How flexible is smart? Determining the flexibility of a smart grid cluster Maas, Wouter Published in: Default journal

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record

Publication date: 2010 Link to publication in University of Groningen/UMCG research database

Citation for published version (APA): Maas, W. (2010). How flexible is smart? Determining the flexibility of a smart grid cluster. Default journal.

Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons). Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

Download date: 16-01-2017

PREFACE This master thesis was written in close collaboration with the three partners within the 'smart grid' pilot project 'PowerMatching City'; KEMA, ECN and Humiq. PowerMatching city is partly funded by the European Commission and is part of the Integral project (FP6-038576). The research was conducted at KEMA GCS in Groningen, which was an inspiring environment to write a thesis. As a member of the smart grid project team of KEMA, I got the opportunity to learn from both experts in the field of 'smart grids' and other students of the University of Groningen and Hanze University. First of all I would like to thank Dr. Albert van den Noort, project manager of PowerMatching City within KEMA, for the time he dedicated to teach me the ins and outs of the 'smart grid'. As my direct supervisor at KEMA he was able to give me proper feedback on my ideas, and helped me to give form to this thesis. I also would like to thank Dr. F. Bliek and ir. I. Bouwman who helped me to further develop my ideas about flexibility. The flexibility of a 'smart grid' is a complex matter, and it was good to have people to share ideas with. Also the contact I had with people at ECN contributed to the overall result of this thesis. Especially the information provided by ir. B. Roossien was valuable to write this thesis. But overall I would like to thank all the members of the PowerMatching City project as they provided me with both information and feedback on my thesis. Special thanks are also in order for my first supervisor within the IVEM research group of the University of Groningen, Dr. René Benders, who provided me with good insights and information how to organize the whole project. He also helped me a lot with the modeling part. Finally I would also like to thank my second supervisor within the university, Prof. Dr. Henk Moll, which contributed critic advice how to improve the overall content of this thesis. Without them I would not have been able to finish this thesis within the set period of five months.

Groningen, September 2010 Wouter Maas

ABBREVIATIONS

APX

Amsterdam Power eXchange: price index for oil and gas

AMR

Automated Meter Reading

BDK

Belasting Duur Kromme (Dutch for Load Duration Curve)

CA

Cluster Aggregator

CHP

Combined Heat and Power

DG

Distributed Generation

DSM

Demand Side Management

DSO

Distribution System Operator

EV

Electric Vehicle

HP

Heatpump

IHD

In-home information display

kW

KiloWatt

MW

MegaWatt

Nm3

Cubic meter of gas under normal pressure

PHEV

Plug-in Hybrid Electric Vehicle

PMC

PowerMatching City

Pte

Day ahead pricing model

PV

Photovoltaic’s

RTP

Real Time Pricing

SDM

Supply Demand Matching

VPP

Virtual Power Plant

WKK

Warmte Kracht Koppeling (Dutch for CHP)

WP

WarmtePomp (Dutch for heatpump)

SUMMARY The introduction of more renewable energy sources into the energy mix increases the mismatch between supply and demand of electricity within the grid. As the storage of electricity is still not very efficient, the concept of a 'smart grid' was developed. A 'smart grid' tries to adjust the electricity demand to match the available supply at that moment and vice versa. This so called supply and demand matching is realized by a constant communication between the producer and costumer of electricity through an advanced metering infrastructure. The extent to which a 'smart grid' cluster is able to deal with unbalance in the electricity grid is commonly referred to in literature as the flexibility of the cluster. Knowledge about flexibility will improve the performance of the ‘smart grid’ cluster when it is combined with supply and demand forecasting. Therefore this research tries to answer the question: how is the flexibility of a ‘smart grid’ determined and which factors will mainly influence the flexibility? Flexibility is defined in this research as 'the rate at which a ‘smart grid’ can take up or deliver electricity to the main electricity grid using this shiftable loads and flexible production'. This definition already implies that flexibility consists of two parameters: the flexibility to produce electricity and the flexibility to consume electricity. This makes that flexibility is hard to describe by a single dimensionless number. The flexibility indicator has to provide not only information about the amount of energy which can be exchanged between the cluster and the main grid, but also tell something about the rate at which this exchange can take place. Therefore this research showed that flexibility is best described by a Load Duration Curve (LDC) with the electrical power of all devices on the y-axis and the maximum runtime of each device on the x-axis. Such a LDC can be constructed for every single moment in time and will provide a snapshot of the actual status of the cluster. The flexibility LDC will then exactly show how much power can be delivered for which period of time and thus also how much energy can be exchanged. This research describes in detail how such flexibility LDC’s can be constructed for several devices. Furthermore it illustrates that not only switching on a device delivers flexibility, but also switching off some devices will deliver 'virtual' flexibility. This accounts for those devices which have already stored energy in their buffer. Because as long as there is still some energy stored in their buffer, it is not necessary to use the device. This means that switching off a device which consumes electricity actually means virtual production and switching off a device which produces electricity actually means virtual consumption electricity which both can provide (virtual) flexibility. This thesis then applies the developed flexibility theory in practice. Data from the ‘smart grid’ pilot project PowerMatching City (PMC) in Hoogkerk was used to show that the presented methods also in practice can give a good overview of the flexibility of a cluster. At the same time a method is presented to follow the change in flexibility over time. This method enables to construct a LDC at the most interesting moments and calculate the exact flexibility at that time. Finally a simple model showed the effect of buffer volume and different configurations of a ‘smart grid’ on the flexibility of the cluster. The results of the model showed that buffer volume was especially limiting for the flexibility of a ‘smart grid’ in the colder months. This is caused by the low heat demand in summer which causes the buffer to fill up no matter how big the volume. The results of the model also suggested that a HP offers less flexibility to the cluster than a µCHP. The methods to calculate flexibility which were presented in this research could be improved by including the heat demand of the household. Also knowledge about the device status (on/off) could improve the virtual flexibility calculations in particular. But already the calculations of the ‘smart grid’ flexibility presented in this research can contribute to improve the overall performance of a ‘smart grid’ cluster.

SAMENVATTING (NL) Door de toename van duurzame energieproductie en de introductie van bijvoorbeeld de elektrische auto, stijgt de mate van onbalans tussen vraag en aanbod van elektriciteit. Aangezien het opslaan van elektriciteit nog steeds erg inefficiënt is, hebben deze ontwikkelingen geleid tot het ‘smart grid’ concept. Een ‘slim netwerk’ is in staat om vraag en aanbod beter op elkaar af te stemmen door het verbruik of juist de productie van elektriciteit iets op te schuiven in de tijd. Dit wordt ook wel ‘supply and demand matching’ (SDM) genoemd. Het ‘slimme netwerk’ maakt hierbij gebruik van een geavanceerd meter systeem wat constante communicatie tussen de gebruiker en producent van elektriciteit mogelijk maakt. De mate waarin het ‘slimme netwerk’ in staat is om vraag en aanbod op elkaar af te stemmen wordt in de literatuur ook wel de flexibiliteit van het cluster genoemd. Inzage in de flexibiliteit van een ‘slim netwerk’ kan ervoor zorgen dat de flexibiliteit optimaal wordt ingezet indien deze kennis wordt gecombineerd met vraag en aanbod voorspellingen. Dit onderzoek heeft daarom als hoofdvraag: hoe wordt de flexibiliteit van een ‘slim netwerk’ bepaald en welke factoren hebben de meeste invloed op deze flexibiliteit? Flexibiliteit is in dit onderzoek gedefinieerd als 'de mate waarin een ‘slim netwerk’ elektriciteit kan afnemen dan wel terugleveren aan het net, door gebruik te maken van flexibele productie en verschuifbaar verbruik van elektriciteit'. Dit geeft eigenlijk al aan dat flexibiliteit lastig te beschrijven is in één enkel dimensieloos getal. Flexibiliteit moet namelijk niet alleen iets zeggen over de hoeveelheid energie die tussen het cluster en het net kan worden uitgewisseld, maar ook iets over de snelheid waarmee deze uitwisseling kan plaatsvinden. Dit rapport beschrijft daarom dat flexibiliteit het beste beschreven kan worden in een belastingduurkromme (BDK) waarin het totaal elektrische vermogen van alle apparaten wordt weergegeven op de y-as en op de x-as de tijd dat deze apparaten nog maximaal aan kunnen staan. Een dergelijke BDK kan op elk willekeurig moment worden bepaald en geeft een goede weergave van de actuele flexibiliteit van het cluster. De BDK geeft in één oogopslag weer hoeveel vermogen het cluster gedurende welke periode kan leveren en dus ook hoeveel energie er in totaal kan worden uitgewisseld. Hoe de flexibiliteit van verschillende apparaten kan worden weergegeven in een BDK wordt in dit rapport uitgebreid beschreven. Daarnaast wordt er dieper in gegaan op het begrip virtuele flexibiliteit. Niet alleen het aanzetten van apparaten geeft een 'slim netwerk' flexibiliteit, maar ook het uitzetten van apparaten levert (virtuele) flexibiliteit. Dit principe geldt voor apparaten die energie kunnen opslaan in een buffer. Immers zolang een apparaat energie heeft opgeslagen in die buffer, is het niet nodig om dat apparaat te gebruiken. Het uitzetten van dat apparaat betekent dat de uitwisseling van de energie tussen het 'slimme netwerk' en het net veranderd wat zal resulteren in (virtuele) flexibiliteit. Vervolgens wordt de ontwikkelde theorie over flexibiliteit toegepast in de praktijk. Data uit de 'smart grid' project PowerMatching City in Hoogkerk is gebruikt om te laten zien dat de ontwikkelde methode ook in de praktijk een goed overzicht van flexibiliteit kan geven. Tevens wordt een methode gepresenteerd om de verandering in flexibiliteit te volgen over de tijd. Deze methode maakt het mogelijk om alleen op de meest interessante momenten een BDK op te stellen om de exacte flexibiliteit op die punten te berekenen. Tot slot wordt in een simpel model aangetoond dat het volume van de buffers voornamelijk limiterend is voor de flexibiliteit van het slimme 'netwerk' in de koudere seizoenen. De warmte vraag in de zomer is dusdanig laag dat de buffers, ondanks een vergroot volume volledig gevuld worden. Daarnaast suggereren de resultaten van het model dat een micro warmtekracht koppeling (µWKK) meer flexibiliteit levert dan een warmtepomp (WP). The methodes om de flexibiliteit te berekenen die worden beschreven in dit onderzoek, kunnen worden verbeterd door de warmtevraag van elk huishouden mee te nemen. Daarnaast zou informatie over de status van elk apparaat (uit/aan) bijdragen aan een beter inzicht in de virtuele flexibiliteit. Deze twee aanbevelingen dragen er aan bij om de instantane flexibiliteit berekeningen uit dit onderzoek uit te breiden met een zekere mate van voorspelling over de flexibiliteit van het cluster in de toekomst. Maar zelfs

zonder deze aanpassingen kan de huidige bepaling van flexibiliteit bijdragen aan het optimaliseren van de prestaties van het 'slimme netwerk'.

