Master Thesis
Hydrogen-based power generation to overcome grid capacity barriers
Alexander van den Brink August 7, 2019
A b s t r a c t
The purpose of this paper is to provide a method for overcoming electricity grid capacity barriers with hydrogen. Hydrogen is being utilized to either generate electricity or store excess electricity in the grid. This is achieved by making use of a fuel cell, an electrolyser, a hydrogen buffer and truck transportation. This paper provides an MILP model to optimize the involved hydrogen supply chain in terms of costs. The model is applied to two case studies. The first case study has a shortage of electricity supply in a grid, where hydrogen needs to be delivered to the grid with trucks for electricity generation. The second case study deals with an excess of electricity supply from a solar park, where the excess electricity is buffered as hydrogen and retrieved from the grid with trucks. The results show that for the case studies, frequent truck transport is favored over having a large hydrogen buffer. Furthermore, the utilization of the equipment is relatively low as it is only active in peaks, and the system could be connected to multiple grids to increase the utilization. The costs for solving the capacity problems with hydrogen is relatively expensive with current price levels, but technical advancements can make the system viable in the future. The model can be used as a comparison and decision-making tool, as it creates insight into the infrastructure cost, supply and requirements if hydrogen were to be used to solve capacity problems in electricity grids.
1 Introduction 2
2 Literature Review 5
3 Model description 8
3.1 Model development . . . . 8
3.2 Mathematical modelling . . . 10
3.2.1 Assumptions . . . 10
3.2.2 Mathematical model for solving electricity shortage . . . 11
3.2.3 Mathematical model for solving electricity surplus . . . 14
4 Case study 16 4.1 Case 1: Electricity shortage in Neerijnen. . . . 16
4.2 Case 2: Electricity surplus in Stadskanaal . . . 20
4.3 Scenarios: . . . 22
4.3.1 Experimental factors for case 1 . . . 22
4.3.2 Experimental factors for case 2 . . . 23
5 Results 24 5.1 Results of case study 1: solving electricity shortage . . . 24
5.2 Results of case study 2: solving electricity surplus . . . 29
6 Discussion 32
7 Conclusion 35
Appendix A 39
Appendix B 40
Appendix C 41
Appendix D 42
Appendix E 43
Appendix F 44
Appendix G 45
1
1 Introduction
Fossil fuels are gradually being replaced by renewable electricity due to the growing issue of global warming. This adoption of renewable electricity has a great impact on the electricity infrastructure, and energy distribution companies have trouble with fulfilling the rising demand in electricity in the grids (Farhangi,2010). The grid needs to be able to deliver high amounts of energy (i.e. charging electric vehicles), and at the same time receive high amounts of energy from distributed sources, like solar panels on houses (Mohd, Ortjohann, Schmelter, Hamsic and Morton, 2008; Roberts and Sandberg, 2011). There are major electricity grid upgrade projects being planned, but these upgrades lag behind demand (Alliander,2018;Mohd et al., 2008). Looking at a densely populated country as the Netherlands for example, expanding the electricity grid is costly and takes years to realize. It can range from one year for low capacity transmission lines to eight years for high capacity transmission networks to install. Dutch energy distribution companies are addressing the grid capacity as the main bottleneck for the energy transition (Alliander,2018).
The current grid is setup with a centralized infrastructure, meaning that there are a few main sources providing all the energy to the end-users, which are for example houses or companies. Due to the inevitable long distances between the source and end-user in this setup, the electricity grid is congested, and power lines have to have high capacities to distribute all the electricity. Furthermore, generated energy from end-users throughout the grid further increases congestion in the grid on irregular moments.Lo and Ansari(2012) summarize the main causes for power congestion in these centralized grids: (1) the rising demand and power generation, (2) insufficient capability of the grid, (3) peak demands (4) long distance demands and (5) a lack of power transmission lines in the grid.
Therefore, the goal of this paper is to research if decentralized electricity generation and buffering can viably overcome electricity grid capacity issues in the near future. Decentralized electricity generation will not be considered as an option to replace the centralized infrastructure in this paper, but rather as a method to deload the congestion in the centralized electricity infrastructure. Decentralized electricity generation brings the energy source and end-user closer, and can enable electricity to be delivered without needing to be transferred through the congested centralized distribution grid. This can bridge the bottleneck for the energy transition until permanent grid expansions are in place. Furthermore, decentralized buffering enables the possibility to buffer local excess energy, and again provide it back to the local grid without congesting the main grid or it could be sold to tank stations. If this solution is viable, grid expansions could be postponed or made unnecessary.
To achieve local energy generation and buffering, an energy carrier is necessary. Z¨uttel, Borgschulte and Schlapbach(2011) propose hydrogen as a viable energy carrier. Hydrogen is one of the most abundant resources on earth, and is gaining popularity as an energy source. It is capable of producing the highest energy per mass for any fuel (Jain,2009), and is increasingly used in energy systems. North Holland just revealed an investment plan of 2,8 billion euro for the next 12 years to integrate hydrogen into the energy systems (Investment Agenda Hydrogen North Holland,2019). Hydrogen is mostly mixed up with other chemical structures and thus needs a production process to be able to get pure hydrogen. The production process can be
performed in a renewable way, with the use of wind and solar power. Furthermore, fuel cells can be deployed to achieve a high efficiency for transforming the hydrogen back into energy.
Therefore, this paper will answer the following research question:
“Is hydrogen-based power generation a viable solution to overcome near-future grid capacity issues?”
