Providing insight into the potential of hydrogen storage as value adding

Technology and Operations Management

Master Thesis

Providing insight into the potential of hydrogen storage as value adding

flexibility tool

Thomas Herwijn - S2719878 Under supervision of:

First assessor: dr. X. Zhu Second assessor: dr. E. Ursavas

June 29, 2019

A b s t r a c t

The purpose of this paper is to provide insight into the operation of hydrogen energy storage applied as arbitraging tool. In this study the storage facility is co-located to a renewable energy source. Energy generated by these mechanisms is sold on the spot market. To analyze this situation, we developed a simulation in which an hourly storage decision is made. With this simulation three experimental sets were tested. One, in which the effect of different types and sizes of generation mechanisms were tested. The second set tested the effect of different storage facility components on the operation. In the third the effect of technological developments on the operation was considered. The most important results found with these experiments are: (1) Due to a more stable generation pattern, wind on sea is, based on generated additional revenues, the best energy source for this operation.

The operation in combination with photovoltaic cells generated the lowest revenues. (2)

When storage technology becomes more efficient the amount of energy stored and the

generated revenues increase. The increase is largest for wind from sea and least for the

operation in which PV cells are used. (3) For wind energy, a larger fuel cell, relative to

the size of the electrolyser, generates most revenues. When solar power is used, most

revenues are generated when the electrolyser and the fuel cell are of the same size. Based

on these conclusions we would advise a potential investor to invest in hydrogen storage in

combination with sea wind. Especially, when the storage technology develops in terms of

efficiency.

Contents

1 Introduction 2

2 Literature Review 4

2.1 Arbitrage operation with hydrogen . . . . 4

2.2 Arbitrage operation with different energy carriers . . . . 5

2.3 Future technological developments of hydrogen storage . . . . 6

2.3.1 Contributions to the literature . . . . 7

3 Methodology 8 3.1 Simulation Model . . . . 8

3.1.1 Model Outputs . . . . 9

3.2 Simulation Policies . . . 10

3.2.1 Revenue generation . . . 10

3.2.2 Model logic . . . 11

3.2.3 Storage Decision . . . 11

3.3 Data . . . 12

4 Numerical Simulation 13 4.1 Generation mechanisms . . . 13

4.1.1 Data analysis . . . 13

4.1.2 Scenario Details . . . 15

4.1.3 Results . . . 16

4.2 Storage system setup . . . 17

4.2.1 Data analysis . . . 17

4.2.2 Scenario Details . . . 17

4.2.3 Results . . . 17

4.3 Technological developments . . . 18

4.3.1 Data analysis . . . 19

4.3.2 Scenario details . . . 19

4.3.3 Results . . . 20

5 Discussion 21 5.1 Limitations . . . 21

5.2 Managerial Implications . . . 22

5.3 Future research suggestions . . . 23

6 Conclusion 24

Appendix A - Assumptions 27

Appendix B - Coefficient of variation 27

Appendix C Generation mechanism - Scenarios and results 28

Appendix D Operation setup - Scenarios and results 29

Appendix E Technological developments - Scenarios and results 30

1 Introduction

Due to the pressure on the traditional energy sources (Rogelj et al., 2016), the consumption of energy generated from coal and gas has diminished and the share of renewable energy sources (RES) in the market increased (Smil, 2016). The plans of governments to further decrease the emission of greenhouse gasses suggests that the share of RES in the energy market will increase even more. In addition to its inhibitory effect on global warming, this transition also brings a technological problem. Due to the weather dependency of RES, the generation pattern of mechanisms using renewable sources is fluctuating strongly. This makes the energy generation from RES very unpredictable. Balancing demand and supply of energy, becomes harder when unpredictable RES become more dominant in the generation mix.

This consequence of the transition towards sustainable energy is also foreseen by Gronin- gen Seaports. This government-owned company is the controller of the Delfzijl haven and Eemshaven. Over 500 MW capacity of RES, in the form of wind turbines and photovoltaic cells (PV cells), is installed on their grounds. These sources generate enough to provide 250.000 average dutch households with energy. To be able to respond to market changes, this company is considering to invest in hydrogen storage. Electricity can be converted into hydrogen and stored until a moment it is needed. With this flexibility tool, an owner of such a system can take advantage of fluctuations in the energy market price. This entails an arbitraging operation, in which energy is stored when the price is low and sold when price increased.

According to experts (Pellow et al., 2015) hydrogen is, compared to other energy carriers, a storage method with a lot of advantages. Compared to batteries, hydrogen storage is a better option to store energy on a large scale. Hydrogen is stored in a tank, which is a cheaper option for large scale storage. Furthermore a part of the energy in a battery is released over time, the amount of self discharge of hydrogen storage is negligible. Therefore hydrogen is better able to store energy for a longer period. Another alternative is pumped hydro storage (PHS), a method in which energy is stored by pumping water from a reservoir to another reservoir on a higher elevation. When electricity is demanded, the water is released and powers several turbines when it flows back to the lower elevation reservoir. The advantage of hydrogen storage over PHS is that it is not limited to any geographical location. Thus the application of hydrogen as arbitraging tool is suitable for any location and for any storage length.

Research (Kousksou et al., 2014; Kloess and Zach, 2014; Fertig and Apt, 2011) showed out that, despite the advantages of hydrogen storage, the operation of this energy carrier as an arbitraging tool is not economically feasible. The initial capital expenditures are high and a lot of energy is lost during the storage operation. The efficiency of the process in which electricity is transformed into hydrogen is low, this causes the losses in energy. Before advantage can be taken from market price fluctuations, by storing energy when the price is low and sell when the price increased, the losses associated with the storage process should be covered. According to the aforementioned research, the difference between the lowest price and the highest price (the price spread) is not of sufficient size to cover for the efficiency losses. Therefore, these researchers conclude that the operation will not generate enough revenues to earn back the initial capital expenditures.

The efficiency of the electrolyser and the fuel cell, the instruments which convert electricity

into hydrogen and back, are respectively, 65% and 60% (Chardonnet et al., 2017). The roundtrip efficiency, so the efficiency of the entire storage process, is the product of the two. Both, the electrolyser and the fuel cell, are not implemented on a large scale in the hydrogen production industry yet. Therefore experts (Hwang, 2013; Schmidt et al., 2017) expect that this technology will develop, in terms of efficiency and cost, in the coming years. The effect of this development on the operation of hydrogen storage as arbitraging tool, is not broadly researched yet. Therefore the goal of this research is to provide Groningen Seaports, and all other potential investors, with insight into the operation of hydrogen storage applied as an arbitraging tool, now and in the future. We will first study the effect of the size of different components, such as the electrolyser, the fuel cell and the amount energy generators, on the arbitraging operation. Then, the effect of future technological developments will be studied. A simulation model will be developed to analyze these effects and to explain in which way they influence the revenues generated by the arbitraging tool.

The remainder of this paper is structured as follows. The next section will give an overview

of the relevant literature. In this section comparable models and the future developments of

technology concerned with hydrogen storage are discussed. After that, model components of

different studies will be combined to create an applicable simulation model for the purposes of

this study. After that, several scenarios will be developed and tested with the model and at last

the results of these simulation will be discussed and summarized.