TABLE OF CONTENT PREFACE .............................................................................................................................................. 1 1

2

3

4

5

Introduction and Background ....................................................................................................... 9 1.1

Introduction ............................................................................................................................ 9

1.2

Problem definition................................................................................................................. 10

1.3

Research objective ................................................................................................................ 10

1.4

Research Questions ............................................................................................................... 10

1.5

Boundary settings.................................................................................................................. 10

1.6

Methodology......................................................................................................................... 10

1.7

Outline of the rest of the thesis .............................................................................................. 11

Intelligent networks: the 'smart grid'.......................................................................................... 13 2.1

Defining a 'smart grid' ........................................................................................................... 13

2.2

Using a 'smart grid'................................................................................................................ 14

2.3

'Smart grid' control method: the PowerMatcher ..................................................................... 15

Flexibility of a 'smart grid' .......................................................................................................... 17 3.1

Introduction .......................................................................................................................... 17

3.2

Flexibility of a µCHP ............................................................................................................ 18

3.3

Flexibility of a HP................................................................................................................. 19

3.4

Flexibility of an EV............................................................................................................... 20

3.5

Flexibility of smart appliances............................................................................................... 21

3.6

Virtual flexibility .................................................................................................................. 21

3.7

Overall flexibility.................................................................................................................. 21

3.8

Conclusion............................................................................................................................ 22

PowerMatching City .................................................................................................................... 23 4.1

The 'smart grid' cluster in Hoogkerk ...................................................................................... 23

4.2

Flexibility of PowerMatching City ........................................................................................ 24

4.3

Data analysis PMC................................................................................................................ 25

4.4

Conclusion............................................................................................................................ 30

Model ............................................................................................................................................ 31 5.1

Introduction .......................................................................................................................... 31

5.2

Model description ................................................................................................................. 31

5.3

Patterns ................................................................................................................................. 32

5.3.1

Electricity demand pattern..................................................................................................... 32

5.3.2

Electricity supply pattern....................................................................................................... 32

5.3.3

Heat demand pattern ............................................................................................................. 33

5.3.4

Tap water pattern .................................................................................................................. 33

6

5.4

Scenario setup and model functioning ................................................................................... 33

5.5

Results .................................................................................................................................. 34

5.6

Conclusion............................................................................................................................ 36

Conclusion, discussion & recommendations ............................................................................... 37 6.1

Introduction .......................................................................................................................... 37

6.2

Conclusions .......................................................................................................................... 37

6.2.1

Flexibility of a 'smart grid'..................................................................................................... 37

6.2.2

Flexibility in practice: PowerMatching City .......................................................................... 38

6.2.3

Varying the cluster configuration .......................................................................................... 38

6.3

Discussion............................................................................................................................. 39

6.4

Recommendations................................................................................................................. 39

Literature ............................................................................................................................................. 41 Appendices ........................................................................................................................................... 43 Appendix I............................................................................................................................................ 43 Appendix II .......................................................................................................................................... 44 Appendix III......................................................................................................................................... 46 Appendix IV......................................................................................................................................... 51 Appendix V .......................................................................................................................................... 51 Appendix VI......................................................................................................................................... 53 Appendix VI......................................................................................................................................... 55

1

Introduction and Background

1.1 Introduction To combat climate change and lower the dependency on fossil fuels, the EU tries to drastically increase the share of Renewable Energy Sources (RES). The Renewable Energy Directive (2009/28/EC) enforces national targets for all EU member states with regard to the share of renewable energy in their energy mix. However the large scale introduction of renewable energy sources such as wind, solar and combined heat and power in households (µCHP), could have a significant impact on our currently top-down oriented electricity system (Van der Veen and de Vries, 2009). At every given moment in time, the total electricity consumption must be equal to the total production (Van der Veen and de Vries, 2009). Conventional demand patterns, however, will not follow the production curves which are caused by solar and wind energy for example, which will cause unbalance in the electricity system. In order for such a system to become completely reliable, the storage of energy will become a crucial factor (Hadjipaschalis et al., 2009). Over the years different technologies to store electricity have been developed, but none of which are very efficient (Lindley, 2010). Therefore the concept of a 'smart grid' was developed. A 'smart grid' is able to balance supply and demand in a more sophisticated way to minimize the need for energy storage. Digital technology in special ('smart') meters can monitor the production and consumption of electricity in a 'smart grid' at any given moment in time. The basic economic pricing model is used to balance supply and demand by making use of smart appliances. In other words appliances like a µCHP will produce electricity in case of a shortage when electricity prices are high and a heatpump (HP) or washing machine will only be switched on when there is a surplus of electricity, and the electricity price is low (EESI, 2009). One of the first pilot projects where a 'smart grid' is brought into practice in 25 households is PowerMatching City (PMC), which is currently performed by KEMA, ECN and Humiq in Hoogkerk (Zoethout, 2010). Roughly half of the households of PowerMatching City are equipped with a µCHP and the other half with a HP. In addition, there are some electric vehicles (EV), PV-cells, a windmill and smart appliances like a smart freezer and a washing machine included in this project. PowerMatching City focuses on the development of a market model that allows for simultaneous in-home optimization, technical coordination and commercial coordination. Currently all of these goals are tried to be realized by the intelligent matching of supply and demand using the flexibility on both the production and the demand side mostly provided by the µCHP's and HP's. Both appliances use a thermal heat buffers to store the produced heat. These buffers enable the decoupling of the electricity production or consumption from the hot water demand which provides flexibility (Hommelberg and Nieuwenhout, 2007). The term flexibility is often mentioned together with the term 'smart grid' but is never described in any detail. In this research flexibility is defined as the rate at which a ‘smart grid’ can take up or deliver electricity to the grid using shiftable loads and flexible production. Flexibility is therefore mainly determined by the buffer capacity of different appliances included in the 'smart grid'.

9

1.2 Problem definition The extensive use of RES will cause variation in the energy output of the electricity system. To deal with this variation, the 'smart grid' concept was developed which enables to balance supply and demand in a more sophisticated way. The extent to which a ‘smart grid’ can deal with unbalance is often referred to as the flexibility of the grid. Flexibility is described as a performance indicator of the 'smart grid', which can be used to optimize the performance of a cluster. The performance of a 'smart grid' is highly dependent on the buffer capacity of the system. Not only the amount of electricity which can be taken up or delivered by these buffers is important, also the rate at which this electrical transfer can take place. But how the flexibility is exactly determined and how it changes over time, has never been researched before.

1.3 Research objective This research seeks answers how to determine the flexibility of a 'smart grid' in order to increase its performance. It tries to develop a theoretical background of flexibility which can directly be applied in practice. This research has not only the objective to determine the flexibility of PowerMatching City; it also investigates the effect of the ‘smart grid’ configuration on its performance.

1.4 Research Questions
How is the flexibility of a 'smart grid' cluster determined and which factors will mainly influence the flexibility? Sub questions
What is the definition of a 'smart grid' for this research?
What is the theoretical flexibility of appliances like a µCHP, HP and EV?
How is the theoretical flexibility applied in practice?
How can the flexibility of a ‘smart grid’ cluster be measured over time?
How is flexibility affected by the configuration of a 'smart grid' cluster?

1.5 Boundary settings In this thesis flexibility is defined as 'the amount of electricity which can be taken up or produced by the 'smart grid' cluster using shiftable loads and flexible production'. The main focus of this research is on the flexibility of a µCHP, HP and EV which can be extrapolated to other 'smart appliances' as well. For the flexibility calculations of PowerMatching City however, only µCHP’s and HP’s are taken into account, as these devices currently provide most flexibility. Heat and electricity losses are not considered in this research. Also time dependent variables like hot water demand, heat demand, etc. are not directly included in the flexibility calculations.

1.6 Methodology First I will describe what encompasses the term ‘smart grid’ and give insight in a method to control such a network, using PowerMatching City (PMC) as an example. Then I will describe a method to calculate the theoretical flexibility of a ‘smart grid’ at a certain moment in time. I will look at the amount of energy which can be stored by several appliances to determine the overall buffer capacity. This buffer capacity will provide absolute data about how much energy theoretically can be taken up or produced by a cluster. Then I will investigate the rate at which this energy exchange can take place. I will try to combine this information with the amount of energy which can be delivered or taken up by the ‘smart grid’ cluster to develop a good performance indicator.

10

Consequently I will analyze the data which is generated in PMC to see how the flexibility of the ‘smart grid’ can be determined in practice. I will look at different simulations which are performed in the cluster to see if the proposed method to describe flexibility actually works in practice. Finally I will construct a simple model to simulate a cluster with fixed unbalance patterns. In the model the configuration of the cluster can be easily changed to look at the effect of different appliances on the performance of the ‘smart grid’. As the performance tells something about the flexibility, the model will provide valuable information how the flexibility is affected by the configuration of the cluster.

1.7 Outline of the rest of the thesis In the following chapter the definition of a ‘smart grid’ is discussed in more detail. It describes how a ‘smart grid’ works and what the benefits are of using a ‘smart grid’. It also described the control method which is used in PowerMatching City. Chapter three described the theoretical flexibility in detail. It will give insight in a method to present flexibility at a certain moment in time. It will illustrate how this method can be applied to a µCHP, HP, EV and other appliances. Also the term virtual flexibility will be introduced here. Chapter four will then apply the theory developed in chapter three in practice. Official data from PMC is used to determine the flexibility of this 'smart grid' cluster. Furthermore a method which enables to follow flexibility over time is described in this chapter. Chapter five will describe a simple model which was constructed in Vensim. It shows the setup of the used model, including the different supply and demand patterns, and the different scenarios which were compared. It will go into more detail what effect different appliances and buffer volume have on the overall performance of a ‘smart grid’. The results from previous chapters will be discussed in chapter six. This chapter will also provide recommendations how to continue the research in this field.

11

12

2

Intelligent networks: the 'smart grid'

2.1 Defining a 'smart grid' The term 'smart grid' refers to any form of intelligence in the electricity grid which allows for advanced supply and demand matching. There is no exact definition of a 'smart grid', as there is still no consensus reached in the literature. But at the core of each definition is an advanced metering infrastructure (AMI), which allows for two-way communication between the energy provider and the costumer (Hledik, 2009). In principle a 'smart grid' can be seen as an internet for energy (Miller and Abbasi, 2009). It is a collection of energy monitoring and control devices in households and businesses, which are able to communicate within the same network. This communication allows for an intelligence which will facilitate functions like (Moslehi, 2010):     

Real-time communication between consumers and utility companies Integration of renewable energy sources (RES) Higher quality of electricity services Autonomous network control Increased network efficiency

The real-time communication provided by the AMI, enables producers of electricity the possibility to offer costumers dynamic price rates for their electricity. Instead of charging the same price for electricity all year round, dynamic pricing will offer costumers lower price rates during off-peaks and higher rates during peak times. This will allow customers to save money on their electricity bill by shifting their electricity consumption to off-peak periods (Hledik, 2009). So called smart appliances will get a price signal and will determine if the price is low enough to perform its duty. The electricity consumption is constantly monitored which will lead to more awareness of the electricity consumption of costumers. The increased awareness will lead to significant progress in energy-efficiency (Miller and Abbasi, 2010). Dynamic pricing of electricity will become crucial for the integration of more RES in the electricity grid. An increase in for example wind and solar energy will result in more peaks in the supply of electricity. A 'smart grid' can deal with these peaks in electricity by shifting the electricity demand in time using smart appliances which avoids inefficient the storage of electricity. The utility companies are able to use the ‘smart grid’ to reduce voltage sags and spikes, which will increase the overall quality of the supply. Furthermore the supply of electricity is less likely to be interrupted or disturbed when more distributed energy sources are connected to the electricity grid. This will increase the overall reliability of the network and will avoid the construction of unnecessary fossil fuel fired generation (Moslehi, 2010). The implementation of a ‘smart grid’ will therefore increase the overall quality of electricity services. With the introduction of a 'smart grid' the network will be operated more and more autonomously. In case of an emergency, like a natural disaster for example, the reaction time to shut down specific areas of the grid will significantly be reduced. This reduces the chance for a total blackout, which increase the security of electricity supply. Finally, a 'smart grid' reduces the change of grid overloads which will allow the operator of the grid to use its maximum capacity. This implies that the introduction of a 'smart grid' on a large scale will allow for a higher integration of renewable energy sources into the Dutch electricity system without large investments in grid expansion.

