Modeling, Simulations and Analysis of a Demand-Side Management Pilot Project

(1)

INOM

EXAMENSARBETE ELEKTROTEKNIK,

AVANCERAD NIVÅ, 30 HP ,

STOCKHOLM SVERIGE 2016

Modeling, Simulations and

Analysis of a Demand-Side

Management Pilot Project

A collaboration together with Vattenfall R&D

ELIN VIBERG

KTH

(2)

TRITA TRITA-EE 2016:135

(3)

Abstract

The increasing number of renewable power sources gives positive outcomes to a more sustainable society. This change does not come without problems, since the power supply from renewable energy sources is intermittent. This leads to uncertainty to keep the balance between production and consumption. One solution could be to make the end-users adapt their consumption to the production.

Vattenfall has encountered this problem with a project called Smart Customer Gotland. The purpose of the project was to give the end-users the opportunity to actively participate in the energy market by remotely steer their space heaters and water boilers. The goals with the project were to investigate if the energy usage and the electricity cost for the end-user got reduced by using a control schedule, without affecting the end-user’s comfort.

The purpose with this master thesis was to answer the project goals with a simulation replica of the Smart Customer Gotland. With the simulation replica could same input parameters be used and the outcomes could easier be compare

when the devices were remotely steered. The simulation replica consisted of a

house model with a space heater and a water boiler module steered with the same control schedule used in the project. To make the model more accurate, a module which models the end-users’ opportunity to override the remote control schedule was designed. This module was designed from a data analysis, showing that the end-users did few overrides and the overrides that were made tend to switch off the device without affecting their comfort.

The results from the simulation replica showed that the end-users only marginally reduce their electricity cost when a control schedule was used. The end-users com-fort seemed not to be affected, since the indoor temperature and the temperature inside the water tank were kept within acceptable levels.

(4)

(5)

Referat

Installationer av förnybara energikällorna ökar kraftigt, vilket ger positiva påföljder till ett mer hållbart samhälle. Nya omställningar i elförsörjningen kommer inte helt utan problem, då eltillförseln från den förnybara energikällorna är oregelbunden. Det leder till en osäkerhet att bibehålla balansen mellan produktion och konsum-tion. Ett sätt att möta den problematiken är att låta slutkonsumenten anpassa sin konsumtionen till produktionen.

Vattenfall har möt problematiken med ett pilotprojekt kallat Smart Kund

Got-land. Syftet med projektet var att få slutanvändarna mer delaktiga i

elmark-naden genom att automatisk styra deras uppvärmning av hus och varmvatten. Målet med projektet var att undersöka om energianvändningen och elkostnader för slutkonsumenten minskade med hjälp av en automatisk styrning, utan att slutkon-sumentens komfort påverkades. Projektet hade även syftet att studera hur effek-tvariationerna i elnätet förändrades och om elkonsumtionen flyttades till timmar med hög produktion.

Syftet med examensarbetet var att försöka svara på projektmålen med hjälp av en simuleringskopia av Smart Kund Gotland projektet. Med en simuleringskopia kunde samma parametrar användas för att enklare kunna jämföra effekterna utav att använda automtisk styrning av enheterna. Simuleringsmodellen bestod av en husmodell som modellerade uppvärmningen av ett hus och tappvarmvattnet för ett hushåll, med samma styrning som i projektet. För att göra modellen verklighet-strogen adderades en modul som modellerar slutkonsumenternas möjlighet till att skriva över styrschemat. Denna överskrivningsmodul gjordes utifrån en dataanalys som visade att majoriteten av slutanvändarna gjorde få överskrivningar och de som gjorde överskrivningar tenderade att stänga av enheten de timmar då deras komfort inte påverkades.

Resultaten från simuleringskopian visade att slutanvändarna endast marginellt minskade sina elkostnader om ett styrschema användes. Slutkonsumentens komfort verkade inte vara påverkat då temperaturen inomhus och i vattentanken inte avvek från acceptabla nivåer.

(6)

(7)

Acknowledgment

I would like to thank everybody at the RD department at Vattenfall, that helped me to do this master thesis. I would like to give a special thanks to my supervisors at Vattenfall, Bjarke Thormann and Mikael Klingvall. With Bjarkes reassurance and belief in me made this work a lot easier. Without Mikaels expertise in data analysis this thesis would not be possible and I am grateful for all the things I have learned from you. I would also want to thank my supervisor at the Royal Institute of Technology (KTH), Danel Brodén. I am thankful for your pep talk during difficult times and the valuable comments.

(8)

(9)

List of symbols

DR Demand Response

SGG Smart Grid Gotland

KTH Kungliga Tekniska Högskolan

SCG Smart Customer Gotland

ACA Automatic Control Algorithm

SCP Smart Customer Price

SMHI Swedish Meteorological and Hydrological Institute

(10)

Introduction

1.1 Background

With the increased number of renewable power sources, the power system will face a challenge to maintain a reliable electricity production that we have today. Re-newable power sources are intermittent and difficult to predict. Without any viable way to store electricity, there may be occasions when the produced amount of elec-tricity is not balancing the amount of consumed elecelec-tricity. Instantaneous balance between production and consumption is crucial to avoid power outages. To address the technical issues with a changing power system the "smart grid" concept has emerged. The purpose of a "smart grid" is to increase the use of communications and information technology to improve efficiency and reliability of the power system. One part of a smart grid concept is to increase communication and informa-tion flow about the power system to end-users, in order to increase the end-user’s participation in the electricity market. With this type of technology the end-users could be part of balancing the consumption against the intermittent renewable power sources and contribute to a reliable power system. This concept is called Demand Response (DR) and the purpose is to shift load from peak load hours to peak production hours. DR user can move, reduce or increase their flexible power consumption, which can be power for heaters or shiftable loads such as power for laundry machines. In the future it can even be necessary that end-users participate in DR solutions since the consumption of electricity is assumed to increase due to increasing numbers of plug-in electrical vehicles. It is estimated that Sweden has total 3300-5500 MW of flexible power, corresponding to 10-20 % of the maximum power output. Where electrically heated homes could contribute to 2000 MW of the total amount of flexible power [1].