To create insight into the costs and supply of using hydrogen for localized power generation and buffering, the whole supply chain must be assessed. A number of papers already exist on this specific topic, with Almansoori and Shah (2006) providing a Mixed Integer Linear Programming (MILP) model that includes this supply chain, so that the production, storage and distribution aspects of hydrogen use are considered. This model considers hydrogen production from large production facilities, uses storage facilities to cover fluctuations and distributes the hydrogen with compressed hydrogen trucks.Dayhim, Jafari and Mazurek(2014) considered the supply chain in a similar fashion, and modelled uncertain demand. The costs and CO2 emissions of different forms of hydrogen for transport and storage are researched in the works ofReuß, Grube, Robinius, Preuster, Wasserscheid and Stolten(2017). These works are all aimed at producing, transporting and using hydrogen.
The novelty of this research is the focus on the end of the supply chain. Where in earlier research the supply chain ended in solely delivering hydrogen, hydrogen will now be modelled to support the electricity grid in a decentralized manner. This means the electricity grid will be considered in the hydrogen supply chain, which is a new extension in the literature. To achieve decentralized production and buffering, a local system of a hydrogen buffer, electrolyser and fuel cell is needed. This system will hereafter be referred to as grid support system. The grid support system is supported by the hydrogen supply chain, which will utilize trucks for the transportation of hydrogen. By assessing the whole supply chain, from large scale hydrogen production to electricity being locally delivered to the grid, a realistic overview of the cost, feasibility and required hydrogen supply can be made to solve capacity problems in the electricity grid.
The mathematical model in this paper takes its basis from the MILP transportation model of Maheut and Garcia-Sabater (2013). They provide a truck transportation model, which we adjusted to hydrogen transportation based on truck data from Yang and Ogden(2007).
Furthermore, certain cost aspects of the supply chain are derived from the paper ofAlmansoori and Shah(2006). We also consider the electricity network, the hydrogen buffer, electrolyser and fuel cell in this model. The model is applied to two case studies in the Netherlands, and relevant data is gathered for these cases. Most importantly, data for electricity demand from houses and electricity generation from solar panels within the grid is needed. For electricity demand, input data is gathered from Enexis, an electricity distribution company in the Netherlands. For electricity generation, an online data tool from the European Commission is used that provides location-specific solar panel electricity generation data. (PVGIS,2017). Furthermore, several scenarios are tested on the two case studies to show the impact of the development of demand and supply in the upcoming years.
The model will be solved with the aim of minimizing the total cost. The grid requires a hydrogen buffer, where either hydrogen is taken from the buffer to create electricity or buffered electricity
in the form of hydrogen is stored. The model will optimize the local buffer size based on cost.
Increasing the buffer size decreases the frequency of trucks used for refilling or depleting the buffer, and vice versa. The results will be used to develop an overview of the cost and supply for hydrogen electricity generation and buffering. With this information, an assessment can be made for the viability of hydrogen to overcome electricity network capacity issues.
The remainder of this paper is structured as follows: Chapter two contains a literature review about the topic. Chapter three introduces the mathematical model, chapter four contains the case studies and in chapter five the results of the scenarios of the case studies are given. Chapter six contains the discussion and finally in chapter seven the conclusion and further research is given.
2 Literature Review
The main focus in this research is to provide a solution for capacity problems in electricity grids.
Therefore, the literature about electricity grids and how the demand and supply is managed in these grids will be discussed first. This research uses hydrogen as an energy carrier to generate or store electricity. For that reason, the main concept of using hydrogen in energy systems will be discussed, which is called the “hydrogen economy”. Continuing within the hydrogen economy, the available literature about hydrogen supply chains will be reviewed. This is done because the model in this paper aims to create insight into the complete supply chain, from the production to the end-use of hydrogen.
Energy grids are used all over the world to provide electricity, however, there are several trends that will require electricity networks to change. The two main drivers are (1) a growing share of renewable energy sources and (2) increasing number of plug-in hybrid electric vehicles (PHEV) and electric vehicles (EV) (Roberts and Sandberg,2011). A method of trying to overcome the shortcomings of the grid, and making it more flexible, is efficiently managing the demand and supply within the grid, which is called the “smart grid” (Farhangi, 2010). Smart grids coordinate the network in a more efficient and effective manner. Wade, Taylor, Lang and Jones (2010) defined the smart grid as a system that integrates all the users in the grid. It connects the electricity sources and consumers to create a stable and economically viable electricity network. On some instances, more electricity is being generated than used. In smart grids, this electricity should be stored to achieve the benefit of local electricity supply for a reduction of the grid load, and to prevent a loss of electricity. Roberts and Sandberg(2011) mention that electricity storage is of great importance in smart grids. Next to electricity generation, the consumption of electricity can also be regulated in smart grids. For example, a case study in Germany researching grid congestion pointed out that the electricity grid cannot support uncoordinated charging of electric vehicles in the upcoming years, even with the planned grid expansions (Staudt, Schmidt, G¨arttner and Weinhardt,2018). The smart grid integrates all users in the grid, andStaudt et al.(2018) researched the effect of coordinating the charging of electric vehicles instead of charging them instantly when they are connected. They found that with coordinated charging, the grid is able to handle the load and also increase its stability.
The local hydrogen buffer in this the grid support system creates the option for storing excess electricity that can be utilized for supplying the fuel cell or long term storage.Zhang and Xiang (2014) researched how a stand-alone micro grid functions with the setup for converting excess electricity to hydrogen and also converting it back to electricity with a fuel cell. Valverde, Rosa, Bordons and Guerra (2016) researched how smart grids with hydrogen as an energy carrier perform under different energy management strategies, where they also utilize an hydrogen electrolyser, buffer and fuel cell. Furthermore,Nastasi and Lo Basso(2016) researched the impact hydrogen could have on both electricity supply and heat supply for the energy transition.
