93
Degree project in Analysis of Demand Response Solutions for Congestion Management in Distribution Networks Daniel Brodén Stockholm, Sweden 2013 XR-EE-ICS 2013:014 ICS Master thesis

Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Degree project in

Analysis of Demand Response Solutionsfor Congestion Management in

Distribution Networks

Daniel Brodén

Stockholm, Sweden 2013

XR-EE-ICS 2013:014

ICSMaster thesis

Page 2: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

AbstractAccording to the 20-20-20 targets set by the European Union, 50 percentof the Swedish electricity share is to be provided by renewable energysources by 2020. The Smart Grid Gotland (SGG) project has emergedas a response to this target. The project aims at demonstrating a proofof concept on how smart grid solutions can be used to integrate largequantities of renewable energy sources in an existing network. The out-comes of the project are intended to pave the way for future renewableenergy integration projects in Sweden.

The Thesis focuses on one of the technical objectives of the SGGproject, i.e. to increase the hosting capacity of wind power on Got-land from 195 MW to 200 MW by using Demand-Response (DR) fromhouseholds and industries. DR consist of shifting peak-loads to peak-production hours. The integration of additional wind power causes arisk of exceeding the transmission capacity of the power export cablebetween Gotland and the Swedish mainland.

The approach considered for this Thesis is to use an Ancillary Ser-vice (AS) toolbox scheme based on multi-agent systems. The AS toolboxconsist of flexibility tools such as DR on long-term, short-term, a bat-tery energy storage system and a wind curtailment scheme. The DRactivity includes space heating and domestic hot water consumptionfrom detached houses on Gotland.

The simulation results indicate that 1900 household participantsare sufficient to balance the additional 5 MW for worst case scenar-ios. Furthermore, it is shown that the DR participation from industriescontributes in some cases to a reduction of 700 household participants.

The findings helped conclude that using an AS toolbox solution onGotland is fully possible from a technical perspective. However, barriersthat stand against its realisation are of economical nature and need tobe investigated in future studies.

Keywords: smart grid, demand side management, demand-response, load shift, wind power integration, distribution net-work, stationary battery

Page 3: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Referat

Enligt EU målen 20-20-20 har Sverige som målsättning att 50 procentav all energiproduktion ska utgöras av förnybara energikällor till år2020. Smart Grid Gotland projektet är ett initiativ som dragits igångför att möta dessa mål. Projektet strävar efter att påvisa hur smartaapplikationer i elnätet kan användas för att integrera stora mängderav förnybar energi i ett befintligt elkraftsystem. Lärdomarna från SGGprojektet kommer att användas som underlag för framtida elnätsprojekti Sverige.

Examensarbetet fokuserar på en av de tekniska målsättningarna förSGG, att öka produktionskapaciteten av vindkraft från 195 MW till 200MW genom att med hjälp av konsumentrespons skifta hushåll och indu-strilaster till timmar av hög produktion. Produktionsökningen medfören ökad risk för energiexportproblem mellan Gotland och fastlandet.Detta problem uppstår pågrund av begränsningar i överföringskapaci-tet.

Tillvägagångssättet för examensarbetet har varit att använda sig avett sidotjänstverktyg som baserar sig på ett multi-agent system. Dettaverktyg består av konsumentrespons på lång och kort sikt, ett batteri-lagringssystem samt vindreducering.

Resultaten påvisar att 1900 hushållskunder är tillräckligt för att ba-lansera 5 extra MW vindkraft under svåra omständigheter. Industrikun-derna lyckades i vissa fall minska behovet av antal hushållskunder medså mycket som 700 hus.

I slutsats kan man påstå att det är teoretiskt möjligt att använda sigav ett sidotjänstverktyg på Gotland för att lösa de exportproblem somkan inträffa. Dock så begränsas sidotjänstverktyget av marknadsstruk-turer och ekonomiska utmaningar. Detta bör vara en utgångspunkt förframtida studier inom ämnet.

Nyckelord: smarta elnät, lasthantering, konsumentrespons,skifta last, vindkraftsintegrering, distributionsnät, stationärtbatteri

Page 4: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Acknowledgment

This Master Thesis study truly captured my interest from the very beginning. Thehigh relevancy of the study with on-going projects in the industry has made merealise the importance of researchers in this field. I believe Smart Grid applicationswill play a very important role for the future of our energy infrastructure, as wewill have to rely on innovative solutions to ensure a sustainable future.

I would like to thank everybody that has assisted me during the Master Thesisstudy. Special thanks to my supervisors from Vattenfall, Erica Lidström and DavidErol, for believing in me and giving me the opportunity to study the challengingissues on Gotland. They have made me feel at ease during my time at Vattenfalland have been a great guidance throughout the study. The work environment atVattenfall has been engaging and I have met some wonderful people that havemade my stay very pleasant. I sincerely hope that the findings of my Thesis will bevaluable for the Smart Grid Gotland project and that I will have the occasion towork with people from the R&D department at Vattenfall in the near future.

Special thanks to Claes Sandels, my supervisor at the Royal Institute of Tech-nology (KTH). He has shown a lot of engagement and enthusiasm in my Thesis.Claes has been more than just a supervisor he has been a true mentor. I sincerelyhope that my Thesis results will be valuable for your research at KTH.

Last but not least, I want to thank all of whom I have interacted with for thepurpose of the Thesis. I felt at all times that I was encouraged to stay creativeand confident of trying alternative routes. Thank you for that, it has been a truelearning experience.

Page 5: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in
Page 6: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

List of Symbols

KTH Kungliga Tekniska HögskolanEU European UnionDR Demand ResponseSGG Smart Grid GotlandDSP Demand-Side ParticipationAS Ancillary ServiceLT Long-Term (day-ahead)ST Short-Term (hour-ahead)BESS Battery Energy Storage SystemPM PowerMatcherPSS/E Power System Simulator for EngineeringRMSE Root Mean Square Error

Page 7: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Contents

Contents

1 Introduction 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Purpose of Master Thesis . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Thesis Goals and Objectives . . . . . . . . . . . . . . . . . . . . . . . 31.4 Delimitation of Study . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Gotland Background Study 52.1 Electric Power System . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.1.1 Power Grid Description . . . . . . . . . . . . . . . . . . . . . 52.1.2 Power Production & Prognosis . . . . . . . . . . . . . . . . . 62.1.3 Consumption Characteristics . . . . . . . . . . . . . . . . . . 62.1.4 Future Objectives and Export Challenges . . . . . . . . . . . 9

2.2 Demand-Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.2.1 What is Demand-Response? . . . . . . . . . . . . . . . . . . . 112.2.2 Demand-Response in Detached Houses . . . . . . . . . . . . . 122.2.3 Demand-Response in Industries . . . . . . . . . . . . . . . . . 14

2.3 Ancillary Service Toolbox . . . . . . . . . . . . . . . . . . . . . . . . 152.3.1 Purpose & Functionality . . . . . . . . . . . . . . . . . . . . . 152.3.2 Flexibility Tools Description . . . . . . . . . . . . . . . . . . 16

2.4 Simulation Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172.4.1 PowerMatcher Technology . . . . . . . . . . . . . . . . . . . . 172.4.2 MATLAB Optimization Toolbox . . . . . . . . . . . . . . . . 19

2.5 Summarized Problem Illustration . . . . . . . . . . . . . . . . . . . . 20

3 Modeling Flexibility Tools 213.1 Demand-Response Participants . . . . . . . . . . . . . . . . . . . . . 21

3.1.1 Detached House Flexibility . . . . . . . . . . . . . . . . . . . 213.1.2 Industry Flexibility . . . . . . . . . . . . . . . . . . . . . . . . 26

3.2 Battery Energy Storage System & Wind Curtailment . . . . . . . . . 263.3 Summarized Model Illustration . . . . . . . . . . . . . . . . . . . . . 28

4 Optimization Method 29

Page 8: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

4.1 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 294.2 Long-Term Cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324.3 Short-Term Cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324.4 Required Input Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 334.5 Optimization Flowchart . . . . . . . . . . . . . . . . . . . . . . . . . 34

5 Simulation Set-Up 355.1 Defining Simulation Scenarios . . . . . . . . . . . . . . . . . . . . . . 355.2 Applying Data Offset . . . . . . . . . . . . . . . . . . . . . . . . . . . 355.3 Applying Production Prognosis Errors . . . . . . . . . . . . . . . . . 375.4 Setting Start Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 38

6 Simulation Results & Analysis 416.1 Scenario 1: Winter days . . . . . . . . . . . . . . . . . . . . . . . . . 41

6.1.1 Expected Power Export . . . . . . . . . . . . . . . . . . . . . 416.1.2 Optimized Consumption . . . . . . . . . . . . . . . . . . . . . 426.1.3 BESS Operation & Wind Curtailment . . . . . . . . . . . . . 456.1.4 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . 45

6.2 Scenario 2: Spring days . . . . . . . . . . . . . . . . . . . . . . . . . 486.2.1 Expected Power Export . . . . . . . . . . . . . . . . . . . . . 486.2.2 Optimized Consumption . . . . . . . . . . . . . . . . . . . . . 496.2.3 BESS Operation & Wind Curtailment . . . . . . . . . . . . . 496.2.4 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . 50

6.3 Scenario 3: Summer days . . . . . . . . . . . . . . . . . . . . . . . . 506.3.1 Expected Power Export . . . . . . . . . . . . . . . . . . . . . 506.3.2 Optimized Consumption . . . . . . . . . . . . . . . . . . . . . 506.3.3 BESS Operation & Wind Curtailment . . . . . . . . . . . . . 516.3.4 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . 51

6.4 Scenario 4: Autumn days . . . . . . . . . . . . . . . . . . . . . . . . 526.4.1 Expected Power Export . . . . . . . . . . . . . . . . . . . . . 526.4.2 Optimized Consumption . . . . . . . . . . . . . . . . . . . . . 526.4.3 BESS Operation & Wind Curtailment . . . . . . . . . . . . . 536.4.4 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . 53

6.5 Scenario Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . 536.5.1 Space Heating Flexibility . . . . . . . . . . . . . . . . . . . . 546.5.2 Export Problem Occasions . . . . . . . . . . . . . . . . . . . 55

7 Discussion 577.1 Model Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577.2 Scenario Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . 587.3 Validity & Reliability of Results . . . . . . . . . . . . . . . . . . . . . 587.4 Thesis Benefits for the SGG project . . . . . . . . . . . . . . . . . . 59

8 Conclusion 61

Page 9: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

8.1 Conclusive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 618.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

Bibliography 65

A Model Calculations 69A.1 Space Heating Slopes . . . . . . . . . . . . . . . . . . . . . . . . . . . 69A.2 Domestic Hot Water Slopes . . . . . . . . . . . . . . . . . . . . . . . 70

B Scenario Result Figures 71B.1 Scenario 2: Spring days . . . . . . . . . . . . . . . . . . . . . . . . . 71B.2 Scenario 3: Summer days . . . . . . . . . . . . . . . . . . . . . . . . 75B.3 Scenario 4: Autumn days . . . . . . . . . . . . . . . . . . . . . . . . 80

Page 10: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Chapter 1

Introduction

1.1 Background

Today’s electricity networks are facing challenges as the pressure increases fromgovernments and political institutions to reduce fossil fuel dependent energy pro-duction and replace them with large amounts of renewable energy production. InMarch 2007, the European Union (EU) set the 20-20-20 targets committing Europeto become a highly energy efficient, low carbon economy and cleaner energy pro-ducer [1]. Most parts of the European electricity networks were built during theearly 20th century, oblivious to the rising environmental concerns that would formnew energy policies. The recent interest of integrating renewable energy sourcesin the existing electricity networks poses technical challenges in maintaining gridstability, matching demand and supply, and transmission capacities. In recent yearsthe "smart grid" concept has emerged as a response to these challenges. The ideahas been to use information exchange and communication technology to improveefficiency and distribution of power in electricity networks. The technology targetsboth producers and consumers where the electricity distribution and usage is opti-mized such that the grid can safely account for renewable energy source integration.The intermittent nature of renewable energy sources makes it difficult to accuratelypredict power production as it is closely tied to complex weather dynamics. Theuncertainty in prognosis and high production variation accentuates the difficultiesin maintaining safe system operation. This is one of many challenges electricity net-works are facing when dealing with mass integration of renewable energy sources.An interesting solution aims at involving the electricity consumer by making themconsume in a proactive manner, for example shift the consumption to periods whenthere is a need from the system to maintain balance. This concept is called Demand-Response (DR) and has become a popular solution. DR solutions has received a lotof attention in the scientific community. One example is a published scientific re-port [2] presenting residential load models for space heating/cooling and tap watersystems. Another example is a study conducted for massive wind power integrationin the Netherlands [3]. The paper shows how DR from households is controlled

1

Page 11: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 1. INTRODUCTION

intelligently to accommodate for mass integration of wind power production. Theresearch has now extended to companies such as Vattenfall AB focusing on imple-menting DR as part of the smart grid applications [4].

The target for the Swedish share of renewable energy by 2020 is approximately50% which has encouraged the development of projects aiming at increasing renew-able energy production. The largest Swedish ongoing project today is the SmartGrid Gotland (SGG) project, which is a collaboration between Vattenfall, GEAB,ABB, Energimyndigheterna, Svenska Kraftnät, Schneider Electric and the RoyalInstitute of Technology (KTH). The project has three overall objectives [4]: Costefficiently increase the hosting capacity for wind power in an existing distributionsystem. Show that novel technology can improve the power quality in a rural gridwith large quantities of installed wind power. And to create possibilities for DR inthe electricity market, in order to shift load from peak load hours to peak productionhours. There are several reasons why Gotland is chosen for a smart grid project.Gotland is in fact an ideal candidate for a small-scale pilot study before extendingthe concept to other parts of Scandinavia. The island is an electrically closed sys-tem, with its own frequency, high wind power production and which only link tothe continent is through an HVDC cable. These characteristics greatly reduces theeconomical and technical complexity of the project and encloses the potential riskswithin the island. The project is a proof of concept of smart grid applications andthe outcomes are intended to pave the way for similar Scandinavian energy projectsin the future.

The Master Thesis focuses on the wind power integration aspect of the SGGproject where 5 MW additional wind power needs to be integrated in the existinggrid without making any changes to the infrastructure. Currently the network onGotland can withstand a maximum installed capacity of 195 MW. Increasing thewind power to 200 MW will cause a risk of exceeding the transmission capacity ofthe export HVDC link from Gotland to the mainland. The study focuses on usingflexibility tools such as DR to balance those 5 MW in the network while ensuringthat the export capacity is not being exceeded. The DR activity includes spaceheating and domestic hot water consumption from detached houses on Gotland.Detached houses are interesting DR candidates as they represent 20% of the totalconsumption on Gotland [16]. Moreover, 75% of the total electricity consumptionin a detached house [5] comes from space heating and domestic hot water activi-ties. The DR activity also includes participation from industries on Gotland wherespecific operation strategies are deployed after receiving requests for DR. Using in-dustries as DR participants can help shift considerable loads and reduce the needfor more household participants.

The approach is to model a technical Ancillary Service (AS) toolbox suitedfor the environmental set-up of Gotland. The AS toolbox is a multi-agent systemconsisting of DR on Long-Term (LT), DR on Short-Term (ST), a stationary BatteryEnergy Storage System (BESS) and a wind power curtailment scheme.

2

Page 12: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

1.2. PURPOSE OF MASTER THESIS

1.2 Purpose of Master ThesisThe main purpose of the Thesis is to study whether it is technically feasible toimplement a technical AS solution on Gotland to balance 5 MW additional windpower capacity.