A problem with DR is that end-users have no incentive to adapt their consump-tion since they have in general bad knowledge about the electricity market and their own consumption. Together with a low fixed electricity price the end-user’s savings on the electricity bill will be low [2]. A solution could be to control the end-users devices remotely and suitable devices are space heaters and domestic water

(14)

12 CHAPTER 1. INTRODUCTION

ers. With this solution the end-users have an illusion of being passive and further knowledge about their power consumption is not necessary.

A development project is ongoing on Gotland to test the smart grid concept, called Smart Grid Gotland (SGG). This project is a collaboration between GEAB, Gotlands Energi, Vattenfall, ABB, Svenska Kraftnät, Schnieder Electric, Royal In-stitute of Technology (KTH) and Energimyndigheten. One of the goals was to give the end-users the opportunity to actively participate in the energy market in a sub-project called Smart Customer Gotland (SCG). The SCG sub-project consisted of a smaller group with their consumption monitored and a larger group of remote con-trolled end-users. The remote control group got a control schedule one day ahead to control the on and off hours for space heating and domestic hot water devices. The end-users had the opportunity to manually override the remote control schedule by using a mobile application. The override option gave the end-users a reassurance that they do not lose control over the remote steered devices. It may even be nec-essary to include an override option to get end-users to participate in DR solutions in the future.

The purpose of the master thesis was to study the override behavior of the end-users and to study the SCG project outcomes in a simulation replica of the project. The master thesis was divided into two parts. In the first part, the overriding behavior was studied with data gathered from the SCG project. Parameters that could affect the override behavior such as outdoor temperature and electricity price were studied. A study of override behavior patterns in the data was also performed. In the second part, the SCG project user behavior was replicated in a simulation used to study the outcome of the project. With the simulation replica the same input parameters such as outdoor temperature and the same amount of domestic hot water could be used, which is difficult to test in the real SCG project. The simulation model consisted of an already existing model of a detached house on Gotland, where the usage of a space heater and a water boiler were modeled. The devices were modified to include remote steering. An override module was also created from the override behavior studies to make the simulation replica more accurate. The outcome studies the changes in the indoor temperature and the temperature inside the water boiler, the change in power usage for both devices and change in electricity costs.

1.2 Purpose of master thesis

The main purpose of the thesis project was to create a simulation replica of the SCG project together with an override module in order to evaluate the following project goals:

(15)

1.3. THESIS GOALS AND OBJECTIVES 13

• Did the power usage get reduced with a low impact on the end-user’s comfort with the usage of the SCG control schedule?

1.3 Thesis goals and objectives

The following master thesis goals and objectives have been formulated: • Literature study

A literature study will be performed to give an overview and a better under-standing of the potential of DR and the SCG project.

• Data analysis of the override behavior

With data gathered from the SCG project, a data analysis of the override behavior will be performed. To give a better overview of the override data, an exploratory analysis will be completed. The data analysis will continue with a deeper analysis in order to find parameters affecting the override behavior, which will be used to create an override module.

• Create a simulation model of the SCG project

A simulation replica of the SCG project will be created in MATLAB. The simulation model will contain a model of a detached house on Gotland with a remote steered space heating device and a remote steered water boiler. An override module will be included in order to mimic the reality of the SCG project.

• Performing simulations

The simulation will compare cases with no remote steering, remote steering without any override module and remote steering with an override module. The parameters that will be evaluated are the change in cost, change in power, indoor temperature and temperature inside the water tank.

1.4 Delimitation of study

The study has been limited in certain aspects to ensure reasonable and qualitative deliverables within the thesis time frame:

• Manual overrides

The study will only study the manual overrides done by the end-user. Auto-matically overrides from security system such as overrides from temperature guards and communication failures between devices were excluded.

• Remote controlled end-users

(16)

(17)

Chapter 2

Background Study

2.1 Demand Response

2.1.1 What is Demand Response?

In traditional power grids the electricity is produced from big central generators and delivered to the distribution level where end-users consume the electricity. In-stantaneous balance between production and consumption is necessary to avoid power outages. To keep the balance, the power production is changed to match the demand of power supply. With installations of intermittent renewable energy re-sources it will become more difficult to change the production to match the demand and the system reliability will be at a higher risk. According to European Net-work of Transmission System Operators it is assumed that installation of renewable power resources such as solar and wind capacity will increase between 133-344 GW respectively 230-430 GW to 2030 [3] in Europe. Therefore, it might be necessary to rethink how the power system matches the balance between production and con-sumption in order to maintain a reliable supply of electricity. One way could be to use demand side management, which is a solution doing the opposite to traditional power grids. The consumption will instead be adjusted according to the production. One example of demand side management is DR, where end-users will change their consumption in response to different signals such as electricity price or cus-tomer payments. End-users can manually, automatically or remotely respond to signals by reducing the demand in order to limit peak demand or increase the de-mand when production is high. End-users can directly participate in DR programs through utility companies or through an intermediary called aggregator. An ag-gregator enlists end-users to participate in different DR programs and sell the load reduction to the market [1], [4].

DR can be confused with energy saving programs, however energy savings can be a part of a DR program depending how the end-user participates. An end-user can participate in DR in different ways [4]:

(18)

16 CHAPTER 2. BACKGROUND STUDY

• Power consumption shifting where the end-users cooling or heating systems are shifted to a period with low demand.