2 Literature Review

The first part of this section will present several articles which studied the operation of an hydrogen storage mechanism as arbitraging tool. In the second part, several papers presenting the same operation with different energy carriers, will be discussed. In both paragraphs we will highlight the differences between the presented paper and this paper. The following paragraph will be devoted to discuss the literature on future technological developments of hydrogen storage technology. Last, we will point out in which way our research differs from the discussed research and which contribution our study provides to the literature.

2.1 Arbitrage operation with hydrogen

Multiple approaches to study the economic value of hydrogen storage, applied as arbitraging tool, are used in past research. Gatzen (2008) presents an overview of numerous energy carriers applied as speculation tool. The goal of his study is to identify the best energy carrier for an arbitraging operation. He suggests that hydrogen storage applied for this purpose is not feasible. In the study of Gatzen (2008) hydrogen storage is directly connected to the grid instead of to a RES. Thus, in setup of that study the variable effect of RES is taken away. When the operation is analyzed with RES, there will be moments in which no energy is available to transform into hydrogen. This will diminish the amount of energy stored. The study presented by Gatzen (2008) is not insightful for investors considering the application of hydrogen storage next to wind turbines or PV cells. In their consecutive study (Loisel et al., 2010) a similar model was developed. In that study the arbitraging value of compressed air energy storage (CAES) and pumped hydro storage (PHS), two other energy carriers, was tested. This study considers two types of value. Besides the economic value of arbitraging as it is discussed before, this study analyzed the economic value of applying storage as flexibility tool in the reserve market.

This entails offering storage capacity to the grid operator, to store energy when there is excess in production. As a consequence, the storage capacity cannot be used for arbitraging purposes.

For that reason this application of hydrogen storage is not considered in this study.

Kloess and Zach (2014) researched the feasibility of several energy carriers in combination with energy from wind farms. According to their study, this operation is not feasible since the efficiency of hydrogen storage technology is too low. Thus, the amount of energy lost during the storage process is very large and the daily price spreads are too small to account for these losses. This study will therefore also consider multiple day storage. This implies that the lowest price of one day can be used to store and if the highest price of that day is not of sufficiently high, the highest price of a following day will be used. Kroniger and Madlener (2014) found the same results as Kloess and Zach (2014). In their study, the application of hydrogen storage as arbitraging tool was, due to the lack of variation in price, infeasible as well. Different from the study by Kroniger and Madlener (2014), this study will also consider the application of hydrogen storage in combination with PV cells.

Research done by Steward et al. (2009) analyzed the competitiveness of hydrogen storage in

comparison to other storage mechanisms (e.g. CAES and PHS). The storage decision policy

used in this study is very straightforward. Three quarters of the day the battery is charged

and the other period, during peak hours, it is discharged. Under this policy, he found that

hydrogen storage is potentially able to compete with other energy carriers, such as CAES and PHS. So he suggests that when the technology associated with hydrogen storage has developed in terms of cost and efficiency, it is a cheaper option than other energy carriers. Different from his study, this research will apply a storage decision policy based on previous prices. Due to the fluctuating nature of prices, the best storage decision varies as well. A static, or time-based policy, doesn’t react to these fluctuations. Therefore the policy used in this study is expected to generate results which are closer to the results of the operation in reality.

In Nojavan et al. (2017) the goal of the study was to find the best storage decision policy in terms of profits. Three types of policies were tested. A fixed pricing policy (1) in which a constant sell and buy threshold is used. A threshold is the price below or above energy is, respectively stored or discharged. Time-to-use pricing (2) distinguishes three price levels, one for low, one for medium and one for peak periods. The third decision policy is real-time pricing. In the last mentioned policy the storage decision is made on the spot, based on future expectations. This policy generated the most optimal outcomes. The setup of the study by Nojavan et al. (2017) differs from this study, since Nojavan et al. (2017) used plug-in electrical vehicles (PEVs) next to hydrogen, as storage location. The study by Nojavan et al. (2017) tests energy generated by PV cells and wind turbines. However, they don’t distinguish between land and sea wind.

2.2 Arbitrage operation with different energy carriers

The following paragraphs will introduce several studies into the economics of storage mecha- nisms. The studies presented in these paragraphs don’t discuss hydrogen storage and therefore the conclusions of these studies are not always relevant. The costs and productivity of these storage mechanisms differ from those of hydrogen. However according to Korpas and Holen (2006), the models and their logic are similar to the studies presented in the previous paragraphs.

To learn from their approach and their applied decision policies, these studies are relevant to discuss.

Lund et al. (2009) describe two ways to analyze CAES. The first is an algorithm used with a perfect foresight hypothesis (PFH), in which the operation is optimized since all upcoming prices are assumed to be known upfront. The other policy used is a backcasting approach. In this policy the decision thresholds are based on the data of the last 24 hours. When the results of the two different methodologies are compared, they found a 10 - 20 per cent difference in profits. The optimization, logically, generated the most optimal outcomes. However, the results obtained with a backcasting decision policy will be closer to reality and therefore more insightful for practical application. Also a study by Sioshansi et al. (2009) compares the value of the operation under PFH with backcasting methods. The decision strategy in this study is based on price data of the last two weeks. They found, due to the highly predictable price patterns of the energy market, that prediction based decisions can realize over 85% of the arbitrage value with perfect foresight. The two week period used in their study is not able to respond to sudden changes in weather conditions or other factors influencing the spot market price of electricity.

Therefore this study will use a similar approach, however with a shorter backcasting period. In

their subsequent research (Denholm and Sioshansi, 2009) the additional revenues of co-locating

storage to energy generation mechanisms is studied. The goal of their study was to verify the

optimal size of the storage mechanism, CAES in this case, relative to the size of the generation

mechanism and the buffer. A similar analysis will be made in this research. However, in this study hydrogen storage is combined with wind and solar energy.

Fertig and Apt (2011) simulated a profit optimizing algorithm to evaluate the feasibility of a CAES system. Their study shows out that the operation can be profitable in years with highly fluctuating prices. However they suggest that a risk averse manager would not invest in this type of operation. Also Abbaspour et al. (2013) studied CAES in combination with a wind farm. He found with a simulation study that a wind farm is 43% more profitable when storage is available. Other research (Nojavan et al., 2018) found an increase in profits of 30% in an optimistic scenario, however a 50% decrease in a pessimistic one.

2.3 Future technological developments of hydrogen storage

Next to a buffer tank, in which hydrogen is stored, a storage facility has two components. An electrolyser, in which electricity is used to convert water into hydrogen and oxygen and a fuel cell which converts these two elements back into water and electricity. The amount of energy when one of these processes is performed, is less than it was before. In addition to oxygen and hydrogen, heat is released during the transformation process. Currently, the roundtrip efficiency, a product of the efficiency of the electrolyser and the fuel cell, is between 35% and 40% (Chardonnet et al., 2017). The application of this technology for the production of hydrogen is relatively new. For this reason, researchers (de F´atima Palhares et al., 2018; Lipman et al., 2004; Hwang, 2013; de F´atima Palhares et al., 2018) expect technological developments. In the following paragraphs several views on this development will be discussed.

