphs have a throughput of approximately 500 kg day -1 while large stations deliver around 1500 kg day -1 ). The cost variability in the literature mostly results from three factors: (i) the design of the station (primarily the amount of on-site storage and the deployment of spare compressors to ensure high station availability); (ii) the cost of each station component; and, (iii) the assumed utilisation factor of the station (i.e. the ratio of actual delivery to potential throughput). The US9R MARKAL model has relatively low costs for both types of station while the UK MARKAL model does not consider the fuelling station cost Representing hydrogen fuelling stations in energy system models Representing fuelling stations in energy system models poses similar difficulties to pipeline simulation. It is necessary to define the fuelling station design(s) and to assume an appropriate utilisation factor which minimises the investment costs while taking into account consumer behaviour (consumers are not generally willing to organise their fuelling station visits so that the maximum utilisation rate is achieved throughout each day). We define four fuelling station designs in this paper, which are summarised in Table 7. We assume that the LH 2 station storage capacity can satisfy demand for six days. The GH 2 -GH 2 stations differ only in the amount of low pressure storage: adding low-pressure storage to a pipeline station improves the station availability during periods of pipeline outages, while on-site generation requires enough storage to meet demand for a full day. Table 7. Hydrogen fuelling station designs in this study. Study LH 2 -LH 2 tanker LH 2 -GH 2 tanker GH 2 -GH 2 pipeline (1) GH 2 -GH 2 pipeline (2) GH 2 -GH 2 on-site Maximum throughput (kg d -1 ) Number of cars per day Dispensers 3 4 High pressure GH 2 storage (kg) N/A 150 Low pressure GH 2 storage (kg) N/A Number of gas compressors N/A required Gas compressor throughput N/A N/A (kg h -1 per compressor) LH 2 storage (litres) 211, , Evaporator unit (kg d -1 ) N/A 1500 N/A N/A N/A Number of cryogenic compressors required Cryogenic compressor power (kw) N/A N/A N/A Table 8 shows the costs of hydrogen storage and compressors from several studies. The high TECHPOL storage costs are based on extrapolations of graphs from Simbeck and Chang (2002) that use out-of-date contemporary data so should be discounted. Some cost variations can result from using larger or smaller-capacity equipment for which the costs are not scalable. Table 8 also shows the costs in 2000 assumed by this project. Much research effort has recently been devoted to developing new cheaper storage technologies, primarily for vehicle applications. The GH 2 storage investment costs are therefore assumed to reduce over time at a rate of 2% per annum, which is consistent with vehicle storage cost estimates (e.g. McKinsey, 2010). Other hydrogen fuelling 19

20 station components are also expected to become cheaper in the future, but at a slower rate, so a 1% annual cost reduction was assumed for these. The costs of the fuelling station components in Table 8 do not include system costs such as permit applications, planning, site clearance and engineering. These indirect costs are accounted for in most studies by applying a percentage uplift to the direct costs. In literature studies this uplift is in the range 10 50% and we assumed a 25% factor in this study. The energy efficiency of both LH 2 and GH 2 fuelling stations generally exceeds 90% (Figure 3). LH 2 fuelling stations are more efficient, despite having higher hydrogen losses due to boil-off, because energy consumption by fuelling stations is dominated by electricity for compressors and the compressor requirements are much greater at GH 2 stations. If the stations had a very low throughput (by being located, for example, in a sparsely-populated area) then the LH 2 losses would be much higher than assumed here. Table 8. Costs of hydrogen storage and compressors from the literature. Storage costs are in US$(2000) kg -1. Compressor costs are in US$(2000) (kg h -1 ) -1. Study GH 2 storage LH 2 storage GH 2 compressor TECHPOL (Krewitt and Schmid, 2004) H2A Production (Steward et al., 2008) H2A Delivery (Steward et al., 2008) NextHyLights (Zaetta and Madden, 2011) 750 This study (costs in 2000) Annual learning curve cost reduction 2% 1% 1% 4 Technology data for energy system models We presented energy systems data for a range of hydrogen delivery technologies in Section 3. In this section, we summarise the currently-modelled technologies in UK MARKAL, recommend revised data for the future and consider uncertainties in the data and the potential for technology learning to reduce investment costs and increase operating efficiencies. 4.1 Summary of technologies in current MARKAL models The hydrogen production technologies currently represented in the UK and US9R MARKAL models are listed in Table 9. Both models include liquid road tanker and pipeline delivery modes but the representations are very different. The US9R model has investment costs for all technologies while the UK model uses a range of investment costs and variable costs or assumes no cost. Neither model includes tube trailers. Both models include on-site hydrogen production but the UK model does not account for the cost of the fuelling station. There are no future technology vintages for delivery technologies in either model. 20