Next to handling demand and supply in a “smart” manner, electricity grid extensions are essential in Europe to handle the rise in (1) renewable energy sources and (2) demand. These grid extensions are necessary to achieve the carbon emission reduction targets (F¨ursch, Hagspiel, J¨agemann, Nagl, Lindenberger and Tr¨oster,2013). In the work ofF¨ursch et al.(2013), a cost
estimation is made for EU grid extensions from 2010 to 2050.
Mohd et al.(2008) mention that there are not enough planning tools and models yet to support the integration of energy storage systems in smart grids. This means that this research, apart from delivering insight into decentralized hydrogen-based electricity production, can simulta- neously provide more in-depth knowledge about integrating storage into the electricity grid.
Furthermore, available research where hydrogen is considered as a decentralized method of electricity production is limited. Zhang and Xiang(2014) researched this, but they focused on the grid itself, and did not consider the hydrogen supply chain.
The pressing matter of global warming forces people to think about their impact on the environment, and there is a very noticeable trend towards sustainability. The term “Hydrogen Economy” was defined for the first time in literature byBockris and Appleby(1972). The concept revolves around using hydrogen as an energy carrier, and is applied for transporting energy from renewable energy sources over large distances (like wind parks on sea). Consequently, the hydrogen is stored and then transformed back to energy or used as a fuel where it can be combusted.Marb´an and Vald´es-Sol´ıs(2007) continued on exploring the hydrogen economy, and describe several paths which could transform the current energy sources to renewables.
Edwards, Kuznetsov, David and Brandon(2008) discuss the potential of hydrogen production, storage and fuel cells. They identify the main scientific and technical challenges that stand in the way of a hydrogen economy. Specifically for hydrogen fuel cells, the main challenges are addressed byEdwards et al.(2008). Furthermore,Tseng, Lee and Friley(2005) researched the impact of hydrogen, and summarized the overall challenges and opportunities based on the MARKAL model. This model analyses the economic aspects of energy systems over multiple decades, and has broad appliances. What they mention as one of the main challenges is creating a viable network of transportation for hydrogen.Bossel(2006) takes a critical stance against the actual viability of a hydrogen economy. His work suggested that only about 25%
of the generated energy is converted to practical use, and argued that it is not fit to be an energy carrier on a large scale. WhereasBall and Weeda(2015) argued mostly in favor of the hydrogen economy, and mentioned that recent energy trends continue to involve hydrogen.
The development and research in hydrogen technology due to the interest in it could very well explain the contrast between the perceived viability of hydrogen in 2006 and 2015.
The hydrogen supply chain generally contains three main components: (1) Hydrogen production plants, (2) Hydrogen storage units, and (3) Hydrogen distribution network (Almansoori and Shah,2006). Several authors have made optimization models for the dynamics of the hydrogen supply chain. Almansoori and Shah(2006) provided a general model, which is used for creating a future scenario for the design and operation of the hydrogen supply chain. It was later further extended to include multiple time-periods in it (Almansoori and Shah,2009). Dayhim et al.
(2014) considered the hydrogen supply chain in a similar fashion, but considered the effect of stochastic demand on the optimal design of the supply chain. Furthermore,Reuß et al.(2017) studied the effect of seasonal storage with hydrogen in the supply chain, and considered in specific the storage of hydrogen in organic compounds. The benefit of storing hydrogen in organic compounds is that it can be stored under ambient conditions, so no extreme temperature or high pressure is required. They found that liquid hydrogen for transportation becomes cost-effective for transportation or distances greater than 500 kilometers. Almansoori and
Betancourt-Torcat(2016) themselves provided a optimization model under emission constraints in a case study for Germany. Regarding distribution within the supply chain, generally two transportation methods are considered: trucks or pipelines (Reuß et al., 2017). The costs difference between these two modes of transportation are researched byYang and Ogden(2007).
They found that the best option is highly dependent on the characteristics of the situation (population density, amount of hydrogen to transport). The Netherlands is a densely populated country, and according toYang and Ogden(2007), a pipeline infrastructure could be viable.
There already exists a pipeline infrastructure in the Netherlands, but this is intended for natural gas. To be able to transport hydrogen with it, the pipelines need to be adapted. This can take years to realize, and thus it is not fit as a near-future solution. The aim of this paper is to provide a near-future solution for the capacity problems in electricity grid, and therefore only hydrogen transportation with trucks will be considered.
This research paper focuses on the hydrogen supply chain. Almansoori and Shah (2006) provided an MILP model for optimizing the hydrogen supply chain, andDayhim et al.(2014) also considered uncertain demand in a similarly modelled hydrogen supply chain. However, until now there is there are no optimization models available that consider the integration of the electricity network into the hydrogen supply chain. This extension will serve as a method to solve near future electricity grid problems with hydrogen. On the topic of smart grids,Roberts and Sandberg(2011) mention storing excess energy is of importance in these types of electricity grids. Nastasi and Lo Basso(2016) considered the use of hydrogen as an energy carrier within smart grids. Furthermore,Zhang and Xiang(2014) researched the performance of an electricity grid that uses hydrogen to buffer and generate electricity. To summarize, there are papers available for both the hydrogen supply chain and hydrogen in smart grids. However, literature that specifically integrates these two concepts is not available.