The three types of electrolysis discussed in literature (Schmidt et al., 2017) are polymer elec- trolyte membrane electrolysis cell (PEMEC), alkaline electrolysis cell (AEC) and solid oxide electrolysis cell (SOED). SOED is a method in which electrolysis is performed with an efficiency approaching 100%. This operation requires temperatures up to 900 degrees (Brisse et al., 2008).

This heat should be distracted from geothermal energy, the associated technology to do that is not developed yet. Therefore, this type of electrolysis is not included in this research. The current roundtrip efficiency of the storage process using PEMEC or AEC is assumed to be between 35% and 40 %. A study from (Steward et al., 2009) expects this efficiency to increase to 41% in the coming decades. This assumption is based on research of Lipman et al. (2004).

Both, the efficiency of the electrolyser and the fuel cell is around 60% (Chardonnet et al., 2017). The most conservative experts conducted in research by Schmidt et al. (2017) think this efficiency will increase to 62%. This implies a roundtrip efficiency of 38.4%. The most optimistic experts conducted in the research by Schmidt et al. (2017) suggest a 82% efficiency, increasing the total storage efficiency to 67%. Hwang (2013) also found a theoretical efficiency above 80%.

However, his realistic expectation is an efficiency of 71%.

Research (Kloess and Zach, 2014; Kroniger and Madlener, 2014) showed that hydrogen storage

applied as arbitraging tool is not economically feasible. Currently, the efficiency losses are

often larger than the daily price spreads. This implies that the spread in price cannot account

for the value of energy lost during the storage process. With a 30% roundtrip efficiency as

assumed by Kroniger and Madlener (2014), only 4% of the days have a price spread which is

sufficiently large, to perform an arbitraging operation. This study will verify, based on the

information provided in the previous paragraph, what the effect of technological developments will be on the arbitrage operation.

2.3.1 Contributions to the literature

The goal of this study is to provide insight in the decisions how to implement hydrogen storage

as an arbitraging tool. This research will consider the co-location of hydrogen storage to three

different generation mechanisms. In the experiments performed, different ratios between the

size of the generation mechanisms, the electrolyser and the fuel cell will be tested. Furthermore,

the effect of technological development on the operation is discussed.

3 Methodology

This section will present the simulation model used in this study. This model analyzes the operation of a hydrogen storage system applied as arbitraging tool. First, we will describe the conceptual model used in this study, then each component of the model will be discussed in greater detail. Next, the simulation policies and the decision rules used in the simulation will be explained. Finally, the input data and source of the input data will be discussed.

3.1 Simulation Model

A simulation is the chosen method of this study for several reasons. A simulation provides the ability to test a decision policy under different price or generation circumstances. The results obtained from these test and the differences between them are tested under different conditions, thereby the conclusion are more robust and better applicable in reality. The proposed model in this study is build based on research conducted by Kroniger and Madlener (2014); Kloess and Zach (2014) and Sioshansi et al. (2009). These articles present the framework for the simulation, the decision policy and the type of input data. These studies are used to validate the model used in this study. Validation is a main element of scientific research (Law et al., 2000) used to make the obtained results more credible and better generalizable Bunge (2012).

Figure 1: Conceptual model

In figure 1 the conceptual model of this study is depicted and in table 1 the inputs of the

simulation are listed. The model has three main components. The first is a renewable energy

source. To provide Groningen Seaports with insight into the effect of different energy generators

on the operation, this study will research the same generators as the ones installed on their

lands. Energy is generated by photovoltaic cells, wind turbines on land and in sea. Each source

has its own generation pattern and characteristics influencing the outputs of the operation.

The energy generated by these RES will either be stored, or directly sold to the grid. This decision depends on the second component of the simulation, the market price of energy. In this research we will use prices from the spot market. This market makes it possible to directly trade energy without any delay. When the price is high, energy will directly be sold to the market. When it’s low and expected to increase, the energy will be stored in a hydrogen storage facility, the last component of the model. When the storage decision is made, a chemical process is used to transform energy into hydrogen and back. In figure 2 the associated process is depicted. Direct current is used by an electrolyser to decompose water, which splits it into hydrogen and oxygen. The hydrogen is compressed and stored under high pressure in a tank.

The oxygen is captured and will be used for the opposite process. When energy is demanded, a fuel cell is used to convert hydrogen back into water. In addition to electricity and hydrogen, heat is released during the conversion processes. As a result, the storage process is subject to efficiency losses occurring with both the conversions. The electrolyser and the fuel cell have limited conversion capacity.

Figure 2: Hydrogen Storage Circuit

Element Parameter Value

Directly sold energy E

MWh

Stored energy E

MWh

Energy price on the spot market P

€

Roundtrip efficiency η

%

Electrolyser efficiency η

%

Fuel cell efficiency η

%

Table 1: Overview of simulation inputs

3.1.1 Model Outputs

In order to find meaningful insight into the operation and to be able to explain the obtained

results, this study will present several outputs. First of all, the additional revenues. These

revenues are additional compared to a base case in which no storage facility are available. With

this output, the effect of different experiments on the operation can be compared. In addition to

the revenues, several other factors will be presented. For example the amount of energy stored

and the utilization of the electrolyser. These outputs will be presented for every experiment and every case to help us to explain in which way revenues are affected by the experimental setup. To explain specific elements or findings, incidentally different outputs will be presented.

3.2 Simulation Policies

3.2.1 Revenue generation

The goal of this simulation is to maximize revenues (π). Compared to a base case in which storage is unavailable, the value of energy delivered to the grid should increase when storage becomes available. To verify whether this is the case, the function describing the revenues of the base scenario is presented in equation 1. This equation will function as the reference amount of revenues. The results of the experiments conducted in this research will be compared to this equation.

π = P

· E

(1)

In this equation, P

is the spot market price for energy and E

the generated energy which is directly sold to the market. In this scenario the generated energy will directly be sold to the grid for the associated price of that moment. This determines the revenues for a given time period. This situation will be compared to one in which storage is available. This function is described in equation 2. In this equation E

is stored energy and η the roundtrip efficiency of the storage process. The roundtrip efficiency describes the amount of energy which is left after energy is stored as hydrogen and transformed back into electricity.

π = P

· E

+ P

· E

· η

(2) The roundtrip efficiency consists of two types of efficiency losses. The energy lost when electrolysis is used to convert electricity into hydrogen, η

and the efficiency lost when hydrogen is converted back into electricity by the fuel cell η

. In 3 this relation is described.

η

= η

· η

(3)

Different arbitraging strategies are discussed in literature. In the article written by Kroniger and Madlener (2014), the application of hydrogen storage for daily arbitrage is considered.

According to this method the price differences of one day, if large enough, will be used to sell and buy energy. The research of Kloess and Zach (2014) considers storage for longer periods. In both methods the price spread must be larger than the efficiency losses of the storage process.

The price spread is defined as the relative difference between the maximum price P

and

the minimum price P

of energy on the spot market. When the roundtrip efficiency of

energy is larger than the price spread, economic value can be created. Equation 4 describes

this relation.

