Certain uncertainties in the design of a
hydrogen supply chain
A simulation study
Thomas G. Agterhuis
MSc. Technology & Operations Management MSc. Supply Chain Management
Supervisors: prof. dr. ir. J.C. Wortmann & dr. M. J. Land
Company supervisor: H. Zwetsloot Groningen Seaports N.V.
Date: 20 June 2020
Adriaan van Ostadestraat 131, 9718RT Groningen T.G.Agterhuis@student.rug.nl
S2732920
Word count (excluding tables, references & appendices): 8,046
ABSTRACT
LIST OF FIGURES AND TABLES ... 5
PREFACE ... 7
1. INTRODUCTION ... 8
2. THEORY ... 10
§2.1 The potentially most substantial problem of the late modern age ... 10
§2.2 Hydrogen ... 10 §2.3 Infrastructure ... 11 §2.4 Buffering ... 12 3. CASE ... 14 §3.1 The case ... 14 §3.2 System overview ... 14 4. METHOD... 16
§4.1 Determining the factors ... 16
§4.2 The simulation model... 18
§4.3 Storage tank costs ... 19
§4.4 Sub-questions... 19
§4.5 Data analysis ... 20
5. RESULT ... 22
§5.1 Core simulation results ... 22
CONTENTS
4
A: Interview protocol ... 45
B: Pipeline location and length ... 47
C: Electrolyzer schematic overview ... 48
D: Production and consumption distributions ... 49
E: Calculations ... 50
F: Model logic ... 52
G: Model assumptions. ... 55
H: Buffer size calculations ... 56
I: Simulation results ... 59
J: Production stops ... 66
5
LIST OF FIGURES AND TABLES
Figure 2.1 Schematic illustration of electrolysis (Zeng & Zhang, 2010) 11 Figure 2.2 Fuel consumption of a fuel station over a week (Reddi et al., 2016) 12
Figure 2.3 Seasonal gas demand pattern (Vattenfall, 2020) 12
Figure 3.1 Supply chain with only chemical customers (a); and various customers (b) 15
Figure 4.1 Model logic 18
Figure 5.1 Buffer behaviour over 1 year (a); 10 years (b) 22
Figure 5.2 Buffer behaviour without flexibility (a); with 5% flexibility (b) 23 Figure 5.3 Effect of flexibility on buffer requirements and standard deviations 23 Figure 5.4 Buffer behaviour with backup suppliers at a 24h (a); 18h (b); and 12h (c)
threshold
24
Figure 5.5 Buffer behaviour with venting 24
Figure 5.6 Buffer behaviour for 1 chemical customer with 1% (a); 2% (b); and 5% overcapacity (c)
25
Figure 5.7 Stable buffer behaviour by means of flexibility (a); overcapacity & venting (b); and backup suppliers & venting (c)
26
Figure 5.8 Stable buffer behaviour by means of 1% overcapacity (a); 2% (b) for fuel stations, and 1% overcapacity (c); 2% (d) for chemical demand
28
Figure 5.9 Buffer behaviour for household demand for 1 year 28
Figure 5.10 Relative buffer size for multiple electrolyzers 29
Figure 5.11 Buffer behaviour for 132 failures of 1 hour (a); 4 failures of 33 hours (b); 2 failures of 66 hours (c)
30
Table 4.1 Factors used in model 16
Table 4.2 Production and consumption distributions. 17
Table 4.3 Model assumptions 18
Table 4.4 Model parameters 19
Table 5.1 Buffer minima per threshold 24
Table 5.2 Buffer requirements per overcapacity level 25
LIST OF FIGURES AND TABLES
6
Table 5.4 Buffer requirements per configuration with associated costs 26 Table 5.5 Buffer requirements with associated costs for fuel stations and chemical firms
with overcapacity
7
PREFACE
1. INTRODUCTION
8
1. INTRODUCTION
In the academic world, there has been consensus about the negative effects of burning fossil fuels (Intergovernmental Panel on Climate Change, 2018). The Dutch government wrote in the Climate Accord that their goal is to reduce national emissions with 49% by 2030 and become climate neutral in 2050 (Ministry of Economic Affairs and Climate, 2019). Apart from environmental effects of burning fossil fuels, their availability is finite. Therefore, a transition to RES like wind and solar is inevitable.
There are, however, problems with RES. For example, intermittency of supply and the non-storability of electricity. Storage is needed to match supply of electricity with demand (Ellis & Nazar, 2012). Converting (excess) electric energy into hydrogen has the potential to solve these problems (Van Wijk & Hellinga, 2018). However, as Rusin & Stolecka (2017: 153) state: “[For a future hydrogen-based economy] the sites of the gas production as well as the methods of storage, distribution and application will have to be determined”. Currently, hydrogen is mostly used as a feedstock in the chemical industry and predominantly produced from natural gas (Mulder, Perey & Moraga, 2019).
This study will provide an assessment of a hydrogen supply chain, with one supplier, and a few customers with varying demand patterns. In this system, there is a challenge to match supply with demand, as both production and consumption will be stochastic in nature. Aiming for a high delivery reliability can lead to exploding buffer costs1. This is where interesting operation
management questions arise as there will be trade-offs between buffer size, customer type, overcapacity and backup suppliers. For example: how much buffer capacity is needed when promising to deliver 100% of the production capacity, how much for 95%, and is this even possible? This will depend on variations in supply and demand. These variations will be amplified if customers are e.g. households instead of chemical firms, due to their differing consumption patterns. The central question to be addressed in this study therefore is:
How do the buffer requirements in a hydrogen supply chain change with variations in supply and demand?
1 The 20MW electrolyzer studied in this research produces 8,200kg 𝐻
2 per day, a 1,000kg storage tank costs ~€1
9
This question will be answered by building a simulation model of a hydrogen supply chain, where data are used from a real life proposal for a large scale hydrogen system based on electricity from RES. First, a base-case will be modelled. Then variation will be inserted by using empirical data to simulate the behaviour of a chain. A cost aspect is added to evaluate designs. For example, when delivering to neighbourhoods instead of chemical firms, buffer costs will increase. By how much does it increase, and how high should the mark-up to account for this be? Data are acquired via interviews with stakeholders of the system.
This paper studies what Kang & Bhatti (2019) call buffer management. They refer to Umble & Umble (2006: 1023) who stated: “the primary concern remains to guard the system against expected (e.g. variability induced due to customer demand), and unexpected (e.g. machine failure) disruptions”. They studied manufacturing settings, where OpEx (holding costs) dominate CapEx (inventory space), but in the hydrogen setting CapEx (tanks) outweigh OpEx (production costs).
The contributions of this study will be twofold. Firstly, (gas) buffering is a well-studied academic subject, but its linkage to a hydrogen supply chain is new. By using simulations with realistic operating conditions, it will be possible to observe the behaviour of the system with different configurations, subject to uncertainty. Secondly, the companies involved in the hydrogen chain will benefit from the managerial guidelines provided by this study. They are still in the design phase of the system and therefore want to be knowledgeable to leverage the system to create most value.
2. THEORY 10
2. THEORY
Nomenclature Acronyms CapEx 𝐶𝑂 𝐶𝑂2 CSS Capital expenditures Carbon mono-oxide Carbon di-oxideCarbon capture & storage
FTL GW(h) MW(h) N𝑚3 OpEx RES SD
Full truck load Giga Watt (hour) Mega Watt (hour) Normal 𝑚3
Operational expenditures Renewable energy source(s) Standard deviation
In this chapter, academic literature will be presented. The problem will be grounded, the role of hydrogen will be explained and the missing links mentioned. The constructs infrastructure and buffers will be clarified.
