• No results found

Flexibility in Energy Systems Validation of a flexibility framework

N/A
N/A
Protected

Academic year: 2021

Share "Flexibility in Energy Systems Validation of a flexibility framework"

Copied!
42
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Flexibility in Energy Systems

Validation of a flexibility framework

Master Thesis

S.W. Rutgers (S1789570)

MSc. Technology and Operations Management 26-06-2017

Supervisors

prof. dr. ir. J.C. Wortmann ing. R.J.H. van der Burg, MSc University of Groningen

Faculty of Economics and Business P.O. Box 800

9700 AZ Groningen

Abstract

The advent of energy grids with large penetration of renewable energy sources (RES) adds complexity to grid balancing. RES generation cannot be fully controlled by their nature of being dependent on e.g. solar or wind energy. Generation can therefor not be fully responsive to variations in demand, which potentially destabilizes energy grids. Grids thus have an increasing demand for flexibility, which could be provided by flexibility service providers. Both providers and clients of these services need a measure to determine the quantities needed, or offered, and in what quantities these can be traded, possibly subject to constraints.

This thesis sets out to verify a flexibility framework currently under development at the University of Groningen. Data on characteristics of elements in a power grid is gathered by performing a

simulation study, and the behavior and flexibility performance in the simulation is compared to predictions made by the flexibility framework. Results show that the framework provides adequate unit of trade, but does not accurately quantify flexibility potential of systems or elements. This inaccuracy is due to nonlinear behavior caused by external influences such as temperature, which is not currently included in the model. Based on these results, it is recommended to increase the value of the flexibility framework as a metric by pinpointing and including nonlinear constraints.

(2)

Table of contents

1. Introduction

6

2. Theoretical background

8

2.1 Flexibility

8

2.2 Flexibility sources

8

2.2.1 Demand-Side Management 2.2.2 Supply aggregation 2.2.3 Storage 2.2.4 Conversion

2.3 Stakeholders

9

2.3.1 Stakeholders in the Dutch energy system 2.3.2 Stakeholders in energy markets

2.3.3 Stakeholders in flexibility markets

2.4 Metrics for flexibility

11

2.4.1 Existing metrics

2.4.2 Flexibility framework under development

2.5 Real-world factors of influence

13

2.5.1 Nonlinear behavior in Lithium-ion batteries 2.5.2 Nonlinear behavior in Heat Storage Vessels

3. Methodology

16

3.1 EnTranCe

16

3.2 Experiments

17

3.2.1 Experiment 1 3.2.2 Experiment 2 3.2.3 Experiment 3

3.3. Simulation model

19

3.3.1 Model operation 3.3.2 Input signals 3.3.3 Control policy

3.3.4 Output signals and KPIs 3.3.5 Limitations of the model

(3)

4.4 Experiment 3

27

5. Discussion

299

5.1 Interpretation of results

29

5.2 Theoretical implications

30

5.2.1 Constraints

5.3 Practical implications

31

5.4 Limitations and further research

31

6. Conclusion

32

7. References

33

8. Appendix

34

8.1 Appendix I: Specifications of simulated equipment

35

8.2 Appendix II: Elía, Belgian TSO imbalance data

35

(4)

List of tables

List of figures

Table Description Page

1 Dimensions of flexibility (Wortmann and Van den Burg, 2017) 12

2 Constraints on flexibility (Wortmann and Van den Burg, 2017) 13

3 Relation between SoC and infeed/outfeed rates 14

4 Grid elements and roles in simulated EnTranCe network 16

5 Overview of experiments, respective experimental factors, expected results and related research questions (RQs)

18

6 KPIs available per grid element 22

7 Calculation of flexibility available per element, per dimension. Calculated with flexibility framework by Wortmann and Van der Burg (2017)

23

8 KPIs of note for experiment 1 24

9 Flexibility performance of system in experiment 1 24

10 KPIs of note for experiment 2 26

11 The effects of temperature on system performance in experiment 2 26

12 Flexibility performance of system in experiment 2 26

13 KPIs of note for experiment 3 27

14 The effects of temperature on system performance in experiment 3 27

15 Flexibility performance of system in experiment 3 27

Figure Description Page

1 Comparison between traditional and future power grids (Verzijlbergh 2016) 10

2 Conceptual model of EnTranCe micro-grid 16

3 Top-level Simulink model for experiment 1 19

4 Input signal and grid state for experiment 1 23

5 State of Charge of LISA (left), and Energy absorbed by HSV (right) 24

6 Input signal and grid state for experiment 2 25

7 State of Charge of LISA (left), and Energy absorbed by HSV (right) 25

8 Input signal and grid state for experiment 3 26

9 State of Charge of LISA (left), and utilization of HSV capacity (right). 27

(5)
(6)

1. Introduction

One of the largest, if not the largest, challenges society faces today is to transition from a system in which the planets’ resources are used up, to a system in which resources are used and reused to achieve sustainability (Verzijlbergh et al, 2016). This is especially true for the way our hunger for energy is fed. namely by burning up fossil fuels which will run out, but which are also responsible for the emission of greenhouse gasses and other harmful emissions. This thesis aims to contribute to this transition by validating a framework. This framework will in the future allow higher levels of penetration of renewable energy sources such as wind and solar power in power grids.

Other technologies for power generations are available, and becoming cost-effective or even more financially viable than conventional power generation technologies. This, in conjunction with the international efforts to increase sustainability and reduce emissions such as the Paris Agreement, has led to growth in renewable energy generation at an increasing rate (Peters et al, 2017). With increasing penetration of Renewable Energy Sources (RES) into power grids, the question of how to deal with peaks and lows in power demand and supply becomes more urgent (Lund et al, 2014). These peaks can be caused by varying wind speeds and sun exposure for the supply side

(Denholm, 2011), and peak hours like moments when large numbers of people turn on their TV’s on the demand side of energy.

It is crucial that a balance in demand and generation is maintained in order to avoid power outages due to overloads or shortages (Lannoye, 2012). In energy grids with no RES present, this balance is maintained by forecasting demand based on historical data, weather, events, etc. and adjusting generation from fossil fuel sources accordingly. The base load is supplied mostly by coal fired power plants, while peaker plants (mainly gas-turbines) deal with sudden and/or short peaks. However, in this situation the generation of power is entirely controllable. Controlling renewable sources is more constrainted, there is variation and uncertainty about their net generation. Thus in a power grid with high penetration of RES there is need for additional measures to keep demand and generation of power in balance (Ulbig and Andersson, 2012). There is a need for flexibility. Flexibility can be generally defined as “the capacity to adapt” (Golden and Powell 2000), which is a very broad and applicable definition but cannot capture the specific characteristics of energy systems. For an energy grid, a definition such as “A power adjustment sustained for a given duration in order to balance supply and demand at a given moment in time” (Eid et al., 2016) is in conjunction with the topic of this thesis. This thesis studies these “power adjustments”, and is focused on validating a framework for quantifying changes in infeed and outfeed of systems. The need for flexibility is studied broadly in the field of engineering, such as the development of energy storing systems or new forms of sustainable power generation. Not one solution has proven to be the best, and it is more likely a myriad of technologies need to be combined and controlled intelligently in a smart grid in order to transit to a fully sustainable energy supply.