1.3 Thesis Goals and ObjectivesThe following Master Thesis goals and objectives have been formulated:

• Choosing Agent ModelA literature study will be performed on two alternative ways of modeling theAS toolbox. The first, using PowerMatcher (PM) simulation tool and thesecond using MATLAB optimization toolbox. The advantages between thetwo will be assessed to justify the choice of agent model.

• Defining Simulation ScenariosFeasible simulation scenarios will be derived on production and consumptionoccasions on Gotland.

• Modeling Flexibility ToolsThe flexibility tools will be modeled either by using PM or MATLAB. Thedeveloped models need to reflect the conditions and prerequisites of Gotland.The DR participants will include industries and detached houses where flex-ibility will be provided from appliances used for space heating and domestichot water. An operation strategy for the BESS will be proposed to absorbthe production prognosis errors.

• Performing SimulationsSimulations will be performed in PM or MATLAB for a set of feasible scenar-ios. The simulation results will serve as reference material to conclude on thetechnical feasibility of implementing an AS toolbox on Gotland.

1.4 Delimitation of StudyThe study has been limited in certain aspects to ensure reasonable and qualitativedeliverables within the thesis time frame:

• Geographical DelimitationThe study is limited to the island of Gotland. However the models can beused and rescaled to study other regions where the conditions are similar. Thestudy will not focus on the influence of individual loads and production units,instead the total load and production of the island will be considered as awhole.

3

Page 13: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 1. INTRODUCTION

• DR ParticipantsThe DR participants considered are detached houses and industries. Onlyindustries with documented consumption activity are eligible to participatein DR. Furthermore, only industries having a large influence on the overallconsumption on Gotland will be studied.

• DR Consumption for Detached HousesSpace heating and domestic hot water are the only two consumption activitiesfor detached houses that will be used for ST and LT DR. This is done to reducemodel complexity,

• Prognosis ErrorsIt is assumed that the network can handle 195 MW of installed wind powercapacity. This include the prognosis errors from electricity production andconsumption. Therefore the only prognosis errors considered in the study willbe on the 5 MW of additional power production for both LT and ST prognosis[4].

• DR time resolutionThe DR participants will operate on an hourly resolution for the simulations.In reality, the SGG project needs faster ST DR and BESS operation, typicallyevery 5 minutes.

• Network SimulatorThe multi-agent model needs to communicate with a network simulator toverify the feasibility of the power flows within the network. The networksimulator of choice is Power System Simulator for Engineering (PSS/E). Thecommunication with the network simulator is important to send readjustmentsignals back to the agent model when power flow limitations are encountered.This process will be performed manually although it is intended to be an auto-mated process. The results from PSS/E will be used internally by Vattenfalland will not be included in the Thesis report.

4

Page 14: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Chapter 2

Gotland Background Study

2.1 Electric Power System

2.1.1 Power Grid Description

The island of Gotland is located in the Baltic Sea, east from the town of Västervik.The power distribution grid on Gotland consist of approximately 300 km of 70kVlines, 100km of 30 kV lines and 2000 km of 10 kV lines [7]. A representation of thenetwork is presented in Figure 2.1.

Figure 2.1. The power distribution grid on Gotland [6].

The electric power system on Gotland is a closed system connected only tothe mainland by two HVDC cables. The current HVDC cables were installed in1983 and 1987 to increase the safety of supply as well as the electricity need on theisland. The HVDC cables separate Gotland and the rest of Sweden into two differentfrequency control areas. The frequency on Gotland is therefore independent of thefrequency changes on the mainland. The HVDC cables are used for import andexport of power and consist of two poles with a rating of 130 MW per pole operatingat 150 kV of rated voltage [8]. The HVDC cables are 100 km long submarine

5

Page 15: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 2. GOTLAND BACKGROUND STUDY

cables stretching from the town of Västervik in the mainland to Ygne on Gotland.Power is imported from the mainland during peak loads of approximately 170 MWand in order to assure N-1 criterion, the HVDC cables do not transmit powersimultaneously in the same direction. The importing HVDC cable runs often on fullcapacity because of the sometimes low local production. This has lead authoritiesto plan the construction of two additional HVDC cables with 2x500 MW powercapacity. The cables are planned to be operational by year 2017 and 2020 [9].

2.1.2 Power Production & Prognosis

The power production on Gotland consists of wind power. There are gas and dieselbased power production on the island with enough installed capacity to cover thetotal demand in case of unexpected outages [10]. In recent years, many small windfarms have been replaced by larger ones, thereby increasing the total installed windpower capacity to approximately 170 MW [12]. Hence, the installed capacity isapproaching the maximum grid limit of 195 MW [4]. The intermittent nature ofwind power makes it difficult to establish a typical yearly power production trend.Generally, wind power production is higher during cold seasons such as winter andautumn because of stronger, more frequent wind blows and greater air density.Figure 2.2 presents the wind power production on Gotland during year 2012.

Rarely, the wind production on Gotland generates at its full capacity. In 2011,the annual wind power production was 340 GWh equating to about 38% of Gotlandstotal consumption [13]. Furthermore, wind power production is difficult to forecast.The production prognosis is based on climatology models taking into account factorssuch as wind speeds, forces and other types of weather conditions. As of today theproduction prognosis error increases rapidly the further ahead in time a prediction ismade. The maximum forecast length of most models today are 48-178 hours ahead[14]. Naturally, a one hour-ahead prognosis is much more accurate and reliable thana day-ahead prognosis. Figure 2.3 presents the wind power production prognosiserror on Gotland for different hours.

2.1.3 Consumption Characteristics

Today, Gotland has approximately 57,000 inhabitants. The population has beenstagnating since 1995 and no forecasts have been made on heavy population growth[15]. This means that the overall consumption profiles on the island will not besubject to any drastic changes during the next coming years. The average totalconsumption since 2002 has been measured to approximately 860 GWh/year, i.e.,100 MWh/h [16]. The power consumption from January 1st to December 31stresembles a valley where the peaks occur during cold seasons such as winter andautumn. Figure 2.4 presents an average of the total daily consumption on Gotlandduring year 2012.

Gotland accounts for approximately 0.65 % of the total Swedish consumption[16]. The high-consuming loads are typically industries, detached houses and col-

6

Page 16: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

2.1. ELECTRIC POWER SYSTEM

Seasonal Average Production

Yearly Average Production

Daily Average Production

Pow

er[M

Wh/h

]

Days

0 50 100 150 200 250 300 3500

20

40

60

80

100

120

140

160

180

Figure 2.2. Daily averages on the total power production on Gotland during year2012 [11]. Day 1 corresponds to January 1st 2012. The yearly average production is47 MWh/h. Examples: winter average production is 51 MWh/h and summer averageproduction is 35 MWh/h. The installed wind power capacity during 2012 was 170MW [12].

lective apartments. Figure 2.5 presents the different load profiles and shares basedon statistics from 2011.

Detached Houses

In 2012 there were 20,590 detached houses on Gotland [16]. According to statisticsfrom Figure 2.5, 20% of the total consumption on Gotland is consumed by detachedhouses. This typically includes high-consuming activities such as space heating,domestic hot water and the use of electric appliances.

Collective Apartments

Collective apartments represent only 4% of the total consumption on Gotland. Theapartments are heated from district heating power plants. The plants are operatedby Gotlands Energi AB (GEAB), the local power utility company, and cover themost populated areas on Gotland, i.e. Visby, Klintehamn, Hemse and Slite [17].

7

Page 17: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 2. GOTLAND BACKGROUND STUDY

Normalized RMSE (hourly)Per

centa

ge

Time [hours]

18.79 %17.79 %

6.12 %

0 4 8 12 16 20 24 28 32 36 40 44 486

8

10

12

14

16

18

20

Figure 2.3. Normalized Root Mean Square Error (RMSE) on hourly wind powerproduction prognosis on Gotland. The graphs are compiled based on hourly windpower production data from Gotland [11]. The production prognosis is estimatedusing a persistence method, where one assumes that the prognosis for time step t + 1will produce as much as the current time step t. The process is repeated for largersteps up to t + 48. The prognosis errors presented are by no means conclusive. Theyare only used for illustrative purposes and a lot more advanced techniques are usedto determine wind power production prognosis errors in reality [14].

Industry

Although Gotland is not an industrial region, the industry accounts for approx-imately one third of the total electricity consumption. The company Cementa,operating in the cement industry, is the main player responsible for this. Theiractivity is estimated to cover 86% of the total industrial consumption. The remain-ing 14% are divided among the companies Arla (5%), Nordkalk (6%) and others(3%). A field study carried out by ABB presents detailed information about theconsumption activities from these three companies [18].

Others

The consumption corresponding to ’others’ comprise commercial entities, the publicsector, transportation and other services. These services represent approximately40 % of the total consumption.

8

Page 18: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

2.1. ELECTRIC POWER SYSTEM

Seasonal Average Consumption

Yearly Average Consumption

Daily Average Consumption

Pow

er[M

Wh/h

]

Days

0 50 100 150 200 250 300 35060

80

100

120

140

160

180

Figure 2.4. Daily averages on the total consumption on Gotland during 2012 [11].Day 1 corresponds to January 1st 2012. The yearly average consumption is 107MWh/h. Examples: winter average consumption is 136 MWh/h and the summeraverage consumption is 87 MWh/h.

Others358 GWh/year 42%

Industries288 GWh/year 34%

Apartments34 GWh/year 4%

Small houses170 GWh/year 20%

Figure 2.5. Electricity consumption per category in GWh/year during 2011 [16](small houses = detached houses).

2.1.4 Future Objectives and Export ChallengesThe installed wind power capacity has increased rapidly during the last years onGotland. In August 2011, there were 158 wind turbines on Gotland with a total

9

Page 19: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 2. GOTLAND BACKGROUND STUDY

installed capacity of 118 MW which during 2010 produced 0.2 TWh of electricity[19]. Today, the installed wind power capacity is close to the maximum grid limit of195 MW. The regional ambition is to produce an annual of 2.5 TWh of electricitywhich means that at least 500 additional wind power plants need to be built. Thechallenges Gotland faces are to integrate the production in the existing distributionnetwork without having to make extensive investments in the infrastructure. Theproblem that arises when the installed capacity exceeds 195 MW in the currentnetwork is that the transmission capacity of the export HVDC link gets overloadedduring hours of high production and low consumption. Historical data suggeststhat consumption has never been as low as to provoke an export problem. Thesefindings are presented in Figure 2.6.

HVDC transmission capacity

Hourly power export 2012

Pow

er[M

Wh/h

]

Time [hours]

0 1000 2000 3000 4000 5000 6000 7000 80000

20

40

60

80

100

120

140

160

Figure 2.6. The exported power between Gotland and the mainland during 2012.The transmission capacity of the HVDC cable is 130 MW [8]. The installed capacityduring 2012 is estimated to around 170 MW [12].

One has to consider a worst case scenario where the production generates at themaximum grid limit of 195 MW and where the consumption is greatly reduced toprovoke an export problem. Figure 2.7 presents observed correlations between timeof the day and export occasions. The figure show that there are no correlationsbetween export occasions and the day of the week. The export occasions are morelikely to occur during night hours (i.e. between 22:00 - 04:00) because of the over-all low consumption. The bar charts complement the accumulated export power

10

Page 20: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

2.2. DEMAND-RESPONSE

Figures for a better statistical analysis.

Accumulated power export occasions per hour of the day (year 2012)

Expor

tocc

asio

ns

Time [hours]

Accumulated power exported per hour of the day (year 2012)

Pow

er[M

W]

Time [hours]

0 3 6 9 12 15 18 21 24

0 3 6 9 12 15 18 21 24

0

20

40

60

0

500

1000

1500

Accumulated power export occasions per day of the week (year 2012)

Expor

tocc

asio

ns

Days Monday-Sunday

Accumulated power exported per day of the week (year 2012)

Pow

er[M

W]

Days Monday-Sunday

1 2 3 4 5 6 7

1 2 3 4 5 6 7

0

5

10

15

20

25

1000

1500

2000

2500

3000

3500

Figure 2.7. The left figures show the correlation between export occasions and thehours of the day during year 2012. The right figures show no or little correlationbetween export occasions and the day of the week.

There are also some issues concerning the import of power during the year.During those occasions the local diesel and gas power plants are operated to avoidoverloading the HVDC cable.

2.2 Demand-Response

2.2.1 What is Demand-Response?

To maintain the stability of an electric power system there has to be a constantbalance between power production and consumption. Traditionally, power produc-ers ensure the balance by either increasing or decreasing their production accordingto demand. Demand side management on the other hand, consists of doing thecomplete opposite, i.e., to adjust consumption according to what is being produced.One increasingly popular example of demand side management is DR where con-sumers change their consumption behaviour after receiving a DR signal. Examplesof DR participants comprise households, industries or the public sector. The in-centives for consumers to engage in DR activities are mostly driven by economicalbenefits. This means that the DR signal sent is often an indication of low electric-ity market prices. DR has become a popular research topic in the field of smartgrid applications thanks to the technical and economical benefits it provides. Oneexample is presented in [22] where DR allows the integration of large quantities ofwind power in an existing grid.

Active DR

Active DR is the process where consumers actively change their consumption basedon requests. In detached houses, active DR includes all consumption that is man-

11

Page 21: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 2. GOTLAND BACKGROUND STUDY

ually controlled, e.g. computers, ovens, stoves, dishwashers, showers, baths, etc.For industries, it is any type of electricity consuming activity that is not operatedby a control system. Being an active DR participant is a sign of engagement andcommitment. These participants have good knowledge of their consumption andare willing to trade comfort for other benefits.

Remote DR

When consumers engage in remote DR they sign up to let their consumption becontrolled by an external entity. For example, it might be the control of appli-ances used for space heating or domestic hot water which offers a certain degree ofconsumption flexibility. The appliances are controlled within a select region whichis judged to have no, or very little, impact on the comfort level of the consumer.Hence, remote DR allows consumers to maintain regular activities with the illusionof being a passive user. The automated process that this type of DR provides makesit easier for large scale implementation. Moreover, the barriers of realization areconsiderably lower than active demand since the consumer commitment is mini-mized. Remote DR offers a lot more reliability than active DR which is crucialfrom a power system stability perspective.

2.2.2 Demand-Response in Detached Houses

The average Swedish detached house uses approximately 55% of its electricity forspace heating, 20% for domestic hot water and the remaining percentage for house-hold equipments [5]. Among these activities both space heating and domestic hotwater can be used as remote DR.

Space Heating

Almost every house on Gotland consume electricity for space heating purposes.The indoor temperature is generally maintained close to a reference temperatureof 20 degrees Celsius. During days of low outdoor temperature the consumptionfor space heating increases to prevent the indoor house temperature from droppingbelow the reference. Figure 2.8 presents the most common energy sources used forspace heating in detached houses on Gotland.

The most abundant form of energy used is a combination of oil, electricity andbio. Details about the methods used for space heating are further explained onthe website of the Swedish Energy Agency [5]. Direct and water carried energyare common space heating methods where the thermal energy spreads through theradiators of the house. These processes can be used as remote DR where a constraintis set for a maximum and minimum allowed indoor temperature. The temperatureboundaries are chosen in such a way to allow room for consumption flexibility whileavoiding experienced consumer discomfort.