13

2.2 Using a 'smart grid' All the advantages provided by a 'smart grid' are based on the principle of adjusting supply and demand in a sophisticated way, commonly referred to as supply and demand matching (SDM). SDM uses the flexible consumption and production to compensate for the variation in the grid causes by the inflexible production and consumption.

Figure 1. Categories in electricity production and consumption.

Figure 1 provides an overview of the different categories in electricity production and consumption which can be considered in a ‘smart grid’. Roughly speaking there are three categories for both production and consumption: inflexible, flexible and flexible with a buffer. Starting with the production side, the inflexible production consists of the renewable energy sources like wind and solar energy for example, but also a coal fired power plant could be placed in this category. The second category is the flexible production which can be adjusted according to the demand of electricity. This category consists of gas fired turbines for example. The last category on the production side which can be distinguished in a 'smart grid' is the flexible production which also produces heat, like a CHP for example. This category is made flexible with the use of a buffer to store the produced heat. That way the heat production is decoupled from the electricity generation. For the consumption side, the inflexible category is the time dependent consumption. This category consists of things like lighting of which the electricity demand cannot be easily shifted in time. As soon as a device is time independent it can be placed in the flexible category. This category includes devices like a dishwasher or washing machine, which can provide flexibility until the moment the appliance is switched on. The last category consists of the devices which consumes electricity to produce heat, like a HP for example. Similar to the last production category these devices also provide flexibility to the system when the heat produced is stored in a water buffer. There are also some categories which can be regarded as an energy source as well as an energy sink. Good examples are the electric vehicle (EV) and electric storage (batteries), which both can provide flexibility for the production and the consumption side. Currently however, the EV is only used for the flexible consumption of electricity with a buffer.

14

2.3 'Smart grid' control method: the PowerMatcher The electricity consumption and decentralized production can be controlled by dynamic pricing. Figure 2 explains in more detail how dynamic pricing works. The line which decreases from a power of 100 to zero represents the demand for electricity while the line which increases from zero to 100 represent the supply curve of electricity in relation to the price. As becomes clear from Figure 2a, the demand for electricity will be very high when the price is low. But there will be no energy supplier who is willing to supply its electricity for free. Therefore the supply of electricity will increase at higher electricity prices. Eventually this will lead to an equilibrium point as supply and demand always have to be in balance.

Figure 2 (a&b). Principle behind dynamic pricing.

Figure 2b shows the situation with an increased electricity supply caused by renewable sources such as wind and solar. The dotted line represents the increased supply of renewable energy, which is placed directly on the conventional power supply curve. The result is a shift of the price equilibrium to a point where electricity will be cheaper and thus the demand for electricity will increase. The determination of this equilibrium point is called supply and demand matching (SDM). SDM can be performed on different levels in a 'smart grid': agent based or distributed control, control on a neighborhood level and a completely centralized control system (Bakker, 2008). The control method which will be considered in this research is the multi-agent system PowerMatcher (Kok et al., 2005). Multi-agent based control of a 'smart grid' has been demonstrated as a suitable solution for the complex control needed to operate a 'smart grid' (Dimineas and Hatziargyriou, 2007; Pipattanasomporn et al., 2009).

15

Figure 3. Schematic outline of a PowerMatcher network.

Figure 3 gives an overview of the multi-agent PowerMatcher network. Every device in the PowerMatcher network is represented by a device agent. The device agent continuously receives a real time price of electricity. The device agent will then determine if it would like to consume or produce electricity at that price. If the electricity price is low enough, it will place a bid to consume or if the price is high it will place a bit to produce electricity (depending at the device considered) at the concentrator agent. The concentrator agent will gather all the bids of the various device controllers and determines the total demand and supply for electricity at that moment. It then places a congregated bid to the auctioneer agent. The auctioneer agent is basically the main frame which gathers all the electricity bids to determine the overall electricity demand. Together with the total electricity supply the auctioneer agent then determines the real time price of electricity. The price will then be communicated back via the concentrator agent to the device agent, after which the process starts again. In Figure 3 also an objective agent is included in the PowerMatcher network. The objective agent does not represent a certain device, but rather a distribution system operator (DSO) or an energy producer like Essent for example. The DSO or energy producer can then use this agent to steer the total ‘smart grid’ into a given e-profile. Therefore objective agent basically turns the 'smart grid' into a Virtual Power Plant (VPP). This means that the cluster can be regarded as one entity or power plant which can be used to either consume or produce electricity. The objective agent can then be used to support grid balancing purposes by virtually demanding or supplying electricity to the cluster. As a reaction the 'smart grid' will then produce or consume less electricity which equals the amount of the objective agent. The extent to which the objective agent can steer the 'smart grid' into a given e-profile is therefore highly dependent on the flexibility of the cluster.

16

3

Flexibility of a 'smart grid'

3.1 Introduction Flexibility is a performance indicator which can be used to optimize the functioning of a ‘smart grid’. A 'smart grid' cluster can be seen as a Virtual Power Plant (VPP) and aggregates smaller energy sources into a single 'virtual' large one. A central control entity (auctioneer agent) monitors the generation and consumption of the cluster and can adapt to the overall electricity supply (Palensky and Bruckner, 2009). This implies that a cluster could either consume or produce energy in order to balance the overall electricity supply. The extent to which the 'smart grid' can take up or deliver energy is determined by the size of the buffers and the capacity of the appliances included in the system. Together with the rate at which these appliances can either take up or produce electricity (power) this will determine the flexibility of the cluster. The definition of flexibility used in this research is therefore: 'the rate at which a ‘smart grid’ can take up or deliver energy to the electricity grid using shiftable loads and production.' The part of the system which is important to determine the flexibility is the electrical exchange between the different appliances included in the 'smart grid' and the main electricity grid. Figure 4 shows that this exchange can take place in two directions: either production or consumption. Therefore the flexibility of a 'smart grid' always consists of two parameters; the flexibility to consume electricity (Fdown) and the flexibility to produce electricity (Fup).

Electricity grid Production (Fup)

Consumption (Fdown)

'Smart grid' Figure 4. System overview of a 'smart grid'

Figure 5 describes in more detail what the advantage is of a known flexibility. The dark parts in the graphs show the unbalance reduction realized by the 'smart grid'. The left side shows the unbalance reduction with a high knowledge of flexibility, and the graph on the right shows the reduction without any knowledge about the flexibility of a ‘smart grid’. From these graphs it becomes clear that when the flexibility is unknown, the cluster will use its flexibility as soon as the unbalance reaches a certain threshold value. This could result however, in the situation that all flexibility is used at a certain moment and the cluster is not able to reduce the high unbalance peaks anymore. However, when the flexibility is known, it can be combined with electricity demand forecasting to save some flexibility for a later moment. The flexibility will then only be used to reduce the highest peaks in the cluster. This will result in overall unbalance which is lower than in the situation were the flexibility is unknown.

Figure 5. Optimization of unbalance reduction with and without knowledge of flexibility.

Flexibility has to provide information about the total amount of electricity which can be produced or consumed by the cluster at a given moment in time and information about the rate at which this exchange can take place. Therefore flexibility is best described in a variation on the conventional load-duration curve (LDC) used in the electric power generation. Instead of the capacity requirements, the sum of the electrical power (in W) is placed on the y-axis and instead of the capacity utilization rate, the total runtime is placed on the x-axis (in s). Such a LDC can be determined for every moment in time and shows how 17

long the cluster can provide a certain electricity exchange. When the power of the graph is then divided by the runtime, the total amount of energy (in J) which can be produced of consumed by the cluster is determined (W/s = J). Therefore the total energy exchange can be written as the integral of the LDC which is: T

E (T ) = ∫ P(t )dt

(1)

To illustrate how the flexibility of a simple 'smart grid' cluster is determined a simple 0 example is provided in Figure 6. The figure shows the LDC of a fictional ‘smart grid’ with five µCHP’s and five HP’s which are represented by the different bars. As electricity production can be regarded as negative consumption, the µCHP’s are given a negative power so that the flexibility can be combined into one single LDC. For simplification each device is given a power of 1W. Furthermore the five µCHP’s and HP's are given equal fill levels to make the LDC symmetrical. The fill levels considered allow for a maximum runtime of respectively 0, 10, 20, 50 and 100 seconds before they are completely filled. Figure 6 shows that the devices with a full buffer (runtime of 0) disappear from the LDC. In addition the LDC shows that the cluster can provide a maximum power of 4W for a period of 10 sec, 3W for 20 sec, 2W for 50 sec and 1W for 100 sec. The integral (or surface) of the LDC provides the total amount of energy which can be consumed or produced by the cluster in Joule. It is important to note that the LDC shows the total amount of possible energy exchange. This means that the energy which is actually used by the cluster will disappear from the Figure 6. Load duration curve for a simple cluster subsequent LDC, but the energy which is not used will still be visible as it still provides flexibility. In this research, the flexibility of a ‘smart grid’ cluster is based on the buffer capacity of the system at a specific moment in time without taking into account future changes. It ignores the change in flexibility caused by the heat demand of the household. The focus is to determine the instant flexibility which can be extended with time dependent variables (like heat demand) in further research. The next paragraphs will describe in more detail how the flexibility of different appliances is derived.

3.2 Flexibility of a µCHP A µCHP uses gas to produce heat and electricity and will thus provide flexibility to produce electricity (Fup). The produced electricity can be delivered back to the grid or used in the households itself. In both cases the µCHP will contribute to overcome a shortage in Electricity Grid electricity supply. A µCHP will, however, always produce heat when switched on which is stored in a water buffer (Figure 7). The buffer enables to decouple the electricity production from Electricity Production the heat demand. As soon as the buffer is completely filled (ActBuffer=100%), the µCHP will be switched off and will not Gas µCHP Heat ActBuffer be able to produce more electricity. This will result in zero flexibility to deliver electricity (Fup=0). Figure 7. System overview of a µCHP

Therefore the amount of electricity which a µCHP still can produce (ECHP) is determined by the remaining capacity in the water buffer times the heat to electricity ratio of the µCHP: 18

E CHP = (1 − ActBuffer ) ∗ C1 ∗

ε CHP (1 − ε CHP )

(2)

Therefore the amount of electricity which a µCHP still can produce (ECHP) is determined by the remaining capacity in the water buffer times the heat to electricity ratio of the µCHP:

E CHP = (1 − ActBuffer ) ∗ C1 ∗

ε CHP (1 − ε CHP )

(2)

With C1 = Cp ∗ V ∗ ρ ∗ (Tmax − Tmin ) In which ECHP is given in Joule, ActBuffer is the actual fill level of the buffer (as a fraction), εHP is the electric efficiency of the µCHP, Cp is the specific heat capacity of water (4180 J/kg °C), V is the volume of the water buffer (in liter), ρ is the energy density of water (1 kg/liter) and Tmax and Tmin are the temperature limits of the CHP buffer (in °C). When the electricity production of a µCHP is divided by its electrical power, the maximum runtime of the µCHP is determined:

t CHP ( s ) =

ECHP Pe CHP

(3)

The maximum runtime together with the electrical power of the µCHP can be used to construct a LDC as illustrated in Figure 6. This LDC will then describe the flexibility of the µCHP at that specific moment in time.