• Load curtailment where the end-user’s energy is reduced during peak hours. • Use on-site power generation, such as solar cells and power banks to make the

end-user less dependent on the main grid.

DR programs can be divided into three categories depending on the signal that initiates a demand reduction [4]:

• Rate-based or price DR programs where end-users respond to a dynamical changing electricity price which follows the periods of power mismatches. • Incentive or event-based DR programs where the DR program rewards the

end-users for their load reduction upon a request.

• Demand reduction bids where the end-users announces their available reduc-tion or increase in demand and a requested price.

Benefits and drawbacks with Demand Response

DR will not only benefit the increased reliability of the system by flattening the demand load profile. Other benefits such as the end-user will become more par-ticipating in the market, which reduces the producers’ power on the market. The electricity bill will probably be reduced for the end-users due to a lower wholesale market price. Moreover, it could be possible to avoid to use the environmentally unfriendly power reserves when power consumptions exceed the production [1], [4]. The drawbacks could be increased costs for the end-users, if they are not commit-ted to change their demand to match a dynamical changing electricity price. This could be solved by steering the end-users demand remotely, however investments in equipment of smart-meters and smart controls will be necessary. The end-users comfort can also be affected due to remote steering. If the remote steering can be overridden, the DR potential could be reduced. Another important drawback that could occur is the kickback effect. When the load of several end-users are switched off, a new demand peak could appear when all the loads are switched on again [1].

DR potential

The amount of flexible DR power in Europe during the year 2012 was estimated to be 800 TWh, which corresponds to 29% of the total electricity consumption. According to [5], the DR potential could almost be completely covered Sweden’s peak load corresponds to 3000 MW. [1] came to the conclusion that Sweden has 3300-5500 MW of flexible power, which corresponding to 10-20% of the maximum power output.

(19)

2.1. DEMAND RESPONSE 17

was drawn in [1]. In Sweden the DR potential in the residential sector is estimated to be 2000 MW.

Figure 2.1: DR potential in Europe for different DR users. [5].

2.1.2 Demand Response in detached houses

The residential sector can be divided into detached houses and apartments. How-ever, the apartment sector has a significantly lower DR potential compared to elec-trically heated houses [1]. In figure 2.2, the devices with biggest DR potential in the residential sector can be seen. Heating systems and electric heaters accounts for the largest part followed by domestic water boilers. [6] showed that space heaters and domestic water boilers stand for 45% of the total consumption for a detached house in Sweden during 2014.

(20)

Power consumption is fluctuating between seasons, weekdays and hours. In Swe-den the power consumption is highest during winters due to heating of households and the low number of installed air conditioners. The power consumption is higher during times when people tend to be at home, in the morning between 5-9 a.m. and in the evening between 16-23 p.m. During weekends the power consumption are slightly higher compared to weekdays. The usage of space heating and domestic hot water boilers are more or less even during the day, meanwhile usage of kitchen devices and lightning are higher during the evening, [2]. This means that the end-user might not be willing to shift the power consumption to other hours, since the power is necessary during that time. With this conclusion the power consumption can roughly be divided into manageable and non-manageable loads, where manage-able loads can further divide into adjustmanage-able and shiftmanage-able loads. The list below describes the difference [7]:

• Manageable loads:

These loads are characters of a certain consumption cycle and divided into adjustable and shiftable loads.

– Adjustable loads:

The energy consumption in these loads can directly be modified and reduced depending on desired comfort level. In this category belongs space heaters, and air conditioners.

– Shiftable loads:

These loads starting time can be programmed by the end-user and can be shifted in time. In this category are washing machines, dishwashers and electrical vehicles found.

• Non-Manageable loads:

These loads are continuously running and little can be done to shift or reduce the power consumption. In this category belongs refrigerators, freezers and lighting.

The adjustable loads are suitable for remote steering in power consumption shift-ing due to the heat exchange has an inertia and a change in these loads are not immediately recognized by the end-user. These loads can also be included in load curtailment depending on which level of comfort the end-user desire.

2.1.3 End-users willingness to participate in DR programs

(21)

2.2. SMART CUSTOMER GOTLAND PROJECT 19

periods of power mismatches. The majority of the household has a fixed price or single tariff, which changes monthly [2].

A study made by [2], studied the households will to participate in different DR programs. The results showed that the households were more willing to steer the space heaters during the mornings. For steering during the evening the households wanted a compensation. The compensation got even higher if the household elec-tricity would be controlled.

2.2 Smart Customer Gotland Project

2.2.1 About

The Smart Customer Gotland (SCG) project was an experiment that started at the end of 2013 and ended in April 2016. The purpose of the experiment was to make end-users more actively participate in the energy market by visualizing and control-ling the energy consumption in the household. Space heater devices and domestic water boilers were remotely controlled by a steering schedule, to control the on and off hours.

The project goals for the project were to:

• Reduce the electricity cost for the end-user.

• Reduce the heating usage with a low impact on the end-user’s comfort.

The target group for the project was end-users with consumption over 8000 kWh/year, with space heaters and water boilers. Around 1600 customers were interested in participating in the project. Unfortunately, due to long term fixed price agreements or missing technical prerequisites some end-users got excluded. A group of 230 par-ticipating end-users were selected for the project with remote steering and today (2016-02-19) 200 end-users are remaining in the SCG-project, [8].

There were several reasons why Gotland was selected for the pilot project. Gotland is an island in the Baltic sea and is an electrically closed system with its own elec-tricity frequency. Gotland has a high wind production with 175 MW installed wind power and only one HVDC link connection to the mainland [9]. For these several reasons Gotland is a suitable place for studying the impact with renewable power sources.