(7)

To be able to trade in the commodity of flexibility, involved parties must answer questions such as:

-

How much flexibility does the grid need?

-

How much flexibility do we, as a flexibility service provider, have on offer?

-

In what quantities are we able to supply? For how long? How quickly?

-

How are those attributes measured?

To answer these questions, a comprehensive operationalization is needed. An operationalization of flexibility consists of a model integrating all characteristics of energy grid elements (such as wind generators, batteries, refrigerators, etc.), expressing their attributes in measurable parameters that can be communicated. At the University of Groningen, research is being performed aiming to develop such an operationalization.

The intermediate results of this study, i.e. the models derived, need to be verified in a real-world situation. This verification can be performed by collecting data from real-world energy systems and comparing their characteristics to model predictions in terms of performance and constraints. This research performs such a verification study using data from EnTranCe. EnTranCe is a testing ground for sustainable energy technologies where experimental installations such as wind

generators, heat pumps, and solar panels are present. A simulation study is performed based on a specific lab, C6, at EnTrance.

This simulation is performed to uncover the effects of temperature and battery characteristics on the flexibility performance of energy systems. These effects are not incorporated in the flexibility framework, and by including these effects in a computer simulation, the flexibility frameworks accuracy can be assessed and recommendations can be made with the goal of improving the value of the model both for a flexibility marketplace and as a metric for assessing the flexibility performance of an energy system.

The development of a framework for, or even a prototype of a marketplace for flexibility can be seen as a long term goal. Such a marketplace can only exist when all parties involved can accurately measure what is supplied or demanded, and the commodity is measured in the same unit of scale. The development of a metric for this end is thus a step on the road to the future in which flexibility service providers satisfy the need for flexibility in sustainable energy grids. Now that such a metric has been developed, it is necessary to evaluate it in a real-world situation, with real-world factors influencing behavior, to assure that stakeholders can utilize a valuable

framework for flexibility.

The main goal of this research is therefor to answer the following research question:

To what extent do the metrics developed in the flexibility framework accurately model flexibility of energy systems under real-world conditions?

This is achieved by answering the following sub questions:

1. Which types of real-world factors are not incorporated in the flexibility framework? 2. How do these real-world factors influence the accuracy of the flexibility framework?

3. Does the actual performance of systems operationalized by the flexibility framework deviate from expected performance?

4. Is it possible to improve system performance by implementing a control policy that reduces the effects of real-world factors?

5. Does the current framework hold value for the stakeholders? Is it usable?

6. Does the framework succeed in modelling the available flexibility in an element or system? 7. How can the frameworks metrics and constraints be improved or extended to increase its

(8)

2. Theoretical background

Between 2002 and 2012, the amount of wind generation globally has increased by 1100%, and the amount of solar generation by 2700% (Lannoye, 2012). Renewable energy is installed at an

increasing rate, and renewable energy penetration in power grids is expected to grow exponentially in the next years (Ellabban, Aby-Rub, Blaabjerg, 2014).

The impact of this transition on energy grids and thus the need for flexibility is becoming apparent (Ulbig and Andersson, 2012). Literature on the subject is therefor not scarce, and covers an array of facets of the subject. First, flexibility is defined and sources of flexibility are uncovered.

Stakeholders relevant to the subject are selected next, and current metrics for flexibility are discussed. From this analysis, it follows that current research and metrics for flexibility is relevant for specific stakeholders but not applicable throughout the complex system of energy grids.

2.1 Flexibility

A general definition for ‘flexibility’ can be obtained from various sources of literature on the topic of business and management. For instance, flexibility can be defined as “the ability to adapt or respond to change” (Vickery et al., 1999). However, general definitions for flexibility lack the specific context of flexibility in energy systems.

Within literature on the topic of energy systems, definitions of flexibility are generally focused on balance within a system or grid. For example: “The ability of a power system to respond to

changes in power demand and generation” (Huber et al., 2014), or “A power adjustment sustained for a given duration in order to balance supply and demand at a given moment in time” (Eid et al., 2016) are definitions in accordance with the overall goal of providing flexibility within a grid with high RES penetration. However, other causes for the need for flexibility exist. Moreover, these definitions refer to flexibility and balance within the boundaries of an entire system or grid whilst the flexibility framework under review is specifically designed to provide metrics for assessing flexibility within parts of a grid or system, and even single elements (such as a single battery or wind

turbine).

Stakeholders, which are further described in section 2.3, are in most cases on one side of the flexibility marketplace and thus are concerned with providing in-feed or out-feed with the goal of achieving balance in the overall system, however not within the system under control by the stakeholder. For instance, a Flexibility Service Provider (defined in section 2.3.2) is not concerned with achieving balance within the boundaries of the system, while operating this system does serve the purpose of achieving balance within a larger context.

Within the smaller scope of such a system, the definition: “the technical ability of a power system unit to modulate electrical power feed-in to the grid and/or power out-feed from the grid over time” (Ulbig and Andersson, 2015) is most appropriate for this thesis since in- and outfeed control are regarded as the main characteristic of flexibility in energy systems.

2.2 Flexibility sources

Four sources or flexibility can be identified, and these sources are very different in terms of technology and characteristics. An aggregation of these sources can therefor be instrumental in providing flexibility at all times.

2.2.1 Demand-Side Management

(9)

electricity consumption pattern modifications by end-use customers that are intended to alter the timing, level of instantaneous demand, or total electricity consumption”.

A substantial number of studies is devoted to the economic, market-side challenges (Eid et al, 2015; Albadi, 2007) of DR, or the technological challenges posed (Lund et al, 2014). These studies acknowledge the need for flexibility, but do not set out to operationalize metrcis for this form of flexibility.

2.2.2 Supply aggregation

Another form of aggregation is in the form of supply aggregators (Moser et al, 1999). Supply aggregators “]…] merge the energy production of small energy producers to take advantage of the wholesale market and increased market power.”. There are not many articles written on this subject yet, but most are focused on Independent Power Producers (IPPs). An example of supply aggregation of IPPs is the Dutch company Vandebron.

Supply aggregation can also be achieved by linking national electricity grids via (undersea) cables. The grids of Germany, Denmark, and the United Kingdom are all interconnected via The

Netherlands, which means that an excess of wind energy in Denmark can be used to fill a demand for energy in the UK, for instance.

2.2.3 Storage

Another key topic of research is on the role of energy storage (Weitemeyer, 2014), and the identification of technologies suited for energy storing elements, or resource aggregators. A large number of technologies are promising, of which examples are: batteries (Diouf, 2014), water reservoirs, capacitors, and conversion to hydrogen or hydrogen carriers.

It is important to note that storage is always limited. Either by the volume of energy contained, the rate at which energy can be stored or supplied, and the time needed to respond to either a supply or demand of energy. All forms of storage are to some extent limited in all of these dimensions and thus a form of resource aggregation might be only suited for a specific application and scope of duration.

Solutions for energy storage are needed for various scopes of duration (Lund, 2014):

-

Very short term (milliseconds to 5 minutes): for managing steep peaks in energy supply or demand.