3.3 Flexibility of a HP When a HP is equipped with a water buffer, it provides flexibility to consume electricity (Fdown). The HP can be used in case there is an oversupply of electricity. Similar Electricity Grid to the µCHP, the flexibility of the HP is thus determined by the total amount of heat which can be stored in the buffer. Again Electricity Consumption when the water buffer is completely filled (ActBuffer=100%), the HP will be switched off and Fdown will drop to zero. Outside Heat

HP

Heat

ActBuffer

Figure 8. System overview of a HP

The total amount of energy which can be consumed by the HP (EHP in J) is thus determined by the available buffer capacity times the inverse of the HP heat-efficiency:

E HP = (1 − ActBuffer ) ∗ C 2 ∗

1

ε HP

(4)

With C 2 = Cp ∗ V ∗ ρ ∗ (Tmax − Tmin ) In which ActBuffer is the actual fill level of the buffer (as a fraction), εHP is the heat efficiency of the HP, Cp is the specific heat capacity of water (4180 J/kg °C), V is the volume of the water buffer (in liter), ρ is the energy density of water (1 kg/liter) and Tmax and Tmin are the temperature limits of the HP buffer (in °C). 19

When the electricity consumption of a HP is divided by the electrical power of that HP, the maximum runtime is determined:

t HP ( s ) =

E HP PHP

(5)

The maximum runtime together with the electrical power of the HP can be used to construct a LDC as illustrated in Figure 6. This LDC will then describe the flexibility of the HP at that specific moment in time.

3.4 Flexibility of an EV An electric vehicle (EV) contains a battery which can act as a buffer within a 'smart grid'. In theory the battery could consume or deliver electricity which would provide flexibility in both directions. Currently, however, the electric vehicle is only used to consume electricity to make sure the battery of the car will be fully charged at the moment the car is needed (tset). Therefore the EV will only provide flexibility to consume electricity (Fdown). Similar to the heating devices the total energy which can be stored in the EV (EEV) is given by the following equation: EEV = (1-SOC) * Cmax

(6)

In which SOC (as a fraction) is the state of charge and Cmax (in J) is the maximum electrical capacity of the battery. As the battery needs to be completely filled at tset, the minimum time to charge the battery (tc in seconds) can be calculated using the electrical power of the charging device (PEV):

tc =

(1 − SOC ) * C max E EV = PEV PEV

(7)

With the minimum time to charge the car battery and electrical power of the charger, the EV can be added directly to the LDC which is described in Figure 6. However the set point (tset) causes the flexibility of an EV to follow a timeline: EEV

t0

t

tc

tset

As the battery needs to be completely filled at tset, the EV can provide a flexibility of EEV during period t from t0 until tc. The flexibility of an EV can be added directly to the overall LDC as long as t0-tc > 0 (see Figure 9). But as soon as t0 = tc (t=0) the flexibility of the EV drops to zero Figure 9. Load duration curve of an Electric Vehicle. (Fdown=0) as at tc it is mandatory to charge the battery. This implies that at tc the block represented in grey (EEV in Figure 9) will completely disappear from the LDC. The minimum time to charge the battery (tc) will, however, change when any amount of energy is consumed. Graphically this would mean that the block shown in the timeline could be divided into smaller 20

blocks. The reduction of the total energy which can be stored in the EV will just result in a reduction of Fdown.

3.5 Flexibility of smart appliances Flexible appliances which consume energy can also provide Fdown in a 'smart grid' cluster when used 'smart'. These appliances include devices like a washing machine, dishwasher or dryer. Similar to the EV the user of these appliances will set a time (tset) at which the appliance needs to be finished. The device itself will then determine the ideal moment to switch on the device, which is mostly during an energy peak. Until the moment the appliance is switched on, it will provide Fdown which is comparable to the EV. The difference with the EV however, is the fact that the amount of energy (and thus tc) is fixed as it is not possible to switch off a washing machine in the middle of its program.

3.6 Virtual flexibility As a 'smart grid' cluster can be seen as a VPP, not only switching on a device delivers flexibility, but also switching off a device. In other words, not having to use a HP will virtually mean an extra production of electricity and switching off a µCHP virtually means consumption of electricity. This virtual flexibility accounts Normal for devices which use a buffer, except the appliances flexibility 1-ActBuffer which have a set ending time (tset). Were normal flexibility is determined by the energy which can still be stored in the buffer (1-ActBuffer), virtual flexibility is determined by Virtual the energy which is actually stored in the buffer flexibility ActBuffer (ActBuffer) as shown in Figure 10. Figure 10. Virtual flexibility vs. normal flexibility

However, virtual flexibility is more complex than normal flexibility which will become clear in the following example. Imagine a situation where all buffers are completely filled. In that situation the virtually flexibility will be maximized. However, only this maximum virtual flexibility will not enable the ‘smart grid’ to balance out an overload of wind energy because no real electrons are used. Only when a device is really producing heat it can be switched off, which will cause a real change in electrons. For these extreme situations it is therefore important to include the status (on/off) of each device in the flexibility calculations. Furthermore it can be argued that only virtual flexibility is realized when there is a heat demand in the household. These factors are however, time-dependent variables which finally should be included in the flexibility calculations, but are beyond the scope of this research.

3.7 Overall flexibility The previous paragraphs describe how the flexibility of several devices can be calculated. The power of each device can be added onto another to derive the overall power of the cluster. As this research only describes the instant flexibility the virtual power provided by a µCHP and HP are included on both sides. The total electrical power to consume electricity is then given by the following equation:

Pup = ∑ PCHP ∗ ε CHP + ∑ PHP + ∑ PEV + ∑ Pappliances CHP

HP

(8)

EV

Similarly the total power to produce electricity can be determined as follows:

Pup = ∑ PCHP ∗ ε CHP + ∑ PHP CHP

(9)

HP

21

Also the energy which can be taken up (Edown) or produced (Eup) by each device can be added up, to calculate the total amount of electricity which can be consumed or produced by the cluster. When the 'virtual electricity consumption' of the µCHP is included, the following overall electricity consumption (Edown) is derived:  1 E down = ∑  (1 − ActBuffer) ∗ C 2 ∗ ε HP  HP

  ε  + ∑  ActBuffer ∗ C1 ∗ CHP 1 − ε CHP  CHP 

  + ∑ (1 − SOC) ∗ C max + ∑ E appliances (10)  EV

Similarly the total electricity production of the cluster (Eup in J) including the ‘virtual production’ of the HP is given by:

 ε Eup = ∑  (1 − ActBuffer ) ∗ C1 ∗ CHP 1 − ε CHP CHP 

  1  + ∑  ActBuffer ∗ C 2 ∗ ε HP  HP 

  

(11)

It is however not possible to calculate an overall runtime as the runtimes of the different appliances cannot just be added onto another. Also dividing Eup and Edown by Pup and Pdown will not provide an overall runtime. Pup and Pdown will therefore not provide too much information as it tells nothing about how long the total power can be delivered. This means that the runtime of every device should be calculated separately. Then the runtimes should be sorted and placed with the corresponding power into an overall LDC. Only such an overall LDC can provide information about the overall flexibility.

3.8 Conclusion Flexibility is a performance indicator which always consists of two parts; the flexibility to consume electricity (Fdown) and the flexibility to produce electricity (Fup). Knowledge of flexibility can be used to optimize the ‘smart grid’ and save flexibility for a later moment. As flexibility of a ‘smart grid’ has to tell something about energy, power and time it is best described by a LDC. The y-axis provides information on the total electrical power the 'smart grid' can consume/produce while the x-axis provides information about the time this power can be delivered to the grid. The surface of the graph is the total energy which can be taken up or produced by the cluster. The LDC can be placed over the normal supply and demand curve of electricity and will constantly change. Especially in the case of smart appliances, the flexibility can disappear within no-time when the device is switched on. Flexibility is therefore a snapshot of the actual status of the ‘smart grid’. As long as a heating device has heat stored in its water buffer, it is not necessary to switch on the device. As a 'smart grid' can be regarded as a VPP, the heat stored in the water buffer will provide 'virtual flexibility'. A HP will therefore not only provide flexibility to consume electricity, but also to (virtually) produce electricity. Similar a µCHP will thus also provide (virtual) flexibility to consume electricity when it is not necessary to use the device. This principle accounts for all devices which are made timeindependent through a buffer.

22

4

PowerMatching City

4.1 The 'smart grid' cluster in Hoogkerk Were the previous chapter has shown how flexibility can be determined in theory; this chapter will apply the theory in practice at PowerMatching City in Hoogkerk. This 'smart grid' pilot is a living lab demonstration project which is performed by KEMA, ECN and Humiq. The individual actors of the cluster are not located near each other, but virtually aggregated in an online cluster. In total there are 25 households which participate in the pilot. In total there are 11 µCHP's and 14 HP's included in the cluster. Only four of the households have PV-cells placed on their roof. The other households are virtually provided with PV-cells which are actually located at KEMA GCS in Groningen.

Figure 11. Overview of the ‘smart grid’ cluster in Hoogkerk.

Figure 11 gives an overview of all the additional locations which are included in the cluster besides the different households. First of all a couple test appliances (HP, µCHP, freezer, washing machine) are located in the DEMO lab of KEMA GCS. Then there is a house located near ECN in Petten, which has some electric batteries to store electricity (E-storage) and a plug-in hybrid electric vehicle (PHEV). Also a wind turbine in the province of Noord-Holland is virtually included in the pilot. Finally two electric vehicles (EV) of Essent are also monitored to see what role the EV can play in the smart grid. The 'smart grid' cluster of Hoogkerk is controlled by the multi-agent system PowerMatcher. Every single household is provided with a concentrator agent which is called the HomeMatcher. In the HomeMatcher 23

lots of data are gathered to provide insight in the functioning of the 'smart grid'. Appendix I provides an overview of the parameters which are collected for each household every five minutes.

4.2 Flexibility of PowerMatching City Figure 12 provides a schematic overview of the current system in Hoogkerk. The flexibility of this 'smart grid' cluster is currently mainly determined by the total number of active µCHP's and HP's. Electricity Grid Production

Gas Outside Heat

Household Heat Demand

Consumption

Device

Heat Electricity

Buffer (heat/ electrical)

Heat Electricity Electricity Use (EV/ Smart Appliances) System Boundary

Figure 12. Schematic overview of the 'smart grid' cluster in Hoogkerk

In the case of a µCHP, the device will consume gas to produce heat and electricity which can be delivered back to the electricity grid. The households which are equipped with a HP will consume electricity from the grid to concentrate outside heat. At any moment in time the flexibility (up & down) of PMC can be represented by a LDC as described in chapter 3. However it is difficult to compare lots of LDC’s to see how flexibility changes over time, which raises the question how flexibility can be measure over time. Therefore another method is proposed to visualize the flexibility of a ‘smart grid’ over time. When only µCHP's and HP's are considered, maximum flexibility is provided when the water buffer of both systems is filled for 50 percent. In that situation the buffer can provide both Fup and Fdown, which is the ideal situation. The µCHP and the HP provide flexibility in two directions. Therefore the sum of Eup and Edown will be constant and equal to the total buffer content. When Eup and Edown are then divided by the total buffer content they can be normalized to a scale ranging from 0 to 1 of which the sum is always equal to 1. The product of these two values will then be maximized at 0.25 when both Eup and Edown are equal to 0.5 (see Figure 13). Using this indicator for the total flexibility is therefore suitable as the value maximizes for the optimal situation when both buffers are 50 0,25 percent full. 0,125

Ftotal

E up

E down = * TotalBufferCapacity TotalBufferCapacity

(12) 0

Figure 13. Ftotal graph

Ftotal does however not provide information about the total amount of energy or power the cluster can deliver. But is does enable to track changes in flexibility over time.