2.2.2 Devices and installations

(22)

turning the device on or off by the automatic remote control schedule or by the end-user. The end-user received a schedule of the on and off hours for the devices, including a visualization of electricity price in an app application one day ahead at 3 p.m. Manual overrides to change the on or off hours of the automatic remote control schedule could be done by the end-user. Overrides could also occur with help of a safety guard systems such as too low indoor temperature registered by a temperature guard or during times when the connection between the devices were lost. An overview of the installation can be seen in figure 2.3. Where the inputs to the automatic control algorithm (ACA) was based on an optimization, which consider the electricity price, wind prognosis and temperature forecasts, [10].

Figure 2.3: SCG house set-up.

The list below shows the devices that were able to be controllable in the project.

Water boiler

(23)

heated stand-alone tanks, storage water heaters and one type that heats water di-rectly when it is used. The storage water collects water from water space heating systems and can also have its own immersion heaters when more water is needed. A water system that heats the water directly is fast, but high power is required and this can get expensive. Therefore, this type is not common in Sweden, but can be present in holiday cottages. The water level is constant in the tank and filled with cold water, which is heated to 60-100°C. It is important that the temperature does not drop below 60°C due to minimize the growth of Legionella bacteria [11].

Space heaters

When the outdoor temperature is colder compared to the indoor temperature the heat will flow out from the house. To keep the indoor temperature at a comfortable level it is essential to heat the house. The Public Health Agency in Sweden recom-mends a temperature at least 20°C and the temperature should not be over 24°C for longer periods [12].

Houses can be heated in different ways. The list below describes the most common space heat devices.

Electric heater

An electric heater (also named electric boiler) is connected to the building’s water system and the water is distributed around the house by a circulation pump and spread via radiators that distributes the heat inside the house. The water inside the electrical heater is heated by immersion heaters connected to the grid. A thermostat is reacting on the indoor temperature and used to regulate the amount of water flow to the radiators. There are different types of electrical heaters such as separated heaters, combi heaters and electrical heaters that also heats the domestic water. Combi heaters can use different types of fuel to heat the water, such as oil, electricity or wood, [13].

Electrical radiator

An electric radiator consists of electrical resistors that produce heat and it is directly connected to the grid, usually connected to the house’s distribution box. There are two types of electrical radiators, one is called open and consists of several areas, where the air around it heats up and rises. The second one is filled with oil and releases heat when air passes the radiator surface. A thermostat is regulating the heat by changing the power. If only an electric radiator is used a separate water boiler is needed, [13].

Water heat pump

(24)

to direct electrical heating and electrical heater systems. Water heat pumps can also be used to heat domestic water, [14].

Air/Air heat pump

Air/Air heat pumps use heat from the outdoor air to heat up the house and therefore the efficiency varies with the outdoor temperature. The pump is driven by electricity. An air/air heat pump is used as a supplement to the current heating systems. When it gets too cold the pump cannot provide any heat and other heating system need to be used. The advantage with air/air heat pump is that it is flexible and can be used as air-condition during the summer or as a dehumidifier. An air/air heat pump cannot provide heat to domestic water, [15].

Underfloor heating

Underfloor heating can be used to heat the house or only used as comfort to avoid cold floors during cold mornings. There are two types of underfloor heating systems, one is electrical floor heating which uses heating coils directly connected to the grid. The other one is waterborne and uses the building’s water systems and the water is heated by an electrical heater or a water heat pump, [16].

Each house is unique and different types of space heating system are more appropri-ate than others. Houses can also have more than one type of heating system with separate or combined domestic water boilers. New type of system can already have some kind of control system, such as the temperature gets lower during the night. The variety of installations was also the case for the SCG project and together with different brands and age of the devices the remote controlled steering installation could be very different. Many of the houses in the SCG project had also a fireplace, which affects the heating of the house. Some installations used different types of relay, which were regulated by a voltage, controlling the opening and closing of the breaker. In other installations the temperature regulator were tricked to turn the device on or off. In some houses only one or some of the devices could be remotely controlled. However, one important aspect to remember was that the remote steer-ing could only switch the device on or off.

The following devices were used for the remote steering in the SCG project. • Water boiler

• Electrical heaters • Electrical radiator

This group contains both electrical radiators and air/air heat pumps, since the installers of the SCG project did not separate all of the device in the app application.

• Water heat pump

(25)

• Electrical floor heating

2.2.3 Smart Customer Gotland algorithm

There were two different generate control algorithms, one for water devices and one for the space heating devices. Due to confidential material the control algorithms cannot be described in detail, however brief explanations of the control algorithms are mentioned.

Control algorithm for the water boiler

The control schedule for the water boiler were generated by dividing the day in dif-ferent sections. Each section had difdif-ferent amount of hours, which the water boiler was switched off, all other hours the device was switched on. The hours that the device was switched off were those hours that were most expensive.

Control algorithm for the space heater

All the different space heating devices used the same control algorithm. This al-gorithm switched off the device during expensive hours. The number of switch off hours was depending on the outdoor temperature. All other hours the space heating devices were switched on.

Figure 2.4 illustrates the generated control schedule for the different devices. The color green represent that the device will be switched ON and red represent that the device will be switched OFF.

Figure 2.4: Generated control schedule

During 2015 different comfort levels were included for both space heaters and water boilers. The different comfort levels had different amount of hours the devices were switch off. The following comfort levels existed and they are order in from soft to hard remote steering.