3 Model description
This section will give an overview of the mathematical model that will be used to answer the research question ”Is hydrogen-based power generation a viable solution to overcome near-future grid capacity issues?”. The research will focus on creating a mathematical model on two variations of capacity issues, namely electricity shortage and electricity surplus. First, a brief description of the system and information about the development of the model will be given. After that, the model for electricity shortage is given, and lastly the variation on the mathematical model for electricity surplus is given.
3.1 Model development
Ideally a model is created that fits within the observations in the reality to create insight into the process behind it.Karlsson(2016) classifies this type of research as descriptive empirical model based research. To ensure a fit with reality, several experts from Enexis, the director of the Centre of expertise energy and the innovation manager of Tennet have been contacted throughout the research period. They were asked for their view on feasibility of the idea of a local hydrogen fuel cell, buffer and electrolyser to solve electricity problems. Furthermore, they provided input that was used for the case studies and scenarios.
Figure 1: Hydrogen based supply chain
First, electricity is generated by a certain source, which could be from solar panels or wind turbines for example. This electricity is then converted to hydrogen by electrolysis and stored in a central buffer. Next, the hydrogen is transported by a truck to deliver it to the local grid support system. The grid support system can use the delivered hydrogen to convert it back into electricity to solve shortage problems in the grid. Furthermore, in case of an electricity surplus in the grid, the grid support system can convert the excess electricity into hydrogen.
The trucks can then be utilized to bring the produced hydrogen back to the central buffer.
Almansoori and Shah(2006) provide the basics of a hydrogen supply chain in their paper.
However, in this paper we also consider electricity generation or buffering within the electricity grid. Figure1above indicates the newly considered aspects of the hydrogen supply chain, compared to the hydrogen supply chain fromAlmansoori and Shah(2006)
Trucks are used as a method of transportation to refill or deplete the local hydrogen buffer of the grid support system. This buffer provides the needed hydrogen for the fuel cell to transform it into electricity for the grid. It is important to notice that the local hydrogen buffer opens up the option for storing excess energy of the grid. However, to realize this, the excess energy will first need to be converted into hydrogen by electrolysis. Though truck transport was included in the model of Almansoori and Shah(2006), it was not modelled for triggering transport based on inventory levels, and the model could not be applied for the purpose of this research.
Therefore, the basics of the MILP truck-transport model byMaheut and Garcia-Sabater(2013) were incorporated to accurately model the needed transport for fulfilling the hydrogen demand.
Table1below summarizes which aspects were derived from the two papers, and what aspects are modelled by the author.
Source: Model component:
Almansoori & Shah (2006) - Cost parameter for buffer and transport Maheut and
Garcia-Sabater (2013)
- Inventory buffer
- Distribution with trucks
Author
- Grid electricity demand and supply - Grid hydrogen demand and supply - Conversion factors
- Requirements of electrolyser and fuel cell
Table 1: Model components
3.2 Mathematical modelling
First, the assumptions on which the model is built will be given. Thereafter the model for solving an electricity shortage is given. Due to the different dynamics of solving an electricity surplus, a different mathematical model will be introduced with minor alterations compared to the first model.
3.2.1 Assumptions
The two variations of the mathematical model are built on the following assumptions:
- The demand and supply of electricity is known.
- The capacity of the electrolyser in kW is determined by the hourly peak of electricity input it needs to be able to handle.
- The capacity of the fuel cell in kW is determined by the highest hourly electricity peak the grid needs.
- Trucks can be used again in the next time period. The aim is not to make a planning for individual trucks, but to indicate how many trucks are needed to deliver or retrieve hydrogen on a certain time instance.
- Once the hydrogen is stored in the buffer, no costs will be incurred over time.
- The hydrogen carrying limit of a truck is 10.000 kWh, based onYang and Ogden(2007).
- Trucks only have one function, either to bring hydrogen to the grid or pickup hydrogen from the grid, depending on the case.
3.2.2 Mathematical model for solving electricity shortage
To solve electricity shortage, a fuel cell is needed to deliver electricity on times where the grid does not have enough capacity. The hydrogen buffer serves as the input for the fuel cell.
Furthermore, an electrolyser is needed to generate hydrogen on time instances where the grid has excess electricity, for example during periods with a lot of sun. The aim is to minimize cost for supplying the needed electricity to the grid, and this is done by optimizing two decision variables: (1) hydrogen buffer size in the grid, (2) transport of hydrogen by a truck.