24

4.3 Data analysis PMC In PowerMatching City different scenarios are tested to look at the performance of the 'smart grid' cluster. Different price profiles are put into the ‘smart grid’, to see how the cluster will react. For the µCHP's and HP's the device agents will translate the given price profile into a requested fill level of the water buffer. This requested fill is constantly compared with the actual fill level and when the actual fill becomes lower than the requested fill, the device will be switched on. One of the first scenarios which was tested in PMC is a normal APX-price profile. The price is a dimensionless number which will follow a price profile of the APX of a couple years ago. As the flexibility of a ‘smart grid’ is calculated from the actual fill level of the buffers (ActBuffer), it is important to first look if the cluster in Hoogkerk functions properly. Figure 14 and Figure 15 give an overview of the average fill levels of the µCHP's and HP's under the APX-price profile. From the figures it becomes clear that the cluster reacts as expected; the requested fill of the µCHP is high when the APXprice is high and the requested fill of the HP is low at that time. On the other hand the requested fill of the µCHP is low when the APX-price is low and the requested fill level of the HP is high. As soon as the actual fill level of the µCHP or HP will drop below the requested fill, the device is switched on and will follow the requested fill level. But as long as the requested fill level stays underneath the actual fill level the graph will slowly fall, which is caused by the heat demand (and losses) of the different households.

APX: Fill level CHP

Actual_CHP Request_CHP 80 APX price 70

100 90 80

60 50

60 50

40

40

Price

Fill level (%)

70

30

30 20 20 10

10 0 16-06-10 4:00

0 16-06-10 16:00

17-06-10 4:00

17-06-10 16:00

18-06-10 4:00

18-06-10 16:00

Date

Figure 14. Average buffer fill levels of the µCHP's under an APX-price profile.

25

APX: Fill level HP Actual_hp 90

Request_hp

80

70

70

60

60

50

50 40 40

Price

Fill level (%)

80

APX price

30

30

20

20

10

10 0 16-06-10 4:00

0 16-06-10 16:00

17-06-10 4:00

17-06-10 16:00

18-06-10 4:00

18-06-10 16:00

Date Figure 15. Average buffer fill levels of the HP's under an APX-price profile.

From the actual fill levels it is possible to calculate the total energy which can be produced (Eup) or consumed (Edown) by the 'smart grid' cluster using equations 2 and 4. Figure 16 shows clearly that the graphs of Eup and Edown are symmetrical. Furthermore it becomes clear that as soon as the price in the cluster rises, Eup goes down and Edown goes up. This is exactly what is expected as the µCHP will be switched on at a high electricity price, which will reduce their available buffer capacity, and the HP’s will be switched off which will increase their buffer capacity due to the heat demand.

APX: Flexibility up and down E_up (MJ) E_down (MJ) 80 APX price 70

30

60 20

50

15

40

Price

Flexibility (MJ)

25

30

10

20 5

0 16-06-10 0:00

10

16-06-10 12:00

17-06-10 0:00

17-06-10 12:00

18-06-10 0:00

18-06-10 12:00

Date Figure 16. Eup and Edown of PMC under an APX-price profile.

26

0 19-06-10 0:00

Figure 17 shows that the total flexibility under the APX-price profile is higher when the price is low. As soon as the price increases the total flexibility goes down. This probably has to do with the minimum runtime of the µCHP’s which is currently set at 30 minutes. As soon as the price is high the µCHP’s will be switched on for a half hour and fill up the most part of the water buffer. It takes a while before the buffer is emptied again, as the heat demand is not very high in summer. Therefore the overall flexibility will go down which is clearly visible in the graph. Flex_total

APX: Flexibility total

APX price

0,25

80 70

0,20

50

0,15

40 0,10

Price

Flexibility

60

30 20

0,05 10 0,00 16-06-10 0:00

16-06-10 12:00

17-06-10 0:00

17-06-10 12:00

18-06-10 0:00

18-06-10 12:00

0 19-06-10 0:00

Date Figure 17. Total flexibility of PMC under an APX-price profile.

27

Figure 18 shows similar results for a different price profile. The price profile which is simulated here is an inverse of an ideal PV price profile. Again the flexibility drops when the electricity price increases.

PV: E-up & E-down E_down

35

E_up

60

Price

50

25

40

20 30 15 20

10

10

5 0 3-07-10 0:00

Price

Energy (MJ)

30

3-07-10 12:00

4-07-10 0:00

4-07-10 12:00

0 5-07-10 0:00

Date Figure 19 shows in more detail that the flexibility to produce energy drops as soon as the price increases. On the other hand the graph shows an increase of the flexibility to consume electricity. At each moment in time the flexibility of these graphs can be describes as a load duration curve. To illustrate this Figure 20 and Figure 21 show two LDC's of PMC under the PV-price profile. Figure 20 shows a LDC when flexibility is maximized which results in a LDC were the upper and lower parts are somewhat symmetrical. This means that the buffers of the appliances are half full. In contrast Figure 21 shows the flexibility at the lowest point in the flexibility graph. It clearly shows that the flexibility to consume electricity at that moment is a lot higher than the flexibility to produce electricity.

28

PV: flexibility

Flex_total Price

0,25

60 50 40

0,15

Price

Flexibility

0,20

30 0,10 20 0,05

10

0,00 3-07-10 0:00

3-07-10 12:00

4-07-10 0:00

4-07-10 12:00

0 5-07-10 0:00

Date Figure 18. Total flexibility of PMC under a PV-price profile.

PV: E-up & E-down E_down

35

E_up

60

Price

50

25

40

20 30 15 20

10

10

5 0 3-07-10 0:00

Price

Energy (MJ)

30

3-07-10 12:00

4-07-10 0:00

4-07-10 12:00

0 5-07-10 0:00

Date Figure 19. Eup and Edown of PMC under a PV-price profile.

29

LDC 4-7-2010 3:00 20000

Power (W)

'

15000 10000 5000 0 -5000 0

500

1000

1500

2000

2500

3000

2500

3000

-10000 -15000 -20000

Runtime (s)

Figure 20. Flexibility of PMC at 04/07/2010 3:00

LDC 4-7-2010 14:40 20000

Power (W)

'

15000 10000 5000 0 -5000 0

500

1000

1500

2000

-10000 -15000 -20000

Runtime (s)

Figure 21. Flexibility of PMC at 04/07/2010 14:40

4.4 Conclusion In PMC the current flexibility is mainly determined by the buffer capacity of the µCHP's and HP's included in the system. These appliances provide maximum flexibility when the water buffers are half full as in those situation they can provide as well Fup as Fdown. It is difficult to follow the flexibility over time as a LDC has to be constructed for every moment in time. Therefore another method has been used in PMC, to show how flexibility changes over time. When only µCHP's and HP's are considered the total buffer capacity will be constant. Therefore Eup and Edown can be normalized to a scale from 0 to 1 by dividing with the total buffer capacity. When these two numbers are then multiplied a dimensionless number is derived which gives an indication of the flexibility over time. This indicator will maximize at 0.25 when the average fill levels of the buffers are for fifty percent full. The results of PMC show that the flexibility is mainly affected by the minimum runtime of the µCHP's at this moment which is set at 30 minutes. With a low heat demand in summer, this minimum runtime will quickly fill the buffer which lowers the flexibility to produce electricity of the cluster.

30

5

Model

5.1 Introduction Previous chapters have shown how flexibility can be quantified and followed over time in theory and in practice. It was also made clear that the buffer capacity and the different devices have an influence on the overall flexibility. To get an idea of the effect of these factors on the flexibility of the grid a simple model was constructed. In the model the performance of a 'smart grid' cluster was tested under various configurations. The results of the model will provide valuable information about the effect of buffer volume and the configuration of a 'smart grid' on the flexibility. The simplified ‘smart grid’ is modeled in the dynamic modeling program Vensim. Vensim is a program developed by Ventana systems inc. which can be used for developing and analyzing dynamic feedback models. This simulation environment allows to graphically construct models or in a text editor1. By varying the input variables it is possible to compare how a simple 'smart grid' cluster deals with a constant amount of unbalance in the electricity grid. It was hereby assumed that the configuration which reduces the most unbalance will have the highest flexibility. This chapter will first provide a brief description of the model which will broadly explain the functioning of the model. Secondly the different patterns which are used in the model will be explained in more detail. Then the different scenario's setups are explained. And finally this chapter will provide the results and the conclusion of the different scenario runs.

5.2 Model description In the model only HP's and µCHP's are simulated during the different seasons (winter, spring and summer). Figure 22 provides a schematic overview of the used model. A detailed overview of the main parts of the model is included in appendix II. Also the exact equations of all the variables including the values of all the constant variables which are used in this research are provided in appendix III. And finally the efficiency curve of the HP which is used in this research is given in Appendix IV, as the efficiency of a HP is dependent on the outside temperature. Initial Unbalance Hot Water Demand

HP

Final Unbalance

Water Buffer

Heat Demand

µCHP Water Buffer

Figure 22. Schematic overview of the model

1

http://www.vensim.com/software.html

31

The modeled 'smart grid' is presented with an initial unbalance pattern which determines if either a HP or a µCHP is switched on. In each season the initial unbalance pattern is defined as the total electricity supply minus the electricity demand. This means that the HP will be switched on in case there is a positive unbalance (thus a surplus of electricity), while the µCHP will be switched on when the initial unbalance is negative. The heat which is then produced by the devices is stored in a water buffer, from which it is extracted for the hot water and heat demand of the household. In case the water buffer is completely filled, the heating device will be switched off and in case the buffer is almost empty, the heating device is switched on. These measures are implemented to safeguard the comfort level of the household. The electricity which is produced by the µCHP's or consumed by the HP's will reduce the initial unbalance. Therefore the final unbalance is determined as the initial unbalance minus the electricity consumption of the HP plus the electricity production of the µCHP. When the integral of the initial unbalance graph is subtracted from the integral of the final unbalance graph, the total unbalance reduction can be calculated. When this number is then expressed as a fraction of the initial unbalance, the percentage of unbalance reduction is determined. The scenario with the highest percentage of unbalance reduction will have the best performance and thus the highest flexibility.

5.3 Patterns Although the production and demand of electricity is a varying process, it will show on average a certain pattern throughout the year, week and day. Also the production of electricity from renewable energy sources like wind and solar energy will follow a certain pattern which can be predicted to some extend using the weather forecast. Even the demand for electricity can be predicted by extrapolating historical human behavior into the future. Predicting the exact supply and production patterns is beyond the scope of this research and also of minor importance to acquire more knowledge about the flexibility of a 'smart grid' under different configurations. Instead most patterns were obtained from historical data generated by PMC. The patterns which are determined are patterns for an average household. All patterns are constructed for one week, on a five minute interval basis. As electricity demand and supply differ throughout the year, different patterns were constructed for different seasons. For the winter season data from February, for the spring/autumn data from April and for the summer data from August was used to construct weekly patterns of the different seasons. The next paragraphs will explain in more detail how the different patterns which are used in the model are constructed.

5.3.1 Electricity demand pattern The electricity demand pattern used in this research was derived from the data generated by PMC. The electricity demand of PMC differs however from the normal electricity demand, as heating devices such as a HP use a lot of energy. Therefore the electricity used by the heating devices was subtracted from the total electricity demand to calculate the net electricity demand of all households in PMC. Only one household was excluded from the calculations as this is not a normal household but a farm. Then the net electricity demand was averaged over all households during four weeks in each season. These four weeks were then averaged again, to derive an average seasonal electricity demand pattern for a single week.