(26)

• Normal • Economy • Economy plus

For this project the end-users could select between two different tariffs, a single tariff or a special time tariff called Smart Customer Price (SCP). The SCP was based on the spot price and added different price parameters to give the end-users an incentive to change their consumption from high load hours to low load hours of the day. The high load hours were defined as working days between 6:00-22:00 dur-ing the months January-March and November-December. An enhancement factor was used to increase the difference in price between high and low load hours during parts of the day. The SCP include a wind discount when the total wind production prognosis for that day exceeded 190 MWh. The end-users received a notice in their smart phones when this occurred. However, the wind discount was limited to a maximum of five days per month, and a maximum of 30 days per year.

The control schedule was delivered at 3 p.m. to the end-users the day before in their mobile application, see figure 2.4. The circle in the middle shows how the price was changing during the day. The red highlighted arrow shows the most ex-pensive hour and the green hour the cheapest hour. The circle in the middle shows the remote control schedule and the outer circle was used for the end-users to man-ually overrides. The indoor temperature was also visualized in mobile application.

(27)

Chapter 3

Override data analysis

The automatically generated control schedule could be overridden by a manual over-ride by the end-user or by an automatic overover-ride of the security system, such as a temperature guard.

There were four different types of manual overrides made by the end-user:

• End-user sets the device to OFF ahead of schedule, any time before the oper-ation time.

• End-user sets the device to ON ahead of schedule, any time before the oper-ation time.

• End-user sets the device to OFF within the scheduled hour. • End-user sets the device to ON within the scheduled hour. There were three different types of automatic overrides:

• Device ON caused by a low temperature in the house, communicated with the temperature guard.

• Device ON, caused by a remote controlled device was not communicating for a certain time.

• Device ON, caused by loss of communicating with the temperature guard. The list below describes some examples were a manual or automatic override could have occurred.

End-user sets the device to ON ahead of schedule, any time before the operation time:

The end-user knew that they would have lodgers the next coming day and they were worried that the amount of domestic water was not enough. Therefore, the end-user did an override after they received the generate remote schedule at 3 p.m. to switch the water boiler ON more hours than the generate remote schedule.

(28)

26 CHAPTER 3. OVERRIDE DATA ANALYSIS

End-user sets the device to OFF within the scheduled hour:

The end-user thought it was hot inside and did an override to turn the device OFF within the schedule ON hour.

Device ON caused by a low temperature in the house, communi-cated with the temperature guard:

During a cold winter day, the temperature inside the house got below the selected minimum indoor temperature. The temperature guard did an auto-matic override to turn the device ON although the hour was scheduled to be OFF.

Device ON, caused by a remote controlled device was not commu-nicating for a certain time:

The smart relay plug connected to a water boiler lost connection to the En-ergy Watch. The device was switched ON for all the next coming hours until the problem was solved.

As it can be seen there could be many different reasons why an override occurred. This data analysis will only focus on the manually override made by the end-user. The data analysis will try to find some clue to why these manually overrides oc-curred, by search for parameters that may affect the overrides, such as temperature or electricity price. The data analysis will also study if there were some special times overrides occurred, such as if the end-users did more overrides during weekends com-pared to weekdays. Data used for these analysis was gathered from the SCG project. The section below will describe the method used to do the data analysis.

3.1 Methods

3.1.1 Data cleansing

In order to do an analysis a cleansing of the data was required. The purpose of the cleansing was to get a dataset to only contain certain aspects suitable for the override analysis. This section describes in detail how the cleansing was made.

Data

The raw data was collected from three different sources;

• A SCG end-user Excel file, created from SCG-admin pages.

• A database from the company Perific, which collects data of the SCG project. • Temperature data from SMHI.

(29)

3.1. METHODS 27

SCG end-user data

Raw data

The SCG admin pages gave the opportunity to see all the end-users that participated in the project. Together with the Perific database and installation files, an Excel file was created containing the following columns:

User ID: Every end-user had a unique ID to make searching in the database

easier. The number was retrieved from the end-user table in the database.

Postal address: Retrieved from SCG admin-pages.

Group code: The end-users were divided into three groups:

– Accepted: End-users that got accepted for remote steering of their

de-vices.

– Energy Watch accepted: End-users that did not have remote control

steering, but got accepted only using an Energy Watch.

– Personal testers: Some end-users worked for Vattenfall, GEAB or Perific

wanted to participate in the project. The personal testers could have remotely steered devices or they used only an Energy Watch.

Steering of water boiler: Yes/No

ID of water boiler: If a water boiler was remotely steered the ID number

of the device was collected from SGG-admin pages, later on double checked with the device table in the database.

Steering of electrical heater: Yes/No

ID of electrical heater: Same principle as for the water boiler. Steering of electrical radiator: Yes/No

ID of electrical radiator: Same principle as for the water boiler. Steering of water heat pump: Yes/No

ID of water heat pump: Same principle as for the water boiler. Steering of floor heating: Yes/No

ID of floor heating: Same principle as for the water boiler.

Living: The end-users lived in different types of buildings: house, terrace

house, agriculture, apartment, industry, company and not permanent living. The information was gathered from installation files, Google maps and Eniro.

Price Tariff: Single tariff or time tariff gathered from admin-pages. Cleansing

(30)

Postal address: Only end-users who lived on Gotland were selected, since

the other data were gathered from Gotland.

Group code: Both personal testers and accepted end-users were used. Only

Energy Watch end-users were removed since they could not do any overrides.

Floor heating: Floor heating devices were removed. It was assumed that

the floor heating mainly was used for comfort and not heating and the number of floor devices were few.

Living: End-users living in house, terrace house and agriculture houses were

selected, since the model used later on represents a house.

Problems with reconfiguration of the devices had occurred and one device could have more than one device-ID. The affected end-users were removed, since the interpretation of the data was difficult. The cleaned SCG end-user data consisted of 180 end-users from a total of 328 in the original file. The SCG end-user data file was further cleansed with the help of the end-user override data, which is explained in the next section.