Set notation:
g ∈ G Grids j ∈ J T rucks t ∈ 1, .., T T ime periods
Decision variables:
ktgj =Hydrogen delivered to grid g by truck j at time t
atgj =1 if truck j is used for delivering hydrogen to grid g on time t, otherwise 0 zg =Hydrogen buffer size for grid g in kWh
btg =Hydrogen buffer level in grid g on time t
Parameters:
Egt =Electricity demand from grid support system g on time t Sgt =Excess electricity supply from grid support system g on time t Wg =Distance from main supply to grid g
L =Hydrogen carrying limit in kWh for a single truck Ng =Amount of trucks available for the supply of grid g TF =Fuel cell conversion factor
TE =Electrolysis conversion factor CF =Cost per kw output for fuel cell CE =Cost per kw output for electrolyser
CA =Average cost per kilometer driven with truck CT =Capital cost per truck
CO =Operations & Maintenance cost trucks (fraction of capital cost)
CH =Cost per kWh hydrogen buffer
Fg =Max kW output required from fuel cell in grid g
Ig =Max kW input capacity required from electrolyser in grid g bstartg =Initial hydrogen level in buffer in grid g on time t
bmin =Lower limit hydrogen buffer tank M =large number
Constraints:
b0g = bstartg ∀ g ∈ G (1)
btg ≥ bmin ∀ g ∈ G, t ∈ T (2)
btg =
J
X
j=1
ktgj+ bt−1g + Sgt∗ TE − Egt
TF ∀ g ∈ G, t ∈ T (3)
J
X
j=1
kgjt ≤ zg− bmin ∀ g ∈ G, t ∈ T (4)
btg ≤ zg ∀ g ∈ G, t ∈ T (5)
kgjt − M ∗ atgj ≤ 0 ∀ g ∈ G, t ∈ T, j ∈ J (6)
kgjt ≤ L ∀ g ∈ G, t ∈ T, j ∈ J (7)
atgj ∈ {0, 1} ∀ g ∈ G, t ∈ T, j ∈ J (8)
kgjt ≥ 0 ∀ g ∈ G, t ∈ T, j ∈ J (9)
zg ≥ 0 ∀ g ∈ G (10)
First, the starting level of each hydrogen buffer in grid g is set by constraint 1. Next, the minimum level of hydrogen in the buffer is given by constraint2. The buffer level on each time instance is set by constraint3, and will now be explained. First, the sum of hydrogen being delivered to the buffer by the trucks in j on time t is taken. Added up to this is the buffer level of the previous time instance. Next, the hydrogen received from the electrolyser on time t is added. Hydrogen received from the electrolyser is defined as the electricity supply (Sgt) times the electrolyser efficiency factor (TE). Lastly, the hydrogen needed for the fuel cell on time t is subtracted from the buffer level. Hydrogen needed for the fuel cell is defined as (Egt) divided by the fuel cell efficiency factor (TF). There is always some hydrogen left in the buffer due to constraint2, and constraint4makes sure that there is not more hydrogen being delivered by trucks in j on a single time instance than the buffer is able to store. Constraint5restricts the buffer level to be smaller or equal than the buffer size at all times. Furthermore,6triggers the binary variable Atgjif there is hydrogen to be transported to the buffer with truck j. The hydrogen carrying limit for a single truck is set with constraint7. Constraint8restricts the variable as binary and finally, constraint9and10are non-negativity constraints.
Objective function:
The objective function (11) aims to minimize several cost aspects related to solving grid capacity issues. Three cost aspects are to be minimized, and these are cost of hydrogen supply by trucks (CT), cost of the buffer sizes (CB) and the costs of the fuel cells and electrolysers (CC). From the results, the viability of decentralized electricity production with hydrogen can be assessed with different modes of supply, and ultimately it can be compared to the cost of grid expansions.
minimize CT + CB + CC (11)
Below, the equations for the different cost aspects are given.
CT = CA ∗
G
X
g=1 T
X
t=1 J
X
j=1
Atgj∗ 2 ∗ Wg+ CT ∗ Ng+ CO ∗ CT ∗ Ng (12) Equation12defines the total costs of using trucks to deliver hydrogen. It determines how many trucks are used on a certain time point, and multiplies this by the distance between the supply point and grid (Wg) times two, to represent a back-and-forth trip. This results in the total distance travelled, and in turn this is multiplied it by the average cost per km. Furthermore, the capital cost and operations & maintenance cost are included based on the amount of trucks used in the model, indicated by Ng.
CB = CH ∗
G
X
g=1
zg (13)
The costs of all the kWh hydrogen buffer required for the grids is given by equation13.
CC = CF ∗
G
X
g=1
Fg+ CE ∗
G
X
g=1
Ig (14)
Finally, the costs of the fuel cells and electrolysers are determined by equation 14. Fg is determined by max{Egt : t ∈ T }, and Ig is determined by max{Sgt : t ∈ T }.
3.2.3 Mathematical model for solving electricity surplus
Dealing with an electricity surplus requires a different approach than solving an electricity shortage. In this case, the only function of the grid support system is to buffer all the excess electricity in the grid, and thus any costs related to the fuel cell are excluded in this model.
Trucks will be used to empty the buffer and transport the generated hydrogen to a hydrogen tank station or central buffer for example. The model will minimize cost based on (1) the size of the hydrogen buffer size and (2) the amount of hydrogen that is depleted from the hydrogen buffer with a truck on each time instance.
For this model, the following changes will be made to the mathematical model from section 3.2.2.
Decision variables
Decision variable kgjt will be replaced by rgjt , and the definition of the binary variable atgj
changes.
rgjt =Hydrogen retrieved from grid g by truck j at time t
atgj =1 if truck j is used for retrieving hydrogen from grid g on time t, otherwise 0
Parameters
This model excludes the parameters Egt and Fg, since there is no electricity demand in this model, and thus no fuel cell will be needed. Furthermore, the following parameters have a different definition.
Ng =Amount of trucks available for retrieving hydrogen of grid g Wg =Distance from grid g to hydrogen delivery point
Constraints
Constraint2will be changed to constraint15below, as there is no need for a safety level if the buffer is only used for storing electricity.
btg ≥ bmin ∀ g ∈ G, t ∈ T (15)
Constraint3will be replaced by constraint16given below.
btg = Sgt∗ TF + bt−1g −
J
X
j=1
rtgj ∀ g ∈ G, t ∈ T (16)
The buffer level is determined by the generated hydrogen from the electrolyser plus the buffer level from the previous time instance. Subtracted from this is the hydrogen being retrieved by means of a hydrogen-carrying truck, indicated with rgjt .
Objective functions:
This model uses the same objective function as the one that is given in (11), only for the cost aspect CC a new equation is introduced, which is given below.