5.3.2 Electricity supply pattern The electricity supply pattern which was used in this research uses 100 percent RES. This avoids the construction of an unbalance pattern for the remaining conventional power supply, which is in fact irrelevant for this research. The renewable supply data are directly collected from the PMC database. PMC is provided with sustainable energy from solar and wind power. Each household is virtually connected to 14m2 of PV-solar 32

panels providing 1590 Wp and 880 kWh (Bliek et al., 2010). This accounts for one fourth of the electricity supply of an average household in the Netherlands2. It was therefore assumed that all energy produced by PV cells was used and supplemented by wind energy. The windmill which is included in PMC has a nominal power output of 2.5 MW. The output of the windmill is therefore scaled down to match the weekly demand of one household. The total electricity supplied to the 'smart grid' will thus equal the total demand, only will the patterns differ in time. The exact supply and demand patterns for each season are included in Appendix V. In addition the initial unbalance, which is the difference between these patterns, is provided in Appendix VI.

5.3.3 Heat demand pattern The heat demand pattern reflects the energy which is needed to heat an average household. The heat demand of a household is determined by the temperature difference between the house and its environment and the insulation of the house itself. In Europe, the insulation of a building is measured in the K-Value. The K-value of a building is obtained by multiplying the form factor of the building with the average U-value of the outward walls of the building. Hereby is the form factor determined as the total inward surface of the outward walls, divided by the total volume of the building and the U-factor is the measure of heat loss through the outward walls. The K-value is therefore always expressed as W·K-1·m-3. For this research an average K-value of 0.45 (which is the norm set by the European Union) and an average volume of 410m3 (derived from CBS statline, 2010) is assumed. The volume and K-value together with the temperature difference will give a good approximation of the heat demand pattern of a household. For simplification the inside temperature of the house is set at 20°C from 7 o’clock till 22 o’clock and at 15°C in the remaining night hours. The outside temperatures are derived from the average temperature measured in PMC.

5.3.4 Tap water pattern The tap water pattern is the hot water demand of a household. This demand pattern consists of the hot water used by appliances such as the shower, a bath and the warm water siphons. All of these appliances have official NEN standards for the rate of flow, temperature levels and filling capabilities. Also the appliances which supply the hot water in households are categorized in classes which comply with the NEN 5128 (EPN) standard, which are referred to as Gaskeur criteria CW/HWww:2003 (GasTec Certification, 2004). All the different classes are provided by GasTec Certification (2004). For this research the CW3 tap water pattern is chosen. This is the pattern for an average medium sized household, which is also assumed in the other patterns. The tap water pattern is transformed into a 5 minute interval and assumed to be equal for each day of the week and in each season.

5.4 Scenario setup and model functioning The input parameters which are varied in the model to test the performance of a ‘smart grid’ under different circumstances are the ratio between µCHP’s and HP’s and the volume of the water buffers. To be able to compare the different scenarios with each other, always a total number of ten households, or ten heating devices, is simulated. The standard scenario is the 5/5 210 scenario which has an even distribution of HP and µCHP’s and a standard buffer volume of 210 liter. Officially the heating devices of PMC are only capable to heat the lower 120 liter of the water buffer, but as no additional heating device (CV) was included in the model, it was assumed that the entire 210 was heated by the HP or µCHP. Table 1 provides an overview of the different scenarios which were tested in this research. Each of these scenarios is tested for all of the three seasons.

2

An average Dutch household uses 3430kWh per year (http://www.nibud.nl/uitgaven/wat-kost/energie.html)

33

Table 1. Scenarios used

Name 1/9 210 9/1 210 5/5 210 5/5 500 5/5 1000

Buffer Volume (l) 210 210 210 500 1000

Ratio µCHP:HP 1:9 9:1 5:5 5:5 5:5

During the simulations all energy losses are neglected. These factors will influence the energy efficiency of the household, but not necessarily the overall flexibility of the system. Therefore neither the heat losses of the water buffer are accounted for, nor the heat losses through ventilation, nor the transmission losses of electricity. Also the heat which is supplied to the household by incoming sunrays is not taken into account in the model. Due to the limit time which was available for this research, the used model does not simulate separate devices but rather one device of which the characteristics like power, buffer size, etc. are scaled up according to the set amount of appliances. This implies that the flexible consumption and production can only be used as one block and not in smaller pieces throughout time, which will make the results of the model more extreme. This was however slightly counteracted by reducing the minimum runtime of the appliances. Normally the minimum runtime is a major limitation as it is set at 30 minutes for a µCHP and at 15 minutes for a HP. But as the devices in the model function as one block, the minimum runtime was divided by the number of devices included in the setup. If for example 10 µCHP's are included in the cluster the minimum runtime will be 3 minutes (=30/10). The three minute peak then represents the energy for the minimum runtime of one device, which makes the simulations more realistic.

5.5 Results First the buffer volume was varied to see how the ‘smart grid’ cluster can handle a certain amount of unbalance. Table 2 shows the initial and final unbalance (in MJ) for the different scenarios. The initial unbalance is higher in the winter season than in spring and summer. This has to do with the amount of wind energy in relation to the solar energy. In winter time the PV-cells produce a lot less electricity than in summer. This means that the amount of wind energy is higher in winter than in summer, as the amount of solar energy is supplemented with wind energy to match the overall electricity demand. Solar energy follows a natural pattern which is closer to a normal household demand pattern. In contrast wind energy is also generated at night, when there is less electricity demand which will result in a higher initial unbalance pattern. For all the scenarios the final unbalance reduction shows an increase when the buffer volume is increased. What is interesting to see is that the final unbalance reduction with a buffer volume of 210 liter (the standard scenario) is somewhat similar for all the different seasons. For the other two scenarios (5/5 500 and 5/5 1000) the unbalance reduction is higher in winter. This means that the buffer volume is in fact limiting in the colder seasons. In contrast the final unbalance reduction in summer does not show the same increase as the unbalance reduction in winter. This can be explained by the low heat demand in this season. When the heat demand is low, eventually the water buffer will be completely filled regardless the increased volume, leaving no flexibility. This makes that the flexibility is lower in summer than in winter.

34

Table 2. Unbalance reduction of the system with different buffer volumes.

Winter Spring Summer

Initial Unbalance (MJ) 2576,88 Initial Unbalance (MJ) 2473,36 Initial Unbalance (MJ) 1814,96

Buffer Volume per hh (l) Final Unbalance Unbalance reduction Final Unbalance Unbalance reduction Final Unbalance Unbalance reduction

210 2422,28 6,00% 2284,92 7,62% 1714,81 5,52%

500 2174,00 15,63% 2238,01 9,52% 1678,22 7,53%

1000 1966,58 23,68% 2222,44 10,15% 1673,64 7,79%

When Eup and Edown are then calculated following the methods presented in chapter three it turns out that especially the flexibility to produce electricity will increase when the buffer volume is increased. This effect is however better visible in summer and spring and less in winter time (see Appendix VII). Then the ratio between µCHP’s and HP’s was varied to see how a different configuration of a ‘smart grid’ is able to deal with a fixed amount of unbalance. Table 3 presents the results of the different scenario runs. In all three scenarios the buffer volume is fixed at a volume of 210 liter. Therefore the first scenario presented below is equal to the first scenario presented in Table 2. It can be noted from Table 3 that the scenario 1/9 210 actually shows more unbalance than the initial situation. This is possible because of the comfort constraints applied to the heating devices. Table 3. Unbalance reduction of the system under different configurations.

Winter Spring Summer

Initial Unbalance (MJ) 2576,88 Initial Unbalance (MJ) 2473,36 Initial Unbalance (MJ) 1814,96

Configuration (µCHP vs HP) Final Unbalance Unbalance reduction Final Unbalance Unbalance reduction Final Unbalance Unbalance reduction

5 vs 5 2422,28 6,00% 2284,92 7,62% 1714,81 5,52%

1 vs 9 2870,32 -11,39% 2532,28 -2,38% 1991,92 -9,75%

9 vs 1 2221,36 13,80% 2233,09 9,71% 1571,11 13,44%

With no heat stored in the water buffer, a heating device will switch on when there is a heat demand. As the final unbalance is defined as the initial unbalance plus the electricity produced by the µCHP’s, minus the electricity consumed by the HP, switching on a device can cause an increase of the final unbalance. This is illustrated in Figure 23, were the 1/9 210 scenario shows more unbalance than the initial unbalance graph especially in the morning. At that time there is a peak demand in hot water for showering which causes the HP to fill its buffer no matter the negative unbalance. The fact that this happens in the scenario with mostly HP's has probably to do with the low temperature range of the device. The upper limit of the HP buffer is set at 35˚C, while the µCHP buffer runs up to 55˚C. This makes that there can be more energy stored in the buffer of a µCHP which enables them to follow the unbalance curve better than the HP.

35

Winter 6

5/5 210 1/9 210

4

9/1 210 Initial Unbalance

Energy (MJ)

2

0 0:00

0:00

-2

-4

-6

Time

Figure 23. Initial and final unbalance graphs for different 'smart grid' configurations during winter.

Because of a high heat demand in winter, more energy from the buffers is used which increases the total energy which can be put into the buffers. This implies that more electricity can be taken up or produced by the cluster in winter which increases the flexibility of the ‘smart grid’. This however does not show from the standard 5/5 210 scenario in Table 3. Again this probably has to do with the comfort constraints to secure the heat demand in case the water buffer is empty.

5.6 Conclusion The simulations in the simple Vensim model have shown the effects of different ‘smart grid’ configurations and buffer volumes on the flexibility of the cluster. The initial unbalance turned out to be higher in winter than in summer due to the ratio between solar and wind energy. It has been shown that the buffer volume is limiting in for the colder seasons, but not in summer. In summertime the heat demand of the household is lower than the flexibility demand, which causes the buffer to fill up no matter how big the volume. Especially the flexibility to produce electricity (Eup) is increased with buffer volume, which can be explained by the higher temperature limits of the µCHP. The fact that this effect is less visible in winter than in the warmer seasons can be explained by the lower efficiency of the HP in winter, which will also increase the flexibility to consume electricity (Edown). The lower unbalance reduction in winter compared to spring for the standard scenario (5/5 210) can be explained by the functioning of the model. When the buffer content is low the heating devices also have to be switched on as soon as the buffer is empty. This can result in amplification of the initial unbalance curve when this happens when there already is a shortage (HP) or oversupply (µCHP) of electricity. As in wintertime the overall heat demand is higher, this is more likely to happen than in other seasons. The comfort constraints to secure the heat demand of the household also explains the lowered unbalance reduction in the 1/9 210 scenario. The buffer content op a HP is a lot lower than that of a µCHP because of the upper temperature limit of the buffer. This implies that the µCHP can provide more flexibility than a HP. 36

6

Conclusion, discussion & recommendations

6.1 Introduction Flexibility is often mentioned together with the term 'smart grid' without any further explanation. Flexibility is assumed to be the ability of a 'smart grid' to balance supply and demand of electricity. But what encompasses the term flexibility of a 'smart grid' is never explained in literature. This research tries to answer the question how the flexibility of a 'smart grid' is determined and which factors will mainly influence the flexibility. Data from PowerMatching City (PMC), one of the first 'smart grid' pilots in Europe, is used to answer the first part of the question and a simple dynamic model is constructed to find out more about the influence of different factors on the flexibility. This chapter will discuss the main findings of this research and provide recommendations for further research.