12

Page 22: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

2.2. DEMAND-RESPONSE

Others 14.7%

District Heating 15%

Heat Pump 7%

Bio 2.3%

Oil+Elec 15.9%

Oil+Elec+Bio 21.6%

Elec.(w) 10.8%

Elec.(d) 9.2%

Figure 2.8. Energy sources used for space heating in detached houses on Gotland(2010) [16]. Elec.(d) is direct electricity whereas Elec.(w) denotes water carried elec-tricity.

Figure 2.9. Direct electricity (Elec.(d)) on the left and water carried electricity(Elec.(w)) on the right. Images are taken from the Swedish Energy Agency website[5].

Domestic Hot Water

Most detached houses which are not part of district heating use water tanks equippedwith a boiler to maintain the hot water at a reference temperature. The hot waterfrom the tank is drained for showers, baths, hand-washing and similar activities.Every time water is drained, the tank temperature drops and the boiler needs toconsume more to prevent the temperature from dropping below 60 degrees C◦. Thisconstraint is set to minimize the occurrence of legionella bacterias in the water [20].The typical temperature interval of a water tank is between 60 and 100 degrees C◦,leaving enough margin for boiler consumption flexibility. One could for example letthe boiler consume more during hours of low water drainage and less during otherhours. This type of flexibility is ideal for remote DR purposes.

13

Page 23: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 2. GOTLAND BACKGROUND STUDY

2.2.3 Demand-Response in IndustriesIndustries on Gotland account for a high share of electricity consumption (see sec-tion 2.1.3), having an industry participate in DR can help shift considerable loads.Industries offer mostly active DR solutions. It might for example involve movingproduction hours from the morning to the evening or utilize existing buffers byconsuming more during certain hours. The following examples illustrate potentialday-ahead DR strategies for high power consuming companies on Gotland [18].

Cementa

The first three steps of the company process flow is presented in Figure 2.10. Thecompany performs stone quarry activity during day shifts, twice a week on weekdays.During times when the production is lagging the demand, the company can be issueda permission from the county for one extra weekday of stone quarry activity. Thestone crushing activity consumes 2.8 MW and the crushed stones are later sent forstorage where they can remain for 4 days. If the company can be issued a similarpermission from the county for DR purposes then power consumption activity canbe utilized.

Figure 2.10. Cementa process flow [18]. Step 1 is the stone quarry activity. Step 2is the stone crushing activity consuming 2.8 MW. Step 3 is the storage of the crushedstones where it can stay for a maximum of 4 days before it moves on to the next step.

Arla

The company operates in the dairy industry and deliver its product to grocerystores. 80% of the electricity consumption goes to fabricate the milk powder. Thisprocess involves 20 hours of drying and 4 hours of washing. As soon as one batchis dried then the washing process begins, followed by the next batch. For hygienicpurposes the washing has to occur within 20 hours. Relative to drying, the washingprocess uses a lot less energy. The DR potential is to relocate the hours when thewashing occurs.

Nordkalk

This company operates in the mineral industry, one of their activity is limestonequarry which occurs twice a day. The stones are crushed on sight through 2x250kW stone crush motors. The stones are transported to the docks where they arecrushed once again in the correct sizes and then categorized accordingly. The DR

14

Page 24: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

2.3. ANCILLARY SERVICE TOOLBOX

possibilities for the company are limited, however loading the boat with stonesconsumes close to 0.5 MW which can be relocated to other hours of the day sincethe boats stays at the dock for 24 hours. The loading activity is estimated to takebetween 6 and 8 hours which opens up a possibility for DR.

2.3 Ancillary Service Toolbox

2.3.1 Purpose & Functionality

The purpose of the AS toolbox is to communicate with flexibility tools to balanceadditional power production in the existing network without overloading the exportHVDC cable. The additional power production is 5 MWwhere the installed capacityhas been increased from 195 MW to 200 MW. A worst case scenario is consideredwhere the wind production generates at installed capacity. The AS toolbox is amulti-agent model using LT and ST production prognosis data as input. The agentsof the AS toolbox are presented in Figure 2.11.

Figure 2.11. AS toolbox communication channels. The agent model operates theflexibility tools every time there is an export problem prognosis. The informationis exchanged with PSS/E to verify the feasibility of the power flow in the network.PSS/E sends readjustment signals to the agent model when power flow limitationsare experienced (refer to section 1.4 for delimitation of study).

The data input to the agent model will be both LT and ST production andconsumption prognosis. The ST production prognosis is much more accurate thanLT prognosis. This was presented earlier in Figure 2.3.

15

Page 25: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 2. GOTLAND BACKGROUND STUDY

2.3.2 Flexibility Tools Description

The model uses flexibility tools such as LT DR, ST DR, a BESS and a wind cur-tailment scheme to balance additional power production in the network when anexport problem occurs.

LT DR

LT DR refers to DR scheduled 24-48 hours in advance (day-ahead). This activityinclude the control of space heating and domestic hot water for detached houseparticipants. Industry participants are also part of LT DR since it is assumed thatthey need at least 24 hours to reschedule and plan their consumption activity forthe upcoming day. Once the day-ahead consumption has been set then the spaceheating and domestic hot water consumption for detached houses will be controlledthe following day and industries will have to consume what they agreed upon.

ST DR

Short-term DR refers to DR scheduled one hour in advance (hour-ahead). Thisactivity include the control of space heating and domestic hot water for detachedhouse participants. It is assumed that industries are not flexible for DR on a hour-ahead basis since most of their high consuming activities are planned well ahead.

BESS

The BESS is used to absorb prognosis errors from wind power production. Eventhough ST production prognosis have high accuracy, the ST DR participants willnot be able to account for all imbalances. The BESS has a capacity of 280 kWhand has the ability to absorb and deliver power to the network. It is assumed thatthe charge rate is 280 kWh. Electricity is consumed when the BESS is absorbingpower and the production is increased when the BESS is delivering power.

Wind Curtailment

Wind curtailment is used as a last resort when other flexibility tools are unableto balance the power in the network. When curtailment is needed a signal is sentto wind turbines to either shut down or run at lower speeds. This process is verycostly and has a damaging effect on the turbines.

16

Page 26: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

2.4. SIMULATION TOOLS

2.4 Simulation Tools

2.4.1 PowerMatcher Technology

PM in Brief

PM is a communication and coordination technology used for supply demand match-ing in electricity networks. It is a multi-agent based system that uses electronicexchange markets to coordinate the supply and demand of a cluster of devices.The multi-agent is used to represent complex networks consisting of devices suchas electricity production units, electricity storage devices (e.g. BESS), electricityconsumption patterns, etc. The agents interact with each other in order to reachan optimal solution. The main purpose of PM is to facilitate the implementation ofsmart grids for conventional and renewable energy sources. Using intelligent clus-tering, small electricity producing or consuming units gain in operation flexibilitywhich can add value to the electricity market or in the case of the thesis, helpbalance additional production capacity in an existing network. [21]

Figure 2.12. Logical tree of PM cluster. Image taken from PM website [21].

PM Structure

The PM technology has a tree structure consisting of agents such as the auctioneer,the concentrator, the device and the objective agent (see Figure 2.12). All agentstry to operate the process associated with its child in an economical optimal way.The information exchanged between different agents is bids and prices which expresshow much agents are willing to pay or to be paid for a certain amount of electricity.

17

Page 27: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 2. GOTLAND BACKGROUND STUDY

The root of the tree is the auctioneer agent which collects the bids and calculatesthe market price. The market price is communicated to the device agents who basedon the information make an assessment on production or consumption of electricity.In some cases neither of the two alternatives is chosen and the device agent waitsinstead until the market price changes. The agents are described in more detail:

Auctioneer agent

It is the root of the logical tree and collects all bids from its children. It forms anequilibrium price based on the price signals received from its children and commu-nicates the market price back to them.

Concentrator agent

It mimics the auctioneer by collecting all bids from the device agents and forms abid price which is communicated to its root.

Device agent

It is a control agent that operates the device in an economical optimal way. Thisagent communicates with all other agents in the cluster by buying or selling theconsumed or produced electricity of the device. The market price and the latestprice of the device agent determine how much electricity that will be produced orconsumed.

Objective agent

It determines the objective of the clusters which most often is to maintain balancebetween electricity production and consumption but additional constraints couldalso be added on network transmission capacity.

PM Limitations

The PM technology was successfully used for a similar study in the Netherlandswhich showed that DR in households could accommodate mass integration of newwind power [22]. However, the PM software presented limitations for the Gotlandstudy. All models used for flexibility tools such as space heating, domestic hot waterwere not configurable from the source. The user is only able to toggle few param-eters such as maximum and minimum capacity. The original source files neededmodification for the models to reflect realistic conditions on Gotland. Furthermore,the selection of wind parks only included those from the Netherlands. Modifying thesource files implies extensive programming and reconfiguration. Unfortunately, thelack of documentation, thesis time frame, and lack of transparency in the softwarelead to re-evaluate the necessity of using PM as an agent model for this study. Thisdecision does not imply that PM was unfit for the study but rather that extensive

18

Page 28: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

2.4. SIMULATION TOOLS

programming and further research is needed to make a rightful assessment on thematter.

2.4.2 MATLAB Optimization Toolbox[23] The MATLAB optimization toolbox is an extension of the standard MATLABversion. The toolbox provides algorithms for optimization problems of any sizes andkinds. The software include linear programming, quadratic programming, binaryinteger programming, nonlinear optimization, nonlinear least squares, systems ofnonlinear equations and multi-objective optimization. These functions can be usedto compute an optimal solution subject to a set of constraints and boundaries. Thetoolbox is of high relevancy for the Gotland study where the power export occa-sions can be formulated as a minimization problem subject to transmission capacityconstraints. The advantage of using MATLAB as a simulation tool is the high de-gree of freedom that is offered. Any data, variables or models are easily modifiablewhich is highly desired, especially for Gotland, having such unique characteristicsin production, consumption and environmental set-up.

19

Page 29: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 2. GOTLAND BACKGROUND STUDY

2.5 Summarized Problem IllustrationThe purpose and approach of the Thesis are summarized in the illustration below.

LONG-TERM

SHORT-TERM

REQUIRES

PROBLEM CAUSE

SOLUTION?

AS TOOLBOX

DOMESTIC

HOT WATER

SPACE

HEATING

INDUSTRY

OPERATION

STRATEGY

DEMAND

RESPONSE

BATTERY

SYSTEM

WIND

CURTAIL.

SIMULATION

POWER

MATCHER

MATLAB

ABSORB

PROGNOSIS

ERRORS

MODELING

EXPORT

OVERLOAD 200 MW

INSTALLED

WIND

CAPACITY

DATA

COLLECTION

SCENARIO

DESCRIPTION

Figure 2.13. Note that MATLAB was chosen as the most suitable software for thisstudy.

20

Page 30: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Chapter 3

Modeling Flexibility Tools

3.1 Demand-Response Participants

3.1.1 Detached House Flexibility

Every detached house has its own set of characteristic such as the house area, thenumber of windows, the exposure to solar radiation, heat losses through surround-ing walls, etc. The difference in characteristics makes it hard to build an accuratemodel to simulate household consumption profiles. One approach explored in [24]uses Markov chain methodology to establish an electricity consumption profile of atypical household. The model takes into account three consumption modules, spaceheating, domestic hot water and the use of electric appliances. Using this method-ology one can produce synthetic electricity demand profiles. The model results werevalidated with empirical data for the consumption of 41 Swedish residents living indetached houses showing high statistical correlation with the actual consumption.These results indicate that the mathematical model used for the three consumptionmodules can be used to simulate the consumption profile for detached houses onGotland when the parameters are adjusted to reflect the conditions on the island.The following subsections will present the mathematical model and the parame-ters used to simulate the electricity consumption activities of detached houses onGotland.

Space Heating Model & Parameters

To use the space heating model proposed in [24], one first has to define detachedhouse parameters. Similar parameters from the study in [24] were chosen to reflectdetached house characteristics on Gotland. The parameter values correspond tothose of an average Swedish detached house. They are summarized in Table 3.1.

These parameters are used to construct the space heating model. The indoortemperature of a detached house will deviate from the reference value due to thefollowing reasons: space heating consumption Qheat(t), outdoor temperature vari-ations Tout(t), solar radiation Qsun(t), presence of occupants Qocc(t), the use of

21

Page 31: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 3. MODELING FLEXIBILITY TOOLS

Table 3.1. Model parameters for a detached house on Gotland [24].

Parameter Value Unit DescriptionAfloor 100 m2 Total floor areaAroof 100 m2 Total roof areaAwall 79 m2 Total wall areaAdoor 4 m2 Total door areaAwindow 20 m2 Total window areaAside,window 5 m2 Total side-window areaUfloor 0.4 W/(m2 · C◦) Transmission factor of floorUroof 0.25 W/(m2 · C◦) Transmission factor of roofUwall 0.3 W/(m2 · C◦) Transmission factor of wallUdoor 1.5 W/(m2 · C◦) Transmission factor of doorUwindow 3 W/(m2 · C◦) Transmission factor of windowdheight 2.5 m Height of one-floor houseVhouse 250 m3 Volume of one-floor houseρair 1.20 kg/m3 Density of air at room temperatureCpair 0.28 W/(kg · C◦) Heat capacity factor of air at room temperatureαred 1 - Heat reduction factorαrc 0 - Heat recycling factorNvent 0.2 h−1 Exchange rate of airτ 100 h Time it takes to drop 3C◦ during a cold snap [25]

electric appliances Qapp(t) and space heating losses QSH,loss(t). There are also twoother sources of loss considered: the transmission losses of heat λtrans and the ven-tilation losses λvent caused by the exchange of heated and surrounding air. Theselosses constitutes the inertia of the system and are calculated from the detachedhouse parameters as followed: λtrans =

∑j∈J

Uj · Aj

λvent = Vhouse · Nvent · Cpair · (1 − αrc)(3.1)

Where the set J includes all building components of the house except the side-window (the side-window is modeled as the window exposed to solar radiation).The overall dynamics of the indoor temperature of a detached house on Gotland attime step t+ 1 is described as:

T(t + 1) = T(t) + 1τ ·(λtrans+λvent) · (Qheat(t) + Qsun(t) + Qocc(t) + . . . [C◦]

Qapp(t) − QSH,loss(t))(3.2)

The expression 1τ ·(λtrans+λvent) denotes the inertia of the system. The reader is

referred to [25] for a better understanding on how the inertia of the system wasdetermined. Having all time dependent variables in hourly resolution allows one

22

Page 32: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

3.1. DEMAND-RESPONSE PARTICIPANTS

to make predictions about the indoor temperature variation for the following hour.Thereby, the relation between space heating consumption and indoor temperaturebecomes predictable. The following list will detail on how the different variablesQ(t) from Equation 3.2 are calculated:

(i) The heat consumption Qheat(t) is equal to the consumed electricity for spaceheating PSH(t). The minimum value that is consumed per household has been setto 0 W while the maximum consumption is calculated from the detached houseparameters as:

PSH,max = (λtrans + λvent) · (Tref − TDUT ) [W] (3.3)

Where Tref is a reference indoor temperature and TDUT the dimensioning win-ter temperature on Gotland [25]. Note that the heat consumption Qheat(t) is theoptimization variable that affects the indoor temperature variation.

(ii) The heat contribution from solar radiation Qsun(t) at time t is expressedas:

Qsun(t) = αred · Psun(t) · Asidewindow [W] (3.4)

Where Psun(t) is the radiated sun heat per square meter for a detached houseat time t (W/m2). The other variables in the expression are detached house pa-rameters.