CC = CE ∗
G
X
g=1
Ig (17)
The costs of the electrolysers are determined by equation17.
Ig is determined by max{Sgt : t ∈ T }.
4 Case study
In this section, two cases are given which will be used as input for the mathematical model.
These cases are located in the Netherlands, where there are two types of capacity issues in the electricity network. These are: (1) an electricity shortage and (2) an electricity surplus.
These capacity issues are a high priority to solve, because they are holding renewable-energy initiatives for the Netherlands back.
4.1 Case 1: Electricity shortage in Neerijnen.
The first case is the electricity grid of Neerijnen, which is located in the province of Gelderland.
The relatively large amount of greenhouse farms in Neerijnen are resulting in a high demand for electricity. Recently, they switched from fossil fuels to electricity as primary energy source.
However, the electricity network there is almost at the limit for electricity supply. Figure2 visualizes the general steps in an electricity network, from production to distribution.
Figure 2: Electricity network components
The problem for Neerijnen lies in the medium-voltage electricity network, which generally operates between 10 and 20 kV. This is the capacity-restricting bottleneck. The low-voltage network beneath it generally has more that 50% of it’s capacity left, according to a distribution system operator from Enexis. Therefore, the grid support system, visible in figure1, which contains the fuel cell, buffer and electrolyser, will be connected to the low-voltage electricity network, after the bottleneck. In this setup, the grid support system can fill in the peaks of shortage in the grid and ensure houses and companies get the electricity they require. A more technical setup, consulted with two distribution system operators from Enexis, is given in figure3below. Important to note is that the electricity output from the fuel cell first needs to be converted before it can be connected to the local 230V grid.
Figure 3: Grid support system overview
Liander is the company responsible for the electrical grid, and they have put a hold on further high-electricity grid connection requests. Current customers are still provided with electricity, but for example company expansions that require extra electricity have to wait until the grid expansion is in place. Liander is working on it and stated that the grid expansion, that is carried out in collaboration with Tennet, can take three to four years for Neerijnen. In the future, Neerijnen can expect electricity shortage if the electricity demand grows. Therefore, a simulation of electricity shortage for this case will be carried out to create input for the mathematical model. First, the data for the simulation will be given.
Data:
Electricity usage for this case is gathered from a low-voltage transformer station that supplies around 150 households belonging to Enexis, which is a network operator in the Netherlands.
Hourly electricity data from February 2018 to February 2019 is available.
To simulate a shortage in the electricity grid, a distribution system operator from Enexis advised to set a certain capacity limit on the transformer station, and to consider electricity consumption above this point as electricity shortage. Based on the data set, the capacity limit is set to 80 kW. Figure4displays the result of the limit on the data set. Most hourly consumption is beneath this point, but there are peaks which represent electricity that cannot be delivered by the transformer station to the network.
Figure 4: Hourly data analysis - 80 kW limit
The electricity generation within the grid comes from solar panels on houses. Due to the unpredictable nature of renewable energy, it can happen that on some moments solar electricity generation exceeds the amount of electricity that needed in the grid. On those instances, the excess electricity needs to be buffered with the electrolyser. There is no exact data available for the number of installed solar panels on houses in Neerijnen, so it is based on the following calculation clarified below.
At the start of 2018 there were around 550.000 houses in the Netherlands that had solar panels installed. On average, there are 12 solar panels installed on each house, with a capacity of 280Wp (Solar Solutions Int., 2018). The total amount of households in the Netherlands is around 7.9 million (Central Agency for Statistics,2018), and this leaves us with an average of 7% of houses with solar panels. This means that in the grid, which contains 150 houses in this case, that there are 11 houses (150*0.07) with solar panels in the grid. This leaves us with a total installed capacity of 36.96kWp of solar panels in the grid for this case. To gather realistic, location-dependent solar production data for Neerijnen, the Photovoltaic Geographical Information System tool from the European Commission is used. It gives hourly electricity generation data for solar panels based on historical solar radiation data (PVGIS,2017). Appendix A contains the specific settings for the electricity generation in Neerijnen.
Simulation for input model
The consumption data from Enexis as well as the solar panel production data is delivered in hourly time instances. As mentioned before, the capacity limit is set to 80 kW. For every hour that electricity consumption is above the limit, the remaining electricity has to be supplied by the fuel cell, so this is the input parameter Egt. For every hour that solar electricity generation is greater than the electricity consumption on that hour, there is an excess of electricity. This excess needs to be buffered, and this is the input parameter Sgt. This simulation is carried out in Microsoft Excel. To limit the computational time of the hourly values for the parameters are combined and converted to daily values (t=1 day).
4.2 Case 2: Electricity surplus in Stadskanaal
The second electricity grid case is located in Stadskanaal, in the province of Groningen. Since 2018, there are feed-in capacity problems in the electricity grids of North-Holland, where Stadskanaal is located. Feed-in means delivering electricity from end-user back to the grid. The problem mainly originates from the sudden increase in solar panels. Farms are recognizing the added value that solar panels can bring, and are placing them in large amounts on their property. Next to that, large scale solar parks are being built all throughout the Netherlands to increase the share of renewable electricity in the Netherlands. The result is that too much electricity is being generated by customers, and the grid is reaching it’s maximum limit on how much electricity can be transported back through the cables. The company which is responsible for the grid, Enexis, is trying to find solutions for this problem together with the province and high-voltage network operator TenneT. Indications on the site of Enexis mention that expansions could take up to 10 years. For Stadskanaal, there is no electricity grid capacity left for new electricity generating sources, and thus plans for new solar parks are put to a halt. In Appendix B a map from Enexis can be found that shows the zones in North Holland that deal with capacity problems when feeding generated electricity back into the grid. In the appendix, it is visible that Stadskanaal is located in a problematic zone. Recently, a plan for the installation of 246 solar panels on the roof of a football club in Stadskanaal was denied because of the capacity restrictions (RTV Drenthe,2018). A simulation will be carried out for this solar park, where there will be a set capacity limit in the grid. The electrolyser and buffer from the grid support system in figure1can store the electricity that exceeds the capacity limit in the grid.