6.2 Conclusions 6.2.1 Flexibility of a 'smart grid' Flexibility is a performance indicator of a 'smart grid'. Knowledge of the flexibility can be used to optimize the 'smart grid' and reserve flexibility for a later moment. A high flexibility implies that the 'smart grid' has a high potential to balance the supply and demand of electricity, or has a high supply and demand matching (SDM) potential. Flexibility will thus tell something about the electrical energy exchange between the main electricity grid and the 'smart grid' cluster itself. This implies that flexibility does not only have to provide information about the amount of electricity, but also the rate at which the electricity can be exchanged (electric power). The exchange can take place in two directions, meaning the flexibility of a 'smart grid' always consists of two elements; the flexibility to consume electricity (Fdown) and the flexibility to produce electricity (Fup). This makes that flexibility is hard to describe by a dimensionless number. Instead, this research showed that flexibility is best described by a Load Duration Curve (LDC). The sum of the electrical power of all devices is placed on the y-axis and the runtime of the separate devices on the x-axis. The runtime of a device is the time the device can be used at full power until its buffer is completely filled. Electricity production can be regarded as negative consumption and can be given a negative power to visualize Fup and Fdown in one LDC. The surface of the LDC then provides the total energy which can be taken up (positive side) and produced (negative side) by the cluster. The electrical power will be constant for each device, but the runtime will be determined by the available buffer capacity. A high buffer capacity will therefore result in a high runtime which means that the cluster can deliver electrical power for a long period of time. A LDC can be constructed at any moment in time and will provide a good overview of the potential energy exchange. The LDC can be placed over the normal supply and demand curve of electricity and will constantly change. Especially in the case of smart appliances, the flexibility can disappear within no-time when the device is switched on. Flexibility is therefore a snapshot of the actual status of the ‘smart grid’. As a 'smart grid' can be regarded as a Virtual Power Plant (VPP), not only the real electricity exchange, provided by switching on a device, but also virtual electricity exchange, provided by switching off a device, will provide flexibility. Virtual flexibility is not determined by the remaining buffer capacity of a device but by the actual energy which is already stored in the buffer. It is for example not necessary to switch on a heating device for as long as there is heat stored in the buffer. A HP will therefore not only provide flexibility to consume electricity (Edown), but also to (virtually) produce electricity (Eup). Similar a µCHP will also provide (virtual) flexibility to consume electricity as long as there is still sufficient heat stored in its buffer. This principle accounts for all devices which are made time-independent with a buffer. 37

6.2.2 Flexibility in practice: PowerMatching City Chapter four of this thesis shows how the theory to calculate flexibility, provided in chapter three, can be applied in practice in the 'smart grid' pilot PMC. Visualizing flexibility in a LDC however, makes it difficult to see how flexibility has changed over time. Therefore a special method is applied in PMC to see how flexibility has changed in the past. Currently the flexibility of PMC is mainly determined by the buffer capacity of the µCHP's and HP's included in the system. These appliances provide maximum flexibility when the water buffers are half full, as in that situation they can provide both Fup and Fdown. When only µCHP's and HP's are considered, the total buffer capacity will be constant. This is caused by the fact that both a µCHP and a HP can deliver normal and virtual flexibility. Therefore Eup and Edown can be normalized to a scale from 0 to 1 by dividing with the total buffer capacity. When these two numbers are then multiplied a dimensionless number is derived which gives an indication of the flexibility over time. This indicator will maximize at 0.25 when the average fill levels of the buffers are fifty percent. This method can be used to get an idea of the change in flexibility of the cluster over time. Consequently the exact flexibility can then be displayed in a LDC at a specific moment of interest. From the data analysis of PMC is can be concluded that the flexibility of the cluster is mainly affected by the minimum runtime of the µCHP's. This is related to the low heat demand in summer. With the current runtime of 30 minutes the water buffer is completely filled in two runs. Therefore the flexibility provided by the µCHP in particular is lowered in summer.

6.2.3 Varying the cluster configuration The dynamic model which was developed in this research has provided more insight in the effect of buffer volume and the ratio between HP's and µCHP's on the flexibility of a 'smart grid' cluster. The results of the scenario runs showed that the buffer volume of the simulated ´smart grid´ is especially limiting for the flexibility in the colder seasons, but not necessarily in summer. In summertime the amount of energy which is produced and stored in the buffer is higher than the heat demand, which causes the buffer to fill up no matter how big the volume. Buffer volume turned out to especially influence the flexibility to produce electricity (Eup) in comparison to the flexibility to consume electricity (Edown) This can be explained by the higher temperature limits of the µCHP compared to the HP. This effect was mainly visible in summer and spring and in less winter time. This probably has to do with the efficiency curve of the HP. As the efficiency of the HP goes down in winter, more electricity is consumed than in summer to produce the same amount of heat. This means that actually the flexibility to consume electricity (Edown) will increase as well. However the standard (5/5 210) scenario showed more unbalance reduction in spring than in winter. This is most likely caused by the functioning of the model. As soon as a buffer is empty the heating device has to be switched on regardless the unbalance at that time, to safeguard to comfort level of the household. This comfort constraint can thus result in an amplification of the initial unbalance curve. As the heat demand is the highest in winter, this is more likely to happen than in other seasons. Also the efficiency of the HP is lower in winter than in summer, which means that it consumes more electricity to produce the same amount of heat. Therefore this can lead to an extra increase in unbalance when the HP is switched on at times there already is a negative unbalance. The comfort constraints for the heat demand of the household also explains the lowered unbalance reduction in the 1/9 210 scenario. The buffer content op a HP is a lot lower than that of a µCHP because of the upper temperature limit of the buffer. Therefore the HP will be switched on more often at times there already is a negative unbalance as there is less heat stored in its buffer compared to a µCHP. This implies that the µCHP can provide more flexibility than a HP.

38

6.3 Discussion The current definition of flexibility has still some limitations. The LDC can describe instant flexibility very well, but is currently only a ´performance snapshot´ of the smart grid cluster. It can be used to see how the ´smart grid´ has functioned over time to get more insight in the performance of the cluster. However the ultimate goal of calculating flexibility is to combine it with supply and demand forecasting. The combination of these two makes it possible to save flexibility for a later moment in time, which will improve the performance of the 'smart grid'. But as soon as demand and supply are forecasted in the cluster it is also necessary to forecast the flexibility. But the LDC as described in this research will not provide such a future forecast. Currently the hot water and heat demand are only indirectly included in the flexibility calculations through the actual fill level of the buffers. By constantly using the actual fill level of the buffers, the change in flexibility by the heat demand and hot water use is included as well. However, when flexibility needs to be predicted, the (future) heat demand needs to be directly included in the calculations. A high heat demand increases the throughput of energy in the water buffer. The higher the energy use, the more energy temporarily can be used for balancing purposes. Therefore the flexibility in winter is presumably higher than in summer. However, when the heat demand becomes too high, this might actually cause a decrease in flexibility causes by the comfort constraints of the household. If for example a heating device is constantly used to fulfill the heat demand of the household, it is not possible to switch off the device which results in zero flexibility. Therefore a very high heat demand can actually cause an increase in unbalance, which as also shown in the model results. But as PMC was not yet completely up and running during the winter months, this still has be tested in practice. Flexibility described by a LDC is difficult to follow in time, which is another limitation. Therefore a different method has been proved to follow the flexibility in time, but this method only works for flexible devices with a buffer. Only when the buffer is half full the flexibility is maximized which implies that this method is also not suitable for buffers which are only being charged by the cluster. These buffers will only appear in one of the equations (Edown) which will influence the outcome. To determine a LDC which describes the flexibility at each moment in time, the minimum runtime of each device needs to be calculated. In this research the runtime was calculated at the highest level in the 'smart grid' (auctioneer agent) by using the fill level (or SOC) of each device and its power. But as the size of the cluster will increase the maximum runtime has to be calculated at the device itself. This will avoid an overflow of data from all the different devices to the auctioneer agent.

6.4 Recommendations The methods which are proposed in this research are new and can be improved in several ways. Therefore several recommendations for further research can be made. First of all it would be interesting to implement some knowledge about the future heat use of a household. Knowledge about the flexibility of a 'smart grid' will be used together with supply and demand forecasting to optimize the performance of the 'smart grid' cluster. However the future heat demand of a household can have a large impact on the flexibility of a ´smart grid´. A higher heat demand can not only increase the flexibility but also decrease the flexibility through comfort constraints. Therefore some knowledge about future heat use should be implemented to be able to improve the future flexibility predictions of a 'smart grid'. Secondly it was assumed in this research that the ´smart appliances´ like a washing machine can only be switched on. This causes the flexibility of these appliances to completely disappear. Therefore it would be interesting to look at the possibilities to switch off these devices for just a little while and maybe save some flexibility for later. 39

Also introduced in this research is the concept of 'virtual flexibility'. As in this research only the instant flexibility was determined, the actual status (on/off) of a device was not included. But the example provided in chapter three already showed that this knowledge is of high importance to get a better idea of the 'virtual flexibility'. Including these parameters in the flexibility calculations will therefore improve the overall quality of the flexibility predictions. Then the simple dynamic model can be improved by modeling the different devices with some sort of preference logic. That way the µCHP´s and HP´s do not function like one block which will improve the overall outcome. Also the losses could be included and different supply patterns.

40

Literature Amelsfoort, A. and R. Zwier (2007). Weg vrij voor duurzame brandstoffen? Onderzoek naar bereidheid consument om over te schakelen op duurzame brandstoffen. EDReC en Wetenschapswinkel Economie & Bedrijfskunde, Rijksuniversiteit Groningen, EC 180. Bliek, F., A. van den Noort, B. Roossien, R. Kamphuis, J. de Wit, J. van der Velde and M. Eijchelaar (2010). PowerMatching City, a living lab smart grid demonstration. Submitted to IEEE conference of Göteborg 1113 October 2010. Dimeas, A.L. and N.D. Hatziargyriou (2007). Agent based control of Virtual Power Plants, presented at the 14th International Conference on Intelligent System Applications to Power Systems (ISAP), November 4-8, 2007, Kaohsiung, Taiwan. EESI; Environmantal en Energy Study Institute (2009). Smart Grid overview. Briefing of Trilliant Inc. and Mission Point Capital Partners. Jan 8, 2009. GasTec Certification (2004). Gaskeur-criteria CW/HRww:2003. Second edition, September 2004. Hadjipaschalis, I., A. Poullikkas and V. Efthimiou (2009). Overview of current and future energy storage technologies for electric power applications. Renewable and Sustainable Energy Reviews 13. 1513–1522. Hledik, R. (2009). How Green is the Smart Grid? Electricity Journal vol. 22. Issue 3. 1040-6190. Hommelberg, M.P.F. and F.D.J. Nieuwenhout (2007). Virtuele elektriciteitscentrale: realiteit voor huishoudens (Virtual power plant: a reality for households). VV+ (Uneto-VNI), November, p.752-755. (in Dutch) Knab, S., Strunz, K. and H. Lehmann (2009). Smart Grid: The Central Nervous System for Power Supply - New Paradigms, New Challenges, New Services. Scientific Series of the Innovation Centre Energy at the Technische Universität Berlin, Vol. 2, University Press, Berlin, Germany. Available at SSRN: http://ssrn.com/abstract=1531655 Kok, J.K., M.J.J. Scheepers and I.G. Kamphuis (2004). Distributed intelligence for supply/demand matching to improve embedding of distributed renewable energy sources. Securing Critical Infrastructures, Grenoble, October 2004 (ECN). Kok, J.K., C.J. Warmer and I.G. Kamphuis (2005). ‘PowerMatcher: Multiagent Control in the Electricity Infrastructure’, International Conference on Autonomous Agents Proceedings of the fourth international joint conference on Autonomous agents and multiagent systems. ISBN:1-59593-093-0. Kok, J.K., C.J. Warmer, I.G. Kamphuis, Z. Derzsi, J. Gordijn, M. Hommelberg and H. Akkermans (2008). ‘Agentbased electricity balancing with distributed energy resources; A multiperspective case study’, Proceedings of the 41st Hawaii International Conference on System Sciences. ISBN: 0-7695-3075-3. Lindley, D. (2010). The energy storage problem. Renewable energy is not a viable option unless energy can be stored on a large scale. Nature vol. 463. 18-20. Miller, E. and D. Abbasi (2009). Smart grid overview. Trilliant, Inc. and MissionPoint Capital Partners. Nieuwenhout, F. et al. (2006). Flexible electricity grids. ECN Flexibel. Pipattanasomporn, M., H. Feroze and s. Rahman (2009). Multi-Agent Systems in a Distributed Smart Grid: Design and Implementation. Proc. IEEE PES 2009 Power Systems Conference and Exposition (PSCE’09), Mar 2009, Seattle, Washington, USA. Palensky, P. and D. Bruckner (2009). Anticipative Virtual Storage Power Plants. Preprint of IECON 2009 Proceedings. 3607-3610. IEEE 2009. Sandholm, T.W. (2000). "Distributed Rational Decision Making", in Weiss (ed.), Multiagent Systems: A Modern Approach to Distributed Artificial Intelligence, MIT Press, Cambridge.