§2.1 The potentially most substantial problem of the late modern age
The notion that humanity needs to move away from fossil fuels has long been known (Fourier, 1827). To do this, we need to increase the use of RES. RES, however, have their problems, namely intermittency and the need for storage (Blarke, 2012): logically, weather does not match energy consumption patterns and electric energy cannot be stored without a medium. Therefore, during a mismatch of supply over demand, electrical energy needs to be converted and stored to lower the costs of oversupply (Su, Kern & Characklis, 2017). Vice versa, during a mismatch of demand over supply, energy will be taken from storage to satisfy demand.
§2.2 Hydrogen
11
Hydrogen can be produced (among others) by reforming natural gas, gasifying coal, and electrolyzing water (Steinberg & Cheng, 1989). When the electricity powering the electrolyzer comes from a RES, it is labelled green. When it comes from a fossil fuel plant, or is made out of gas or coal, it is labelled grey. When the 𝐶𝑂𝑥 emitted during the production of hydrogen is captured and stored (CCS) or
deployed in useful applications, it is called blue (Mulder et al., 2019). See figure 2.1 for an illustration of electrolysis adapted from Zeng & Zhang, 2010.
Hydrogen has a lot of potentials, nonetheless, there are issues to be solved. At the production side, only green hydrogen will solve our emitting problem as grey is made from fossil fuels, CSS can only capture between 60-90% of emissions (Bui et al., 2018), and green hydrogen production is still small scaled and requires a lot of investments (Mulder et al., 2019). Green hydrogen does not mean the electrons flowing in are produced by a RES. With the production of green energy, so-called green certificates are created which are sold to energy consumers (Bergek & Jacobsson, 2010).
Tackling the obstacles of RES has been studied intensively. Researchers have developed ingenious ways to build (stand-alone) energy systems with renewable production and hydrogen for storage of oversupply (e.g. Anifantis, Colantoni & Pascuzzi, 2017; Ishaq, Dincer & Naterer, 2018).
§2.3 Infrastructure
When building a hydrogen supply chain, infrastructure requires high investments: according to Minnee (2019) infrastructure, transportation and storage costs can be considered two to three times higher than the electrolyzer’s costs. Moreno-Benito, Agnolucci & Papageorgiou, (2017) stress the importance of aligning energy conversion, distribution, and storage. For the infrastructure, the specific components from supply to demand in this study will be the electrolyzer, pipelines, storage units, and customers.
Electrolyzers are not a new invention, they were already used on an industrial scale during WW1 to produce poisonous Chlorine gas by Germany (Everts, 2015). Pino, Valverde & Rosa
FIGURE 2.1
2. THEORY
12
(2011) studied the effects of operational factors on the hydrogen production of an electrolyzer and found that thermal inertia, operating temperature, response time to changes in input power influence the production rate. Other operating factors, like failures, maintenance and degradation also influence the rate. It is therefore clear that the output of an electrolyzer is subject to variation. High pressure storage of hydrogen is mostly done in tanks. Different types of tanks exist which are selected based on costs and technical aspects as weight, pressure, and portability (Barthelemy, Weber & Barbier, 2017).
The types of customers studied in this paper are chemical firms, hydrogen fuel stations (for mobility), and households (for heating). Each type has different demand patterns. Chemical firms are relatively stable in demand, fuel stations follow a very specific distribution as found by Reddi, Mintz, Elgowainy & Sutherland (2016) (figure 2.2). Household consumption distributions are high in the morning and evening, and show a clear seasonal pattern (figure 2.3).
§2.4 Buffering
Buffering in a gas supply chain can be compared to warehousing in a goods supply chain. Warehouses fulfil their role by keeping inventory to manage variations in supply and demand (Sainathuni, 2014). One can reduce chances of running out of stock by keeping high inventories, but as said, for a hydrogen supply chain, this leads to substantial capital expenditures.
When operating a gas infrastructure, there is short term variability from supply and demand. In a natural gas system, these fluctuations are handled by the high inventory of gas in the pipes (Alabdulwahab, Abusorrah, Zhang & Shahidehpour, 2015). Also, because of the high number of consumers, the variability they pose cancel each other out and the overall demand is very predictable. When this would not be the case, small changes in supply or demand without buffering could lead to malfunctions. Additionally, there is long term variability from demand, for example, the winter spike in demand from buildings for heating. To be able to keep a steady supply to customers, buffering is needed to overcome the variation.
FIGURE 2.2
Fuel consumption of a fuel station over a week (Reddi et al., 2016)
FIGURE 2.3
13
Apart from storage in tanks, there is line packing. This is the amount of gas in the pipelines minus the cushion gas, which is the gas needed to achieve the minimal pressure (Tomasgard, Rømo, Fodstad, & Midthun, 2007). Large scale buffering can be done in empty salt caverns. This is currently already done to handle the winter natural gas demand (Mulder et al., 2019). They state that a salt cavern can hold 150GWh of hydrogen, which is around 55 million N𝑚3, or two year’s production from a 20MW electrolyzer. Kruck, Crotogino, Prelicz, & Rudolph (2013) estimate the costs hereof to be around €35 million. The lifetime costs per storage unit of salt caverns is deemed much lower than tanks (Michalski et al. 2017). Also, opportunities arise with the de-scaling of natural gas extraction in The Netherlands. The vast amounts of high capacity transmission pipes in the north are being phased out and can be used for storage of hydrogen.
Furthermore, there is another variant of buffering taken into account: backup suppliers. When supply is low for an extended time, backup suppliers can be used to acquire hydrogen. This can be done via compressed gas trucks (tube trailers) or liquid hydrogen trucks. In this study, only compressed gas trucks will be considered as handling liquid hydrogen requires high capital investments (Demir & Dincer, 2018).
Many studies integrate hydrogen in a hybrid energy system to increase efficiency or lower levelized costs (e.g. Eriksson & Gray, 2017). However, most studies do not take into account variations in performance of the system itself, and “to enable optimum performance, uncertainties should be considered during the design stages of such systems” (Giannakoudis, Papadopoulos, Seferlis, & Voutetakis, 2010: 874). The assumption that an electrolyzer has an overall equipment effectiveness of 100% is often made and the problem studied is to match RES production patterns with demand. As Baghaee, Mirsalim & Gharehpetian. (2017: 1758) state: “Unfortunately, there is no information about reliabilities of electrolyzers”, and therefore they neglect it. This assumption is dropped in this study. Overall, a hydrogen supply chain has variations in supply. Also, there are different types of demand, all with their own patterns and variation. Buffering is needed to overcome this, but is expensive. What then, are the effects of variations in supply, combined with stochastic demand patterns on system design and associated buffer costs?
3. CASE
14
3. CASE
§3.1 The case
In the North of The Netherlands, a coalition has emerged with government, businesses and academics (Noordelijke Innovation Board, 2017). Their aim is to convert the harbours of Delfzijl and Eemshaven into a hydrogen hub. Recently, the European Union appointed a subsidy of €11 million to the hydrogen pilot project DJEWELS from Nouryon and GasUnie. They will build a 20MW electrolyzer which will be used to produce hydrogen as a feedstock for local chemical factories. This electrolyzer will be the largest in Europe (Dagblad van het Noorden, 2020). The connection to factories nearby will be provided by a network operator. The pipe network will connect up to six chemical firms in the short term and more on the long run to use hydrogen as a feedstock. The backbone will be used as a pilot to consider extending the network to more firms, to hydrogen fuel stations, and to neighbourhoods to be warmed with hydrogen. The supply chain that will arise is the unit of analysis in this research. It offers a unique chance to answer the research question with empirical data and expert opinions.