Similar to the case in Neerijnen, the bottleneck is also caused by the medium-voltage network.
For a clear overview of the complete network, see figure2. Most solar parks are connected to the medium voltage network, and seeing as the electrolyser uses the electricity from the solar park, the grid support system will be connected to the medium voltage network.
Data:
Like with the first case, the Photovoltaic Geographical Information System tool from the European Commission (PVGIS,2017) provides location-dependent solar electricity generation data for these 246 solar panels. The installed solar panel capacity in the grid comes to a total of 68.88 kWp, based on the average solar panel having a capacity of 280 Wp (Solar Solutions Int., 2018). Appendix A contains the specific settings for the electricity generation in Stadskanaal.
The electricity network in Stadskanaal does have some feed-in capacity left, only there is not enough capacity left to facilitate complete solar parks. To simulate this situation, a limit will be set to the solar production data as to represent the grid having limited capacity. Everything above the limit will have to be stored as hydrogen with the electrolyser. Based on the data set, the capacity limit is set to 30 kW, displayed in figure5below. The grid support system is meant to handle capacity problems due to peaks in electricity supply or demand, and this limit represents significant peaks, especially in the summer, that have to be buffered.
Figure 5: Hourly data analysis - 30 kW limit
Simulation for input model
As before, the solar production values is available in hourly instances. At each hour, the solar electricity production was checked if it was above or below the limit of 30 kW. If the solar production was below the limit, no action has to be taken as the grid can handle these amounts of electricity. If on a certain hour the electricity production exceeds this limit, the remaining electricity above the limit needs to be buffered by the electrolyser. This data is the input for the parameter Sgtfor the mathematical model for electricity surplus. The simulation is carried out in Excel. Like before, the hourly values for the parameters are combined and converted to daily values (t=1 day) to limit the computational time of the model.
4.3 Scenarios:
There are multiple factors that influences the outcome of the grid support system on the cases above. Grid expansions can take up to 10 years to complete, and it is interesting to create insight into the possible developments of technology prices, electricity demand and installed solar capacity in this time period. In this section, scenarios will be given to create interesting insights for the possible developments in 10 years. The two cases will have different experimental factors, and the experimental factors will be discussed per case.
4.3.1 Experimental factors for case 1
The first experimental factor is the price of technology. The development of hydrogen technol- ogy needs to be considered, because technological advancements are likely to decrease cost over time. This can make the use of hydrogen more viable. Price predictions for the fuel cell, electrolyser and hydrogen buffer for the years 2020, 2025 and 2030 are given in Appendix G.
The values for the electrolyser are derived from the works ofSaba, M¨uller, Robinius and Stolten (2018), the values for the fuel cell fromChardonnet, Giordano, De Vos, Bart and De Lacroix (2017), and finally, the cost values for hydrogen buffering are derived fromHydrogen Europe (2018). It is important to mention that these values represent the purchase value. The case study however covers one year, and therefore the price per year is calculated, based on the expected lifetime of the equipment. Furthermore, operation and maintenance costs also need to be taken into account. The complete of the cost factors can be found in Appendix G, and the resulting yearly costs for the case study are summarized in table2below.
Parameter Description 2020 2025 2030
Cf Cost per kW fuel cell € 630 € 525 € 315 Ce Cost per kW electrolyser € 131,25 € 115,50 € 78,75 Ch Cost per kWh buffer € 1,84 € 1,58 € 1,05
Table 2: Yearly cost prediction of hydrogen technology in 2020, 2025 and 2030
The second experimental factor is the number of solar panels on houses. In 2018, the amount of installed panels increased with 20 percent, mainly due to subsidies. Alliander(2018) mentions in their yearly report that they are expecting a continuous increase of solar panels. The assumption is made that on average the installed solar energy capacity will increase with 15% per year until 2025. Any further increase on the installed solar capacity will change the capacity problem from shortage into surplus. This is not desirable, as the surplus problem is discussed in the second case study. Therefore, the installed solar capacity in 2030 will not be considered here.
Year Solar panel installed capacity 2020 Current installed capacity (100%) 2025 200% of the installed capacity in 2020
Table 3: Predictions about solar panels
The last experimental factor is the growth of electricity demand. Electrical vehicles and the energy transition will create increased electricity demand. The values for expected growth are derived from the scenarios from the forecast report (2018 until 2027) fromLiander(2017), and are given in table4.
Year Expected growth of electricity demand 2020 0% (current electricity demand)
2025 5%
2030 10%
Table 4: Predictions about electricity consumption
The complete set of scenarios for case 1 is given in Appendix C.
4.3.2 Experimental factors for case 2
The first experimental factor for the second case study is also the price of hydrogen technology.
Since there is no fuel cell in the grid support system of the second case study, only the cost for the hydrogen buffer and the cost for the electrolyser are considered, The values for the electrolyser are derived fromSaba et al.(2018) and the cost values for hydrogen buffering are derived fromHydrogen Europe(2018). Like with the first case, the lifetime and the operation and maintenance cost of the equipment are taken into account. The resulting values are given in table5below, while the total overview for the cost factors can be found in Appendix G.