-

Short term (minutes to hours): to cope with other fluctuations in energy supply caused by for instance gusts of wind, clouds, etc.

-

Medium term (hours to days): to cope with variations in demand and generation such as peak hours and power station outage.

-

Long term (several weeks or months): to manage seasonal fluctuations. 2.2.4 Conversion

Forms of energy can be converted to other forms, via conversion. For instance electrical energy can be converted into heat and chemical energy can be converted into electrical energy.

Sometimes these processes are reversible. Conversion can thus be a source of flexibility by converting a form of energy of which there is a surplus, to another form for which there is demand. Some forms of energy are also more convenient to store, such as heat in the form of hot water. Thus via conversion additional energy storage methods can be deployed within a system.

2.3 Stakeholders

(10)

2.3.1 Stakeholders in the Dutch energy system

The Dutch energy system consists of energy producers, which are private parties. These

producers deliver to the nationwide high-voltage net, the transmission system, which is operated by Transmission System Operator (TSO) Tennet, which is fully owned by the Dutch government. Tennet is responsible for the balance between supply and demand of electricity in the Netherlands. Regional distribution network operators then deliver electricity to homes, business and other

customers, but between the system operators (such as Enexis) and the customer, three parties are responsible for parts of the electricity supply:

• The ‘programmaverantwoordelijke partij’, or PV-party, is responsible for ensuring delivery of electricity to the customer.

• The ‘meetverantwoordelijke partij’, or MV-party, is responsible for measuring the actual delivery to customers.

• The suppliers (such as Nuon, Eneco, etc) are responsible for the delivery of electricity by buying from producers and selling to customers.

A supplier can include the PV- and MV-parties, and might be the same company as an energy producer, but this is not necessarily the case.

2.3.2 Stakeholders in energy markets

According to Painuly (2000) “Stakeholders may include RET [Renewable Energy Technology] industry (manufacturers of plant, equipment and appliances, owners of plant), consumers, NGOs, experts, policy makers (government), and professional associations.” This account of stakeholders is incomplete since for instance network operators are not represented. For this thesis, only

stakeholders directly involved with the generation, distribution and consumption of energy are of relevance.

Verzijlbergh (2016) lists generation, transmission, distribution and consumption as stakeholders or elements in a traditional energy distribution network. Ulbig (2012) adds storage, and combines transmission and distribution into one element.

In future power grids, Ulbig (2012) divides consumers in controllable and non-controllable loads, to account for demand response. Also, Power Generation is divided into Variable Generation (VG) and Conventional/Firm-RES Generation, in which firm-RES generators are renewable energy sources that can be controlled. Verzijlbergh (2016) moves on to describe a network with high RES penetration, nuclear or other generation for back up, as well as storage.

It is clear that a transformation is taking place, in which traditional power grids are becoming more complex by adding RES, interconnections and sources of flexibility such as DR and storage. Such a power grid will need a flexibility framework to be able to quantify and exchange flexibility.

A schematic comparison of traditional and future networks are illustrated in figure 1.

(11)

2.3.3 Stakeholders in flexibility markets

The network described by Verzijlbergh (2016) and associated stakeholders is the starting point in identifying the stakeholders in a flexibility market. Flexibility is present in all areas of a grid, be it in the power generation, the transmission, the distribution or at the consumer via DSM. In a modern energy grid, flexibility may be supplied or demanded by:

Producers of energy, either by RES, conventional generation, or other,

Local/private RES (IPPs, see Moser et al, 1999),

Transmission System Operators (TSOs),

Distribution System Operators (DSOs),

Consumers via DSM,

Storage providers,

Energy conversion systems,

Operators of (international) interconnections.

Note that multiple stakeholders might be represented by one company or organization. For the purpose of this thesis, the list above is divided into stakeholders that will have a demand for flexibility and stakeholders that may provide flexibility.

In the Netherlands - and in fact in most energy grids - the TSO is responsible for maintaining balance in the energy grid, which is why this stakeholder is regarded as the demand-side of

flexibility markets. A supplier of flexibility can take many forms, such as a DSM-system or a storage system. These can be grouped as Flexibility Service Provider (FSP): a service that can be

contracted to provide a certain quantity of flexibility, regardless of which source of flexibility is utilized, and possibly aggregating multiple sources of flexibility in a single system.

The flexibility marketplace will in this thesis be regarded as the system of trade between TSOs and FSPs.

2.4 Metrics for flexibility

2.4.1 Existing metrics

For each source of flexibility, there might be one or several metrics to measure flexibility. These metrics are relevant for a subset of stakeholders and are only applicable within the field of aggregation under review.

Models such as Insufficient Ramping Resource Expectation (IRRE), or Flexibility Assessment Tool (FAST), focus on planning of conventional generation resources in order to account and plan for generation from renewable sources and associated risks. These models have a system-wide scope and do not consider partial systems or individual elements. Moreover, these models do not quantify the available flexibility in such a way that an accurate measurement of flexibility is

calculated, nor does it provide the means to flexibility service providers to quantify the amount of flexibility on offer.

Other metrics focus on a single type of element within power grids, such as the flexibility provided by conventional generators (Oree, 2016). Oree identifies four flexibility attributes:

I. Operational range, the range of output between Pmin and Pmax in which stable generation is ensured.

II. Ramping capabilities, or Ramp Up/Down Rate, or the speed at which generation output can be increased and decreased,

III. Start-up and shut-down times,

(12)

It can be concluded that current models for flexibility do not provide a quantification of all sources of flexibility in a grid, system or element. Furthermore, metrics for flexibility that treat flexibility as a resource or commodity that can be exchanged and compared do not exist.

2.4.2 Flexibility framework under development

At the University of Groningen, research is being conducted by Wortmann and Van den Burg on a measure that allows flexibility to be modelled and measured for all types of elements within an energy grid. This allows for comparison between elements, as well as an estimation of the flexibility either provided or needed by an energy grid, from elements that might not be designed for the purpose of providing flexibility to the grid. This in turn will allow for flexibility aggregation to take place.

2.4.2.1 Dimensions of flexibility

The model divides the flexibility of grid elements in dimensions first by either infeed or outfeed flexibility, and second by either infeed/outfeed rate, ramp up/down rate, and volume or time. Table 1 summarizes these dimensions.

A. Infeed flexibility

Dimension Description Unit

Infeed rate Speed MW

Ramp rate up/down Acceleration or deceleration MW/h

Volume or Time Quantity or sustained period of time MWh or T

B. Outfeed flexibility

Dimension Description Unit

Outfeed rate Speed MW

Ramp rate up/down Acceleration or deceleration MW/h

Volume or Time Quantity or sustained period of time MWh or T

Elements in a grid are modelled according to these dimensions, and subject to constraints and parameters depending on the technical attributes of the specific element. The advantage of this approach is that virtually any energy grid element can be assessed in terms of flexibility. This allows the operator of an energy grid to assess the flexibility needed in very specific terms. The same holds for potential flexibility service providers, and thus a starting point is created for a marketplace in which flexibility is exchanged between grid operators and flexibility service providers.