Currently, the chemical customers acquire their hydrogen via parties who produce it with steam reforming, which is delivered by truck. The fuel station is supplied by a local chlorine factory where the residual product is hydrogen. The neighbourhoods are mostly warmed with natural gas, coming in via pipelines.
§3.2 System overview
15
FIGURE 3.1
Supply chain with only chemical customers (a); and various customers (b)
households in an extended system. In figure 3.1a a schematic overview of the initial chain is given. In 3.1b the extended chain is shown.
4. METHOD
16
4. METHOD
To answer the research question, a simulation model has been built to study the behaviour of the proposed hydrogen supply chain and assess its buffer requirements. Literature and expert interviews were used to determine the factors to include in the simulation model. In this model, real performance data of an electrolyzer was used to simulate hourly hydrogen production. To simulate demand, interview, literature and historical data were used. Total yearly demand was set equal to production. Whilst the case has certain characteristics, multiple values for design parameters were used to create various configurations to increase the generalizability of the results (see §4.2 for these parameters). By adding financial measures to the system’s design, cost effects were studied. The output of the study is quantifications of the effects of various customer types and the effects of stabilizing mechanisms on buffer requirements.
Following the line of reason of Bertrand & Fransoo in Karlsson (2016: 311) a simulation is the most appropriate method when “the problem is too complex to solve with mathematical analysis” which holds for this case.
The method consists of five phases. Primo, literature and additionally interviews with practitioners are combined to determine the factors in the system. Secundus, the simulation model and assumptions are proposed. Tertium, the costs of storage tanks are given. Quartus, sub-questions are provided. Ultimatim, the data analysis is covered.
§4.1 Determining the factors The modelled factors were derived from literature and verified by experts. Additionally, the experts were asked to propose additional factors and determine where to neglect them to
make certain
assumptions. The
TABLE 4.1 Factors used in model
Category Factor Value Source
Pipeline* Dimensions* 122mm x 5km Avery et al. (1992) Pressure range* 20-42 bar Mukherjee et al. (2015) Flow capacity* 0.175 N𝑚3
/s/bar Avery et al. (1992)
Electrolyzer Nominal power 20MW Buttler & Spliethoff (2018) Output load See appendix D Idem
Overload Not possible Idem
Customers Demand load See appendix D Avery et al. (1992) Flexibility 0-5% Network operator
Buffer Size Unlimited Idem
Backup supplier
Delivery time 24h Kim, Lee & Moon (2008) Capacity 13,200 N𝑚3 Idem
Increments FTLs Idem
Overall Time Hour Avery et al. (1992)
17
interview protocol is given in appendix A. In table 4.1, an overview of the categories and factors used is given.
The factors with an asterisk are not modelled but when the flow capacity becomes a constraint, this is mentioned. The dimensions and pressure range are used for calculations after running the simulations. To verify the factors, two interviews were held. One with a pipeline expert (49 minutes) and one with the sales manager (71 minutes), both from the network operator.
The pipeline to be constructed will be a circle in the chemical park (appendix B). Given its dimensions, the total volume will be 58𝑚3. The maximum pressure is 42 bar with a minimum cushion pressure of 20 bar. As the hydrogen flows with 15m/s, 0.175 N𝑚3/s/bar can be transported. The electrolyzer is a 20MW alkaline electrolyzer. See appendix C for a schematic overview of this system. The electrolyzer produces 3,000 tons per annum. Data on the operational output rate of an electrolyzer is acquired via the network operator. The listed output rate is 4,110 N𝑚3 per hour, however, the operational data shows that the total effectiveness is not 100% Therefore, the producer can only promise to deliver the real output. Overloading this type of electrolyzer is not possible.
Information on the chemical firms is also acquired via the network operator. These firms do not have long term variability in demand as they have a 24/7 continuous process with a steady demand (apart from accidents and production stops). According to the network operator, 132 yearly hours
of unplanned downtime can be considered realistic. Chemical firms are expected to be able to offer flexibility in their demand. This is not possible for households and fuel stations. Demand patterns from households are based on the average gas consumption patterns of 10,000 households in The Netherlands (Liander, 2009). Reddi et al. (2016) is used for demand patterns of fuel stations. As supply and demand are set equal, the production and consumption patterns all have an hourly mean of xxxx N𝑚3. In table 4.2, an overview of the standard deviations per pattern is given. In appendix D the distributions are visualized. All data are per hour.
The buffer is modelled as one large tank installed on the backbone. The location where the gas enters is not taken into account. The economic lifetime of the tank is considered to be 20 years based on Boudries (2018) which is confirmed by the pipeline engineer to be an industry standard.
Type Standard deviation
Electrolyzer 954
Chemical 444
Fuel station 2,223
Households 4,200
TABLE 4.2
4. METHOD
18
Delivery conditions for backup suppliers will be very case specific and therefore have a low generalizability. When transporting compressed gas in trucks, the capacity depends on the pressure. According to Lahnaoui, Wulf, Heinrichs, & Dalmazzone (2018), there are 4 main design pressures for composite tube trailers: 250, 350, 500, and 540 bar. These trucks contain 720, 907, 1,100, and 1,350 kg respectively. This is confirmed by Reddi, Rustagi, & Gupta (2018) who found that 1,100kg is the maximum. Therefore, in this study, a capacity of 1,100kg (13,200 N𝑚3) is used.
In appendix E an overview of used calculations is provided.
§4.2 The simulation model
The model is built in Python using the simulation tool SimPy. SimPy is a discrete-event framework. Therefore, hydrogen will be simulated as units in real numbers. In figure 4.1, the model logic is visualized.
The initial buffer level is set to 10 million to make it never go below zero as the software does not support a negative buffer. In the analysis, this is subtracted again. The buffer capacity is set to infinity.
From the electrolyzer data, a random production amount is picked each hour. The customers all have their own patterns and for the chemical firms these are also randomized. The residual of supply and demand is taken from, or put into the buffer. Then an optional decision will be made to acquire additional hydrogen from a backup supplier or to vent the surplus. Customer flexibility determines to what extent demand can be lowered when inventory is low (or increased when it is high). The buffer size will vary over time and is the output of the model. See appendix F for an in-depth explanation.
While constructing the model, certain assumptions are made for simplification, scope and due to a lack of data. These assumptions are given in table 4.3 and motivated in appendix G.
TABLE 4.3 Model assumptions
(1) The hydrogen behaves as a perfect gas in the system; (2) The locations of parts are irrelevant;
(3) Supply and demand occur at the exact same moment;
(4) There are no additional infrastructural costs per customer type; (5) The tank faces no scalability limits.
19
In table 4.4, an overview of the changing experimental variables and their values is given. Not all possible configurations will be tested, as this would lead to 4,608 different configurations, without varying the amount of hydrogen going to the various types of customers. Selections are made to study the isolated effect of the factors given in the sub-questions in §4.4.