(iii) The heat contribution from house occupants Qocc(t) is assumed to be onaverage 100 W/person. It is also assumed that the average household in Swedenconsist of 3 persons/household. The assumptions are based on the study in [24].

(iv) The heat contribution from electric appliances in the house Qapp(t) is dif-ficult to estimate, it is assumed that the consumed electricity is on average 380W/household. The assumptions are based on the study in [24].

(v) The total space heating loss QSH,loss(t) of the system at time step t isexpressed as:

QSH,loss(t) = (T(t) − Tout(t)) · (λtrans + λvent) [W] (3.5)

Using the expressions formulated in (i)-(v) one can solve Equation 3.2 and thusestimate the indoor temperature at the next time step. Some of the expressions in(i)-(v) require outdoor temperature or solar radiation data for calculations. Table3.2 summarizes the space heating parameters derived and the comfort interval forindoor temperature in a detached house based on the study made in [24].

23

Page 33: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 3. MODELING FLEXIBILITY TOOLS

Table 3.2. Space heating parameters and base assumptions for a detached house onGotland [24].

Variable Value Unit Descriptionλtrans 154.7 W/C◦ Transmission lossesλvent 14 W/C◦ Ventilation lossesPSH,min 0 kWh Minimum space heating consumptionPSH,max 4.61 kWh Maximum space heating consumptionTref 20 C◦ Reference indoor temperatureT0 20 C◦ Initial indoor temperatureTmin 18 C◦ Minimum indoor temperatureTmax 22 C◦ Maximum indoor temperatureTDUT -9.7 C◦ Dimensioning winter temperature on Gotland [26]

Domestic Hot Water Model

The domestic hot water module is modeled as a simplified tank with a water inflow ofconstant temperature Tinlet and a consumer controlled outflow of assumed constanttemperature Toutlet. The outflow depends on every day activities such as showering,baths and hand-washing. It is assumed that the water level in the tank is alwayskept at a constant level, i.e., the water inflow equals outflow at all times. Thewater dynamics in the tank are more easily understandable for DR purposes whenexpressed in terms of energy produced and energy consumed. The energy producedin the tank is the energy provided by the boiler. The energy consumed from thetank is Qdrain,i(t) from showers, baths and hand-washing where i denotes eitherweekday or weekend activity. The tank is also drained on energy from losses, whereQDHW,loss(t) denotes the losses from the surrounding walls of the tank. The wateroutflows are expressed in terms of consumed energy through the transformation inEquation 3.6:{

Qdrain,i(t) = Vflow,i(t) · Cpwater · (Toutlet − Tinlet) [W]QDHW,loss(t) = λtank · (Ttank(t) − Tamb(t)) [W] (3.6)

Where Vflow,i(t) denotes the drained water from the tank at time t in litreswith i indicating whether it is a weekday or a weekend, Cpwater the heat capacityof water, λtank the insulation factor of the water tank and Tamb(t) the ambient airtemperature at time t. The heat capacities in the water tank are calculated fromEquation 3.7:

Mmin(t) = Vtnk · Cpwater · (Ttnk,min − Tinlet) [W]Mmax(t) = Vtnk · Cpwater · (Ttnk,max − Tinlet) [W]Mref (t) = Vtnk · Cpwater · (Ttnk,ref − Tinlet) [W]

(3.7)

Where Vtnk denotes the volume of the tank and Ttnk,min and Ttnk,max theminimum and maximum allowed tank temperatures. The model parameters andconstants are summarized in Table 3.3.

24

Page 34: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

3.1. DEMAND-RESPONSE PARTICIPANTS

Table 3.3. Parameters and constants for domestic hot water model for a detachedhouse on Gotland [24].

Variable Value Unit DescriptionTmin 60 C◦ Minimum tank temperatureTmax 100 C◦ Maximum tank temperatureTref 80 C◦ Reference tank temperatureTamb 20 C◦ Ambient air temperatureTinlet 10 C◦ Inlet water temperature to tankToutlet 40 C◦ Outlet water temperature from tankVtnk 300 L Volume of water tankCpwater 1.17 Wh/(kg C◦) Heat capacity of waterλtnk 1 W/C◦ Insulation factor of water tankMmin 31.6 kWh Minimum heat capacity in water tankMmax 17.6 kWh Maximum heat capacity in water tankMref 24.6 kWh Reference heat capacity in water tankPboil,max 3 kWh Maximum boiler consumption capacity

The consumption of the domestic hot water tank Pboil(t) should be equal to thesum of the losses from the tank and the drained water at all time steps to maintainthe tank temperature at reference level, i.e.

Pboil(t) = QDHW,loss(t) + Qdrain,i(t) (3.8)

Note that energy drained can exceed the maximum boiler capacity Pboil,max. Inthat case the boiler consumes at maximum capacity, i.e. Pboil(t) = Pboil,max.

Required Input Data

The detached house model requires input data for the calculation of some of thespace heating and domestic hot water parameters. The required input data ispresented in Table 3.4.

Table 3.4. Required input data for detached house model. Note that the resolutionof the data is on hourly basis and that all vectors are of same length (minimumallowed length 24x1)

Data Input Consumption Activity Unit DescriptionQsun(t) Space heating W Solar radiationQocc(t) Space heating W Occupant heat contributionQapp(t) Space heating W Electric appliance consumptionQSH,loss(t) Space heating W Heat lossesQdrain,weekday(t) Domestic hot water W Weekday hot water drainedQdrain,weekend(t) Domestic hot water W Weekend hot water drainedQDHW,loss(t) Domestic hot water W Heat losses to surrounding tank walls

25

Page 35: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 3. MODELING FLEXIBILITY TOOLS

Qdrain,i(t) and Vflow,i(t) are interchangeable and in some cases it is easier toget a hold of domestic hot water data in Watt. In that case the first expression inEquation 3.6 is omitted.

3.1.2 Industry FlexibilityCementa Consumption Model

The DR consumption of Cementa is assumed to be operational only during week-days and day shift hours, i.e. hours between midnight and noon. The hourly DRconsumption Pcementa(t) becomes:{

Pcementa(t) = 2.8 MW for all t = 1, . . . , 12Pcementa(t) = 0 MW otherwise (3.9)

Where t denotes the hour of the day. Note that the industry participation will onlyoccur if there is an export problem prognosis during day shift weekdays.

3.2 Battery Energy Storage System & Wind CurtailmentThe BESS is modeled as a simple battery where the discharge of energy increasesthe power export due to more power flowing in the network. The charging on theother hand decreases the export as a consequence of consuming more power.TheBESS has a capacity of 280 kWh and all power excess that is not absorbed by theBESS is cut through the wind curtailment scheme. The proposed operation strategyis summarized in flowchart 3.1.

The BESS operation will result in a number of battery cycles depending on dailyconditions. A critical cycle consist of charging the battery close to it’s maximumcapacity and fully discharging it in a very short time interval. Battery cycles,especially critical ones, have a damaging effect on the BESS by reducing its lifeexpectancy. It is desired to have as low battery cycles as possible because of thehigh cost associated with replacing a BESS.

26

Page 36: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

3.2. BATTERY ENERGY STORAGE SYSTEM & WIND CURTAILMENT

No

Yes

step step

No

No

Start

Export problem

after hour t?

Wind

curtailment

until export

problem solved

BESS fully

charged?

Charge BESS

as much as

possible

Discharge

BESS as much

as possible

while avoiding

new export

problems

Export

problem

solved?

BESS

empty?

Yes

No

Yes

t = t + 1

t <= 24 Stop

Yes

No

Yes

Start at t=0

Figure 3.1. Proposed BESS and wind curtailment operation strategy.

27

Page 37: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 3. MODELING FLEXIBILITY TOOLS

3.3 Summarized Model IllustrationThe modeling of the flexibility tools are illustrated below.

AS TOOLBOX REQUIRES

MODELING

Detached house

flexibility

Industry

flexibility

Water

tank

~

𝑄𝐷𝐻𝑊,𝑙𝑜𝑠𝑠(𝑡) 𝑃𝑏𝑜𝑖𝑙(𝑡)

𝑃𝑆𝐻(𝑡)

𝑄𝑑𝑟𝑎𝑖𝑛,𝑖(𝑡) House

parameters

𝑇𝑖𝑛,𝑚𝑖𝑛 ≤ 𝑇𝑖𝑛(𝑡) ≤ 𝑇𝑖𝑛,𝑚𝑎𝑥

𝑀𝑡𝑛𝑘,𝑚𝑖𝑛 ≤ 𝑀𝑡𝑛𝑘(𝑡) ≤ 𝑀𝑡𝑛𝑘,𝑚𝑎𝑥

Production

12 24

2.8 MW Offers flexibility

during weekdays

and dayshifts!

t

BESS and wind

curtailment

t

280 kW

Charge/discharge

strategy for BESS.

Wind curtailment used

as a last resort!

𝑇𝑡𝑛𝑘,𝑚𝑖𝑛 ≤ 𝑇𝑡𝑛𝑘(𝑡) ≤ 𝑇𝑡𝑛𝑘,𝑚𝑎𝑥

Figure 3.2. Summarized illustration of flexibility tools modeling for the AS toolbox.

28

Page 38: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Chapter 4

Optimization Method

4.1 Problem FormulationThe operation of flexibility tools from the AS toolbox requires optimization whileconsidering a series of technical constraints. The objective of the Thesis can beformulated as an optimization problem:

Maximize The sum of DR consumption for detached housesduring hours of export problem prognosis.

Subject to Transmission capacity constraintLoad shifting balance constraintSpace heating constraintsDomestic hot water constraints

The DR from Cementa is not included in the objective function of the opti-mization problem. The reason for this is that industry consumption activity is notremotely controlled and thus no optimization is possible. It is however considered asan optimization parameter. For detached houses, the space heating and domestichot water consumption are considered as remote DR and therefore require opti-mization. The optimization problem formulation is the same regardless of LT or STDR, the only difference is in input data. Mathematically, the linear optimizationproblem is formulated as:

Maximize fTx

Subject to Ax ≤ bAeqx = beq

Upper/lower lb ≤ x ≤ ubbounds

(4.1)

All variables formulated in Equation 4.1 are described below:

29

Page 39: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 4. OPTIMIZATION METHOD

The objective function fTx

The number of optimization variables as well as the hours simulated will determinethe size of the vector fT . Each hour containing an export problem will be assignedthe value 1, otherwise the element will be assigned the value 0. The optimizationvariables in x are:

x = [PSH(t) Pboil(t) Mtnk(t) Ttnk(t) Tin(t)] (4.2)

Where the size of each element of x correspond to the number of hours simulated.The optimization variables are dimensioned per house and need to be multiplied bythe total number of detached houses participating in DR to prevent export problemsfrom occurring.

Inequality constraints Ax ≤ b

There are two inequality constraints formulated and they apply for time steps t =1, 2, . . . , N , where N is the number of hours simulated. The matrices A and b areconstructed from the inequalities below.

Transmission capacity constraint

proddati(t) − consdati(t) ≤ transcap (4.3)

The wind power production data proddati(t) and the total consumption dataconsdati(t) are inputs to the AS toolbox model where i denotes whether productionand consumption is LT or ST data. The data differ in the sense that LT datahas a higher probability of prognosis errors. The constant transcap is equal to 130MW and denotes the transmission capacity of the export HVDC cable. The totalconsumption is divided into consumption of detached houses participating in DR,Cementa and other activities:

consdati(t) = n · (PSH(t) + Pboil(t)) + PCementa(t) + Pothers(t) (4.4)

Where n denotes the number of detached house DR participants. InsertingEquation 4.4 in 4.3 and isolating the optimization variables will yield elements ofthe A and b matrix.

Load shifting balance constraint24∑t=1

PSH(t) + Pboil(t) ≤24∑t=1

PSH,reference(t) + Pboil,reference(t)

The total daily consumption for a detached house participating in DR must be equalor less to its consumption without DR. This condition ensures that the load duringthe day is shifted.

30

Page 40: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

4.1. PROBLEM FORMULATION

Equality constraints Aeqx = beq

There are three equality constraints to consider and they apply for t = 1, 2, . . . , N ,whereN is the number of hours simulated. The matrices Aeq and beq are constructedfrom the equations below:

Space heating constraint: indoor temperature vs consumption

Tin(t) = Tin(t− 1) + kSH(t) · (PSH(t) − PSH,reference(t)) (4.5)The indoor temperature at hour t is the sum of the temperature from the previ-

ous hour with the difference between DR and no DR consumption for space heating.The change is assumed to be linear with slope kSH(t) expressed in C◦/kW. Thecalculations on how the slope is determined for each time step is found in AppendixA1. The findings show no difference in slopes for the first four decimals betweenthe different time steps. Thus, one can say that the kSH = 0.06 C◦/kW.

Domestic hot water constraint: tank temperature vs consumption

Ttnk(t) = Ttnk(t− 1) + kDHW (t) · (Pboil(t) − Pboil,reference(t)) (4.6)The tank temperature at hour t is the sum of temperature from the previous hourwith the difference between using DR and no DR consumption for domestic hotwater. The change is assumed to be linear with slope kDHW (t) expressed in C◦/kW.The calculations on how the slope is determined for each time step is found inAppendix A2. The findings show no difference in slopes for the first four decimalsbetween the different time steps. Thus, one can say that the kDHW = 2.85 C◦/kW.

Domestic hot water constraint: tank content balance

Mtnk(t) = Mtnk(t− 1) + Pboil(t) − Qdrain,i(t) − QDHW,loss(t) (4.7)The hourly energy in the water tank is equal to the sum of the energy in the tankduring the previous hour with the consumption of the boiler subtracted by thedrained energy for water from the house and the water tank losses.

Upper/lower bounds lb ≤ x ≤ ub

The upper and lower bounds are defined as

lb = [PSH,min(t) Pboil,min(t) Mmin(t) Ttnk,min(t) Tin,min(t)]

ub = [PSH,max(t) Pboil,max(t) Mmax(t) Ttnk,max(t) Tin,max(t)]The upper and lower bounds are set according to model specifications for space

heating and domestic hot water activity (see previous chapter on flexibility toolsmodeling). Each element of the vectors lb and ub are constant for every time stept, e.g. the maximum allowed indoor temperature does not change from one hour tothe next.

31

Page 41: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 4. OPTIMIZATION METHOD

4.2 Long-Term Cluster

The LT cluster consist of a group of detached houses and the industry Cementathat offers to participate in DR on day-ahead (i.e. 24-48 hours ahead). Cementahas limited participation where it can only contribute with increased consumptionif there is an export problem prognosis during day shifts and weekdays. The opti-mization process is executed sequentially where the LT cluster consumption is firstoptimized upon receiving LT specific data on production and consumption. Anillustrative example is presented in Figure 4.1 where the optimization problem hasbeen solved for a day-ahead prognosis.

Figure 4.1. Example of how the export problem hours are minimized during aday when the LT cluster consumption has been optimized. The file optLT.m is thefunction implemented in MATLAB to runt the long-term optimization.

4.3 Short-Term Cluster

The ST cluster consist only of detached houses where there main objective is tominimize the prognosis errors from the LR DR cluster. A requirement for STparticipants to consume optimally is for them to know ahead of time how the LTcluster consumes for the same time resolution. Hence, the optimization for the STcluster occurs after the LT cluster optimization. The ST cluster operates on a hour-ahead basis which means that more accurate production prognosis data is availableat that time. Therefore the optimization algorithm become slightly different wherethe input data becomes a mix of LT and ST prognosis for each time step. Anillustrative example of the ST optimization is presented in Figure 4.2.