2.4.2.2 Constraints

The model includes several constraints on available flexibility in addition to the six dimensions that form the backbone of the model. These constraints are either of a technical nature or set by the owners or users of a system, limiting the flexibility available in one of the six dimensions, or time-related (for instance an EV not plugged in all the time). Table 2 provides an overview of the constraints, grouped by constraint category.

(13)

Constraint Unit Technical constraints

Min/max infeed and outfeed rate MW

Min/max ramp up and down rate MW/h

Min/max volume MWh

Time-related constraints

Minimum up/down time Hours

Maximum up/down time Hours

Reaction time Hours

Availability Specific instance of time t*

2.4.2.3 Losses

Other limiting factors of flexibility are losses. An example are storage losses, for instance a battery losing charge over time, or a heat storage vessel losing heat to its surroundings. Losses can also be due to conversion, since for instance burning gas to generate electricity is not 100% efficient.

2.5 Real-world factors of influence

The flexibility framework discussed in section 2.4 assumes ideal behavior for all types of grid elements and systems. This implies that elements have identical and linear behavior, regardless of the type of element or the conditions in which it is operating. Research sub question 1 of this thesis is;

RQ1: Which types of real-world factors are not incorporated in the flexibility framework? this question is answered in this section.

Different types of equipment can exhibit vastly different behavior depending on the technology used and associated internal and external factors of influence. As will be discussed in chapter 3, the system on which the simulation model in this thesis is based involves a Lithium-Ion Storage Array (LISA) and a Heat Storage Vessel (HSV). Thus, literature on nonlinear behavior and external factors of influence will be discussed regarding these technologies.

2.5.1 Nonlinear behavior in Lithium-ion batteries

Rechargeable lithium-ion batteries are promising candidates for building grid-level storage

systems because of their high energy and power density, low discharge rate, and decreasing cost“ (Xu et al, 2016). Batteries exhibit nonlinear behavior under influence of many factors, and vary between different technologies and chemistries. Literature on this topic is very extensive, however for a large part based on empirical studies and rules of thumb. Based on theory, general

assumptions are made on charging and discharging profiles, and the effects of temperature and degradation on capacity and in/outfeed rates. These assumptions form a basis for demonstrating in a simulation the effects these factors have on flexibility performance.

2.5.1.1 Nonlinear charging and discharging profiles

Maximum infeed and outfeed of batteries is measured in a fraction of maximum capacity. For instance C/1 indicates that a 6 kWh battery can charge and discharge with a power of 6 kW, while

(14)

C/3 indicates that a battery with the same capacity can have a maximum infeed and outfeed of 2 kW. When capacity changes due to effects such as temperature or aging, max infeed and outfeed change accordingly (Erdinc et al, 2009; Chen, 2006).

Infeed and outfeed rates are also dependent on state of charge, SoC, the fraction of maximum capacity currently used. At lower SoCs, charge rate is significantly higher, while discharge rate drops. At higher SoCs the inverse holds. The changes in maximum outfeed and infeed rates in relation to SoC are very different for different types of battery technology and chemistry. Table 3 shows assumed values for the purpose of this thesis. These values are given as fractions of rated infeed/outfeed values.

SoC < 10% 10% - 90% > 90%

Charge 110% 100% 90%

Discharge 90% 100% 110%

2.5.1.2 The effects of ambient temperature

The effects of temperature on battery capacity are well studied, see for instance Erdinc et al, 2009. In general, lithium-ion batteries have an operating temperature between 20 °C and 25 °C

(Bandhauer et al, 2011). Higher temperatures lead to a slight increase in capacity at the expense of battery life (Xu et al, 2016), while lower temperatures decrease the capacity linearly down to around -10 °C at which point capacity drops dramatically. For the purpose of real-world simulated behavior, a linear relationship between temperature and capacity is assumed with 100% of rated capacity at operating temperature and 50% of rated capacity at 10 °C.

2.5.1.3 Relating depth of discharge to battery degradation

Lithium-Ion batteries are, just like other battery technologies and chemistries, prone to degradation over time. This degradation is in part due to chemical reactions occurring at the lowest and highest depth of discharge (DoD) levels. Avoiding these DoD levels will significantly enhance battery life time and thus economic lifetime. As a rule of thumb, avoiding DoD levels above 80% and below 20% will yield the optimal balance between capacity utilization and lifetime (Xu et al, 2016). This, however, will reduce available storage volume to 60% of the maximum available volume, which reduces flexibility in terms of volume. Additionally, since in- and outfeed rates of batteries are related to capacity, a reduction in available flexibility in terms of infeed/outfeed is implied. 2.5.2 Nonlinear behavior in Heat Storage Vessels

Technologies for heat storage are just as plentiful as for electrical energy storage in the form of batteries, see for instance Pinel et al, 2011. The most common technologies involve heat storage in the form of an electrical heating system for water. As discussed in chapter 3, the Heat Storage Vessel (HSV) in this experiment also utilizes the excellent heat storage capacity of water.

2.5.2.1 The effect of ambient temperature on capacity

The equation for calculating the required energy for heating up a given volume of water is:

𝑄 = 𝑚 ⋅ 𝐶 ⋅ Δ𝑇

With 𝑄 the required energy, 𝑚 the mass of the medium, 𝐶 the specific heat of the medium, and Δ𝑇

the temperature difference between initial temperature and target temperature.

Although the target temperature is constant, and usually 90 °C, the initial temperature can vary with ambient temperature, and thus the capacity of a HSV increases with decreasing ambient temperature. For the purpose of this study, it is assumed that there is a direct and constant relation between ambient temperature and capacity, ignoring the effect of a partially filled and heated storage vessel.

(15)

2.5.2.2 Nonlinear infeed profile (heating)

The heating profile of the medium can be approximated by Newtons law of cooling (or heating), which is applicable for heating via convection. This law states that the rate of heating decreases proportional to the difference in temperature between an object and it’s surroundings, and it is given by the equation:

𝛿𝑄

𝛿𝑡 = ℎ ⋅ 𝐴 ⋅ Δ𝑇

With 𝛿𝑄

𝛿𝑡 the required energy, ℎ the heat transfer coefficient of the medium, 𝐴 the surface area

through which the heat is transferred, and Δ𝑇 the temperature difference between initial temperature and target temperature.

The absorbed heat thus decreases when the temperature in the HSV approaches its target temperature. In effect, this means that the infeed of energy decreases when the HSV approaches it’s maximum capacity.

To answer research sub question 1:

RQ1: Which types of real-world factors are not incorporated in the flexibility framework? Factors non incorporated in the framework but relevant for equipment in the simulation are:

• For batteries:

-

the effect of State of Charge on infeed and outfeed rates,

-

the effect of ambient temperature on capacity,

-

the effect of Depth of Discharge on degradation.

• For Heat Storage Vessels:

(16)

3. Methodology

To answer the research questions posed in this thesis, a simulation study is performed. For this purpose a model is built in the Matlab/Simulink software package. Three experiments take place to acquire results on the basis of which the research questions can be answered. This chapter first discusses EnTranCe and the lab that is represented in the simulation. The experiments are then proposed after which the simulation model is presented.