§4.3 Storage tank costs
The costs of buffer tanks are based on work from Kalinci, Hepbasli & Dincer (2015), Baldwin (2017), and Kharel & Shabani (2018). These authors found $574, $590, and $438 per kg of hydrogen capacity respectively. Logically, a linear formula for tank costs will not hold in reality, but for calculating purposes, $500 per kg of tank capacity is used.
§4.4 Sub-questions
To answer the research question, several sub-questions have been formulated (partly based on questions posed by the case company). The first six are the core of this study:
1. What is the effect of various demand patterns on buffer requirements?
a. What is the effect of having flexibility in chemical demand on buffer requirements? 2. What is the effect of using backup suppliers on buffer requirements?
3. What is the effect of venting hydrogen on buffer requirements? 4. What is the effect of production overcapacity on buffer requirements? 5. What is the effect of production stops on buffer requirements?
6. What are the cost effects of the above factors?
a. How much should the buffer tank mark-up per customer type be?
Four more have been formulated for the sensitivity analysis:
1. What is the effect of variations in demand from chemical firms on buffer requirements? a. What is the effect of having multiple chemical customers on buffer requirements?
TABLE 4.4 Experimental variables
4. METHOD
20
b. What is the effect of having multiple electrolyzers on relative buffer requirements?2 2. What is the effect of variations in demand from fuel stations on buffer requirements? 3. What is the effect of variations in demand from households on buffer requirements? 4. What is the effect of correlated downtimes on buffer requirements?
§4.5 Data analysis
The production and consumption amounts lead to a varying buffer size over time. The model keeps track of all factors and writes them to a file. Then, the behaviour of the buffer is studied. The results are 20 runs per configuration. In appendix H, a visual explanation of the buffer size assessment is given. To determine how large the buffer needs to be, the minimum and maximum per run are determined. As the initial buffer level will influence the buffer size required, this needs to be accounted for. The assumption is made, that the most suitable initial value is a half-full buffer. Therefore, the fitting buffer size is the largest of the maximum or absolute minimal value times two, as that largest value will be the half-full level to be able to manage the most negative or most positive values. Standard deviations are used to calculate the distribution of the buffer sizes between the boundaries.
However, with production overcapacity and venting, the system does not aim for a half-full buffer. The overcapacity will make the buffer fill over time. To counter this, venting is used to release hydrogen at a certain point. The system will thus aim for a full buffer, and the above method for calculating the buffer size is inadequate. In those configurations, if the venting level is larger than the absolute most negative buffer value, there is no need for twice the maximum in buffer capacity. This is because the buffer will, on average, go up to the venting level and release the surplus from the backbone. Therefore, the appropriate buffer size is the venting level plus the lowest buffer level (if this is negative).
Additionally, as the buffer starts at zero in each run, a warm-up period of 1,000 hours is included for the numerical analysis, in the graphs this is illustrated with a red vertical.
All simulations are run with 5% demand variation (unless stated otherwise) to add realism. When system stability has to be assessed, the runtime is increased from 8,760 to 87,600 hours. The inter-year results are then compared (appendix I). If these do not significantly differ (or do not
21
increase), stability can be assumed. Visually this can be done by observing if the lines in the graphs diverge as time goes on. Scales are kept identical for graphs that have to be compared.
A sensitivity analysis is performed to determine the vulnerability of the results to certain variables. Finally, a storage opportunity in a phased out natural gas transmission pipe will be researched.
5. RESULT
22
5. RESULT
In this chapter, the main results will be elaborated on. For an overview of all simulation outcomes, see appendix I. When interpreting the results, some caution is necessary. The output of the simulations are the sizes of the buffer per run in N𝑚3. Which is visualized in figure 5.1a for a single electrolyzer and one chemical customer. Each colour depicts one of 20 runs. As the buffer size is the result of various stochastic variables (hydrogen production & demand), simulations would result in unlimited positive and negative buffers as runs and time go to infinity, with most of the runs lying in between the extremes. This behaviour can be observed when the simulation is run for 10 years ceteris paribus
(figure 5.1b). This means that in order to create a stable system, there must be feedback mechanisms (which will be discussed later). In the graphs, the buffers become negative, this is not possible and means that the most negative value should at least have been the initial buffer level.
The values are rounded to 1,000 N𝑚3.
§5.1 Core simulation results
In this section, output of the simulations will be provided per configuration.
Buffer sizes per customer type. For chemical demand and fuel stations the buffer requirements are in the same order of magnitude. For households it is extremely high. When the three customer types are combined (i.e. each customer gets one third of the production), the effects of households dominate the others.
FIGURE 5.1
23
Buffer sizes for chemical customers with flexibility. It is assumed that chemical firms can offer flexibility in their demand, especially when there would be a financial incentive to do so. When increasing the flexibility, multiple changes occur. The effects of no and 5% flexibility are visualized in figure 5.2. It is however not correct to compare specific results of a non-stable to a stable system. Therefore, when comparing 1 to 5% flexibility, the buffer requirements decrease from 228,000 to 72,000 N𝑚3. Also, the standard deviations of the runs decreases massively, from around 66,000 to 5,000 N𝑚3.
These decreases hold also for 2% flexibility, as can be seen in figure 5.3. When comparing the three graphs, flexibility clearly stabilizes the system.
Backup suppliers. According to the pipeline engineer, it is very case and contract specific what lead times and cost aspects for ordering hydrogen by trucks are. In accordance with her, a lead time of 24 hours is
proposed. As the backup is used to counter short term variability, it not used in the household simulations. When this would be implemented, the program will order a lot of the winter demand at backup suppliers, which would ask tremendous capacities. In the simulation, no costs are considered, so the ordering of trucks will not be an economic decision. When the buffer level drops beneath a certain threshold in hours of production capacity, the system will order that deficit of hydrogen with increments of full truck loads. Backup supply will lead to deviations between overall production and consumption, as these are approximately equal and the backup hydrogen will be an addition to the production. This means that the chances of ordering a backup truck will decrease with each one requested. The simulations show, that when using a demand threshold equal to the lead time, i.e. 24 hours, backup orders are never placed more often than once in a run. When
FIGURE 5.2
Buffer behaviour without flexibility (a); with 5% flexibility (b)
FIGURE 5.3
5. RESULT
24
lowering the threshold to 18 and 12 hours, the amount of runs with usage of backup increases, the number of times ordered increases, and the amount ordered each time decreases. The amount ordered is based on the threshold and truck capacity. One 20MW electrolyzer produces 4,110 N𝑚3 per hour, and one FTL contains approximately 3 hours of production (13,200 N𝑚3). Therefore, when the buffer drops beneath the thresholds 24, 18 or 12 hours, 8, 6 or 4 trucks will be ordered respectively. In graph 5.4, the effect of backup suppliers is given for runs where orders took place. It is clearly visible that it stabilizes the minima in the system.
In appendix I, there is an overview of the effects of backup suppliers on multiple electrolyzers. As the amount of hydrogen produced increases with the electrolyzers, the chances of reaching the threshold drop. Therefore, when having multiple electrolyzers, the necessity of backup suppliers decreases, but its usage will increase as the effect of a FTL on total production becomes lower.
In table 5.1, the effect on the lowest values of the buffers are given for the thresholds studied. These values indicate the minimum initial buffer level required to make these runs possible. Therefore, with a low threshold, more trucks are ordered, and less initial buffer is needed, but the costs of backup will increase.