Perfic database

Data in Perific database extended from 2013-05-31 to 2015-12-17. The data for the year 2015 was not complete, since the copy of the database was made 2015-12-17. The tables which were used from the database are listed and described below:

Schedules Raw data

This table was showing both the remote control schedule and the end-user override schedule for the highlighted day. A description of the data structure in the table is shown.

Unix time: Contains the date and time of the measurement in Unix time

format.

Remote control table: Consisted of 24 values of 1 or 0 generated from the

remote control algorithm. Each value represents the on/off state. 1: Device ON

0: Device OFF

End-user override table: Consisted of 24 values of 2, 3, 4, 5, 7, 8 , 9 or -,

(31)

3.1. METHODS 29

3: End-user sets the device to ON ahead of schedule, any time before the operation time.

4: End-user sets the device to OFF within the scheduled hour. 5: End-user sets the device to ON within the scheduled hour.

7: Device ON, caused by a low temperature in the house, communicated with the temperature guard.

8: Device ON, caused by a remote controlled device was not communi-cating for certain time.

9: Device ON, caused by loss of communicating with the temperature guard

-: This state indicates no change in the original remote control schedule.

Time stamp: Showing the date when the schedule was used. Cleansing

The values 7,8 and 9 were changed to not a number (NaN) in the end-user override table. These are those hours the system did an automatic override and the end-users themselves could not do any manual override. Since this analysis is focusing only on the manual override the values 7,8 and 9 were not of interest. The largest override dataset for a year was found and contained data from 2014-05-01 to 2015-04-30. The data contained 1033680 values with a NaN percentage of 6.4%. End-users that had a dataset spanning from the start date to the end date were used for further studies. End-users individual NaN percentage was calculated and end-users with a NaN percentage over 20 % were removed. After this cleaning 107 end-users were remaining of 180.

The remote control schedule table was also further separated to only contain data for the whole year for the period 2014-05-01 to 2015-04-30.

Smart Customer Price

Raw Data

The raw data consisted of a spot price, which was re-calculated to the SCP. The spot price contained a complete set of hourly time interval data (8760 values) with the unit in öre/kWh.

Cleansing

(32)

SMHI

Data for the outdoor air temperature was collected from Swedish Meteorological and Hydrological Institute (SMHI) from a weather station called Fårösund, [17]. This weather station was selected since it had the least number of missing values for the period 2014-05-01 to 2015-04-30. The weather station is also close to Visby, where the majority of the end-users lived. The other weather stations at Gotland had more missing values and therefore they were excluded, see table 3.1. An interpolation was made to fill in the 9 missing values.

Table 3.1: Number of missing values for period 2014-05-01 to 2015-04-30 at weather stations on Gotland.

Weather station Number of missing values Distance from Visby (km)

Roma 977 19.8

Hemse 7665 52.8

Fårösund 9 56.5

Östergarnsholm 22 53.7

3.1.2 Exploratory data analysis

An exploratory data analysis is a process of exploring the data and the purpose is to get a better understanding what the data is containing and determine if the questions that are sought can be answered. An exploratory data analysis includes examining the components of the dataset and visualizing data using a graphical representation, [18].

The exploratory data analysis will focus on the following points: • How many end-users were doing an override?

• How many manual overrides did the dataset consist of?

• How many overrides belonged to the different manual override options? • How many different types devices were there?

• How many manual overrides were there between the different devices? • How many different manual override options were there between the different

devices?

• How many manual overrides did each individual end-user?

3.1.3 Deeper analysis

(33)

3.1. METHODS 31

behavior such as the outdoor temperature and the SCP price. In order to find pa-rameters affecting the override behavior a correlation analysis and a comparison of averages values were performed.

The deeper analysis included a time series analysis and the purpose was to study if the number of the override occurrences were changing between different time series. Different time series were studied and the following questions were sought for:

• Were there any difference between the number of overrides between weekends and weekdays?

• Were there any difference between the number of overrides between the hours in a day?

• Were there any difference between the number of overrides between different seasons?

These analyses were performed by studying a standard override week and a stan-dard override day, which is described in more detail in this chapter.

For these analyses were "OFF ahead of schedule" and "OFF within the scheduled hour" grouped together and the same for "ON ahead of schedule" was grouped to-gether with "ON within the scheduled hour". Table 3.2 shows an example how the different override options were illustrated for each end-user. 1 represent that an override occurred to switch the device ON/OFF for that hour.

Table 3.2: Example how the different override options were illustrated for each end-user.

Date _{Override OFF}End-user 1_{Override ON} _{Override OFF}End-user 2_{Override ON} _{. . .}. . ._{. . .}

2014-05-01 00:00 1 0 0 0 ... ... 2014-05-01 01:00 0 1 0 1 ... ... 2014-05-01 02:00 0 0 0 0 ... ... .. . ... ... ... ... ... ... 2015-04-30 23:00 0 1 1 0 ... ... Correlation

(34)

Figure 3.1: Examples of positive and negative correlation examples.

• The end-users did overrides to turn the device OFF when the price or the outdoor temperature was high.

• The end-users did overrides to turn the device ON when the price or the outdoor temperature was low.

For this analysis the value of the correlation coefficient was classified according to table 3.3.

Table 3.3: Correlation classifications.

Correlation coefficient Classification

-0.29 to 0.29 No correlation

-0.59 to -0.3 and 0.3 to 0.59 Weak correlation -1.0 to -0.6 and 0.6 to 1.0 Strong correlation

The correlation analysis was performed by calculating the sum of all end-users’ overrides for each hour. Table 3.4 gives an example how this sum was calculated for a dataset with three end-users. "P

Overrides" is the sum of all the end-users’ overrides and x the number of end-users.

Table 3.4: Example of the summation of the override option ON.