η

> P

P

(4)

3.2.2 Model logic

The model has two main inputs, energy prices and energy generation. In the simulation each hour a price and an amount of energy is generated. Based on these height of the price, a decision is made in which energy is either stored, or sold to grid. When energy is directly sold, it leaves the simulation. When energy is stored, the model will first verify whether the storage tank is full. When it is full, the energy is sold to the grid and leaves the simulation. When it is not full, the simulation will verify whether the electrolyser has enough capacity to convert the entire amount of energy into hydrogen. If not, the surplus is sold directly to the energy market and the full capacity of the electrolyser is used to produce hydrogen. When the electrolyser has enough capacity, the entire amount is used to produce hydrogen. When the price is sufficiently high to account for the transformation losses, hydrogen is discharged. The amount of hydrogen that is converted into energy depends on the amount in the buffer and the capacity of the fuel cell. If the amount of energy in the buffer is larger than the capacity of the fuel cell, the full capacity of the fuel cell is utilized. If the amount of energy in the buffer is smaller than the fuel cell, all energy is transformed. The simulation could leave a remaining amount of energy in the buffer at the end of the simulated period. To determine the value of this energy, the storage decision will be made until all the energy in the storage is sold to the market, with a maximum of one year. The hydrogen tank is subject to self-discharge, so energy is lost while it is in the storage tank, however the amount is very small and therefore neglected in this study (Pellow et al., 2015). This assumption is listed together with all others in Appendix A.

3.2.3 Storage Decision

In the literature, storage decisions are based on different principles. This paragraph explains various choices and discusses the expected effect on the results of the study. In Nojavan et al.

(2017)’s study, three different types of storage decisions are applied. Despite the fact that all these were developed for an optimization method, they can function as inspiration for this simulation model. The first decision strategy proposed is a constant and predetermined threshold, above and below which energy is, respectively, sold or stored. When this threshold is determined in advance for a longer period, the operation of the buffer will be far from optimal in some periods. Simply since this method does not allow to correct for unforeseen price fluctuations and seasonal or daily variation. The second policy suggested is time-of-use pricing.

In Nojavan et al. (2017) three different price levels are set. The idea of this strategy is to assign varying thresholds to market price patterns which are repeatedly observed (Hatami et al., 2011).

For example daily, weekly or seasonal schedules are assigned with specific threshold values.

The last decision policy use in the study is real-time pricing. This optimizes the profits of the

operation, since it determines the best moment to sell and buy, however in their research perfect

foresight is assumed. Conclusions of the article written by Sioshansi et al. (2009) suggest perfect

foresight is not required to achieve close to optimal decisions. They argue that fluctuations

in the energy market are very predictable and therefore price thresholds can be determined

very accurately based on the prices of previous days. Therefore this study will apply such

a backcasting method. Different from Sioshansi et al. (2009), this study will use a seven day backcast method. Compared to two weeks, this method is more flexible and will be able to react in a more accurate way to sudden price changes. The maximum price of the last seven days will be used as the expected maximum price of that day. Thus, the storage threshold of that day is the expected maximum price multiplied by the roundtrip efficiency of the storage process. All energy stored for a lower price than this threshold is expected to generate a profitable turnover.

A maximum storage threshold is set in order to prevent an extremely high price, occurring just occasionally, influencing the price unjustly high. When energy is stored, the associated price is used to determine the discharge threshold for the next days. The average price of all previously stored energy is divided by the roundtrip efficiency, this number will be used as the upper threshold of that moment. So when the price is above this discharge threshold, energy will be released and delivered to the grid.

3.3 Data

Generation

Three types of generation data are used in this study: solar panels, off- and onshore wind.

To make the case most insightful for Groningen Seaports, available generation patterns from Northern Netherlands are used for this study. The data describes hourly percentages of the total annual production of each source in the given region. Data is available from 2012, 2013 and 2017, which is a total of 26280 observations per generation method. These percentages are used to generate absolute data points in kWh. First, a quantity of installed capacity is determined, a fraction (20%) of which is taken since turbines and panels are not always utilized. This amount is multiplied by the hourly percentages to get absolute generation in MWh.

Price

Data on the electricity price is gathered from the Northwestern European energy market operator Nord Pool

. Nord Pool offers the possibility to trade on the spot market between different countries, mainly in northwestern Europe. Hourly data on the spot market price in Euro’s is collected for the past 3 years. So data from 2018 up to and including 2016. The prices in this data set are from the three Baltic states and several areas in Norway, Denmark and Sweden. This data is chosen, simply due to availability reasons. This data will be used as input data for the simulation.

1

https://www.nordpoolgroup.com/

4 Numerical Simulation

As discussed in the introduction, the goal of this simulation study is to provide insight into the investment decision in hydrogen storage, applied as arbitraging tool. This insight will be given in three ways. First, the effect of the type of generation mechanism and its size is tested. Then the setup of the storage facility is analyzed. This implies we will test the effect of different buffer, electrolysis and fuel cell sizes on the operation. After that, the effect of future technological developments on the operation will be discussed. An overview of the experimental sets is given in table ??.

First a data analysis is conducted. With this analysis, expectations are formed on the effect a varying element has on the operation. Based on the expectations, we will develop several experiments to test whether the expectations hold. To measure and explain the observed relations, we will present several simulation outputs. We will provide the additional revenues of each experiment. In addition, the utilization rate of the electrolyser and the amount of energy stored will be presented. When needed, we will provide several other simulation outputs. These outputs will help us to understand and explain the obtained results.

4.1 Generation mechanisms

In this paragraph, the effect of the type and the size of different storage mechanisms on the arbitraging operation will be tested. The energy generators used in these tests are PV cells, wind turbines on land and wind turbines in sea. The size of the wind farm and PV field will vary between 50MW and 500MW.

4.1.1 Data analysis

In figure 3 the spot market price of energy is depicted. This figure shows the price of two days in February 2018. In this figure a clear and generalizable pattern in the price can be observed.

This pattern has two characteristics. First of all, the two daily peaks, one between 08:00 and 11:00 and the other between 17:00 and 20:00. The other identified characteristic is the daily minimum price during the night. The arbitraging operation stores energy when the price is low and sells to the market when price is high. Therefore we expect that energy is stored during the night and sold during one of the two price peaks.

In figure 4 a histogram of the relative daily price spread is depicted. The relative price spread is

the ratio between the highest and the lowest price of a day. A ratio of 1 implies that the lowest

and the highest price are equal and a ratio of .5 implies that the highest price is two times the

size of the lowest price. The histogram in figure 4 depicts the amount of times a price spread

occurred in three years (2016, 2017 and 2018). Only the largest spread of a day is included, so

1095 observations are depicted in this figure. As discussed in the previous section, the spread

should be of sufficient size to cover for the efficiency losses. Thus, when the efficiency of the

storage process is low, a large spread in the price is required to cover for the losses. Currently,

hydrogen storage has an efficiency around 35%. With the data used in this study, the spread

of 5% of the days can be used for the operation. In this study we will also consider multiple

days, or even seasonal storage. So the energy stored on one day can be sold to the market on

Figure 3: Spot market price two days

another day. We expect to observe storage for longer periods more often when the efficiency of storage is low.