After every hour the BESS operation strategy is deployed to absorb the prognosiserrors and the optimization problem is reduced to account for the remaining hoursof the day. When the last hour of the day is reached there is no longer anythingleft to optimize.

32

Page 42: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

4.4. REQUIRED INPUT DATA

Figure 4.2. Minimizing LT prognosis errors by optimizing the consumption of theST cluster with a mix of LT and ST prognosis data as input. The figure illustrateswhat happens after the first hour. The shadowed export and consumption curvescorrespond to the prognosis and optimization made for the previous time step. Thefile optST.m is the function implemented in MATLAB to run the ST optimization.

4.4 Required Input DataTable 4.1 summarizes the hourly input data required to solve the optimizationproblem for both the LT and ST cluster. These data inputs were briefly introducedin the previous sections of this chapter. The data inputs are vectors on hourlyresolution where the length of the vectors are always a multiple of 24 (i.e. thenumber of hours in a day). Therefore consumption data for an unlimited number ofconsecutive days can be optimized. However, the more days that are simulated withexport prognosis problems the more difficult it becomes for the system to maintainbalance since the DR participants have very little recovery periods.

Table 4.1. Required hourly input data for the AS toolbox to get an optimizedconsumption output. *It is assumed that the system is already capable of balancingthe consumption prognosis errors (refer to section 1.4 for delimitation of study).Therefore, Consdat, ConsdatLT and ConsdatST are identical data sets.

Data Input [MW] Descriptionproddat Actual productionconsdat Actual consumptionproddatLT LT production prognosisconsdatLT LT consumption prognosis*proddatST ST production prognosisconsdatST ST consumption prognosis*

33

Page 43: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 4. OPTIMIZATION METHOD

4.5 Optimization Flowchart

No

No

Yes

No

Repeat

Yes

No

No

Step

Select next day

Repeat

Yes

Start

No Export prognosis

problem?

Data input

Yes

Include DR

participation

from Cementa

Include DR

participation

from detached

houses

Deploy

consumption

strategy if

possible

Optimize

consumption of

controllable

appliance

Increase LT

cluster size

Day-ahead

(24-48 hours)

Start at t=0

t<=24?

Export prognosis

problem after

optimization?

Yes

hour-ahead

(1 hour)

Export prognosis

problem?

Include DR

participation

from detached

houses

Optimize

consumption of

controllable

appliance

Yes

Export prognosis

problem after

optimization?

Increase ST

cluster size

Export problem

after hour t?

BESS operation

strategy

t=t+1

Stop

- LT, ST and real wind production data + offset (hourly)

- LT, ST and real consumption data + offset (hourly)

- Solar radiation data (hourly)

- Outdoor temperature data (hourly)

- Water drainage data (hourly)

Choose time period for data

Figure 4.3. Summary flowchart of LT and ST optimization process. To step in the"BESS operation strategy" box see Figure 3.1

.

34

Page 44: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Chapter 5

Simulation Set-Up

The set-up is intended to go through all requirements that are needed before startingthe simulation with the AS toolbox scheme. These requirements include for exampledefining simulation scenarios, setting offsets to provoke export problems, applyingproduction prognosis errors and setting start parameters.

5.1 Defining Simulation ScenariosThe power export challenges on Gotland during 2012 were presented in the back-ground section 2.1.4. The simulation scenarios chosen are three consecutive daysfor the seasons of 2012. These days are chosen at different times of experiencedpower export peaks. Figure 5.1 presents the export occasions on Gotland during2012 and the periods that have been chosen for simulation.

The simulation days were carefully selected such that frequent and evenly dis-tributed export peaks occurs during the specified period. The reason for defining asimulation scenario for each season is that the seasonal variation has an influenceon the consumption flexibility for detached houses participating in DR. Each seasonhas a unique consumption and production pattern as well as a typical outdoor tem-perature and solar radiation. Hence, there is an expected correlation between theseasons simulated and the space heating consumption. There is also an expectedcorrelation with the frequency of hourly export problems.

The simulation scenarios together with specifications are presented in Table 5.1.

Cementa will only participate in DR if there is a long-term export problemprognosis during day shift weekdays. The simulation scenarios include at least oneweekday where Cementa could participate in DR.

5.2 Applying Data OffsetHistorical observations show that an export problem has never been close of caus-ing an export problem at 170 MW installed wind power capacity. The hourly wind

35

Page 45: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 5. SIMULATION SET-UP

HVDC transmission capacity

Hourly power export 2012

Pow

er[M

Wh/h

]

Time [hours]

0 1000 2000 3000 4000 5000 6000 7000 80000

20

40

60

80

100

120

140

160

Figure 5.1. The exported power between Gotland and the mainland during 2012.The arrows indicate the periods of each season chosen for simulation.

Table 5.1. Simulation scenarios and specifications

Season Time Period (dd-mm-yy) DaysWinter 12-01-12 to 14-01-12 Thursday-SaturdaySpring 09-05-12 to 11-05-12 Wednesday-FridaySummer 06-08-12 to 08-08-12 Monday-WednesdayAutumn 06-10-12 to 08-10-12 Saturday-Monday

power production during 2012 is evenly scaled to represent 200 MW of installedwind power capacity. The transmission capacity is now close to being exceeded.The historical data is by no means an indication of the absence of export problemsfor an installed capacity of 200 MW. The yearly production and consumption canchange drastically from year to year mostly because of changing weather dynamics.Thus, in order to determine if the network can balance 5 additional MW one has toprovoke an export problem by setting an offset on the hourly production and con-sumption for the days simulated. Applying an offset means increasing the hourlyproduction and decreasing the hourly consumption such that the export marginincreases and provokes an export problem. The consumption and production char-acteristics are preserved by evenly scaling all values. The procedure used to set anoffset is summarized in the following bullet list:

36

Page 46: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

5.3. APPLYING PRODUCTION PROGNOSIS ERRORS

• A unique offset is set for each day simulated. One offset for the productionand one for the consumption.

• The production offset for the whole day is set by multiplying the production athour t by the ratio between the maximum installed capacity and the maximumhourly production of the day, i.e.

proddatoffset(t) = proddat(t) · 200proddatmax

• The consumption offset is set such that the maximum hourly export during theday is approximately 135 MW, meaning 5 MW above the allowed transmissioncapacity:

consdatoffset(t) = consdat(t) · consdatrefconsdatmin

exportoffset(t) = proddatoffset(t) − consdatoffset(t) ≈ 135MW

The value consdatref is tuned such that no more than 5 MW per hour needsto be balanced in the network.

Table 5.2 summarizes the offset values set for the different simulation scenar-ios. Positive offsets are set for production and negative offsets for consumption toprovoke the export problems. The average offset on consumption during summertimes is considerably lower than the offset during the winter. This is because theconsumption is generally lower during the summer.

5.3 Applying Production Prognosis ErrorsPrognosis errors are applied for the production of wind power for the 5 MW ad-ditional installed wind capacity. It is assumed that the prognosis errors for theremaining 195 MW of wind power as well as the consumption are already handledby the system. The normalized RMSE values for day- and hour-ahead wind powerproduction were presented in the background section. The prognosis errors are ran-domized through a normal distribution N(µ, σ2) where µ denotes the mean value,i.e. the production at hour t and σ the standard deviation, i.e. the normalizedRMSE at hour t. Thereby, the day- and hour-ahead production prognosis will vary,thus replicating the real prognosis errors occurring when performing wind powerproduction forecasts. It is important to note that the hourly production prognosiscan never exceed the installed wind power capacity of 200 MW. For example, if theproduction (with offset) at hour t is set to 200 MW then only a pessimistic produc-tion prognosis on LT and ST can be made (i.e. proddatLT(t), proddatST(t) ≤ 200).The basic idea is that a rational player does not make a prediction on a powerproduction value that exceeds the maximum installed wind power capacity.

37

Page 47: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 5. SIMULATION SET-UP

Table 5.2. Production and consumption offset for the simulated days for each season.The average offset is calculated for each simulation scenario.

Scenario Season Day Production Offset Consumption Offset1 Winter 12-01-12 +40% -46%

13-01-12 +40% -51%14-01-12 +40% -48%

Average +40 % -48%2 Spring 09-05-12 +48% -40%

10-05-12 +40% -35%11-05-12 +39% -40%

Average +42 % -38%3 Summer 06-08-12 +83% -32%

07-08-12 +42% -32%08-08-12 +38% -16%

Average +54 % -27%4 Autumn 06-10-12 + 84% -31%

07-10-12 +64% -34%08-10-12 +76% -27%

Average +48 % -31%

5.4 Setting Start ParametersTable 5.3 lists the start parameters used for the simulations:

Table 5.3. Initial conditions and start values for the simulation. *Cementa is notincluded initially, however nind = 1 for sensitivity analysis to determine the effect ofindustry participation.

Parameter Value Unit DescriptionTin(0) 20 C◦ Initial indoor temperatureTtnk(0) 80 C◦ Initial tank temperatureMtnk(0) 24.6 kWh Initial tank heatnLT,ref 1600 - Reference long-term cluster size for all seasonsnST,ref 300 - Reference short-term cluster size for all seasonsnstep 100 - Step size when changing cluster sizesnind 0* - Cementa not included initially

The start conditions for space heating and domestic hot water are set accordingto the model specifications presented in chapter 3. The cluster sizes have been setto a reference value which ensures that the optimization is solved for all simulationscenarios. The reference values were determined after having simulated all scenariosby iterating the cluster sizes until the optimization was solved. For one of thesimulated days in a season, nLT,ref and nST,ref correspond to the minimum clustersize required to solve the optimization problem. Consequently, this implies that

38

Page 48: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

5.4. SETTING START PARAMETERS

smaller cluster sizes could have been chosen for some of the other simulations andstill solve the optimization problem. The participation of Cementa is only includedfor sensitivity analysis to see how the LT cluster size decreases when Cementaincreases consumption to reduce an export problem prognosis.

39

Page 49: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in
Page 50: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Chapter 6

Simulation Results & Analysis

This chapter presents the results and analysis for scenario 1 to 4 correspondingto the different seasons of the year. For each scenario, the expected power exportoutcomes are presented for day- and hour-ahead resolutions. Then the analysisextends to the consumption characteristics of the DR clusters as well as the BESSand wind curtailment operation. The simulations have been executed according tothe set-up described in the previous chapter. Also included, is a sensitivity analysison the LT and ST cluster size. Finally, the scenarios are compared to analyse theimpact of seasonal variation in DR flexibility. The results of the first scenario willbe thoroughly analysed to help the reader understand the sometimes overwhelmingdynamics behind the optimization process. The remaining scenarios will containless detail specific explanations and focus solely on the findings. The reader is freeto read the result section for each scenario independently of each other.

6.1 Scenario 1: Winter days

6.1.1 Expected Power Export

See Figure 6.1

The actual and expected LT and ST power export curves for three consecutivewinter days are presented. The LT and ST prognosis error is causing a small hourlyvariation between the power export curves. There is a total of 14 LT export problemprognosis for the three days simulated. Most of the hourly export problems areconsecutive. Two critical export problem hours are identified: one at hour 10 andone at hour 50. A critical export problem hour is defined as an hour where one of theprognosis on either LT or ST is contradicting what is actually going to happen. Athour 10, the LT prognosis predicts a power export problem when the actual exportis in fact below the transmission capacity. At hour 50, the LT export prognosis failsto predict a power export problem which is bound to happen.

41

Page 51: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 6. SIMULATION RESULTS & ANALYSIS

Transmission capacity

ST export prognosis

LT export prognosis

Real exportPow

er[M

W]

Time [Hours]

0 12 24 36 48 60 7280

90

100

110

120

130

140

150

160

Figure 6.1. LT and ST export prognosis compared to the real export outcome.All hours when the export is above the transmission capacity are considered exportproblem hours.

6.1.2 Optimized Consumption

Total consumption after DR (see Figure 6.2)

The total consumption during those three winter days are presented as well as theoptimized consumption when LT and ST DR has participated. The optimized con-sumption curve increases during hours of export problem prognostics. The increaseis a mix of activity from the LT and ST DR clusters. At the critical export prob-lem hour 10, no consumption increase was in fact necessary since the power exportoutcome was below the transmission capacity. This shows how the prognosis errorscan sometimes lead to pessimistic LT export prognosis.

Total power to balance per cluster (see Figure 6.3)

The power to balance in the network for the LT and ST clusters respectively ispresented. The optimized consumption curve represents the change in total spaceheating and domestic hot water consumption from the reference. The LT clustermanages to successfully balance the excess in the network. The load shifting phe-nomena is well observed, where the consumption increases during hours of exportproblem prognosis and decreases during hours when there is no danger of exceedingthe transmission capacity. The ST cluster adjusts its consumption during persistingor new export problem hours that arise from the LT prognosis errors. Thus, the

42

Page 52: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

6.1. SCENARIO 1: WINTER DAYS

Optimized consumption

Consumption (no DR)Pow

er[M

W]

Time [hours]

0 12 24 36 48 60 7250

55

60

65

70

75

80

85

90

Figure 6.2. The optimized consumption after LT and ST DR.

power to balance for the ST cluster becomes either greater, less or equal to whatwas previously predicted. The positive peaks in the ST cluster Figure indicateshours where the optimized LT consumption was not sufficient to avoid the exportproblem (see hour 5, 8 and 30). Hour 50 is a critical export problem hour where theST prognosis has identified a potential export problem that was not seen in the LTprognosis. The LT cluster had in fact consumed more than it’s reference during thathour which led the ST cluster to consume less at that peak. This phenomena is tiedto the complex dynamics of the systems where all constraints need to be satisfiedat all times. This is one explanation on why the LT cluster consumed more at hour50 even though there were no LT export problem prognosis. The negative peaksin the ST cluster Figure indicates hours where the optimized LT consumption wastoo ambitious. The ST cluster can in those circumstances reduce it’s consumptionuntil the power exported is just below the transmission capacity.

Space heating and domestic hot water consumption per cluster (seeFigure 6.4)

A comparison between the space heating and domestic hot water consumption perhousehold is presented. The figures represent one household participating in LTDR and one in ST DR. It is difficult to draw any definite conclusions on whichhousehold activity that offers more flexibility for the days simulated. The reasonfor this is that both consumption curves are varying a lot compared to their reference

43

Page 53: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 6. SIMULATION RESULTS & ANALYSIS

Optimized consumption change (ST cluster)

Power to balance (ST/LT prognosis)

ST Cluster

Pow

er[M

W]

Time [hours]

Optimized consumption change (LT cluster)

Power to balance (LT prognosis)

LT Cluster

Pow

er[M

W]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

!1

0

1

2

3

!10

0

10

20

Figure 6.3. Hourly power to balance in the network for the LT and ST clustersrespectively.

consumption. For space heating consumption the indoor temperature boundary isnarrow but the inertia of the system is slow while the opposite situation applies todomestic hot water consumption.

Indoor and tank temperature changes per household (see Figure 6.5)

The indoor and tank temperature change in time are presented for one householdparticipating in LT DR and one in ST DR. These figures show how the constraintsvary in time when the consumption has been optimized. There is a clear relationbetween the optimized consumption and the rise in temperatures at export problemhours. During the winter the solar radiation is very low as well as the outdoor tem-perature. This is why the indoor temperature curves have a tendency of decreasingin time. The tank has a high statistical water drainage during evening hours, i.e.hours 18-23, hours 42-47 and hours 66-71 which explains the recurrent temperaturedips during those hours. The water drainage is even higher during weekends whichis seen for hours 48-72 which is a Saturday. Because of the high water drainageduring these hours, the consumption is increased shortly before, to avoid fallingoutside the allowed boundaries.