Venting. One can imagine that when the buffer is full and production is higher than demand, hydrogen has to be vented. An (emergency) venting
system will always be present in a gas infrastructure (appendix C). In figure 5.5, the effect of venting can be seen. By letting gas out of the system, a very specific top on the buffer is instigated. Apart from directly venting hydrogen, selling ad-hoc and turning down the
Threshold Minima
24h (-98,640 N𝑚3) -106,000
18h (-73,980 N𝑚3) -72,000
12h (-49,320 N𝑚3) -53,000
TABLE 5.1 Buffer minima per threshold
FIGURE 5.4
Buffer behaviour with backup suppliers at a 24h (a); 18h (b); and 12h (c) threshold
25
electrolyzer will also be possible. These are not considered in this study, as details on their suitability and possibility are unknown.
Overcapacity. It is clear that venting can stabilize the filling of the buffers. However, this does mean that the longer the timeframe gets, shortages will still be a problem. Therefore, the
system is only half-stable. The bottom part can be stabilized by production overcapacity. Overcapacity means that the electrolyzer still runs at full capacity, but less is sold than produced to lower the chances of shortages. In table 5.2, various degrees are compared. The specific effect of venting is dependent on the level at, and increments with which it is done and therefore these numbers can only be compared to each other. It is clear that the average minima increase with overcapacity. Also, the buffer requirements and the standard deviations decrease. With 5% overcapacity the buffer equals the venting level (49,315 N𝑚3) as the buffer never becomes negative after the warm-up period. Also, it is clear that stability can be achieved with 1% overcapacity.
The effect is illustrated in figure 5.6 and should becompared tofigure5.5.
§5.2 Stable systems
As could be seen in the first paragraph, certain factors increase or decrease the buffers and affect the variability therein. However, some of them are not stable. In this section the mechanisms to achieve stability will be discussed. In table 5.3 the mechanisms and their effects are given.
Ideally, one would compare the available mechanisms for a particular supply chain and optimize which combination would minimize a certain performance indicator. But, as said before,
Overcapacity SD Buffer size Minima
1% 18,000 118,000 -15,000
2% 7,000 57,000 9,000
5% 3,000 49,315 28,000
TABLE 5.2
Buffer requirements per overcapacity level
Mechanism Affects Backup suppliers Minima Venting Maxima Overcapacity Minima Customer flexibility Both TABLE 5.3 Stability mechanisms FIGURE 5.6
5. RESULT
26
the terms, costs and possibilities for backup suppliers depend on lots of factors. The suitability of venting, selling ad-hoc or turning down the electrolyzer is case (and time) specific and depends mostly on the buffer size, market factors and electrolyzer characteristics. The flexibility of customers has been studied by Minnee (2019) and Radstaak (2019) for this specific case. According to them, chemical firms are not keen on offering this. However, during the interviews it became clear that certain firms should be capable of offering flexibility.
In figure 5.7, it is shown that customer flexibility, backup suppliers and overcapacity (combined with venting) lead to stable systems. The graphs visualize simulations which ran for 10 years. In appendix I, it can be seen that 1% flexibility, or 1% overcapacity with venting are already enough to attain system stability. Using backup suppliers does require larger buffers to cover for the thresholds and lead times.
§5.3 Buffering costs
Mark-up for chemical firms with flexibility. When comparing the buffer requirements for a single electrolyzer with one chemical customer and add the tank cost to it, the following becomes clear (see table 5.4). The buffer requirements are in N𝑚3. There are 12 N𝑚3 in a kg, and a kg capacity costs $500. The costs per N𝑚3 are the tank costs divided by the net 20-year production. The value per N𝑚3 in the last column is the marginal value of the additional flexibility. It can be seen that the increases from 1 to 2% is worth 0.56 cent per N𝑚3 and from 2 to 5% 0.48 cent. From the buffer requirements, 1,285 N𝑚3 can be subtracted, as this is the amount of linepack capacity, and saves $54,000 in tank costs.
Flexibility Buffer size Tank costs Costs per N𝒎𝟑
Value per N𝒎𝟑
1% 228,000 $9,449,000 $0.0152 -
2% 146,000 $6,012,000 $0.0096 $0.0056
5% 72,000 $2,964,000 $0.0048 $0.0048
TABLE 5.4
Buffer requirements per configuration with associated costs FIGURE 5.7
27
Mark-up for chemical firms and fuel stations. To be able to compare chemical demand to fuel stations, flexibility is not suitable as fuel stations cannot offer this. Also, backup supplier costs are too situation specific. Therefore, to make comparisons, a stable configuration with overcapacity and venting are studied to find amounts vented, and buffers required. The results are given in table 5.5 and their behaviour is visualized in figure 5.8. One could argue that the chemical system with 1% shows occasional extreme minima (figure 5.8c). Stability has however been proven with 1% overcapacity (see appendix I). The variance from chemical demand is lower than from fuel stations. However, if there are no failures at the chemical firm, they consume at full capacity. Therefore, when the electrolyzer has problems, chemical demand will continue. For fuel stations, the demand always drops during the night and the electrolyzer has time to fill the buffer again, and thus less overcapacity is needed. For fuel stations, more overcapacity does not decrease the buffer size, as can be seen in the second column. The variance decreases slightly. For chemical demand, 2% is needed to get the buffer to the same level as for fuel stations. 2% overcapacity also reduces the standard deviation by factor 2.
When considering the costs of overcapacity, the costs of missed sales will have to be determined, which is pitifully impossible at this stage as marginal costs and sales price are unknown. A percentage point of overcapacity corresponds to approximately 316,000 N𝑚3 in production capacity. It can be seen that the amount vented increases with that amount.
TABLE 5.5
Buffer requirements with associated costs for fuel stations and chemical firms with overcapacity
Customer, % overcapacity Buffer size SD Average amount vented
Tank costs Costs per
5. RESULT
28
Mark-up for households. For households, the total amount of buffer for one electrolyzer would be 13 million N𝑚3. (Note that to be able to produce the total demand, the supply chain is to be initiated in the middle of April, or to have an initial buffer to meet future demand). In figure 5.9 the buffer behaviour for a year is given. See appendix H for the buffer calculation.
A buffer able to hold 13 million N𝑚3 of
hydrogen is very extreme. For example, when storing at 500 bar (which is extremely high), more than 10 Olympic sized swimming pools of storage is required. Household seasonal storage should therefore be done in salt caverns. According to Kruck et al. (2013), a typical €35 million salt cavern holds 55 million N𝑚3. This would thus only be necessary when using 4 ∗ 20𝑀𝑊 electrolyzers. The buffer costs would then be €35 million spread over 4 times 31 million N𝑚3 hydrogen produced
-10000000 -5000000 0 5000000 B uf fe r si z e in N m 3 Time FIGURE 5.8
Stable buffer behaviour by means of 1% overcapacity (a); 2% (b) for fuel stations, and 1% overcapacity (c); 2% (d) for chemical demand
Figure 5.9: Buffer behaviour for household demand for 1 year.
FIGURE 5.9
29
per electrolyzer times 20 years, which equals €0.014 per N𝑚3 (storage in tanks would cost $0.87143 per N𝑚3, which is factor 62 more).
The 4 electrolyzers needed to fully use a salt cavern can produce the non-electrical energy demand of 25,000 households (appendix E). This will thus be the minimum amount for using a cavern of this size.