Figure 4: Maximum daily price spread

Due to the unpredictable nature of wind, it is hard to depict a typical pattern. In figure 5 the

energy generated from turbines on land is depicted. In this figure, three days of the same date

in April, each from a different year (2012, 2013 and 2017) are depicted. As one can observe, generation varies strongly, therefore it’s hard to make expectations on the data.

A way to compare wind from land with wind from sea, is to compute the variation in the generation of the two mechanisms. To do this, the coefficient of variation is calculated. This is a measure to determine the relative variation of a data set. In Appendix B the formula and the background of this measure is provided. The coefficient of variation of turbines on sea is smaller than on land, respectively 0.67 and 1.02. When the generation pattern is more stable, the amount of energy generated is more equally divided over time. When a storage decision is made, the chance that energy is produced on that moment is larger. This will increase the utilization rate of the buffer when the storage decision is made. Therefore, we expect more energy is stored when the generation pattern is more stable.

Figure 5: Wind energy generation of three days

Unlike energy generated from wind, solar energy generation follows a very specific pattern.

In figure 6, three days in July in three different years (2012, 2013 and 2017) are depicted with the spot market price of one day in July 2017. We can observe that the lowest prices of energy occur during the night and that energy generated from solar is concentrated around noon. For the operation this would imply that on the best moment to store, there is no energy generation.

Therefore we expect that the amount of energy stored in the operation with PV cells is lower than for wind.

4.1.2 Scenario Details

In Appendix C, the values of the used parameters and the developed scenarios are summarized

and provided. The efficiency of the electrolyser and the fuel cell are, respectively 65% and

60%, based on Chardonnet et al. (2017) and Hwang (2013). The capacity of the electrolyser is

assumed to be 20MW, according to literature this will be the size of an electrolyser in which the

economies of scale are maximized (Chardonnet et al., 2017). All larger capacities will simply be

Figure 6: Solar energy generation

a multiple of this electrolyser and everything smaller is more expensive per MW. Thereby the largest electrolyser currently in development for the Delfzijl Haven has a capacity of 20MW.

The size of the fuel cell will be twice the size of the electrolyser. In order to isolate the effect of the type of generation mechanism and its capacity on the operation, the buffer is assumed to be infinitely large. The varying elements in the scenarios are the generation type (solar, sea and land wind) and its capacity (50, 250 and 500 MW). This implies nine tests will be performed.

4.1.3 Results

In Appendix C, the results of the first nine scenarios are presented. The results are averages of nine runs. In the experiments with sea wind, the amount of energy stored is highest. Thereby the amount of additional revenues is also largest for this generation mechanism. The operation in combination with solar energy generation produces the least additional revenues.

When we compare the amount of times the full capacity of the electrolyser is used, between the scenario (2) and (5), the reason for the difference in the amount of energy stored between sea and land wind becomes clear. In experiment 2, the electrolyser is, when used, in 69% of the times fully utilized. In scenario 5 this rate is only 54%. This implies that the energy generator couldn’t fill the electrolyser in 46% of the cases, thereby less energy is stored. This implies that the stability of the generation mechanism is very important for the results of the operation.

We we study the results of the tests with energy from PV cells, we can see the importance

of the stability of the generation mechanism. The utilization rate of the buffer, connected to

photovoltaic cells, is very low and therefore the amount of energy stored small.

4.2 Storage system setup

This paragraph is dedicated to test the effect of the setup of the storage facility on the operation.

Several scenarios will be developed in order to test the effect of the relative size of the buffer, electrolyser and fuel cell on the generated revenues. Similar to the previous section, first expectations will be made. These expectations are based on a data analysis and on the results of the previous set of experiments. Then scenarios will be presented and finally we will discuss the results.

4.2.1 Data analysis

In figure 7, the price of seven days in June 2017 is depicted, a similar pattern as in figure 3 can be observed. From this overview, which shows the price for a period of seven days, it becomes clear that the minimum price is relatively stable and the daily price peak varies considerably.

Furthermore, when a price peak occurs which is large enough to account for the efficiency losses of storage, it is just for a very short period. This implies that if such a peak occurs, as much advantages as possible should be taken from it. This applies to a lesser extent for low prices, since the periods in which they occur are more constant and the price is around the minimum for several hours. Therefore we expect that the operation will generate more revenues when the fuel cell is larger than the electrolyser.

In table 6 in Appendix C, the maximum amount of storage used per scenario is provided.

When the size of the buffer is larger than the maximum amount used, a part is not utilized and capital expenditures will be unnecessarily high. When it’s too small, the buffer will be full from time to time and energy has to be sold directly to the grid instead of stored in the buffer. We expect that lowering the buffer size will therefore diminish the additional revenues generated for each generation type.

4.2.2 Scenario Details

Table 7 in Appendix D presents the values of the parameters used. Table 8 presents the experiments performed and the values of the varying elements. Each single scenario will be tested with all three the generation mechanisms. The first nine experiments will test the effect of the relative size of the electrolyser, compared to the size of the fuel cell, on the operation.

To isolate this effect, the buffer size will be constant and infinitely large in these scenarios.

The following twelve experiments will be used to verify the effect of the buffer size on the operation. Therefore the other inputs are held constant. In the last two scenarios the relative size of the electrolyser and the buffer varies in combination with a limited buffer size.

4.2.3 Results

The goal of the first nine experiments was to test the effect of the relative size of the electrolyser

compared to the fuel cell on the operation. We expected to observe higher profits when the

fuel cell was larger than the electrolyser. However, the results show that the operation is most

productive, when the fuel cell and the electrolyser are of the same size. This holds for each

generation mechanism. When the electrolyser is larger than the fuel cell, the amount of energy

Figure 7: Spot Market price one week

stored is more than the fuel cell can discharge. Therefore the amount of storage left after one year is different from 0 in scenario 7 and 8. Thereby the total additional revenues in these scenarios is lower than in the other experiments.

The goal of the following twelve experiments was to verify the effect of the buffer size on the operation. As expected a larger buffer will increase the additional revenues. For solar energy, the marginal effect diminishes strongly. So the positive effect an increase in the buffer size has on revenues, diminishes when the buffer becomes larger. The effect on the operation with wind turbines is less straightforward. The increase in additional revenues, when the storage capacity increases from 1000 to 2000 MWh, is very small, 9.8% for sea wind and 16.3% for land wind. When the storage capacity increases from 2000 to 3000 MWh the increase is revenues is, respectively, 35.9% and 35.6%.

In the last six experiments, the effect of the relative size of the electrolyser and the fuel cell is tested. Different from the first nine experiments a limited buffer capacity is set. Under this condition, more revenues are generated when the electrolyser is smaller than the fuel cell. This only holds for an operation in combination with wind. For solar generation the revenues are slightly lower.

4.3 Technological developments

The last paragraphs of this section are devoted to verify the effect of future technological

developments on the operation. The efficiency of the electrolyser and the fuel cell will vary in

this scenario set.

4.3.1 Data analysis

The effect of an increase in the efficiency is expected to have a large impact on the revenues.