44

Page 54: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

6.1. SCENARIO 1: WINTER DAYS

Domestic hot water consumption change (ST cluster)

Space heating consumption change (ST cluster)

Domestic hot water consumption change (LT cluster)

Space heating consumption change (LT cluster)

DR consumption per household

Pow

er[k

W]

Time [hours]

DR consumption per household

Pow

er[k

W]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

!4

!2

0

2

4

6

!4

!2

0

2

4

6

Figure 6.4. Comparison between space heating and domestic hot water consumptionper household and cluster.

6.1.3 BESS Operation & Wind Curtailment

See Figure 6.6

The hourly BESS charge/discharge level and required wind curtailment is presented.The BESS level never reaches maximum capacity and therefore no wind curtailmentis needed for the days simulated. There is no circumstance where the ST hourlyprognosis error is significant enough to charge the BESS at maximum capacityduring one hour. Although rare, this type of circumstances might occur which inthat case would require wind curtailment. During hour 8, the ST prognosis errorhappens to be 0 %. The BESS does not charge nor discharge it’s energy when thistype of situations occur. One should note that the BESS does not discharge if itcontributes to a new export problem. Finally, there are three battery cycles for thissimulation. The BESS charges over 100 kW during one hour and discharges thatamount during the next hour. This type of activity is damaging for the battery onthe long-run.

6.1.4 Sensitivity Analysis

The sensitivity analysis are intended to show how the consumption per householdvaries for a change in LT and ST cluster size. The optimization is solved on a dailybasis meaning that each day simulated has it’s own minimum required LT and ST

45

Page 55: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 6. SIMULATION RESULTS & ANALYSIS

Comfort interval

Reference temperature

Tank temperature (ST cluster)

Tank temperature (LT cluster)Tank temperature change/household

Tem

per

ature

[C]

Time [hours]

Comfort interval

Reference temperature

Indoor temperature (ST cluster)

Indoor temperature (LT cluster)Indoor temperature change/householdTem

per

ature

[C]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

60

80

100

120

18

20

22

24

Figure 6.5. Indoor and tank temperature variation for one LT and ST householdrespectively during DR.

Max BESS capacity

Wind curtailment

BESS level

Pow

er[k

W]

Time [hours]

0 12 24 36 48 60 720

50

100

150

200

250

300

Figure 6.6. The hourly BESS charge/discharge levels as well as required windcurtailment when using DR.

46

Page 56: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

6.1. SCENARIO 1: WINTER DAYS

cluster size. The cluster sizes are increased from the minimum values to see howthe consumption per household changes. The cluster step size is 100 households atthe time.

The sensitivity analysis also include the impact of industry DR on LT clustersize, i.e. the participation of Cementa. The days simulated are between a Thursdayand a Saturday. There are hours of export problems during the first 12 hours onThursday and Friday, meaning that the LT cluster will include the participation ofCementa during those hours.

Influence of cluster size on total load shift per household (see Figure6.7)

The total load shift required per household naturally decreases with the size ofthe cluster. The difference between the curves is related to the number of exportproblems (consecutive or not) and the amount of power that has to be balancedthroughout that day. For the LT cluster, the export problems are larger and morerecurrent on Saturday. This is also verified with Figure 6.1. For the ST cluster,the largest total load shift to balance occurs on Thursday. This means that the LTprognosis errors resulted in more hours where the LT consumption was not sufficientto solve the actual export problems. This was a lot more apparent on Thursdaythan on Friday for example.

Saturday

Friday

Thursday

Total load shift/day and household

Pow

er[k

W]

ST cluster size

Saturday

Friday

Thursday

Total load shift/day and household

Pow

er[k

W]

LT cluster size

0 500 1000 1500 2000

1000 1500 2000 2500 3000

0

2

4

6

8

0

5

10

15

Figure 6.7. Total daily power to balance per household with changes in cluster size.

47

Page 57: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 6. SIMULATION RESULTS & ANALYSIS

Influence of Cementa on cluster size (See Figure 6.8)

Two simulations are presented. One without Cementa nor any cluster size restric-tions, i.e. the minimum LT and ST cluster size per day required to solve the exportproblems are used. The second simulation includes the participation of Cementa.The export problem prognosis and the DR strategy of Cementa only allows par-ticipation during Thursday and Friday. On Thursday the cluster sizes decreaseswith a total of 700 households (600 LT + 100 ST). On Friday only the LT clustersize decreases by 700 households. Having Cementa participate in DR modifies theoptimized consumption pattern for each individual household and thereby also thechange in indoor and tank temperature. A consequence of this is that on Saturdaythe LT cluster size is instead increased by 100 households to ensure that constraintsare fulfilled. In reality the number of households per cluster will be fixed and there-fore the worst case scenario will always be used as reference. In this case the worstcase scenario for those three days increased the LT cluster by 100 households.

Figure 6.8. The impact of DR from Cementa on the reduction of the minimum LTand ST cluster sizes. ’C’ denotes that Cementa is participating in DR whereas ’NC’denotes that it is not.

6.2 Scenario 2: Spring days

6.2.1 Expected Power Export

Figure B.1 analysis

• 4 LT export problem prognosis on Wednesday

• 2 LT export problem prognosis on Thursday

• 4 LT export problem prognosis on Friday

• Hour 19 is a critical export problem hour

48

Page 58: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

6.2. SCENARIO 2: SPRING DAYS

6.2.2 Optimized Consumption

Figure B.2 analysis

• The load is successfully shifted to hours of export problem prognosis hours

Figure B.3 analysis

• The LT cluster successfully balances the power in the network

• The ST cluster successfully minimizes the LT export prognosis errors.

• There are hours when the consumption increases even though there is no ex-port problem prognosis, e.g. hours 36-43. In general this happens to maintainthe constraints in either indoor or tank temperatures. In this specific casethere is a high water drainage between hours 43-46 and an export problem athour 48 for which some load needs to be shifted. Therefore the consumptionbetween hour 36-43 needs to increase such that the tank temperature doesnot fall out from the lower boundary.

Figure B.4 analysis

• The variation in space heating consumption seems to be higher than for do-mestic hot water for both the LT and ST cluster. Thus, it seems that spaceheating offers more flexibility because of the export problems occurring athours of high water drainage.

Figure B.5 analysis

• The indoor temperature varies slightly around the reference temperature. Forevery day around noon one can see a natural increase in indoor temperaturedue to solar radiation. There is still a large margin left before encounteringconstraint errors which means that the limiting factor for flexibility is theinertia of the system as well as the upper bound on the maximum hourlyspace heating consumption.

• The tank temperature dips more during hours of high water drainage for smallconsumption changes, i.e. hours 18-23, hours 42-47 and hours 66-71. The tanktemperature for the LT cluster drops close to the lower boundary at hour 53-66 due to the decrease in consumption few hours before the expected exportproblem.

6.2.3 BESS Operation & Wind Curtailment

Figure B.6 analysis

• The prognosis errors on ST are significant enough to require wind curtailment

49

Page 59: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 6. SIMULATION RESULTS & ANALYSIS

• The wind curtailment needed is 37 kW at hour 15, 193 kW at hour 17, 173kW at hour 18 and 108 kW at hour 66.

• There are two BESS cycles for the days simulated.

6.2.4 Sensitivity AnalysisFigure B.7 analysis

• Friday is the most challenging day to balance power in the network. To solvethe export problems the LT cluster needs to include at least 1600 households.In terms of energy that needs to be balanced on a day, Wednesday is themost significant although a smaller cluster size is required to solve the exportproblems. When the export problems occur on a day and whether they areconsecutive or not is an important aspect determining the cluster size needed.

• Increasing the LT cluster size will reduce the need for load shifting per house-hold. Including 2500 households would for example imply that only around 5kWh would be load shifted per household.

• Including at least 1000 ST households would make the load shift needed perhousehold barely noticeable.

• An increase in ST cluster size also implies that variations in indoor and tanktemperatures are minimized.

• From the ST cluster plot one can see that Wednesday is the day with themost significant power to balance per household. This is directly tied to theLT prognosis errors which were more important for that particular day.

6.3 Scenario 3: Summer days

6.3.1 Expected Power ExportFigure B.8 analysis

• 1 LT export problem prognosis on Monday

• 1 LT export problem prognosis on Tuesday

• 3 LT export problem prognosis on Wednesday

• No critical export problem hours

6.3.2 Optimized ConsumptionFigure B.9 analysis

• The load is successfully shifted to hours of export problem prognosis hours

50

Page 60: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

6.3. SCENARIO 3: SUMMER DAYS

Figure B.10 analysis

• The LT cluster successfully balances the power in the network

• The ST cluster successfully minimizes the LT export prognosis errors

Figure B.11 analysis

• The variation in space heating and domestic hot water consumption are verylow in this simulation scenario because of the few export problem hours.

Figure B.12 analysis

• The indoor temperature remains stable for most hours of the day and increasesslightly during the last day. There is still a large margin before encounteringconstraint errors.

• The tank temperature dips faster at hours of high water drainage, i.e. hours18-23, hours 42-47 and hours 66-71. The tank temperature for the LT clusterdrops close to the lower boundary between hours 46-50 as a consequence ofshifting the load to hours of export problem prognosis.

6.3.3 BESS Operation & Wind CurtailmentFigure B.13 analysis

• The prognosis errors on ST are not significant enough to require wind curtail-ment. The BESS almost charges up to full capacity before discharging shortlyafter.

• There is one BESS cycle for the days simulated.

6.3.4 Sensitivity AnalysisFigure B.14 analysis

• Increasing the cluster size will reduce the need for load shifting per household.Including 2500 households would for example imply that on average 2 kWhper household would be needed for load shifting throughout the day.

• The LT prognosis errors are more significant on Wednesday as seen from theST cluster plot. If the ST cluster included at least 1500 households then theconsumption per household would be barely noticeable.

• There are no prognosis errors on LT export problems for Monday, thereforethe households can consume as intended prior to DR.

• An increase in cluster size implies that the variations in indoor and tanktemperatures are minimized.

51

Page 61: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 6. SIMULATION RESULTS & ANALYSIS

Figure B.15 analysis

• DR from Cementa is only deployed on Wednesday. The LT cluster size isreduced by 400 households when Cementa is participating in DR.

6.4 Scenario 4: Autumn days

6.4.1 Expected Power Export

Figure B.16 analysis

• 1 LT export problem prognosis on Saturday

• 7 LT export problem prognosis on Sunday

• 3 LT export problem prognosis on Monday

• Hour 28, 52 and 53 are critical export problem hours

6.4.2 Optimized Consumption

Figure B.17 analysis

• The load is successfully shifted to hours of export problem prognosis hours

Figure B.18 analysis

• The LT cluster successfully balances the power in the network

• Since there are many hours of consecutive export problems on Sunday thenthere is a dip in consumption during hours of the day where the load has beenshifted, e.g. hours 43-44.

• The ST cluster successfully minimizes the LT export prognosis errors

Figure B.19 analysis

• There is a high variation in space heating and domestic hot water consumptionfor this simulation scenario. The reason for this is the frequent and consecutiveexport problem prognosis.

Figure B.20 analysis

• The indoor temperature varies for most of the time below the reference tem-perature but is nowhere near falling outside the boundaries. The temperatureincreases above the reference for the hours of numerous consecutive exportproblems.

52

Page 62: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

6.5. SCENARIO COMPARISON

• The tank temperature is close to fall outside the lower boundary during thehours where the load has been shifted. Also during the hours on Sunday wherethere is 7 consecutive export problem hours the tank temperature almostexceeds the upper boundary.

6.4.3 BESS Operation & Wind CurtailmentFigure B.21 analysis

• The prognosis errors on ST are significant enough to require wind curtailment

• The wind curtailment needed is 29 kW at hour 30, 292 kW at hour 33 and423 kW at hour 35

• There are no critical BESS cycles for this simulation.

6.4.4 Sensitivity AnalysisFigure B.22 analysis

• Saturday is the most challenging day to balance power in the network. Thisis quite obvious judging that there are 7 consecutive export problems duringthat day. To solve the export problems the LT cluster needs to include atleast 1200 households.

• Increasing the LT cluster size will reduce the need for load shifting per house-hold. Including over 2500 households would for example imply that a con-sumption close to 10 kWh would be load shifted per household.

• Including at least 500 ST households would make the load shift needed perhousehold participating in the ST cluster barely noticeable. However this isnot true for Sunday which seems to still require participation for larger clustersizes because of the numerous export problems during that day.

• An increase in ST cluster size also implies that variations in indoor and tanktemperatures are minimized.

Figure B.23 analysis

• Cementa is only deployed on Monday where it’s DR participation contributeto a decrease of the LT cluster size by 400 households.

6.5 Scenario ComparisonThe scenarios are comparable from a couple of perspectives. The first perspective ison space heating consumption flexibility. The simulated days all have outdoor tem-perature and solar radiation that are unique for their respective seasons. Therefore,

53

Page 63: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 6. SIMULATION RESULTS & ANALYSIS

this implies that the space heating consumption flexibility will be different. The do-mestic hot water flexibility can not be compared in the same way as space heating.The reason for this is that the water drainage input data does not consider seasonalvariation. Ideally, unique water drainage data should be used for the specific daysimulated.

The second perspective is on the seasonal power production and consumptionpatterns. The power consumption characteristics varies between seasons and sodoes the power production. This results in a difference in number of hourly exportproblems encountered for each scenario after applying an offset. The comparisonis by no means conclusive since wind power production patterns vary from year toyear but it gives a fair indication on seasonal flexibility during the periods simulatedin 2012.

6.5.1 Space Heating FlexibilityThe indoor temperature variation for a LT household is analysed. Figure 6.9presents the indoor temperature change for a LT DR household participant forthe different scenarios/seasons simulated.

Autumn

Summer

Spring

Winter

Indoor temperature change for a LT household participant

Tem

per

ature

[deg

ees

C]

Time [hours]

0 12 24 36 48 60 7216

17

18

19

20

21

22

23

24

25

Figure 6.9. Indoor temperature variation for a LT household DR participant forscenario 1 to 4.

For all scenarios, the results show that in the worst case, there is a margin ofat least one degree before falling outside the boundaries. The margin between thepeaks and the boundaries do not entirely translate into consumption flexibility. One

54

Page 64: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

6.5. SCENARIO COMPARISON

should also take note of the maximum hourly consumption limit for space heatingwhich prevents rapid temperature increases. The results show that the flexibility inspace heating per household is largest for scenario 3 (i.e. Summer days), followedby scenario 2, 4 and 1 (i.e. Spring, Autumn and Winter days). The result onflexibility is directly related to the number of export problems and amount of powerto balance for each scenario. Hence, the flexibility is higher during the summerbecause of the production and consumption patterns of that season resulting in fewexport problem occasions.

6.5.2 Export Problem OccasionsThe total power to balance per day and scenario is presented in Figure 6.10. Thetotal number of hourly export problems per day and scenario simulated is presentedfor comparison in Figure 6.11.

Figure 6.10. Total power to balance per day and scenario

It is clear from the Figures that scenario 3 (i.e Summer) has the lowest totalnumber of hourly export problems and power to balance for the period simulated. Itis followed by scenario 2, 4 and 1 (i.e. Spring, Autumn and Winter). These resultsare in accordance with the space heating flexibility presented earlier. This showsthe strong influence between the seasonal production and consumption pattern withavailable flexibility. These results are also determinant of the number of requiredLT household participants where less export problems and power to balance willresult in less household participants needed. These results were expected knowingthe fact that Autumn and Winter seasons usually have stronger wind blows.