3.1 EnTranCe

Since at present an experiment using real equipment is not possible at EnTranCe, it serves as a basis on which to model the simulation. This simulation model can in turn serve as a basis for a real on-site experiment in the near future. The following types of grid elements are available in lab C-6 at EnTranCe:

-

Lithium-Ion Storage Array (LISA),

-

Combined Heat and Power generator (CHP),

-

Heat Storage Vessel (HSV),

-

Tri-phase generator.

This equipment is to be interconnected and controllable. Furthermore, the tri-phase generator can be controlled from a computer where several generation profiles can be loaded to simulate

different situations.

The simulation model will consist of the tri-phase generator in the role of an unbalanced grid. This simulated grid needs flexibility which is provided by a controlled system aggregating the sources of flexibility. Table 4 lists the grid elements grouped by the stakeholder role fulfilled in an experiment.

Flexibility Sources Aggregator TSO

CHP-generator (CHP) Control systems Tri-phase generator Lithium-Ion Storage Array (LISA)

Heat Storage Vessel (HSV)

A conceptual model of the system is shown in figure 2. Two-way transmission of power is possible between the Grid and the Flexibility Aggregator, as well as between the Aggregator and LISA. Note that the flexibility aggregator is not a physical piece of equipment but takes form within the control system that controls the flows of power between the equipment. The CHP generator can only supply while the HSV can only store energy. The conceptual model is a basis on which the simulation model is developed.

Grid (Triphase) Flexibility aggregator

Lithium-Ion

storage

array

CHP

Heat

storage

Table 4: Grid elements and roles in simulated EnTranCe network.

(17)

3.2 Experiments

Before conducting experiments, it is necessary to calculate the theoretically available flexibility. This flexibility is calculated from the technical specifications (See appendix I) of the equipment, utilizing the metrics provided in the framework. Results from this first step are presented in section 4.1.

The input signal, discussed in section 3.3.1, will be scaled in such a way that the theoretically available flexibility will be sufficient to completely balance the simulated grid. This means that in volume, in- and outfeed, and ramp-up and down rates, the simulated system can absorb deviations from a balanced state resulting in a balanced grid.

Results of the experiments are measured in the flexibility performance of the system deployed in the respective experiments. Flexibility performance is expressed as the extent to which the system can meet the flexibility requirements of the input signal in terms of the six dimensions proposed in the framework.

3.2.1 Experiment 1

The first experiment includes only the dimensions and constraint proposed in the framework. The framework predicts that the flexibility aggregator can supply a certain amount of flexibility to the grid. In a Simulink simulation, this flexibility will almost certainly be available when only

characteristics currently modelled by the model are incorporated in the simulation. Flexibility performance is thus expected to be 100%. The results of the first experiment will therefor be useful to compare the results of the other experiments with, and to verify the simulation model.

3.2.2 Experiment 2

The second experiment aims to answer research questions 2 and 3:

2. How do the real-world factors influence the accuracy of the flexibility framework?

3. Does the actual performance of systems operationalized by the flexibility framework deviate from expected performance?

To answer these questions, an experiment with ‘simulated real-world’ behavior is performed. To this end factors of influence (discussed in section 2.5) not currently included in the flexibility framework are incorporated into the model for the second experiment. The results of this

experiment will provide insight on the effects of real-world factors on the flexibility performance of the system. Information on certain KPIs of the system (discussed in section 3.3.2) will reveal which factor of influence caused constraints to be met and performance to be sub-optimal.

3.2.3 Experiment 3

Experiment 3 has the goal of answering research question 4:

4. Is it possible to improve system performance by implementing a control policy that reduces the effects of real-world factors?

In a realistic situation real-world behavior, with the influence of factors discussed in section 2.5, will be known and forecastable. This means that by applying a control policy that accounts for

expected factors of influence can reduce or mitigate their negative effect. In experiments 1 and 2, a very basic control policy is in place. For experiment 3 a more advanced control policy is

implemented. Control policies in the simulation are discussed further in section 3.3.3.

(18)

This set of experiments will allow for comparison of flexibility performance between simulations with or without additional real-world factors of influence, and different control policies, and thus to answer research questions 5, 6, and 7:

5. Does the current framework hold value for the stakeholders? Is it usable?

6. Does the framework succeed in modelling the available flexibility in an element or system? 7. How can the frameworks metrics and constraints be improved or extended to increase its value

as a unit of trade in flexibility marketplaces?

The experiments and their respective experimental factors are summarized in table 5.

Constraints Control policy Expectation RQs

Experiment 1: Base Case

Flexibility framework only (section 2.4)

Basic control policy (Section 3.3.3)

100% Performance, useful for debugging model

Experiment 2: Simulated real-world Additional constraints and factors of influence (Section 2.5)

Basic control policy (Section 3.3.3) Less than 100% performance 2,3 Experiment 3: Optimized control policy Additional constraints and factors of influence (Section 2.5)

Advanced control policy, aiming to reduce effects of additional factors. (Section 2.5 and 3.3.3) Less than 100% performance, however improvement compared to experiment 2 results. 4

(19)

3.3. Simulation model

To conduct the experiments proposed in the previous section, a simulation model is developed in Matlab/Simulink. This software package provides an excellent suite in which to develop

interconnected subsystem with continuous calculated signals, which resemble real-time operating and responding systems.

Provided below is an image of the top-level Simulink model for experiment 1. It closely resembles the conceptual model from section 3.1, with the same elements in place. However, arrows now symbolize signals (i.e. flows of information) rather than flows of energy, and thus might be in a different orientation. This is because each block represents a subsystem, communicating information about its state, and it’s in- and outputs with the Aggregator, which in turn communicates to the grid how much the grid state is balanced.

An overview of the subsystem model for each element, and information on the functionality of the simulation model, is provided in appendix III.

3.3.1 Model operation

The model operates by reading the input signal (Section 3.3.2) from a spreadsheet, while running for a selected period of time. The default setting is to load one day of imbalance grid data and let the experiment run for one day. Via the aggregator and its control policy (Section 3.3.3), the model simulates a day of grid imbalance where any shortage or overage is dealt with by the aggregator which automatically stores energy in the battery and the heat storage vessel, or draws from the battery and CHP on demand. The output signal (Section 3.3.4) is read either in a plot or as spreadsheet data, which is interpreted to yield experiment results.

3.3.2 Input signals

Three input signals are processed by the simulation model. The first signal, relevant to all

experiments, is the grid imbalance signal. Since external factors are in effect in experiments 2 and 3, an ambient temperature signal is processed as well. A demand for heat is simulated as a third input signal.

3.3.2.1 Grid imbalance signal

The main input signal for the model is an imbalance signal, such that it is zero when the grid is in balance and no flexibility is needed, it is larger than zero when there is an overage, and the signal is below zero when there is a shortage of power in the grid. The signal will thus be measured in kilowatts, where a signal of 20 kW means that at that instance of time the grid needs to deliver 20 kW of power to the flexibility aggregator. A grid imbalance signal of -7 kW indicates that the flexibility aggregator will need to supply the grid with 7 kW of power at that moment of time.

In order to simulate real-world fluctuations in grid balance, historical data is chosen as input for the model. The dataset originates from the Belgian TSO, Elía, which publishes grid imbalance data on a real-time basis with data points on each exact minute of the day. Data from the first of May, 2017 has been chosen, which is recent enough to be relevant. An excerpt from the dataset supplied by Elía is provided in Appendix II.