A higher efficiency will raise the value of the lower threshold, thereby the amount of energy stored will increase. The size of the required price spread for this energy is lower, since the amount of energy lost during the storage process is lower. This will increase the amount of times that storage is used and thereby the generated revenues. Thus, when we compare two scenarios, one with a higher roundtrip efficiency than the other, the revenues will increase since more energy is stored. In addition to this effect, the value of the energy which is also stored with a lower efficiency increases. Due to the higher roundtrip efficiency, the amount of energy lost is lower. This ’extra’ energy is sold to the grid as well and thereby increasing the revenues. This effect will increase the additional revenues even more.

In figure 8, the moments of storage and discharge, on a yearly basis, are depicted. The positive amount on the vertical axis is stored and the negative amount is discharged. On the horizontal axis the number of the month is given. This figure describes the output of a run of scenario 2 in the first experimental set. In this figure, we observe that storage is mostly used for weekly, or even seasonal periods. When the efficiency increases, the required price to sell stored energy, is lower. Thereby we expect that the storage length will decrease, when the efficiency increases. This implies we expect to observe daily storage more often in the scenarios with higher roundtrip efficiency.

Figure 8: Moments of storage and discharge in one year

4.3.2 Scenario details

In the first three scenarios the efficiency of the electrolyser and the fuel cell are respectively, 65% and 60%. This scenario is the current technological state. With this setup, each generation mechanism is tested. In the following experiments the efficiency of both tools increase with 5% to, respectively 70% and 65%. For the last scenarios this step is taken again. The values for efficiency in these scenarios are based on the realistic expectations discussed in the literature review. Each scenario is tested in combination with all three generation mechanisms.

The size of the generation capacity is 250 MW. To isolate the effect of technological developments

we assume that the buffer is infinitely large. Based on the results of the previous scenario set,

the size of both, the electrolyser and the fuel cell, is 30MW. For each generation method this was the size with the largest additional revenues.

4.3.3 Results

From the results presented in Appendix E we can conclude that the revenues indeed increase with technological efficiency. The amount of additional revenues in the operation with energy from sea wind increased with 85.5%, when the efficiency of both, the electrolyser and the fuel cell increased with 5%. The next increase in efficiency, increased the additional revenues with 63.1%. This effect was slightly less for wind turbines installed on land. The revenues of the operation with solar energy increased the least. The additional revenues increased with 18%

with the first efficiency increase and with the second 30.3%.

In figure 9 the moments of storage and discharge, on a yearly basis, is depicted. This figure describes the output of a run of scenario 7 in the third experimental set. When we compare this figure to figure 8, we can see that the amount of energy stored has increased significantly.

The time between storage and discharge has diminished. However, storage is still mainly used for longer periods. This implies that the operation is used to take advantage of price trends instead of daily spreads.

Figure 9: Moments of storage and discharge in one year with a high roundtrip efficiency

Figure 10 depicts the accumulated amount of energy stored for each hour of the day for a whole

year. The numbers presented in this figure are the outputs from a run in scenario 2 and 8 from

the third experimental set. The pattern in the moments of storage is similar when the efficiency

is increased. The peaks are at the same moments of the day in both experiments. In figure

11 the time when energy is discharged are depicted from the same experiments and runs as

depicted in figure 10. The same conclusions apply here. The peaks of discharge are at the same

moment of the day and in both experiments all energy is discharged between 05:00 and 23:00.

Figure 10: Moment on the day of storage of land wind

Figure 11: Moment on the day of discharge of land wind

5 Discussion

5.1 Limitations

Past research applied different methods to encounter this type of analysis. Each attempting to create the best arbitraging model. The simulation models proposed can, according the researchers of the models (Lund et al., 2009) (Sioshansi et al., 2009), realize up to 80-85% of the optimized value. Whether this model is of similar accuracy as the two aforementioned models is uncertain. However, we are certain that the value created by an arbitraging expert making real time decisions will generate greater revenues than this model (Nojavan et al., 2018).

Another limitation of this study is the lack of future price data. The energy market price is

the most important factor on which the findings of this research are based. To increase the

credibility of this exploratory research, expectations on future price data should be used in this study. The market price patterns are expected to change when the penetration of renewable energy in the market increases. According to economic market mechanisms, prices will increase when the generation of energy is relatively high, compared to the consumption. When total supply fluctuates more with the generation of RES, prices will fluctuate accordingly. Intuitively, this would increase the value of the proposed operation positively. Therefore, a study utilizing econometric price forecasts will be very insightful and meaningful for this operation and the literature writing about it.

5.2 Managerial Implications

This section will discuss the implications of our study for parties considering to invest in hydrogen storage. The content of this section is based on the actual advice we gave to Groningen Seaports, on the 20th of June 2019, and their feedback. This advice was given to Eertwijn van den Dool and Herbert Colmer, both project directors specialized in the energy and oil sector and to Henk Zwetsloot, the manager digital innovation. Each generation mechanism is provided with its own insight and implications.

Wind from sea

The results from this study show that the highest additional revenues are generated when the operation is performed with sea wind as generation mechanism. Under 80% of the tested circumstance, sea wind was the best generation mechanism. We expect that a stable generation pattern is the reason for this observation. When the simulation identifies a price below the storage threshold, the chance energy can be delivered by a generator is larger when the generation pattern is more stable.

We expected that the additional revenues would have been higher when the electrolyser was larger than the fuel cell. However, the results of the simulation show that, when storage is limited, the revenues are the highest when the fuel cell and the electrolyser are of the same size. When storage is infinitely large, more revenues are generated when the fuel cell is larger.

This could imply that this also holds for a buffer larger than 3000 MWh, since this is the largest buffer we tested in this study.

The revenues of the operation increase when the technology associated with hydrogen storage becomes more efficient. More energy is stored and additional revenues increase significantly.

When the efficiency of both, the electrolyser and the fuel, increase with 5% the operation generates 85.5% more revenues. Thereby the obtained results in this study are similar to the conclusion of Kloess and Zach (2014) and Kroniger and Madlener (2014). When the roundtrip efficiency is higher, the performance of the operation increases significantly. As Kloess and Zach (2014) already concluded, the time that energy is in the buffer, is often longer than one day. Even when the roundtrip is increased to 52.5%, we observe seasonal storage more often.

Wind from land

The additional revenues of the operation with energy generators on land is, after the operation

with sea wind, generating most revenues. On average the revenues are 13.5% lower compared

to the operation with wind turbines in sea. As discussed before, we conclude that this difference

can be explained by the variation of each generation mechanism. The implications for the operation with wind on land are the same as the ones we provided for the operation with wind from sea.

Solar

The operation in combination with energy produced by PV cells is, in terms of revenues, least productive. The patterns in the market price and the patterns of energy generation explain this conclusion. The lowest market prices occur during the night and the price peaks are in the morning around 09:00 and in the evening around 19:00. For the optimal operation this implies that energy is stored during the night and sold in either one of the price peaks. The energy generation pattern of PV cells is concentrated around noon and before 05:00 and after 22:00 barely any energy is generated. Thus, during the period with the lowest prices of the day PV cells are not generating any energy. Thereby the amount of energy stored is several times smaller than the amount of energy stored when wind generation is used. This decreases the additional generated revenues. In this technological state we can conclude, PV cells is the worst energy source, from the three energy generators discussed in this research. In addition, the effect of the increase in efficiency is smallest for this operation. The increase in revenues is only 18.2%, compared to the 85.8% increase for wind from sea. Therefore, we can conclude that the operation in combination with solar energy generation is the least productive setup, now and in the future.