Production stops. Misaligned production stops between the electrolyzer and chemical firms will lead to high costs. If one were to buffer a full week of production, the tank costs would be in the same order of size as the complete project ($24 versus €30 million) and it would be around 600,000 N𝑚3 in size. Using backup suppliers would take 45 FTLs. Therefore, alignment would be the logical choice (see appendix J for an elaboration).
§5.4 Sensitivity analysis
Number of electrolyzers and customers. A full sensitivity analysis is given in appendix K, a summary of the results follows below.
When delivering to multiple chemical customers, while keeping the amount produced equal, the effects are low and not consistent. This makes sense, as the variation from the electrolyzer is more
than twice as high as that from the chemical customer(s), the standard deviations are 954 and 444 respectively.
Therefore, when increasing the amount of electrolyzers and thus the amount produced, the relative buffers decline. In figure 5.10, these effects are shown. Going from 1 to 2 electrolyzers leads to half the relative buffer sizes. Increasing it more, has less effect. The standard deviations per unit go down with (divided by) approximately√𝑛.
Variance. The effect of variance in chemical demand has been simulated with 1-5% flexibility and is low. For households it is completely offset by the seasonality effect. For fuel stations, one could
3 13 million N𝑚3, with 12 N𝑚3 per kg, and $500/kg spread over 31 million N𝑚3 per year for 20 years.
FIGURE 5.10
5. RESULT
30
expect more variation as it is subject to more factors e.g. station location, weather. Therefore, 10 and 20% variation have also been simulated with 1% overcapacity and venting. No conclusions can be drawn from the results as they are not concise. This makes sense, as the variation induced by the electrolyzer and the demand pattern itself are larger than 20%. The electrolyzer has a standard deviation of 954 and the fuel station 2,223 with a mean of xxxx, which are xx% and xx% respectively.
Correlated downtimes. In the previous simulations, the downtime of chemical customers was uncorrelated. This means that the 132 hours of downtime appeared randomly. It is possible that this is not realistic, but data are lacking and would be firm-specific. In figure 5.11 the effects of 132-1 hour, 4-33 hour, and 2-66 hour downtimes per year are visualized (note that these values are chosen arbitrarily) with 1% chemical demand flexibility. It can be seen that the behaviour of the buffer changes tremendously, the required buffers triple and the standard deviations become even four times as high in the 33 and 66 hour cases compared to the 1 hour case. This shows that customer failures have an extremely large effect on the buffers.
The effect will be dependent on the amount of chemical flexibility. When the chemical customer is down, the buffer will fill, to counter this the customer needs to increase its demand, but as it is down it cannot do this. The flexibility mechanism increases demand when the firm is up again, so with a higher flexibility the buffer deflates faster. This does however not lower the peak. It makes sense that it will when there are multiple customers, they could then consume more while one has a failure.
§5.5 Obsolete pipelines
As said in §2.4 storage opportunities arise by the out phasing of transmission pipelines in the North of The Netherlands. These pipes were used to transport natural gas from the drilling sites to the
FIGURE 5.11
31
national gas highways. These pipes are relatively large and can offer a cheap alternative for hydrogen tank storage. One of these pipes has been studied as such a storage opportunity.
6. DISCUSSION
32
6. DISCUSSION
The central question to be addressed in this study was how the buffer requirements in a hydrogen supply chain change with variations in supply and demand. It became clear that operating conditions influence the required buffers and this varies per customer type. Although the specific requirements will be case dependent, some general guidelines can be formulated. First of all it is clear that there is a positive relationship between variation and buffer requirements. Chemical firms are subject to failures and fuel stations have a distinct, but predictable, profile. Households follow a clear seasonal pattern which makes it unsuitable for tank storage.
By promising to sell up to the production capacity, the system will be unstable. This will lead to infinite buffer requirements, as production and consumption are stochastic. Hydrogen systems therefore need stabilizing mechanisms. Demand flexibility can, even in small percentages, lead to major buffer savings. In other words, with only a small percentage of flexibility, a high resource utilization can be achieved. Flexibility is, however, not possible for all customers. The suitability of backup suppliers will be contract and case-specific but can clearly stabilize the system by lowering the chances of shortages. However, the goal of hydrogen as an energy carrier is to reduce emissions. Backup suppliers will produce their hydrogen from methane, which is not green. To solve the problem of a full storage system, venting can be used to release hydrogen when the system is full. This goes hand in hand with production overcapacity, as these together have a similar effect as demand flexibility. However, in terms of costs, the overcapacity also leads to more venting, which should be taken into consideration.
Unplanned downtime can lead to massive buffer requirements. Firms that do not have their maintenance in order and thus have a low reliability of their production process could disrupt the flow of hydrogen. A firm-specific mark-up should be in place to account for this variation and uncertainty. As seen in §5.3, when facing electrolyzer disruptions, a continuous high demand can also lead to higher buffer requirements. With 1% overcapacity, the chemical firm with lower variance than the fuel station, still needed more buffer. So, more demand variance leads to more buffer variance, but continuous high (or low) demand can lead to extremes, which influence the buffer size.
33
awkward to consider building a system with only one of each. The variability posed should be distributed, or pooled, over multiple units.
Production stops should clearly be aligned between chemical firms and the electrolyzer(s), as not doing so will make all other buffer saving mechanisms unnecessary. Shutting down an electrolyzer for a week leads to a buffer requirement which will almost double the full CapEx of the project.
7. CONCLUSION
34
7. CONCLUSION
§7.1 Academic contributions
Research has focussed on hydrogen as a solution to the intermittency problem of RES. Due to a lack of data, these could not take realistic operational conditions into account. (e.g. Baghaee et al., 2017). This study fills that gap by performing simulations with data from a real hydrogen system proposal. It can be concluded that differences in buffer requirements between customer (types) are prevalent and an equal sales price for hydrogen is not justified. A mark-up for variation and uncertainty should be incorporated based on stability in demand and failures. Stabilizing mechanisms (e.g. demand flexibility, backup suppliers, venting, and production overcapacity) can be used to achieve a high resource utilization. However, this will only be suitable for short-term variation, as the buffering scope and costs for household heating, with its seasonality pattern, will be a large barrier.
This study confirms the conclusion drawn by Giannakoudis et al. (2010) that taking uncertainty into account during the design phase increases the appropriateness of the system to various realistic variations. The assumption that an electrolyzer operates at the listed capacity is inaccurate, whilst its effect can be vast.
§7.2 Managerial implications
It is of utmost importance that the system shows a steady behaviour. Without venting, ad-hoc selling, customer flexibility, or backup suppliers, this will not be possible in combination with a high resource utilization. Therefore, parties creating a hydrogen supply chain should consider which works best, given system configurations, constraints and development expectations. These mechanisms and their value should be taken into account during the investment decision.
Flexibility can lead to system stability without having to vent a large amount of hydrogen and have unutilized capacity. It therefore has a clear value which could lead to a discount for a customer. A measure of demand variability (based on stability and failures) should be determined per firm to assess a fair buffer mark-up.
35
According to AirLiquide (a backup supplier) the minimal amount bought and required delivery time will have a large effect on the costs. Therefore, its economic fitness is hard to determine.
§7.3 Limitations
In this simulation study, some assumptions are made and factors neglected, which lead to limitations.
The input data are from an electrolyzer which is currently active at the chemical park. It is possible that the to-be-build electrolyzer has different operational conditions and therefore the effects are amplified or compressed. At this moment this is still unknown, as it is not operational yet. The data used for chemical firms is not based on real firms, as this data will be very case specific it is not suitable to use one firm. Also, firms are not keen on sharing this data.