55

Page 65: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 6. SIMULATION RESULTS & ANALYSIS

Figure 6.11. Total number of hourly export problems per day and scenario. Notethat most of the problems that occur are during consecutive hours.

56

Page 66: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Chapter 7

Discussion

7.1 Model Limitations

The model used for this study present certain limitations that the reader should beaware of before drawing any hasty conclusions:

No space cooling was considered in the consumption model for detached houses.The assumption is that detached house occupants take measures during hours ofhigh solar radiation to maintain a reference indoor temperature of 20 C◦. Thesemeasures include for example opening windows or shadowing the interior of thehouse. In reality, it would be very unlikely for the occupants to let their consumptionincrease during the hours of the day having high solar radiation. During the summer,this period includes more or less the whole afternoon. Therefore, it would be veryunlikely for households to offer consumption flexibility during those hours. Thiswould act as a constraint for the optimization process and make load shifting moredifficult. The model used for simulation has however neglected this aspect, meaningthat the space heating consumption results for the summer scenario might not reflectthe entire reality of things.

The water drainage data used in the model is a statistically averaged drainagedata in hourly resolution. There is a data sample for every hour of the week. Theonly distinction that is made is between weekday and weekend drainage. Thereforethe data is not unique to a specific time of the year and no comparison can be madeon the consumption flexibility between seasons. The data used in the model was theonly data available. Ideally, one should however consider more time specific data.

The LT and ST production prognosis errors are based on the persistence methodwhere the guess is that the production for the up coming hour is equal to the one ofthe current hour and so forth. In reality, there are powerful wind power predictiontools such as WPPT that uses weather input data to make much more precisepredictions. An example is an hour ahead production prognosis error (normalisedRMSE) for the offshore wind park Lillgrund where the WPPT reduced the prognosiserror by over 10% [14].

57

Page 67: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 7. DISCUSSION

7.2 Scenario Limitations

The simulation scenarios were chosen in such a way to apply low offsets and toreflect the characteristics of each season. However, it is hard to identify if the pro-duction and consumption for these scenario periods reflect typical seasonal patterns.The reason for this is that the study was limited to only a year of data which isnot sufficient to identify seasonal production and consumption trends. Therefore,conservative conclusions should be drawn on the consumption flexibility per season.

The reference LT and ST cluster sizes were chosen as the worst minimum clus-ter size required between the scenarios. This ensured that all scenarios would besolvable for the same cluster size to allow comparison. There is however no guaran-tee that the worst minimum cluster size required would not increase if the chosensimulation days were slightly different. The number of consecutive export prob-lems and the power needed to balance dictates the cluster size needed. Thereforea different configuration in choosing the simulation days might generate differentresults on required cluster sizes. It is important to note that the consecutive daysof each scenario were carefully selected to reflect the worst days during that season.Therefore, the chances of the reference cluster size being larger than what was usedis small but not unfeasible.

7.3 Validity & Reliability of Results

The mathematical model used for the study has been validated for the consumptionof 40 Swedish residents living in detached houses [24]. The optimization problemformulation gives a very clear indication of the expected results from the simulation,e.g. the optimized consumption needs to increase during export problem hours.Thus, intuitively it becomes very easy to verify the results since there is already avery clear image of what is expected. By comparing the result figures one can alsounderstand the logic behind the optimization. This serves as an additional checkregarding the reliability of the results.

Concerning the optimized consumption results from the simulations. It mustbe noted that the local distribution grid on Gotland has not been included as aconstraint in the simulations. Therefore the results presented are only theoreticaluntil a network simulator such as PSS/E is used to simulate the power flows withinthe network. PSS/E takes into account the transmission capacities of the local gridas well as the geographical location of the loads and the wind farms.

Truly, the results need to be interpreted for what they are i.e. simulation resultsfrom a detached household and industry consumption model. The mathematicalmodel has used assumptions to reflect realistic conditions. Despite this, a modelwill always differ from the reality. To ensure that no false guidance is presentedfor the SGG project, the assumptions as well as the simulation set-up has alwaysstrived to reflect worst case scenarios. One should stress that the results presentedare pessimistic in regards to several technical aspects.

58

Page 68: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

7.4. THESIS BENEFITS FOR THE SGG PROJECT

Firstly, all results are provided in hourly resolution which means that the powerflows within the hour are not considered. Each elements of the export data used isan averaged value of the power flows within that hour. This means that there arepower flow peaks within the hour that are larger and require less offset to provoke anexport problem. Using a smaller offset percentage means that less export problemhours will occur during the day and therefore a smaller LT cluster size is required.

Finally, the ST cluster has the ability to operate on a 5 minute resolution inreality. In the simulations only hour-ahead operation was considered. Operatingat lower resolutions implies that there are recovery periods where the householdconsumption would not need to be optimized. These recovery periods provide moreconsumption flexibility for households when facing up coming export problems. Byincluding these aspects in the analysis, one would greatly reduce the LT and STcluster size and increase the consumption flexibility for detached houses.

One should also note that no economical barriers have been included and in thatregard the results from the model are optimistic.

7.4 Thesis Benefits for the SGG projectThe Thesis work has led to a number of findings and results that are of great valuefor the SGG project. Among these, the most important categories are listed andbriefly discussed below:

1. Operation StrategiesThe Thesis has presented numerous strategies on how to optimize the con-sumption during export problem events. This includes the optimization strate-gies for the LT and ST cluster as well as the operation strategy for the BESSand wind curtailment scheme. A DR strategy for the industry Cementa wasalso proposed which presented a significant influence on the required LT clus-ter size. These strategies could serve as a starting point for the real DRimplementation on Gotland.

2. Simulation ModelA simulation model was developed to predict the space heating and domestichot water consumption for detached houses on Gotland. This model has a highdegree of complexity, using input data such as outdoor temperature, waterdrainage and solar radiation data. The relation between the consumption andthe indoor and tank temperature are established, making the model capable ofsimulating consumption flexibility for detached houses. The simulation modelcan be used for the SGG project as a comparative tool once the model hasbeen validated with real consumption data for Gotland.

3. Simulation ResultsThe simulation results can be used as a worst case reference for the SGGproject. The interesting results are on the reference cluster sizes (1900 house-holds: 1600 LT + 300 ST) as well as on the reduction of up to 700 households

59

Page 69: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 7. DISCUSSION

with Cementa as a DR participant. This gives an indication on how much DRhouseholds are needed for the SGG project to balance the additional 5 MWof wind power production in the existing network. The results also providean idea of how the indoor and tank temperature varies when the consumptionflexibility is used. Furthermore, the seasonal variation presents the patternson potential power export problems and the available consumption flexibilityduring that time period. The SGG project can use a lot of these results forcomparison during the real implementation phase.

60

Page 70: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Chapter 8

Conclusion

8.1 Conclusive Summary

The increasing interest of integrating additional intermittent energy sources such aswind power to the existing electricity grid on Gotland has created a risk of powerexport challenges to the Swedish mainland. The transmission capacity of the HVDCexport cable and the low local consumption does not ensure safe operation at 200MW installed wind power capacity. These issues are one of many left to solve forthe SGG project. The Thesis analyzed an AS toolbox scheme solution, based onmulti-agent systems to counter these issues. The proposed solution would preventexport problems by operating flexibility tools such as DR on LT and ST, a BESSand a wind curtailment scheme.

After having introduced the Thesis, the second chapter focused on literature andbackground studies about the Gotland situation. This included the electric powersystem and the power production and consumption characteristics of the islandduring 2012 for an approximate installed capacity of 170 MW. The power exportedduring the year peaked during all seasons but not remotely close to the maximumtransmission capacity. These results persisted even after scaling the productionto 200 MW of installed wind power capacity. The concept of DR was introducedand the most important DR potential on Gotland was identified as space heat-ing and domestic hot water consumption for detached houses. The DR potentialfrom industries were also explored by studying the DR consumption strategies ofCementa, Arla and Nordkalk. Ultimately, Cementa was chosen because of the sim-plicity in modeling it’s DR activity and the large share of consumption it couldcontribute with compared to other industries. Furthermore, the AS toolbox schemewas presented putting emphasis on the flexibility tools it uses to balance the addi-tional power production. The agent model the AS toolbox uses was implementedin MATLAB using the Optimization Toolbox. Alternative agent model candidateswere researched such as the PM Simulation Tool. Unfortunately, the limitationsencountered with PM and the high degree of freedom offered in the MATLAB en-

61

Page 71: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 8. CONCLUSION

vironment justified the choice of agent model.The third and forth chapter focused on modeling the AS Toolbox mathemati-

cally and formulating the optimization problem. The flexibility tools were modeledbased on the literature study carried out where high emphasis was put on the DRconsumption activity for detached houses. The concept of LT and ST clusters weredefined and strategies were proposed for these as well as for the BESS operation.

The final chapters included information about the simulation set-up and the re-sults of four simulated scenarios corresponding to the different seasons of the year.The results showed that 1900 households (1600 LT + 300 LT) could balance 5 addi-tional MW of installed wind power capacity in the existing distribution grid for allscenarios simulated. Some scenarios could even reduce their cluster size to a total of1000 households (i.e. scenario 3: Summer 900 LT + 100 ST). The changes in indoortemperature for a DR household participant did never come close of falling outsidethe allowed comfort interval during DR. The tank temperature on the other handhad stronger variations indicating that domestic hot water consumption uses almostall of it’s consumption flexibility. The BESS cycles for the three days simulated werenever more than three and the wind curtailment scheme was only deployed for sce-nario 2 and 4 (i.e. Spring and Autumn) reaching a maximum of 423 kWh of windpower curtailed. The simulation scenarios showed that the export problem patternsfor each season are the main influence on the available consumption flexibility fromDR participants. It is also a determining factor for the LT and ST cluster sizes. Thesimulation scenarios concluded that Summer days offered the highest space heatingflexibility, followed by Spring, Autumn and Winter days. No such conclusion couldbe drawn for the domestic hot water consumption since the input data used did notreflect on the specific time period. Finally, DR from Cementa showed significantinfluence on the total cluster size. For one of the simulation scenarios the totalcluster size was reduced by 700 households.

Installing an AS toolbox on Gotland is technically feasible depending on how manyhouseholds that are willing to participate in DR. According to results, at least 1900households is needed to balance the additional 5 MW of wind power productionfor the simulated scenarios. This represents roughly 5% of all detached houses onGotland. In reality this number is probably a lot less since the simulation have beenworst case scenarios. The main challenges on Gotland are not so much of technicalnature but rather economical. The households on Gotland will for example onlyengage in DR if it helps reduce their electricity bill. This means that there is aneconomical constraint associated with load shifting, more specifically, the hourlyelectricity market price for SE3. These are not the only economical concerns, thereare also installation costs associated with the control systems in detached houses,battery costs and many more. The control system installations need to be relativelyeasy to execute and repeatable in order to integrate this technology in the thousandsof detached houses participating in DR.

62

Page 72: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

8.2. FUTURE WORK

8.2 Future Work

This Master Thesis has succeeded in investigating technical solutions to integraterenewable energy sources in an existing power grid. The results will be used asreference for the SGG project but it’s value stretches far beyond the simple case ofGotland. The developed methods and optimization strategies for DR can be usedin any system facing similar challenges in it’s network. There are however technicalimprovements and economical investigations that needs to be carried out to finalizethe feasibility of the AS toolbox concept.

On the technical side, the agent-model used for the AS toolbox needs to communi-cate with a network simulator such as PSS/E to verify that the local grid can handlethe load shifts in consumption. The mathematical model needs to be improved ona few aspects. One of them being the model input data. In reality there is a degreeof uncertainty when predicting outdoor temperature, solar radiation, consumptionor production. Therefore, all of these inputs to the model should be associated withprognosis errors that are absorbed by the BESS. As of now the BESS only handlesproduction prognosis errors on ST. This is because the prognosis errors on outdoortemperature and solar radiation are negligible while the consumption prognosis er-rors are assumed to be handled by the current system regardless of new wind powercapacity installations. Furthermore, the optimization methods for the ST clustershould be redesigned such that the households can participate in DR every 5 min-utes instead of on a hourly basis. This will allow the BESS to discharge withinthe hour and households to increase their consumption flexibility during recoveryperiods.

On the economical side, there is a lot of investigations that needs to be done inorder to implement an AS Toolbox scheme. Typical questions that need answeringare: how to incentivize detached houses to participate in DR? What type of elec-tricity contracts would be proposed? What impact would this have on the currentelectricity market? Does the implementation and installation costs for the smartgrid systems exceed the cost of re-building the grid or installing new HVDC cables?How much does the price of the BESS play in given its life expectancy and itsoperation strategy? These questions are only few of those concerning the econom-ical aspect of the proposed AS toolbox. Furthermore, a more thorough statisticalstudy should be performed on the probability of occurrence of an export problemfor 200 MW of installed capacity. This study should include as much historical dataas possible to establish accurate power production and consumption trends for awhole year. This would ensure accurate predictions on the risk of having an exportproblem. Knowing the chances of this happening, contracts can be formed wherefor example the risk is sold to DR participants in exchange for economical benefits.

Conclusively, the Master Thesis has served as a proof of concept on the idea ofusing a technical AS toolbox scheme on Gotland to balance additional power pro-duction in the network. The AS toolbox made use of flexibility tools such as DR

63

Page 73: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

CHAPTER 8. CONCLUSION

from detached houses which in reality is closely tied to economical constraints. Thefuture work in the field includes perfections on the technical aspect of the AS tool-box but also on economical investigation regarding the feasibility of implementingsuch a system in today’s electricity market structure.