5.3 Future research suggestions

The logical follow-up of this study is to determine the value of energy accommodation. Within

these frameworks, two types of values can be considered which the storage of over-produced

energy can bring. On the one hand the economic value and on the other the social and

ecological value. The second is hard to measure and therefore complex to model. However

to capture the positive externalities, from an ecological perspective, in the market, policies

should be developed. As discussed in the literature review, the aim of this study was, different

from Loisel et al. (2010), not to analyze the operation of the buffer as balancing tool. The

opportunity costs for this operation would be the value created by speculation. However,

when the market becomes fully dependent on RES, the usage of flexibility tools is practically

inevitable. According to this, in combination with the ambitious plans (Bodansky, 2016) of

governments to repress global warming, one can expect regulators have to create policies which

give incentives to accommodate renewable energy. In Hirth and Ziegenhagen (2015), several

articles are discussed estimating the price of balancing when RES penetrate the market. A

consecutive study could either build a model in which the day-ahead decision is made whether

it participates in the reserve market or applies hydrogen storage as arbitraging tool. Otherwise

a model can be made in which a provision is paid for the energy stored and released in specific

periods of the day or the year. Both would be exploratory studies to provide insight in the

future of flexibility operations. Kroniger and Madlener (2014) researched the application of

hydrogen storage for different tools.

6 Conclusion

In this paper, an attempt was made to provide Groningen Seaports and other investors, with

insight into the operation of hydrogen storage applies as arbitraging tool. In the literature, the

feasibility of the operation is broadly researched, however no research has been conducted

into the elements influencing the operation. For that reason, we studied the effect of the type

and the size of different generation mechanisms on the operation. In addition, the best way

to design a storage facility, for this purpose, was studied and finally we gave insight in the

operation when the associated technology develops in terms of efficiency. These insights were

used to advise Groningen seaports on their investment decision and contributes to the literature

by describing and explaining elements an application of hydrogen storage.

References

Abbaspour, M., Satkin, M., Mohammadi-Ivatloo, B., Lotfi, F. H., and Noorollahi, Y. (2013).

Optimal operation scheduling of wind power integrated with compressed air energy storage (caes). Renewable Energy, 51:53–59.

Abdi, H. (2010). Coefficient of variation. Encyclopedia of research design, 1:169–171.

Bodansky, D. (2016). The paris climate change agreement: a new hope? American Journal of International Law, 110(2):288–319.

Brisse, A., Schefold, J., and Zahid, M. (2008). High temperature water electrolysis in solid oxide cells. International journal of hydrogen energy, 33(20):5375–5382.

Bunge, M. (2012). Scientific research II: The search for truth. Springer Science & Business Media.

Chardonnet, C., Giordano, V., De Vos, L., Bart, F., and De Lacroix, T. (2017). Study on early business cases for h2 in energy storage and more broadly power to h2 applications. FCH-JU:

Brussels, Belgium , 228.

de F´atima Palhares, D. D., Vieira, L. G. M., and Damasceno, J. J. R. (2018). Hydrogen production by a low-cost electrolyzer developed through the combination of alkaline water electrolysis and solar energy use. International Journal of Hydrogen Energy, 43(9):4265–4275.

Denholm, P. and Sioshansi, R. (2009). The value of compressed air energy storage with wind in transmission-constrained electric power systems. Energy Policy, 37(8):3149–3158.

Fertig, E. and Apt, J. (2011). Economics of compressed air energy storage to integrate wind power: A case study in ercot. Energy Policy, 39(5):2330–2342.

Gatzen, C. (2008). The economics of power storage: theory and empirical analysis for central europe, volume 63. Oldenbourg Industrieverlag.

Hatami, A., Seifi, H., and Sheikh-El-Eslami, M. K. (2011). A stochastic-based decision-making framework for an electricity retailer: Time-of-use pricing and electricity portfolio optimiza- tion. IEEE Transactions on Power Systems, 26(4):1808–1816.

Hirth, L. and Ziegenhagen, I. (2015). Balancing power and variable renewables: Three links.

Renewable and Sustainable Energy Reviews, 50:1035–1051.

Hwang, J.-J. (2013). Sustainability study of hydrogen pathways for fuel cell vehicle applications.

Renewable and Sustainable Energy Reviews, 19:220–229.

Kloess, M. and Zach, K. (2014). Bulk electricity storage technologies for load-leveling operation–

an economic assessment for the austrian and german power market. International Journal of Electrical Power & Energy Systems, 59:111–122.

Korpas, M. and Holen, A. T. (2006). Operation planning of hydrogen storage connected to wind power operating in a power market. IEEE Transactions on Energy Conversion, 21(3):742–749.

Kousksou, T., Bruel, P., Jamil, A., El Rhafiki, T., and Zeraouli, Y. (2014). Energy storage:

Applications and challenges. Solar Energy Materials and Solar Cells, 120:59–80.

Kroniger, D. and Madlener, R. (2014). Hydrogen storage for wind parks: A real options evaluation for an optimal investment in more flexibility. Applied energy, 136:931–946.

Law, A. M., Kelton, W. D., and Kelton, W. D. (2000). Simulation modeling and analysis, volume 3.

McGraw-Hill New York.

Lipman, T. E., Edwards, J. L., and Kammen, D. M. (2004). Fuel cell system economics: comparing the costs of generating power with stationary and motor vehicle pem fuel cell systems.

Energy Policy, 32(1):101–125.

Loisel, R., Mercier, A., Gatzen, C., Elms, N., and Petric, H. (2010). Valuation framework for large scale electricity storage in a case with wind curtailment. Energy Policy, 38(11):7323–7337.

Lund, H., Salgi, G., Elmegaard, B., and Andersen, A. N. (2009). Optimal operation strategies of compressed air energy storage (caes) on electricity spot markets with fluctuating prices.

Applied thermal engineering, 29(5-6):799–806.

Nojavan, S., Najafi-Ghalelou, A., Majidi, M., and Zare, K. (2018). Optimal bidding and offering strategies of merchant compressed air energy storage in deregulated electricity market using robust optimization approach. Energy, 142:250–257.

Nojavan, S., Zare, K., and Mohammadi-Ivatloo, B. (2017). Application of fuel cell and electrolyzer as hydrogen energy storage system in energy management of electricity energy retailer in the presence of the renewable energy sources and plug-in electric vehicles. Energy conversion and management , 136:404–417.

Pellow, M. A., Emmott, C. J., Barnhart, C. J., and Benson, S. M. (2015). Hydrogen or batteries for grid storage? a net energy analysis. Energy & Environmental Science, 8(7):1938–1952.

Rogelj, J., Den Elzen, M., H¨ohne, N., Fransen, T., Fekete, H., Winkler, H., Schaeffer, R., Sha, F., Riahi, K., and Meinshausen, M. (2016). Paris agreement climate proposals need a boost to keep warming well below 2 c. Nature, 534(7609):631.