Since the data relates to the entire Belgian high-voltage grid, it is measured in megawatts, which is orders of magnitude larger than the lab setup at EnTranCe. Thus, it has been scaled down to remain within the constraints of the theoretically available flexibility of the micro-grid in the lab, while keeping the characteristics of the original data with respect to ramp-up/down rates, relative volumes, and relative powers.

3.3.2.2 Ambient temperature signal

The temperature profile has been chosen to represent an average daily cycle from colder

temperatures at night to warmer temperatures at noon. This signal is modelled as a sinusoid with a period of 1 day, with a high at 12.00 of 20 °C and a low at 00.00 and 24.00 of 10 °C.

(20)

3.3.2.3 Heat demand

Heat is being stored in the Heat Storage Vessel (HSV) in the form of hot water. There must therefor be an outflow as well, to avoid overflow of the HSV after the first loading cycle. It is assumed that the HSV can supply heat to the EnTranCe heat network at a rate of 1 kW. There is demand for hot water from a certain temperature, meaning that outflow can only occur when the HSV is loaded to 70% of its capacity. Furthermore, it is assumed that outflow is replaced by an inflow of water at ambient temperature.

3.3.3 Control policy 3.3.3.1 Basic control policy

For experiments 1 and 2, a very basic control policy is set. This is realized within the aggregator sub-system of the Simulink model. The aggregator is set to feed all power to the battery storage, as long as it is within the specifications of the battery, any excess power due to higher loads, higher ramp-up rates or when the battery reaches capacity, will be used to heat the volume of water in the HSV using an electrical heating element.

In case of a shortage in the grid, any power is delivered by the battery storage, while any higher loads are consequently delivered by the CHP generator.

This basic control policy is set to demonstrate that the system behaves very differently when under idealized circumstances, and when real-world factors are simulated. It is expected that the

idealized circumstances in experiment 1 will lead to the system delivering exactly the amount of flexibility as calculated, while in experiment 2 the effects of additional factors of influence (see section 2.5),

3.3.3.2 Advanced control policy

For experiment 3, a more advanced control policy is simulated. This policy demonstrates that a control policy enables the system to deliver flexibility closer to the theoretical value, even under real-world circumstances in which additional factors are influencing the performance of the system and the elements in the system.

The control policy implemented in the simulation model for experiment 3 has the following properties:

-

Infeed is directed to the Heat Storage Vessel (HSV), within the boundaries of its specifications,

-

Outfeed is supplied to the Combined Heat and Power (CHP) generator, within specifications,

-

Infeed or outfeed outside CHP or HSV capabilities is covered by the Lithium-Ion Storage Array

(LISA).

-

The state of charge (SoC) of LISA is kept at 50% minimum by charging the batteries from the CHP as long as CHP outfeed has surplus capacity.

-

The SoC of LISA is kept between 20% and 80% to ensure optimal life time by limiting battery degradation (Section 2.5.1.1). This also ensures linear behavior of the infeed and outfeed characteristics by staying within the linear part of the charge and discharge curves (Section 2.5.1.3).

By preferring the CHP and HSV for most of the in- and outfeed of the system, equipment with the highest flexibility in terms of capacity is used. LISA then serves as a cushion to cover the delay caused by CHP response time, the limited ramp-up rate of the HSV, and any infeed and outfeed or ramp-rate limits on the CHP and HSV. LISA must therefor be always available to either supply or absorb power for short durations of time, which is why a SoC of 50% is preferred.

Within the scope of this thesis it is not possible to develop and test out a myriad of control policies, however this is a very interesting topic for future research. Knowledge on this subject will be

instrumental in designing optimal control policies for flexibility service providers (FSPs), with regard for equipment, internal and external factors of influence on performance, and performance

(21)

3.3.4 Output signals and KPIs

The model outputs several types of relevant data from which experiment results can be obtained. The first type of data is the grid state, which is the output signal of the simulation. The input signal of imbalance is, as best as possible, balanced by the flexibility aggregator, and the output signal is thus a balanced grid. This signal has a constant zero as optimal value, indicating that the grid is now in balance. A nonzero output signal indicates that the aggregator is not able to balance the grid.

3.3.4.1 Flexibility Performance

Flexibility performance is measured by integrating the absolute input signal and absolute grid state signal. The integral of the absolute input signal denotes the total volume of power either required to be stored into, or drawn from the flexibility aggregator. The integral of the absolute grid state sums the total amount of power that could not be fed in or drawn from the flexibility sources. The ideal state of the absolute grid state integral is zero, indicating that all requested flexibility could be handled by the flexibility aggregator. The fraction of the absolute grid state integral of the absolute input signal integral is used as the main performance measure of flexibility in this thesis. In formula:

𝐹𝑙𝑒𝑥𝑖𝑏𝑖𝑙𝑖𝑡𝑦𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒 = ∫ |𝐺𝑟𝑖𝑑𝑠𝑡𝑎𝑡𝑒|

∫ |𝐼𝑛𝑝𝑢𝑡𝑠𝑖𝑔𝑛𝑎𝑙|

In order to identify whether the missed flexibility occurred during infeed or outfeed, the integral of only positive or only negative values of both the input signal and the grid state is calculated. 3.3.4.2 Flexibility state

The second type of output is the flexibility state. This signal indicates how much flexibility is in use and how much is available, at any point in time, expressed in the six main dimensions (volume, ramp up/down, power). This data is also available from each of the elements deployed by the aggregator. It forms the experimental value, since it indicates whether the system is providing the theoretical flexibility as derived using the flexibility framework.

3.3.4.3 KPIs

A third output of the system is a set of KPIs of the system. Metrics on all elements can be read from plots or exported as spreadsheet data. Table 6 lists the KPIs available per element.

Grid LISA CHP Generator Heat Storage Vessel

Input imbalance signal (kW) Charge state (kWh) Power output (kW) Power input (kW) Output grid state signal

(kW)

Power output (kW) Total energy generated (kWh)

Total energy stored (kWh)

Power input (kW) Ramp rate (kW/min) Relative capacity due to temperature (%)

Ramp rate of in -and outfeed (kW/min)

Ramp rate (kW/min)

Total energy supplied (kWh)

Total energy stored (kWh) Relative capacity due to temperature (%)

(22)

3.3.5 Limitations of the model

The simulation model has a limitation because of mathematical loops. Mathematic loops occur when a system’s output is fed back as an input in the same system. For instance the output of the battery subsystem is an input for the aggregator subsystem, which in turn outputs a signal that is an input for the battery sub-model again. This leads to problems for the simulation software when compiling and running simulations.

Mathematical loops can be avoided by including delays. These delays hold signals for a set amount of time steps, causing the input signal of a loop to be the output signal of the previous time-step in which case the software is able to run the simulation without problems.

Because of the delays included, very high ramp-rates in the input signal are not communicated quickly enough through the simulated system. This in turn may cause imbalance in the output signal. These effects occur for ramp-rates far exceeding the specifications of the system modelled in this simulation, but will also occur when switching on equipment. In future use of this model for simulations of other systems, this limitation could further influence results when very high ramp-rates occur or when equipment is switched on and off in quick succession.