Venting is now used to release hydrogen when the system is full. Therefore, if the buffer does not get negative due to overcapacity, the buffer size equals the venting level. If the venting level is increased, less overcapacity is required. This is not entirely realistic, because when aiming for a half-full buffer, other mechanisms should be in place to control the buffer fill. Options are, for example, to turn down the electrolyzer, sell hydrogen ad-hoc, or reduce the price to deflate the buffer. These mechanisms should work when the buffer is above the half-way point instead of full. Choosing when to initiate these mechanisms will depend on the electrolyzer’s functionality and cost/selling prices.
The additional demand chemical customers have, on top of the hydrogen supplied by the electrolyzer is neglected. If firms have flexibility in this amount coming in, this could solve part of the buffering problem in the short run.
7. CONCLUSION
36
§7.4 Future research
Follow-up research is needed to answer the questions arising from this study. More stabilizing possibilities should be studied, e.g. a dynamic pricing strategy to align demand with supply. A combination of mechanisms could be studied to find which set could create most value. As demand for hydrogen increases, market opportunities can arise to buy hydrogen when there is a surplus, and sell when there is a shortage.
Queueing theory has long found that a high utilization leads to exploding waiting times (e.g. Adan & Resing, 2001). In the hydrogen system, this means that as demand approaches supply capacity, malfunctions will arise or buffer requirements will rapidly increase. In this study, stationarity of the system was assumed. This means that supply and demand were equal or did not vary over time. As a hydrogen system grows, it will not be stationary. For example, in the beginning there will be supply overcapacity, demand will then grow, and then a second electrolyzer will be built. During these phases the utilization will change. This will have implications for the buffer requirements, as they grow with the utilization. And, if there would be too much buffer capacity, it will be hard to allocate these costs to specific customers. The effects of a developing hydrogen market on buffer requirements should therefore be studied.
The Dutch government has decided to fuel public transportation in the North with hydrogen. This can be an interesting case as refuelling busses and trains can lead to very high variations in demand, or act as a large stabilizing mechanism. Can this be a flexibility opportunity?
37
REFERENCES
Adan, I., & Resing, J. 2001. Queueing theory. Department of Mathematics and Computing Science, Eindhoven University of Technology, The Netherlands.
Alabdulwahab, A., Abusorrah, A., Zhang, X., & Shahidehpour, M. 2015. Coordination of interdependent natural gas and electricity infrastructures for firming the variability of wind energy in stochastic day-ahead scheduling. IEEE Transactions on Sustainable Energy, 6(2), 606-615.
Anifantis, A. S., Colantoni, A., & Pascuzzi, S. 2017. Thermal energy assessment of a small scale photovoltaic, hydrogen and geothermal stand-alone system for greenhouse heating. Renewable energy, 103, 115-127.
Atkins, P. W., & De Paula, J. 2006. Atkins' physical chemistry (8th ed.). New York City, NY: Oxford University Press.
Avery, W., Brown, G., Rosenkranz, J., & Wood, R. 1992. Optimization of purchase, storage and transmission contracts for natural gas utilities. Operations Research, 40(3), 446-462.
Baghaee, H. R., Mirsalim, M., & Gharehpetian, G. B. 2017. Multi-objective optimal power management and sizing of a reliable wind/PV microgrid with hydrogen energy storage using MOPSO. Journal of Intelligent & Fuzzy Systems, 32(3), 1753-1773.
Baldwin, D. 2017. Development of high pressure hydrogen storage tank for storage and gaseous truck delivery (No. DOE-HEXAGON-GO18062). Lincoln, NE: Hexagon Lincoln LLC.
REFERENCES
38
Bergek, A., & Jacobsson, S. 2010. Are tradable green certificates a cost-efficient policy driving technical change or a rent-generating machine? Lessons from Sweden 2003–2008. Energy Policy, 38(3), 1255-1271.
Bertrand, J. W. M., & Fransoo, J.C. 2016. Modelling and simulation. In Karlsson, C. (2nd ed.) Research Methods for Operations Management: 165-197. New York City, NY: Routledge.
Blarke, M. B. 2012. Towards an intermittency-friendly energy system: Comparing electric boilers and heat pumps in distributed cogeneration. Applied Energy, 91(1), 349-365.
Bossel, U., Eliasson, B., & Taylor, G. 2003. The future of the hydrogen economy: bright or bleak? Cogeneration and Distributed Generation Journal, 18(3), 29-70.
Boudries, R. 2018. Techno-economic study of hydrogen production using CSP technology. International Journal of Hydrogen Energy, 43(6), 3406-3417.
Brandsma, R. 2020, January 22. Bouw waterstoffabriek (de grootste van Europa) in Delfzijl start nog dit jaar. Dagblad van het Noorden. Retrieved from:
https://www.dvhn.nl/economie/Bouw-waterstoffabriek-Delfzijl-nog-dit-jaar-25264764.html
Bui, M., Adjiman, C. S., Bardow, A., Anthony, E. J., Boston, A., Brown, S., & Hallett, J. P. 2018. Carbon capture and storage (CCS): the way forward. Energy & Environmental Science, 11(5), 1062-1176.
39
Certifhy. 2015. Overview of the market segmentation for hydrogen across potential customer groups, based on key application areas. Deliverable No. 1.2. Retrieved from: http://www.certifhy.eu/images/D1_2_Overview_of_the_market_segmentation_Final_22_ June_low-res.pdf
Demir, M., & Dincer, I. 2018. Cost assessment and evaluation of various hydrogen delivery scenarios. International Journal of Hydrogen Energy, 43(22), 10420-10430.
Ellis, B. L., & Nazar, L. F. 2012. Sodium and sodium-ion energy storage batteries. Current Opinion in Solid State and Materials Science, 16(4), 168-177.
Eriksson, E. L. V., & Gray, E. M. 2017. Optimization and integration of hybrid renewable energy hydrogen fuel cell energy systems–A critical review. Applied Energy, 202, 348-364.
Everts, S. 2015. When Chemicals Became Weapons of War. Chemical & Engineering News, 93(8). Retrieved from: http://chemicalweapons.cenmag.org/when-chemicals-became-weapons-of-war/
Fédération Internationale De Natation. 2015. FINA facilities rules 2015-2017, 5. Retrieved from: https://www.fina.org/sites/defailt/files/finafacilities_rules.pdf
Fourier, J. 1827. Mémoire sur la température du globe terrestre et des espaces planétaires. Mémoires de l'Académie des sciences de l'Institut de France. 569–604.
Giannakoudis, G., Papadopoulos, A. I., Seferlis, P., & Voutetakis, S. 2010. Optimum design and operation under uncertainty of power systems using renewable energy sources and hydrogen storage. International Journal of Hydrogen Energy, 35(3), 872-891.
REFERENCES
40
global response to the threat of climate change, sustainable development, and efforts to eradicate poverty V. Masson-Delmotte, P. Zhai, H. O. Pörtner, D. Roberts, J. Skea, P.R. Shukla, A. Pirani, M. Moufouma-Okia, C. Péan, R. Pidcock, S. Connors, J. B. R. Matthews, Y. Chen, X. Zhou, M.I. Gomis, E. Lonnoy, T. Maycock, M. Tignor, T. Waterfield (eds.). In Press.