Schmidt, O., Gambhir, A., Staffell, I., Hawkes, A., Nelson, J., and Few, S. (2017). Future cost and performance of water electrolysis: An expert elicitation study. International journal of hydrogen energy , 42(52):30470–30492.

Sioshansi, R., Denholm, P., Jenkin, T., and Weiss, J. (2009). Estimating the value of electricity storage in pjm: Arbitrage and some welfare effects. Energy economics, 31(2):269–277.

Smil, V. (2016). Energy transitions: global and national perspectives. ABC-CLIO.

Steward, D., Saur, G., Penev, M., and Ramsden, T. (2009). Lifecycle cost analysis of hydrogen

versus other technologies for electrical energy storage. Technical report, National Renewable

Energy Lab.(NREL), Golden, CO (United States).

Appendix A - Assumptions

• The hydrogen buffer is not subject to self discharge. So the amount of hydrogen entering the storage is equal to the amount leaving it.

• The electrolyser and the fuel cell are not subject to ramp up rates. So both machines will function the moment they are demanded.

• The energy market is always capable of absorbing the energy this model supplies it with.

• The amount of energy this model supplies the market with is assumed not to influence the market or its price.

• The electrolyser and the fuel cell can handle all levels of input to maximum. So no minimal amount of hydrogen or energy is required.

• The productivity of each generation mechanism is assumed to be 20% and set for all experiments in the entire study. It is known to fluctuate among the different sources however it is assumed not to influence the relative obtained results.

Appendix B - Coefficient of variation

The coefficient of variation is the ratio between the standard deviation and the average of a data set. It is, different from the variance, a relative measure of the variation in a data set. The numbers presented in table 3 are based on distributions of generation for each mechanisms.

Therefore the averages are the same, however since this is a relative measure for the variation it suffice (Abdi, 2010).

Coefficient of variation formula: c

= σ µ

Standard Deviation Average CoV

Sea Wind 0.0076555 0.011416 0.6706

Land Wind 0.011677 0.011416 1.0229

Solar 0.017930 0.011416 1.5706

Table 3: Coefficient of variation

Appendix C Generation mechanism - Scenarios and results

Input Unit Value

Elektrolysis Capacity MW 20 Fuel Cell Capacity MW 40

Buffer Capacity MWh ∞

Elektrolyser efficiency % 65 Fuel cell efficiency % 60

Table 4: Parameter values of the first set

Input Unit (1) (2) (3) (4) (5) (6) (7) (8) (9)

Installed generation

Capacity MW 50 250 500 50 250 500 50 250 500

Renewable Energy Source Type Sea Sea Sea Land Land Land Solar Solar Solar

Table 5: Varying elements of the first scenario set

Unit (1) (2) (3) (4) (5) (6) (7) (8) (9)

Additional

yearly revenues € 37253 51011 56411 33435 43816 46424 3933 11078 15258 Amount of

energy stored MWh 3382 5812 6249 2913 4690 5178 887 1536 1696

Utilized

when available % 96.9 96.9 96.9 86.0 86.0 86.0 30.3 30.3 30.3

Fully utilized

when available % 6.5 69.1 80.5 19.4 54.0 63.1 5.6 18.4 21.6

Maximum

Storage used MWh 1837 3564 3929 1596 2704 3047 435 751 857

Table 6: Results of the first set of scenarios

Appendix D Operation setup - Scenarios and results

Input Unit Value

Installed capacity MW 250 Electrolyser efficiency % 65 Fuel cell efficiency % 60

Table 7: Parameter values of the second set

Input Unit (1-3) (4-6) (7-9) (10-12) (13-15) (16-18) (19-21) (22-24) (25-27) Buffer

Capacity MWh ∞ ∞ ∞ 500 1000 2000 3000 2000 2000

Electrolyser

Capacity MW 20 30 40 30 30 30 30 40 20

Fuel Cell

Capacity MW 40 30 20 30 30 30 30 20 40

Table 8: Varying elements of the second scenario set

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Additional

yearly revenues € 51011 43816 11079 57531 54940 14564 2953 14334 12648 Amount of

energy stored MWh 5812 4690 1536 8167 6514 2118 9673.7 8096 2611 Utilized when

available % 96.9 86.0 85.4 96.9 86.0 88.0 92.4 85.7 89.6

Fully utilized

when available % 69.1 54.0 18.4 61.6 47.3 15.5 51.8 42.0 13.6

Storage left

after one year MWh - - - 84 - - 1279 886 -

Table 9: Results of the second set of scenarios (1-9)

(10) (11) (12) (13) (14) (15) (16) (17) (18) Additional

yearly revenues € 18814 16872 4899 26546 26734 7494 29147 31092 11308 Amount of

energy stored MWh 2414 2269 1300 3054 2917 1793 4047 3915 2567 Utilized when

available % 34.0 36.5 96.0 41.1 44.5 93.7 51.9 56.5 89.9

Fully utilized

when available % 17.2 15.8 6.6 23.0 20.8 9.1 31.5 29.1 13.3

Storage left

after one year MWh - - - - - - - - -

Table 10: Results of the second set of scenarios (10-18)

(19) (20) (21) (22) (23) (24) (25) (26) (27)

Additional

yearly revenues € 39633 42136 12648 24060 25684 11308 39055 36079 11079 Amount of

energy stored MWh 5022 4709.6 2611 3677 3564 2567 4230 3749 1536 Utilized when

available % 64.0 66.0 89.6 40.2 42.1 89.9 74.3 71.5 85.4

Fully utilized

when available % 39.2 35.2 13.6 20.5 19.3 13.3 51.3 43.2 18.4

Storage left

after one year MWh - - - - - - - - -

Table 11: Results of the second set of scenarios (19-27)

Appendix E Technological developments - Scenarios and re- sults

Input Unit Value

Installed capacity MW 250

Buffer capacity MWh ∞

Elektrolyser capacity MW 30 Fuell cell capacity MW 30

Table 12: Parameter values of the third set

Input Unit (1-3) (4-6) (7-9) (10-12) (13-15) Electrolyser

efficiency % 65 70 75 75 65

Fuel cell

efficiency % 60 65 70 60 70

Table 13: Varying elements of the third scenario set

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Additional

yearly revenues € 57531 54940 14564 106744 83079 17214 174137 132749 22438 Amount of

energy stored MWh 8167 6514 2118 15530 13332 4691 31047 25538 9751 Utilized when

available % 96.9 86.0 88.0 97.5 88.4 83.1 97.5 88.7 81.4

Fully utilized

when available % 61.6 47.3 15.5 65.1 50.4 17.1 65.5 50.0 18.8

Table 14: Results of the third set of scenarios (1-9)

(10) (11) (12) (13) (14) (15) Additional yearly revenues € 90337 78004 16056 98143 84785 18589 Amount of energy stored MWh 15875 13735 4813 14421 12379 4356 Utilized when available % 97.4 88.2 83.2 97.5 88.4 83.1 Fully utilized when available % 65.8 50.4 17.0 65.1 50.4 17.1

Table 15: Results of the third set of scenarios (10-15)