Date Override ON end-user 1 Override ON end-user 2 Override ON end-user 3 Px

(35)

3.1. METHODS 33

The "P

Overrides" table was joined with the SMHI outdoor temperature or the SCP table and the correlation coefficient was calculated from the paired table. An example of how the joined table could look like is shown in table 3.5.

Table 3.5: Example of data table used for correlation calculations.

Date Px

i=1(OverridesON )x Temperature [°C]

2014-05-01 00:00 1 -3.0 2014-05-01 01:00 3 -3.5 2014-05-01 02:00 0 -3.5 .. . ... ... 2015-04-30 23:00 2 -5.0

Comparison of average values

Comparison of average values can be used to study each end-user individually. The average price when an override occurred to average price when an override did not occurred can be compared for each end-user. It can be assumed that the end-users, by themselves, wanted to improve the remote steering schedule by trying to opti-mize the ON and OFF hours according to the SCP, even more than the remote steering algorithm. The end-users would then do an override to switch the device OFF when the price was high and do an override to switch the device ON when the price was low. The ON and OFF overrides were separated into two groups and the two following comparisons were made:

1. The average value of the SCP when an override to switch the device OFF, ¯

p1OF F , was compared to the average value of the SCP when no overrides

were made, ¯p0OF F .

2. The average value of the SCP when an override to switch the deceive ON

, ¯p1ON , was compared to the average value of the SCP when no overrides

were made, ¯p0ON .

Table 3.6 shows an example of how the average values were calculated. The average value, ¯p1OF F , was calculated from a dataset, which was created by picking out the

SCP values for the hours an override to turn the device OFF occurred. The average value ¯p0OF F was calculated from another dataset, which was created by picking

out the SCP values for the hours an override did not occurred in the override OFF dataset. The average values for the second comparison, ¯p1ON and ¯p0ON , were

(36)

The difference should be vise versa for the average SCP value when the end-user switched on the device compared to the average SCP value case when no overrides were made, PON < 0.

POF F = (¯p1OF F − ¯p0OF F) > 0 (3.1)

PON = (¯p1ON − ¯p0ON) < 0 (3.2)

Table 3.6: Example of how the datasets for calculating ¯p1OF F, ¯p0OF F, ¯p1ON and

¯

p0ON were selected.

Date _Override End-user 1 . . .

OF F SCP OverrideON SCP . . . . 2014-05-01 00:00 1 SCP₀ 0 SCP₀ ... ... ... ... 2014-05-01 01:00 0 SCP1 0 SCP1 ... ... ... ... 2014-05-01 02:00 0 SCP2 1 SCP2 ... ... ... ... .. . ... ... ... ... ... ... ... ... 2015-04-30 23:00 1 SCP_n 1 SCP_n ... ... ... ...

x = 1_nPni=1SCPi : p¯1OF F p¯0OF F p¯1ON p¯0ON ... ... ... ...

Time series analysis

In order to study if the number of the overrides were changing between different time series a standard override week and a standard override day were calculated for each of the override options OFF and ON. A standard override week will illustrate how many end-users that will do an override for each hour of a week. Meanwhile a standard override day will illustrate how many end-users that will do an override for each hour of a day. A standard override week was calculated by summing the overrides occurrences for each hour of a week. I.e. all numbers of overrides ON/OFF occurrences for all Mondays at 00:00, Mondays at 01:00, ..., Sunday at 23:00 were separated summed up. Table 3.7 and 3.8 shows an example of the calculations for

override option OFF.Px

i=1(OverridesOF F )x was the sum of all overrides from all

(37)

3.1. METHODS 35

Table 3.7: Example of the summation of override OFF occurrences used to calculate the standard override week.

Date Px i=1(OverridesOF F )x Mon: 2014-05-05 00:00 5 Mon: 2014-05-05 01:00 0 .. . ... Sun: 2014-05-11 22:00 10 Sun: 2014-05-11 23:00 2 .. . ... Mon: 2015-05-20 00:00 1 Mon: 2015-05-20 01:00 3 .. . ... Sun: 2015-05-26 22:00 4 Sun: 2015-05-26 23:00 2

Table 3.8: Dataset used to plot a standard override week for override option OFF.

Date Average of overrides OFF during each hour of a week

Mon: 00:00 _x1(_n1 Pn

i=1 (Override OFF Mon 00:00)i)

Mon: 01:00 _x1(_n1 Pn

i=1 (Override OFF Mon 01:00)i)

..

. ...

Sun: 22:00 1_x(_n1Pn

i=1 (Override OFF Sun 22:00)i)

Sun: 23:00 1_x(_n1Pn

(38)

3.2 Results override data analysis

3.2.1 Exploratory data analysis

From the dataset containing 107 end-users and spanning from 2014-05-01 to 2015-04-30 is the following result of the exploratory data analysis presented.

Figure 3.2 represent a box plot for the end-users individual NaN percentage. It can be seen from the figure that the majority of the end-users’ individual NaN percentage was located above the median. This states that the end-users had a high value of NaN percentage and this might cause misleading results.

NaN percentage [%] 0 2 4 6 8 10 12 14 16 18 20

Box plot for end-users individual NaN-percentage

Figure 3.2: Box plot for end-users individual NaN percentage. The end-users got divided into the two following groups:

• OU group: Remote steered end-users that did an override at some point • NOU group: Remote steered end-users that chose not to do any override at

all.

Table 3.9 shows the number and percentage of end-users in the different groups. Table 3.9: Classifications of the end-users.

Group Number of End-users Percentage of End-users [%]

Remote Steered End-users (RU) 107 100

Override End-users (OU) 80 74.8

No Override End-users (NOU) 27 25.2

(39)

3.2. RESULTS OVERRIDE DATA ANALYSIS 37

tending to turning off the device ahead of schedule more than turning the device on. One reason for this could be that the control schedule was too soft and there was potential to control the devices even more. Another explanation could also be that the end-user wants to reduce the electricity costs even further.