3.4 Data analysis

3.4.1 Data preparation

All data on outputs and KPI’s of the simulation are stored in a data file in Matlab/Simulink, with one data point for every time step which is one minute. The first step in the preparation of this data is to export to a spreadsheet which can be opened in Microsoft Excel. A template for Excel has been made in which the output and KPI data can be imported, where after experiment data is

automatically calculated and graphs are generated. 3.4.2 Data analysis

(23)

4. Results

4.1 Flexibility calculations

Based on specifications of equipment (See appendix I), the theoretically available flexibility of the aggregated system can be calculated by adding values for all dimensions and constraints into overall figures. The calculated values are assumed constant by the flexibility framework and are not affected by any factor other than the specifications of the equipment. The table below summarizes these values.

Constraint LISA CHP HSV Overall system

Technical constraints

Max infeed rate 6 kW - 6 kW 12 kW

Max outfeed rate 6 kW 15 kW - 21 kW

Max Ramp-rate infeed 6 kW/min 0,6 kW/min 6,6 kW/min

Max Ramp-rate outfeed 6 kW/min 1,5 kW/min 7,5 kW/min

Max volume absorbed 14 kWh 12 kWh 26 kWh

Max volume delivered 14 kWh * - **

Time-related constraints

Reaction time - 2 minutes - -

4.2 Experiment 1

In the first experiment, the basic control policy (section 3.3.3.1) is in effect, and no additional factors of influence (section 2.5) are incorporated in the simulation. A comparison of a plot of the input signal and a plot the grid state shows that the flexibility aggregator is able to almost

completely balance the grid (figure 4).

Table 7: Calculation of flexibility available per element, per dimension. Calculated with flexibility framework by Wortmann and Van der Burg (2017)

* The CHP has no volume restrictions since it is assumed that an unlimited supply of natural gas is available. ** The overall system outfeed volume therefor has no volume restrictions.

(24)

A notable exception occurs at t = 367 minutes when there is a peak in the grid state that indicates that the flexibility aggregator cannot handle the infeed demands of the grid. This is caused by the Lithium-Ion Storage Array (LISA) reaching its maximum capacity. At that point any infeed power is directed to the Heat Storage Vessel (HSV). A plot, in figure 5, of the volume usage of both

elements shows how LISA reaches capacity and the HSV takes over.

The peak in the grid output graph is due to a limitation of the model discussed in section 3.3.5. A one-period delay causes the HSV response to lag leading to one period in which the aggregator is not absorbing any power from the grid. This is a consequence of the simulation and is not due to shortcomings in the flexibility framework.

Another notable event is that the CHP is only switched on at a very late moment in the simulation when the battery is depleted and can no longer provide power to the grid. A peak is visible in the grid state plot at t = 1340 minutes, which is again caused by a delayed response of the CHP to LISA cutting out caused by the lag within the simulation.

Table 8 shows the percentage of flexibility utilized. Ramp rates and outfeed keep well within the system’s capabilities, while infeed utilization is very high and reaches capacity of LISA. From the flexibility performance in table 9 it can be concluded that the system can meet nearly all requested flexibility. The small fraction of failed flexibility request is due to simulation limitations.

KPIs

Max ramp rate kW/min Utilization %

CHP 0,21 14%

HSV 0,14 24%

Max infeed kW Utilization %

LISA 6,00 100%

HSV 5,25 88%

Total 11,25 94%

Max outfeed kW Utilization %

LISA 3,88 65% CHP 5,25 35% Total 9,14 44% Flexibility performance Total absolute In Out Requested 44,18 26,45 17,73 Failed 0,61 0,53 0,13 Fraction failed 1% 2% 1%

Fig 5. State of Charge of LISA (left), and Energy absorbed by HSV (right).

Table 8: KPIs of note for experiment 1.

(25)

4.3 Experiment 2

In the second experiment, the basic control policy (section 3.3.3.1) is in effect, and additional factors of influence (section 2.5) are now incorporated in the simulation. A comparison of a plot of the input signal and a plot the grid state shows that the flexibility aggregator is not able to

completely balance the grid (figure 6).

The moment at which the flexibility aggregator is not able to provide enough flexibility occurs at t = 346 minutes when LISA reaches 100% capacity indicating that it’s flexibility in the volume

dimension is depleted. As shown by the same peak as in experiment 1, the HSV starts absorbing power one period late, but in contrast to the results obtained in experiment 1, an imbalance in the grid remains visible afterwards.

The imbalance starting at t = 346 minutes remains due to the reduced infeed capabilities of the HSV as it reaches capacity. Recall that in this experiment, infeed characteristics are modelled nonlinear as discussed in section 2.5.2.2. Furthermore, the ramp-rate limitations of the HSV are met by the infeed demand of the grid causing additional imbalance in the grid state signal. Table 11 shows that the effects of temperature on system performance is significant. It is clear from the KPIs in table 10 ramp rates of both HSV and CHP are fully utilized and the infeed of the HSV reaches 94% of maximum available infeed capacity.

The flexibility performance figures in table 12 show that 2,19 kWh or 5% of total flexibility demand has not been met. The most significant drop in flexibility performance occurs when the grid

experiences and overage of power and has to supply excess power to the flexibility aggregator. 7% of this excess power cannot be absorbed. It can be concluded that performance has significantly declined by including the nonlinear behavior and additional factors introduced in section 2.5 of this thesis into the model.

Fig 6. Input signal and grid state for experiment 2.

(26)

KPIs

Max ramp rate kW/min Utilization %

CHP 1,5 100%

HSV 0,6 100%

Max infeed kW Utilization %

LISA 4,72 79%

HSV 5,64 94%

Total 10,36 86%

Max outfeed kW Utilization %

LISA 1,31 22%

CHP 2,72 18%

Total 4,03 19%

Temperature effects

Capacity 10 °C 20 °C Fraction best/worst (%)

LISA 9,8 14,0 70%

HSV 14,9 13,0 87%

Max infeed 10 °C 20 °C Fraction best/worst (%)

LISA 3 6 50%

HSV 6 6 100%

Total 9 12 75%

Max outfeed 10 °C 20 °C Fraction best/worst (%)

LISA 3 6 50% CHP 15 15 100% Total 18 21 86% Flexibility performance Total absolute In Out Requested 44,18 26,45 17,73 Failed 2,19 1,96 0,21 Fraction failed 5% 7% 1%

Table 10: KPIs of note for experiment 2.

Table 12: Flexibility performance of system in experiment 2.

(27)

4.4 Experiment 3

The third experiment also includes additional factors of influence (section 2.5) in the simulation, however the advanced control policy (section 3.3.3.2) is now in effect. A comparison of a plot of the input signal and a plot the grid state shows that the flexibility aggregator is able to completely balance the grid (figure 8).