Parameter Description 2020 2025 2030
Ce Cost per kW electrolyser € 131,25 € 115,50 € 78,75 Ch Cost per kWh buffer € 1,84 € 1,58 € 1,05
Table 5: Yearly cost prediction of hydrogen technology in 2020, 2025 and 2030
For further test how the optimal costs and supply of the grid support system of the second case study change over different situations, an increase of size the solar park will also be considered in the scenarios. The different percentage increases of the solar park size are given in table6.
Solar park size Size in kWp (Current size in case)100% 68.88
105% 72.32
110% 75,77
Table 6: Different capacities solar park
5 Results
In this chapter, the results of the mathematical model applied to the case studies will be discussed. The chapter is split up into two parts, where the first part discusses the results of the first case study, which deals with electricity shortage. The second part discusses the results of electricity surplus case study. The specific input parameters for case study 1 and 2 can be found in Appendix D.
5.1 Results of case study 1: solving electricity shortage
The base scenario for Neerijnen (scenario 1 in Appendix C) represents the present-day solution for solving the simulated electricity shortage, and will be discussed first. The results from running the optimization model showed that the cost for covering shortage for a year with electricity shortage is €73.348,50, and the optimal buffer size is 1927 kWh. Figure6shows the behaviour of the buffer over the whole simulated year.
Figure 6: Buffer level over whole year, scenario 1
There are sharp peaks in the beginning and end of the year. These peaks represent the buffer being filled with hydrogen by trucks. After the buffer is filled, it decreases over time until the buffer nears its lower limit of 500 kW and truck transportation is used again to refill the buffer. The lower limit of the buffer can cover two high demand days of electricity so that the buffer can function for a certain time period without needing trucks. The buffer size is about four times the size of the lower limit, and not meant for covering long periods of time. The cost-effective option for this case is to use a smaller buffer size and incur more trucks over the year to re-supply the hydrogen tank. Noticeable is the fact that the buffer continuously reaches it’s peak capacity, meaning that in the situation where hydrogen is being delivered to the buffer, it will likely be refilled to it’s maximum capacity. From day 70 to day 200 in figure6
the buffer level only increases and decreases by small amounts. This time period is in the spring and summer, and electricity consumption is generally less than the electricity consumption in colder months. This can be seen in figure4, where the grid capacity limit of 80 kW per hour is rarely reached in this time period. Furthermore, the spring and summer months have high sun-hours, and this means that the solar panels will produce more electricity. If there is any electricity shortage in this time period, it will likely be matched with the electricity from the solar panels. Towards the end of the year, the demand grows and leads to sharp decreases of the hydrogen buffer level. If the lower limit is almost reached, the buffer is refilled again.
The electrolyser, which is powered by the solar panels, generates only small amounts of hydrogen during the summer months. This is because of the usage prioritization on generated electricity from solar panels. Due to the efficiency losses of the electrolyser, it makes sense to first use the electricity to meet the demand of the grid. The simulation is set-up so that generated solar panel electricity will only be buffered with hydrogen if it exceeds the electricity demand in that time period.
The base case has a moderate amount of solar panels installed in the grid, and therefore it will not occur frequently that the generated solar electricity exceeds electricity demand. In total, the electrolyser is activated for 71 days (19% of total time), with an average electricity input of 9.3 kWh for a complete day when it is used. Logically, if the electrolyser is used, it is usually in high sun-hours, which are for example between 10am and 3pm. For the base case, the electrolyser is capable of transforming 20kW electricity to hydrogen each hour, based on the peak supply to the electrolyser. This means that is the electrolyser has a low utilization rate, and if it is used, it is generally used far below it’s capacity. The fuel cell however, is activated for 271 days (74% of the total time) and delivers an average of 150 kWh on the days it is used.
The installed capacity of the fuel cell is 64 kW, and thus can deliver 64 kWh per hour. The fuel cell, as well as the electrolyser, is meant to handle peaks in the grid, which occur on irregular intervals and cause lower utilization rates. However, the fuel cell is still utilized most of the days, and runs for multiple hours on usage days.
Influence of hydrogen technology price
The results in table 15, which can be found in Appendix F, show that the predicted lower hydrogen technology prices of 2025 and 2030 cause lower total costs, as expected. When looking at the cost reduction caused by the hydrogen technology prices across the different installed solar capacities and electricity demand, the cost reductions in percentages are very close to each other. This buffer size results in more variation when dealing with lower technology prices, and the results are displayed below, where figure7displays the scenarios with an installed solar capacity of 100% and figure8displays the scenarios with 200% installed solar capacity.
∆Y in the figures represents the considered year of hydrogen technology prices.
Figure 7: Influence of hydrogen technology prices, with a solar capacity of 100%
Figure 8: Influence of hydrogen technology prices, with a solar capacity of 200%
The buffer size generally increases for all scenarios that includes an installed solar capacity of 100%. The buffer size increases with larger increments for ∆Y=2025 and ∆Y=2030, mainly because of the lower price per kWh buffer compared to ∆Y=2020. In contrast with the previous graph, the buffer sizes decrease for scenarios with an installed solar capacity of 200%. This will be further explained in the sub-chapter below. The important aspect to note here the comparison of the patterns between ∆Y=2020, ∆Y=2025 and ∆Y=2030. In both graphs7and8, the buffer sizes across the scenarios seem to increase and decrease largely in a similar way.