41

Warmer C.J., M.P.F. Hommelberg, B. Roossien, J.K. Kok, J.W.Turkstra (2007) ‘A field test using agents for coordination of residential micro-chp’, presented at the 14th International Conference on Intelligent System Applications to Power Systems (ISAP), November 4-8, 2007, Kaohsiung, Taiwan. Warmer C.J., Hommelberg M.P.F., Roossien B., Kok J.K., Kuijper F.J., Turkstra J.W. (2007). ‘Geintegreerde microwkk’s als virtuele centrale; First trial smart power systems’, Juli 2007. Van der Veen, R.A.C. and L.J. De Vries (2009). The impact of microgeneration upon the Dutch balancing market. Energy policy 37. 2788-2797. Zoethout, T. (2010). Gronings slim net wordt slimmer naarmate het groeit. (The smart grid of Groningen becomes smarter when it grows). Energiegids 3.4. April 2010. (in Dutch).

42

Appendices Appendix I Gas and e-meters AMR_AGGREGATED_ACTIVE_POWER AMR_GAS_USAGE_M3 AMR_WATER_USAGE_M3 AMR_COLD_USAGE_GJ AMR_WARMTH_USAGE_GJ AMR_HOUSEHOLD_REMAINING_ELECTRIC_USED_POWER AMR_KWH_DELIVERED_NORMAL_RATE AMR_KWH_DELIVERED_LOW_RATE AMR_KWH_USED_NORMAL_RATE AMR_KWH_USED_LOW_RATE Heating System (HS) HS_ACTUAL_FILL_LEVEL HS_FILLING_LEVEL_REQUEST HS_BW_TEMP_SETPOINT_LL HS_BW_TEMP_SETPOINT_UL HS_ELECTRIC_USED_ENERGY HS_ELECTRIC_USED_POWER HS_ELECTRIC_USED_CURRENT HS_ELECTRIC_USED_VOLTAGE HS_GAS_USAGE HS_HEAT_CV_ENERGY HS_HEAT_CV_FLOW_TEMP HS_HEAT_CV_RETURN_TEMP HS_HEAT_CV_POWER HS_HEAT_CV_VOLUME HS_HEAT_DHW_ENERGY HS_HEAT_DHW_FLOW_TEMP HS_HEAT_DHW_POWER HS_HEAT_DHW_RETURN_TEMP HS_HEAT_DHW_VOLUME HS_REASON_NOT_AVAILABLE HS_TEMP_CH HS_TEMP_DHW HS_TEMP_LROOM HS_TEMP_OUTSIDE HS_TEMP_SETP_CH HS_TEMP_SETP_LROOM PhotoVoltaics PV_VIRTUAL_ELECTRIC_DELIVERED_CURRENT PV_VIRTUAL_ELECTRIC_DELIVERED_ENERGY PV_VIRTUAL_ELECTRIC_DELIVERED_POWER PV_VIRTUAL_ELECTRIC_DELIVERED_VOLTAGE

Total Total Total Total Total Total Total Total Total Total

Electrical Power Gas Usage Water Usage Cold Usage Warmth Usage Remaining Electrical Power Electricity Delivered (Normal) Electricity Delivered (Low) Electricity Used (Normal) Electricity Used (Low)

Units kW m3 m3 GJ GJ kW kWh kWh kWh kWh

Actual Fill Level Requested Fill Level Bandwidth Temperature Setpoint LL Bandwidth Temperature Setpoint UL HS Electricity Usage HS Electric Power HS Electric Current HS Electric Voltage HS Gas Usage Energy use of heat meter at CV Temperature of water leaving CV Temperature of water entering CV Power of CV Volume of CV Energy use of heat meter at hot water Temperature of leaving hot water Temperature of returning hot water Power of heat meter at DHW Volume of DHW system Error message CV water temperature Domestic Hot Water temperature Temperature living room Outside temperature CV water temperature setpoint Temperature living room setpoint

% % °C °C kWh W A V m3 kWh °C °C Watt m3 kWh °C °C Watt m3

Delivered current of the PV cells Delivered electricity of the PV cells Delivered power of the PV cells Delivered voltage of the PV cells

A kWh Watt Volt

°C °C °C °C °C °C

43

Appendix II Vensim model overview sheet 1: input variables Manual input CHP CHP Power

HP HP Power

CHP buffer volume (l)

HP buffer volume (l)

House Volume

CHP Min buffer temp

HP Min buffer temp HP Max buffer temp

Inside temperature

CHP Max buffer temp

House properties K-value

CHP E-efficiency

HP E-efficiency

CHP Min runtime

HP Min runtime

Number CHP's

Number HP's

Number of Households

Excel inputs E-demand per hh (J)

Total E-demand (J)

E-supply per hh (J)

Unbalance

Total E-supply (J)

Outside temperature Tap water (J)

Vensim model overview sheet 2: µCHP

Electricity production

CHP

Buffer temp range Hot water demand CHP (J) Max buffer Content CHP (J)

Actual Fill CHP (%)

CHP switch

CHP Initial buffer content (J)

44



Heat capacity water

Vensim model overview sheet 3: HP

Hot water demand HP (J)

HP Buffer temp range





Actual Fill HP (%)

HP switch



Vensim model overview sheet 4: output variables.



Unbalance MJ

Final Unbalance MJ





Overall buffer content

Max buffer Content HP (J)

Flexibility total E-CHP

E-HP



Edown

Eup



45

Appendix III Inputs CHP Power= ~

300000 Joule

"CHP buffer volume (l)"= 210 ~

liter

CHP Min buffer temp= 20 ~

Degree

CHP Max buffer temp= 55 ~

Degree

"CHP E-efficiency"= (1/6) ~

Dmnl

CHP Min runtime= 30 ~

Minute

Number CHP's= 5 ~

Dmnl

Number HP's= 12 ~

Dmnl

Number of Households= Number CHP's+Number HP's ~

Dmnl

HP Power= 300000 ~

Joule

"HP buffer volume (l)"= 210 ~

liter

HP Min buffer temp= 20 ~

Degree

HP Max buffer temp= ~

35

Degree

"HP E-efficiency"= "HP E-efficiency"= WITH LOOKUP (Outside temperature, ([(-40,0)-(35,6)],(-22.3394,3.02632),(-12.4771,2.78947),(-6,2.5),(-4,2.3),(-1.92661\

46

,1.97368),(-0.0917431,1.21053),(1.95719,2.21053),(3.66972,2.52632),(5.25994,2.86842\ ),(8.19572,3.44737),(9,3.7),(11,4.1),(13,4.5),(15,4.7),(18,4.9),(20,5),(22,5),(24,5\ ),(26,5),(28,5),(30,5),(32,5) )) ~

Dmnl

HP Min runtime= ~

15

Minute

House Volume= 410 ~

m3

"K-value"= 0.45 ~

Joule/Degree/m3

"E-demand per hh (J)":= GET XLS DATA('Inputs.xls','Month','A', 'D4') ~

Joule

"E-supply per hh (J)":= GET XLS DATA('Inputs.xls','Month','A', 'L4') ~

Joule

Outside temperature:= GET XLS DATA('Inputs.xls','Month','A', 'O4') ~

Degree

Inside temperature= GET XLS DATA('Inputs.xls','Month','A', 'O3') ~

Degree

Heat capacity water= ~

4180

Joule/Degree/liter

"Tap water (J)":= GET XLS DATA('Inputs.xls', 'Month', 'A', 'P4') ~

Joule

"Total E-demand (J)"= "E-demand per hh (J)"*Number of Households ~

Joule

"Total E-supply (J)"= ~

"E-supply per hh (J)"*Number of Households

Joule

Stopping mechanism for minimum runtime CHP/HP (layers 5 and 6) CHP initial state= ~

0

Dmnl

47

CHP should run= IF THEN ELSE(CHP switch >0,1,0) ~

Dmnl

CHP is on= INTEG (starting CHP-stopping CHP, CHP initial state) ~

Dmnl

starting CHP= MAX(0,(CHP should run-CHP is on)/TIME STEP) ~

1/Minute

last start time CHP=SAMPLE IF TRUE(starting CHP>0,Time,Time) ~

Minute

stopping CHP= IF THEN ELSE( Time-last start time CHP > (CHP Min runtime/Number CHP's) , (CHP is on-CHP should run)/TIME STEP, 0 ) ~

1/Minute

HP initial state= 0 ~

Dmnl

HP should run= IF THEN ELSE(HP switch>0,1,0) ~

Dmnl

HP starting= MAX(0,(HP should run-HP is on)/TIME STEP) ~

1/Minute

HP stopping= IF THEN ELSE( Time-last start time HP > (HP Min runtime/Number HP's), (HP is on-HP should run)/TIME STEP, 0 ) ~

1/Minute

HP is on= INTEG (HP starting-HP stopping, ~

HP initial state)

Dmnl

HP= IF THEN ELSE(HP is on>0, (HP Power*Number HP's)/TIME STEP,0) ~

Joule/Minute

last start time HP= SAMPLE IF TRUE(HP starting>0,Time,Time) ~

Minute

CHP model (layer 2) CHP=

IF THEN ELSE(CHP is on>0, (CHP Power*Number CHP's)/TIME STEP,0) ~

Joule/Minute

Electricity production= ("CHP E-efficiency" * CHP)*TIME STEP ~

48

Joule

Heat=

((1-"CHP E-efficiency")*CHP) ~

Joule/Minute

Buffer CHP= INTEG (+Heat-Heat demand hh CHP-"Hot water demand CHP (J)", "CHP Initial buffer content (J)") ~

Joule

Buffer temp range= CHP Max buffer temp-CHP Min buffer temp ~

Degree

"CHP Initial buffer content (J)"= ~

CHP Min buffer temp*"CHP buffer volume (l)"*Heat capacity water

Joule

"Max buffer Content CHP (J)"= Buffer temp range*"CHP buffer volume (l)"*Heat capacity water ~

Joule

C1= Heat capacity water*"CHP buffer volume (l)"*(CHP Max buffer temp-CHP Min buffer temp) ~

Joule

"Actual Fill CHP (%)"= (Buffer CHP/"Max buffer Content CHP (J)")*100 ~

Dmnl

"Hot water demand CHP (J)":= (("Tap water (J)"*Number CHP's)/TIME STEP)*0.4 ~

Joule/Minute

Heat demand hh CHP= IF THEN ELSE(Outside temperature