Figure 9 shows that power infeed is directed to the LISA at an early stage, and from table 13 it can be concluded that this happens because the HSV ramp rate reaches its maximum. Infeed

utilization of the system is high for all elements, but stays well within constraints. Outfeed of LISA only occurs to supply outfeed while the CHP is starting up, since it has a 2 minute delay. After this delay, all outfeed power is supplied by the CHP. Temperature effects (table 14) are identical to the

effects measured in experiment 2, and the capacity of the LISA is limited because the SoC must be kept between 20% and 80%. However, from the flexibility performance in table 15 it can be

concluded that the system performs identical to the system in experiment 1. This shows that prioritizing the CHP and HSV and using the LISA as backup system for delays dramatically improves flexibility performance. The system is more robust and able to cope with the effects of temperature and other additional factors, showing that a control policy is instrumental in optimizing a system’s flexibility performance.

Fig 8. Input signal and grid state for experiment 3.

(28)

Temperature effects

Capacity 10 °C 20 °C Fraction best/worst (%)

LISA 5,9 8,4 70%

HSV 14,9 13,0 87%

Max infeed 10 °C 20 °C Fraction best/worst (%)

LISA 3 6 50%

HSV 6 6 100%

Total 9 12 75%

Max outfeed 10 °C 20 °C Fraction best/worst (%)

LISA 3 6 50% CHP 15 15 100% Total 18 21 86% Flexibility performance Total absolute In Out Requested 44,18 26,45 17,73 Failed 0,63 0,46 0,17 Fraction failed 1% 2% 1% KPIs

Max ramp rate kW/min Utilization %

CHP 0,11 7%

HSV 0,60 100%

Max infeed kW Utilization %

LISA 3,99 67%

HSV 4,13 69%

Total 8,13 68%

Max outfeed kW Utilization %

LISA 0,22 4%

CHP 3,88 26%

Total 4,10 20%

Table 13: KPIs of note for experiment 3.

Table 15: Flexibility performance of system in experiment 3.

(29)

5. Discussion

In this chapter, these research sub questions will be answered by interpreting the results presented in chapter 4. Research sub question 1 is an exception, since it was answered in section 2.5 as part of the theoretical background study.

5.1 Interpretation of results

The performance difference in terms of flexibility between experiments 1 and 2 shows that available flexibility may be significantly lower when the system is subject to nonlinearities and factors currently not included in the framework (section 2.5). Flexibility performance is decreased from 99% to 95% (figure 10). This decrease is caused by the effects of low ambient temperature on battery storage capacity and feed rates, and the decrease of heat storage available when the initial temperature of the water is higher.

Moreover, since the flexibility performance drops in experiment 1 are explained by limitations of the simulation model. It can thus be assumed that the system is performing 100% effective under ideal circumstances. This conclusion increases the significance of the performance decrease measured in experiment 2 since it shows that the flexibility framework is no longer accurate when the effects of section 2.5 are included in a simulation. These effects will, among other factors, certainly be of influence in a real-world situation and operationalization of the framework.

Based on these performance results research sub question 3;

RQ3: Does the actual performance of systems operationalized by the flexibility framework deviate from expected performance?

is answered. It is clear that actual performance of systems will be lower than expected from operationalization by the flexibility framework.

Research sub question 2;

RQ2: How do these real-world factors influence the accuracy of the flexibility framework? is answered by comparing KPIs from the results of experiment 1 with the results of experiment 2. Capacity of the HSV and LISA can decrease by 13% and 30%, respectively. This in turn limits infeed/outfeed rates and as a consequence HSV infeed ramp rate reduces as well. The effects lead to significantly higher utilization of available flexibility in all dimensions. Performance is reduced to such an extent that flexibility demand by the grid cannot be met even though it could meet the same demands under ideal circumstances in experiment 1.

(30)

In experiment 3, the conditions and factors of influence are identical to experiment 2 conditions, however a more advanced control policy for the aggregator is in place. The results of this experiment answer research sub question 4:

RQ4: Is it possible to improve system performance by implementing a control policy that reduces the effects of real-world factors?

Flexibility performance of the system in experiment 3 is similar to the performance results of experiment 1, while temperature effects are identical to the measurements in experiment 2. Utilization of flexibility increases in all dimensions, however more intelligent use of the flexibility sources mitigates the performance decrease from the factors discussed in section 2.5. The effects of a control policy are discussed further in section 5.3.

5.2 Theoretical implications

The flexibility framework provides a powerful tool to quantify both the flexibility demands of a grid in a state of imbalance, and the available flexibility of a system that aims to balance it, by expressing dimensions of flexibility in equal units. In this thesis, the available flexibility was calculated by inputting the specifications of the equipment in the framework. The resulting flexibility figures could be directly compared to the flexibility requirements of the grid. It can be concluded that the

framework is very useful for calculating theoretically available flexibility. However, this conclusion does not fully answer research sub question 6:

RQ6: Does the framework succeed in modelling the available flexibility in an element or system? The results of experiment 2 show that in a simulation in which constraints on the system due to external factors and internal nonlinearities are modelled, the actual available flexibility decreases. From these results it can be concluded that the framework does not succeed in modelling the real-world available flexibility in an element or system. This limits the practical use of the framework, and implies that in its current form it is only applicable to calculate idealized values for available flexibility.

5.2.1 Constraints

The current flexibility framework consists of dimensions and constraints which are not related to operating conditions or form of flexibility source. From the results of experiment 2, it is clear that constraints are not necessarily constant or even linear and can depend on operating conditions and equipment characteristics.

The behavior of constraints can vary significantly between flexibility sources. For a Heat Storage Vessel (HSV), decreasing temperatures increase the available flexibility in the volume dimension while decreasing battery (LISA) volume capacity. Infeed and outfeed rates as well as ramp rates may vary accordingly, but not necessarily.

The answer to research sub question 7;

RQ7: How can the frameworks metrics and constraints be improved or extended to increase its value?

is therefor: in order to increase the value of the flexibility framework, constraints need to be variable and nonlinear. This implies a twofold recommendation:

I. Available flexibility for systems or elements cannot be calculated without defining operating conditions or ranges for which the comparison holds true.

Referenties

GERELATEERDE DOCUMENTEN

Appendix A Transformation Matrix and Momentum Biased Dynamics A.1 Inertial to Orbit Coordinates Transformation matrix The relation between the two coordinate systems is shown in

Antwi, Bansah en Franklin (2018) se navorsing ondersteun Agyei (2013) se navorsing, want die resultate van hulle studie oor IKT in hoërskole binne ’n metropolitaanse gebied van Ghana

Daarom zal van een dwingend voorschrijven voorloopig géen sprake kunnen zijn. Bovendien zou dit ook allerminst gewenscht zijn. Indien kosten en ruimte geen bezwaren zijn, zal men

Before discussion of infrared spectra of crystalline alkali tungstates is under- taken, a few remarks, which will also apply to the spectra discussed in sub-

The following suggestions are taken from Baptist et al. These are both model and data analysis studies. A shoreface nourishment has the following two effects on the coastal system.

No part of this publication may be reproduced, stored in a retrieval system of any nature, or transmitted in any form or by any means, electronic, mechanical, now known or

In terms of response time A, the results show that most services start immediately after a customer request (video on-demand, roadside assistance, energy flexibility, water supply,

Based on both the literature on the management of on-demand services platforms and the case study of an aggregator offering flexibility services, this study was able to derive various