Ishaq, H., Dincer, I., & Naterer, G. F. 2018. Development and assessment of a solar, wind and hydrogen hybrid trigeneration system. International Journal of Hydrogen Energy, 43(52), 23148-23160.
Kalinci, Y., Hepbasli, A., & Dincer, I. 2015. Techno-economic analysis of a stand-alone hybrid renewable energy system with hydrogen production and storage options. International Journal of Hydrogen Energy, 40(24), 7652-7664.
Kang, P. S., & Bhatti, R. S. 2019. Continuous process improvement implementation framework using multi-objective genetic algorithms and discrete event simulation. Business Process Management Journal, 1-33.
Kharel, S., & Shabani, B. 2018. Hydrogen as a long-term large-scale energy storage solution to support renewables. Energies, 11(10), 2825.
Kim, J., Lee, Y., & Moon, I. 2008. Optimization of a hydrogen supply chain under demand uncertainty. International Journal of Hydrogen Energy, 33(18), 4715-4729.
Koninklijk Nederlands Meteorologisch Instituut (KNMI). 2009. Winter 2008-2009 (December, januari, februari) koud, zeer zonnig en droog. Retrieved from:
41
Kruck, O., Crotogino, F., Prelicz, R., & Rudolph, T. 2013. Assessment of the potential, the actors and relevant business cases for large scale and seasonal storage of renewable electricity by hydrogen underground storage in Europe. HyUnder Deliverable No. 3.1.
Lahnaoui, A., Wulf, C., Heinrichs, H., & Dalmazzone, D. 2018. Optimizing hydrogen transportation system for mobility by minimizing the cost of transportation via compressed gas truck in North Rhine-Westphalia. Applied Energy, 223, 317-328.
Liander. 2009. Dagprofielen gas. Retrieved from:
https://www.liander.nl/partners/datadiensten/open-data/data
Michalski, J., Bünger, U., Crotogino, F., Donadei, S., Schneider, G. S., Pregger, T., & Heide, D. 2017. Hydrogen generation by electrolysis and storage in salt caverns: Potentials, economics and systems aspects with regard to the German energy
transition. International Journal of Hydrogen Energy, 42(19), 13427-13443.
Ministry of Economic Affairs and Climate. 2019. Klimaatakkoord. Retrieved from https://www.klimaatakkoord.nl/binaries/klimaatakkoord/documenten/publicaties
/2019/06/28/klimaatakkoord/klimaatakkoord.pdf
Ministry of Infrastructure and Water Management. 2018. Various Pipeline Reports. Retrieved from https://www.risicokaart.nl/ specifically from http://prdcrs.risicokaartinvoer.nl/ Viewer.aspx?Idx=94&argument=rBvf9-74vOslaPnmGt8vQ-Ti5ZrIdWi3DmpY63lzHuD PXEngAp3wj0hd1iTleVC7-OISQgkd6gGqRKErNNQeUn1mSb0Oum2HO*6CRa4MQ9 s=
REFERENCES
42
Moreno-Benito, M., Agnolucci, P., & Papageorgiou, L. G. 2017. Towards a sustainable hydrogen economy: Optimisation-based framework for hydrogen infrastructure development. Computers & Chemical Engineering, 102, 110-127.
Mukherjee, U., Elsholkami, M., Walker, S., Fowler, M., Elkamel, A., & Hajimiragha, A. 2015. Optimal sizing of an electrolytic hydrogen production system using an existing natural gas infrastructure. International Journal of Hydrogen Energy, 40(31), 9760-9772.
Mulder, M., Perey, P., & Moraga, J.L. 2019. Outlook for a Dutch hydrogen market :
economic conditions and scenarios. Centre for Energy Economic Research
(CEER). Policy Papers (5). Retrieved from:
https://www.rug.nl/research/portal/files/77687340/ceer_policypaper_5_web.pdf
Noordelijke Innovation Board. 2017. Groene Waterstof Economie in Noord-Nederland.
Retrieved from: http://profadvanwijk.com/wp-content/uploads/2017/04/NIB-BP-NL-DEF-webversie.pdf
Nouryon & Gasunie. 2019. Eindrapportage MegaWatt Design Delfzijl. Retrieved from:
https://www.gasunienewenergy.nl/projecten/djewels-20-mw-elektrolyse-installatie/$1352/$1353
Pino, F. J., Valverde, L., & Rosa, F. 2011. Influence of wind turbine power curve and electrolyzer operating temperature on hydrogen production in wind–hydrogen systems. Journal of Power Sources, 196(9), 4418-4426.
43
Reddi, K., Elgowainy, A., Rustagi, N., & Gupta, E. 2018. Techno-economic analysis of conventional and advanced high-pressure tube trailer configurations for compressed hydrogen gas transportation and refueling. International Journal of Hydrogen Energy, 43(9), 4428-4438.
Reddi, K., Mintz, M., Elgowainy, A., & Sutherland, E. 2016. Building a hydrogen infrastructure in the United States. In Compendium of hydrogen energy, 293-319. Cambridge, United Kindom: Woodhead Publishing.
Rusin, A., & Stolecka, K. 2017. Hazards associated with hydrogen infrastructure. Journal of Power Technologies, 97(2), 153-157.
Sainathuni, B., Parikh, P. J., Zhang, X., & Kong, N. 2014. The warehouse-inventory-transportation problem for supply chains. European Journal of Operational Research, 237(2), 690-700.
Steinberg, M., & Cheng, H. C. 1989. Modern and prospective technologies for hydrogen
production from fossil fuels. International Journal of Hydrogen Energy, 14(11), 797-820.
Su, Y., Kern, J. D., & Characklis, G. W. 2017. The impact of wind power growth and hydrological uncertainty on financial losses from oversupply events in hydropower-dominated systems. Applied Energy, 194, 172-183.
Tomasgard, A., Rømo, F., Fodstad, M., & Midthun, K. 2007. Optimization models for the natural gas value chain. In Geometric Modelling, Numerical Simulation, and Optimization, 521-558. Heidelberg: Springer.
REFERENCES
44
Van Wijk, A., & Hellinga, C. 2018. Hydrogen the key to the energy transition retrieved from: http://profadvanwijk.com/wp-content/uploads/2018/05/Technical-Report- Hydrogen-the-key-to-the-energy-transition.pdf
Vattenfall. 2020. Energierapport, elke maand inzicht in je energieverbruik. Retrieved from: https://www.vattenfall.nl/service/energierapport/
45
APPENDIX
A: Interview protocol
First of all, I would like to thank you for helping me with my research. Do you agree that this interview is recorded?
I am performing a study to find effects of variation on required buffer sizes. Additionally, I will look at cost effects of different designs. To do this, I will make a simulation model of the whole supply chain. I started in the academic literature to determine which factors should be included in the model. The goal of this interview is to verify these factors and identify additional ones.
---a table with the categories and factors from literature is presented and systematically walked through with the interviewee---
Pipeline Could you provide me general information on pipe? What are the dimensions used in this infrastructure? What is pressure range of the pipe?
What will be the operational pressure of the pipe, is this variable? What is the flow capacity of this pipe?
How large will the pipeline be? (in different scenarios)?
Electrolyzer Could you provide me general information on electrolyzer? What is the nominal power of the electrolyzer?
What is the output load distribution of the electrolyzer? What is the output pressure range (Bar)?
Is it possible to temporarily overload it? Pros Cons? What is the response time of the Electrolyzer?