64

Page 74: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Bibliography

[1] 20-20-20 Goals, European Commission Website, url:http://ec.europa.eu/clima/policies/package/, (Last update: 31/10/2013)

[2] Development of Physical-Based Demand Response-Enabled Residential LoadModels, Shengnan Shao, Manisa Pipattanasomporn, and Saifur Rahman, IEEEtransactions on power systems, vol. 28, no. 2, (May 2013)

[3] Mitigation of Wind Power Fluctuations by Intelligent Response of Demandand Distributed Generation, Pamela MacDougall, Cor Warmer, and Koen Kok,(2011)

[4] Smart Grid Gotland Project, GEAB, Vattenfall, ABB, SVK, Schneider Electric,Energimyndigheterna, KTH, url: http://www.smartgridgotland.se/, (2013)

[5] Average Consumption of a Swedish Household, Energimyndigheterna,url: http://energimyndigheten.se/sv/Hushall/Din-uppvarmning/, (Last update:2012-04-25)

[6] Pioneering Smart Grids on Gotland, Sweden, ABB, image fromurl:http://www.abb.com/cawp/seitp202/077f92def9668579c1257a400037425b.aspx,(Last update: 2012-07-23)

[7] Gotland HVDC light transmission - World’s first commercial small scale DCtransmission, Urban Axelsson, Anders Holm, Christer Liljegren, Kjell Erikssonand Lars Weimers, Vattenfall, GEAB, ABB, (May 1999)

[8] The Gotland HVDC Link, ABB, url:http://www.abb.com/industries/ap/db0003db004333/8e63373c2cdc1cdac125774a0032c5ed.aspx, (2013)

[9] Samrådsunderlag: planerad stamnätsförbindelse mellan Gotland och fastlandet,SVK, PDF: http://www.svk.se/PageFiles/47517/SvK-Samradsunderlag-ej-bilagor.pdf, (February 2012)

[10] Business Models for an Aggregator, Master Thesis by Quentin Lambert, pp.20, XR-EE-ICS 2012003, (2012)

65

Page 75: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

BIBLIOGRAPHY

[11] Hourly Production and Consumption Data 2012, provided by GEAB,accessible on projectplace with log in details under SGG project, url:http://projectplace.com, (2013)

[12] Gotland blir labb för nya nät, Nyteknik, url:http://www.nyteknik.se/nyheter/energi_miljo/energi/article3753691.ece,(2013-11-02)

[13] The power of the everlasting breeze, Official website of regional Gotland, url:http://www.gotland.se/54547, (September 2013)

[14] Forecasting and Optimization for the Energy Sector, ENFOR WPPT model de-scription, url: http://www.enfor.eu/wind_power_prediction_tool_wppt.php

[15] Gotland i siffror 2010-2012, Brochure on Gotland Statistics, url:http://www.gotland.se/1353, (2012)

[16] Regionfakta Gotland, Website with Regional Statistics in Sweden, url:http://www.regionfakta.com/Gotlands-lan/, (2013)

[17] Business Models for an Aggregator, Master Thesis by Quentin Lambert, XR-EE-ICS 2012003, (2012)

[18] ABB Corporate Research Notes, Carl-Fredrik Lindberg, Mötesanteckningarefter besök på Nordkalk, Arla och Cementa, (2012-11-09)

[19] Vindkraft på Gotland, Official website of regional Gotland, url:http://gotland.se/1729, (January 2013)

[20] Varmvattenberedare, Energimyndigheterna, url:http://www.energimyndigheten.se/sv/Hushall/Varmvatten-och-ventilation/Vatten-och-varmvattenberedare/Varmvattenberedare/, (Lastupdate: 2012-04-24)

[21] PowerMatcher Smart Grid Technology, developed by TNO, url:http://www.powermatcher.net/get-technical/

[22] Mitigation of Wind Power Fluctuations by Intelligent Response of Demandand Distributed Generation, Pamela MacDougall, Cor Warmer, and Koen Kok,(2011)

[23] Mathworks Documentation, MATLAB Optimization Toolbox,url:http://www.mathworks.se/products/optimization/, (2013)

[24] Modeling household consumer electricity load profiles with a combined physicaland behavioral approach, Claes Sandels, Joakim Widen, and Lars Nordström,ICS Department KTH, (2013)

66

Page 76: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

[25] Marginaler i fjärrvärme system, Patrik Selinder and Heimo Zinko, ZW Ener-giteknik AB, Forskning och utveckling, ISSN 1402-5191, (2003)

[26] Dimensionerande vinterutetemperatur, DVUT, Boverket Dimensionerande vin-terutetemperatur, url: http://www.nollhus.se/Pages/dvut.aspx, (2009)

67

Page 77: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in
Page 78: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Appendix A

Model Calculations

A.1 Space Heating SlopesThe indoor temperature of a detached house was expressed in Equation 4.5 as:

Tin(t) = Tin(t− 1) + kSH · (PSH(t) − PSH,prognostic(t))

Where kSH (expressed in C◦/MW) is the slope between the indoor tempera-ture and the space heating consumption. This relation is assumed to be a linearincreasing function. The indoor temperature equation was expressed in 3.2 as:

T(t + 1) = T(t) + 1τ ·(λtrans+λvent) · (Qheat(t) + Qsun(t) + Qocc(t) + . . .

Qapp(t) − QSH,loss(t))

This equation is simplified on the form:

T(t + 1) = C1 · T(t) + C2(t) + 1τ ·(λtrans+λvent) · PSH(t)

WhereC1 = 1 − 1

τ

C2(t) = 1τ · (λtrans + λvent)

· (Qsun(t) + Qocc(t) + Qapp(t) +Tout(t) · (λtrans +λvent))

We assume that initially T(0) = Tref . Thus, the temperature for the next timestep T(1) can be calculated for different space heating consumption values rangingfrom 0 to PSH,max (the maximum allowed value). Plotting T(1) against a spaceheating consumption vector within the allowed range, shows that the function islinear with kSH being the slope for that specific time step. This process is repeatedfor the remaining time steps to determine their respective slopes. Note that theslope is dependent on the input data of the model, as seen from the expression ofC2(t). Nonetheless, observations have shown that the changes in slope between thetime steps are barely noticeable.

69

Page 79: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

APPENDIX A. MODEL CALCULATIONS

A.2 Domestic Hot Water SlopesThe temperature of a domestic hot water tank was expressed in Equation 4.6 as:

Ttnk(t) = Ttnk(t− 1) + kDHW · (Pboil(t) − Pboil,prognostic(t))

Where kDHW (expressed in C◦/MW) is the slope between the tank temperatureand the boiler consumption. This relation is assumed to be a linear increasingfunction. The tank temperature heat content equation was expressed in 4.7 as:

Mtnk(t) = Mtnk(t− 1) + Pboil(t) − Qdrain,i(t) − QDHW,loss(t)

The heat content from the previous expression is rewritten in terms of tanktemperature from the equation formulated in 3.7. Assuming the heat content is ata reference energy level at t = 0, the expression for the first time step becomes:

Ttnk(1) = Pboil(1) + Mref

Vtnk · Cpwater+ Tinlet − Qdrain,i(1) − Qloss(1)

Vtnk · CpwaterThus, the tank temperature for the first time step Ttnk(1) can be calculated

for different boiler consumption values ranging from 0 to Pboil,max (the maximumallowed value). Plotting Ttnk(1) against a boiler consumption vector within theallowed range, shows that the function is linear with kDHW being the slope forthat specific time step. This process is repeated for the remaining time steps todetermine their respective slopes. Note that the slope is dependent on the inputdata of the model, as seen from the expression ofMtnk(t). Nonetheless, observationshave shown that the changes in slope between the time steps are barely noticeable.

70

Page 80: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

Appendix B

Scenario Result Figures

B.1 Scenario 2: Spring daysThis section includes the simulation results from scenario 2, i.e. three spring days.The analysis on these Figures are found in the corresponding result and analysischapter.

71

Page 81: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

APPENDIX B. SCENARIO RESULT FIGURES

Transmission capacity

ST export prognosis

LT export prognosis

Real export

Pow

er[M

W]

Time [hours]

0 12 24 36 48 60 72

0

50

100

150

200

Figure B.1. LT and ST export prognosis compared to the real export outcome.All hours when the export is above the transmission capacity are considered exportproblem hours.

Optimized consumption

Consumption (no DR)

Pow

er[M

W]

Time [hours]

0 12 24 36 48 60 7250

55

60

65

70

75

80

Figure B.2. The optimized consumption after LT and ST DR.

72

Page 82: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

B.1. SCENARIO 2: SPRING DAYS

Optimized consumption change (ST cluster)

Power to balance (ST/LT prognosis)

ST Cluster

Pow

er[M

W]

Time [hours]

Optimized consumption change (LT cluster)

Power to balance (LT prognosis)

LT ClusterPow

er[M

W]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

!1

0

1

2

!5

0

5

10

15

Figure B.3. Hourly power to balance in the network for the LT and ST clustersrespectively.

Domestic hot water consumption change (ST cluster)

Space heating consumption change (ST cluster)

DR consumption per household

Pow

er[k

W]

Time [hours]

Domestic hot water consumption change (LT cluster)

Space heating consumption change (LT cluster)

DR consumption per household

Pow

er[k

W]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

!4

!2

0

2

4

6

8

!4

!2

0

2

4

6

8

Figure B.4. Comparison between space heating and domestic hot water consump-tion per household and cluster.

73

Page 83: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

APPENDIX B. SCENARIO RESULT FIGURES

Comfort interval

Reference temperature

Tank temperature (ST cluster)

Tank temperature (LT cluster)Tank temperature change/household

Tem

per

ature

[C]

Time [hours]

Comfort interval

Reference temperature

Indoor temperature (ST cluster)

Indoor temperature (LT cluster)Indoor temperature change/household

Tem

per

ature

[C]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

60

80

100

120

18

20

22

24

Figure B.5. Indoor and tank temperature variation for one LT and ST householdrespectively during DR.

Max BESS capacity

Wind curtailment

BESS level

Pow

er[k

W]

Time [hours]

0 12 24 36 48 60 720

50

100

150

200

250

300

350

Figure B.6. The hourly BESS charge/discharge levels as well as required windcurtailment when using DR.

74

Page 84: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

B.2. SCENARIO 3: SUMMER DAYS

Friday

Thursday

Wednesday

Total load shift/day and household

Pow

er[k

W]

ST cluster size

Friday

Thursday

Wednesday

Total load shift/day and household

Pow

er[k

W]

LT cluster size

0 500 1000 1500 2000

1000 1500 2000 2500

0

5

10

15

0

5

10

15

Figure B.7. Total daily power to balance per household with changes in cluster size.

B.2 Scenario 3: Summer daysThis section includes the simulation results from scenario 3, i.e. three summer days.The analysis on these Figures are found in the corresponding result and analysischapter.

75

Page 85: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

APPENDIX B. SCENARIO RESULT FIGURES

Transmission capacity

ST export prognosis

LT export prognosis

Real exportPow

er[M

W]

Time [hours]

0 12 24 36 48 60 72

!50

0

50

100

150

200

Figure B.8. LT and ST export prognosis compared to the real export outcome.All hours when the export is above the transmission capacity are considered exportproblem hours.

Optimized consumption

Consumption (no DR)

Pow

er[M

W]

Time [hours]

0 12 24 36 48 60 7245

50

55

60

65

70

75

80

85

90

Figure B.9. The optimized consumption after LT and ST DR.

76

Page 86: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

B.2. SCENARIO 3: SUMMER DAYS

Optimized consumption change (ST cluster)

Power to balance (ST/LT prognosis)

ST Cluster

Pow

er[M

W]

Time [hours]

Optimized consumption change (LT cluster)

Power to balance (LT prognosis)

LT ClusterPow

er[M

W]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

!1

0

1

2

3

!5

0

5

10

Figure B.10. Hourly power to balance in the network for the LT and ST clustersrespectively.

Domestic hot water consumption change (ST cluster)

Space heating consumption change (ST cluster)

DR consumption per household

Pow

er[k

W]

Time [hours]

Domestic hot water consumption change (LT cluster)

Space heating consumption change (LT cluster)

DR consumption per household

Pow

er[k

W]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

!4

!2

0

2

4

6

!4

!2

0

2

4

6

Figure B.11. Comparison between space heating and domestic hot water consump-tion per household and cluster.

77

Page 87: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

APPENDIX B. SCENARIO RESULT FIGURES

Comfort interval

Reference temperature

Tank temperature (ST cluster)

Tank temperature (LT cluster)Tank temperature change/household

Tem

per

ature

[C]

Time [hours]

Comfort interval

Reference temperature

Indoor temperature (ST cluster)

Indoor temperature (LT cluster)Indoor temperature change/household

Tem

per

ature

[C]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

60

80

100

120

18

20

22

24

Figure B.12. Indoor and tank temperature variation for one LT and ST householdrespectively during DR.

Max BESS capacity

Wind curtailment

BESS level

Pow

er[k

W]

Time [hours]

0 12 24 36 48 60 720

50

100

150

200

250

300

350

Figure B.13. The hourly BESS charge/discharge levels as well as required windcurtailment when using DR.

78

Page 88: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

B.2. SCENARIO 3: SUMMER DAYS

Wednesday

Tuesday

Monday

Total load shift/day and household

Pow

er[k

W]

ST cluster size

Wednesday

Tuesday

Monday

Total load shift/day and household

Pow

er[k

W]

LT cluster size

0 500 1000 1500 2000

1000 1500 2000 2500

0

2

4

6

8

0

5

10

15

Figure B.14. Total daily power to balance per household with changes in clustersize.

Figure B.15. The impact of DR from Cementa on the reduction of the minimumLT and ST cluster sizes. ’C’ denotes that Cementa is participating in DR whereas’NC’ denotes that it is not.

79

Page 89: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

APPENDIX B. SCENARIO RESULT FIGURES

B.3 Scenario 4: Autumn daysThis section includes the simulation results from scenario 4, i.e. three autumn days.The analysis on these Figures are found in the corresponding result and analysischapter.

80

Page 90: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

B.3. SCENARIO 4: AUTUMN DAYS

Transmission capacity

ST export prognosis

LT export prognosis

Real exportPow

er[M

W]

Time [hours]

0 12 24 36 48 60 72

!50

0

50

100

150

200

Figure B.16. LT and ST export prognosis compared to the real export outcome.All hours when the export is above the transmission capacity are considered exportproblem hours.

Optimized consumption

Consumption (no DR)

Pow

er[M

W]

Time [hours]

0 12 24 36 48 60 7255

60

65

70

75

80

85

90

95

Figure B.17. The optimized consumption after LT and ST DR.

81

Page 91: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

APPENDIX B. SCENARIO RESULT FIGURES

Optimized consumption change (ST cluster)

Power to balance (ST/LT prognosis)

ST Cluster

Pow

er[M

W]

Time [hours]

Optimized consumption change (LT cluster)

Power to balance (LT prognosis)

LT ClusterPow

er[M

W]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

!2

0

2

4

!10

0

10

20

Figure B.18. Hourly power to balance in the network for the LT and ST clustersrespectively.

Domestic hot water consumption change (ST cluster)

Space heating consumption change (ST cluster)

DR consumption per household

Pow

er[k

W]

Time [hours]

Domestic hot water consumption change (LT cluster)

Space heating consumption change (LT cluster)

DR consumption per household

Pow

er[k

W]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

!5

0

5

!5

0

5

Figure B.19. Comparison between space heating and domestic hot water consump-tion per household and cluster.

82

Page 92: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

B.3. SCENARIO 4: AUTUMN DAYS

Comfort interval

Reference temperature

Tank temperature (ST cluster)

Tank temperature (LT cluster)Tank temperature change/household

Tem

per

ature

[C]

Time [hours]

Comfort interval

Reference temperature

Indoor temperature (ST cluster)

Indoor temperature (LT cluster)Indoor temperature change/household

Tem

per

ature

[C]

Time [hours]

0 12 24 36 48 60 72

0 12 24 36 48 60 72

60

80

100

120

18

20

22

24

Figure B.20. Indoor and tank temperature variation for one LT and ST householdrespectively during DR.

Max BESS capacity

Wind curtailment

BESS level

Pow

er[k

W]

Time [hours]

0 12 24 36 48 60 720

50

100

150

200

250

300

350

400

450

Figure B.21. The hourly BESS charge/discharge levels as well as required windcurtailment when using DR.

83

Page 93: Analysis of Demand Response Solutions for Congestion …681429/FULLTEXT01.pdf · 2013-12-19 · Degree project in Analysis of Demand Response Solutions for Congestion Management in

APPENDIX B. SCENARIO RESULT FIGURES

Monday

Sunday

Saturday

Total load shift/day and household

Pow

er[k

W]

ST cluster size

Monday

Sunday

Saturday

Total load shift/day and household

Pow

er[k

W]

LT cluster size

0 500 1000 1500 2000

1000 1500 2000 2500

0

1

2

3

4

5

0

10

20

30

Figure B.22. Total daily power to balance per household with changes in clustersize.

Figure B.23. The impact of DR from Cementa on the reduction of the minimumLT and ST cluster sizes. ’C’ denotes that Cementa is participating in DR whereas’NC’ denotes that it is not.

84