93% 7%

Total percentage of overrides

No Override Override 73% 23% 3% 2%

Percentage of the different override options

Override option OFF ahead of schedule Override option ON ahead of schedule Override option OFF within scheduled hour Override option ON within scheduled hour

Figure 3.3: To the left: The total override quantity in the dataset. To the right: Quantity of the different overrides options in the dataset.

The number of the different devices for the 107 end-users can be seen in figure 3.4. There were totally 120 different controlling devices and water boilers were more common as a controlling device. Some end-user had both a water boiler and a space heating device controlled. The percentage if they only had one device con-trolled or both, are listed in table 3.10.

Table 3.10: Percentage of steered devices.

Remote steered device Percentage [%].

Water boiler 45.8

Space heater 43.9

(40)

Water Boiler Electrical Heater Electrical Radiator Water Heat Pump

Number of devices 0 10 20 30 40 50 60 70 61 27 13 19

Number of different control devices

Figure 3.4: Number of different control devices for the 107 end-users.

Figure 3.3 was further separated into the different devices in figure 3.5. The figure shows the percentage of overrides versus no overrides in the dataset within different devices. It can be concluded, from the figure, that overrides were more frequent for electrical radiator and electrical heater.

Hot Water Boiler Electrical Boiler Electrical Radiator Water Heat Pump

Percentage [%] 0 10 20 30 40 50 60 70 80 90 100

Percentage of number of Overrids and No Overrides at each device in the dataset

Overrides No Override 6.9% 93.1% 90.4% 90.0% 96.9% 3.1% 10.0% 9.6%

Figure 3.5: Percentage of the difference between override and no override for each device.

(41)

water boiler more than the other devices. This can be due to the fact that the water boiler is depending on the usage, meanwhile space heating usage is determined by thermodynamic laws.

Water Boiler Electrical Heater Electrical Radiator Water Heat Pump

Percentage % 0 10 20 30 40 50 60 70 80 90

Percentage of different override options for each device.

Override Option OFF ahead of schedule Override Option ON ahead of schedule Override Option OFF within schedule hour Override Option ON within schedule hour 0.8% 8.7% 2.1% _0.7% 9.2% 9.8% 3.5% 18.1% 2.9% 1.0% 62.0% 35.7% 1.6% 88.4% 77.5% 78.0%

Figure 3.6: Percentage of different override options for each device.

Figure 3.7’s x-axis is showing the number of overrides divided into percentage in-tervals, meanwhile the y-axis is showing the number of end-users. The percentage interval consists of steps of 1 % intervals from 0-48.9. All 80 OU end-users’ override percentage were counted separately and grouped together if the override percentage fell within the same interval. From figure 3.8 can the override quantity in percent-age for each individual end-user can be seen. The conclusion for these two graphs is that it is most common to a very small quantity of overrides, 0-0.99 %. The two end-users that were doing almost 50 % overrides were studied in more detail and extra information can bee seen in table 3.11. This two end-users confirms figure 3.6, since the water boiler was switched off more than the electrical heater.

Table 3.11: End-users doing almost 50 %.

Override percentage interval Remote steering device Overrides OFF [%] Overrides ON [%] NaN[%]

(42)

Override percentage intervals

0.0-0.9 1.0-1.9 2.0-2.9 3.0-3.9 4.0-4.9 5.0-5.9 6.0-6.9 7.0-7.9 8.0-8.9 9.0-9.9 10.0-10.9 11.0-11.9 12.0-12.9 13.0-13.9 14.0-14.9 15.0-15.9 16.0-16.9 17.0-17.9 18.0-18.9 19.0-19.9 20.0-20.9 21.0-21.9 22.0-22.9 23.0-23.9 24.0-24.9 25.0-25.9 26.0-26.9 27.0-27.9 28.0-28.9 29.0-29.9 30.0-30.9 31.0-31.9 32.0-32.9 33.0-33.9 34.0-34.9 35.0-35.9 36.0-36.9 37.0-37.9 38.0-38.9 39.0-39.9 40.0-40.9 41.0-41.9 42.0-42.9 43.0-43.9 44.0-44.9 45.0-45.9 46.0-46.9 47.0-47.9 48.0-48.9 Number of end-users 0 2 4 6 8 10 12 14 16 18 20 17 9 8 3 4 4 3 3 5 0 1 2 3 1 2 0 0 1 0 0 0 2 0 1 2 1 0 0 0 2 0 0 0 1 0 0 0 0 0 1 0 1 0 1 0 0 0 1 1

Number of end-users in different overrides percentage intervals

Figure 3.7: Number of end-users in different overrides percentage intervals

End-user ID 10 20 30 40 50 60 70 80 Percentage of overrides 0 5 10 15 20 25 30 35 40 45

Percentage of overrides for each End-user

Override Option OFF ahead of schedule Override Option ON ahead of schedule Override Option OFF within scheduled hour Override Option ON within scheduled hour

(43)

3.2.2 Deeper analysis

The deeper analysis was divided into two sections, one for the water boiler and one for the space heating device. The space heating device; electrical heater, electrical radiator and water heat pump were combined to a space heating module, since they used the same remote steering schedule and the results were similar to each other, which will be shown in this section.

Water boiler

Overrides for the water boiler was further analyzed by 44 end-users doing override at some point.

Correlation

The results of the correlation analysis, described in methods on page 32, are shown in table 3.12. No correlation between the different override options and the tem-perature or the SCP could be found. In the appendix 8.1.1 can respective scatter plot be seen.

Table 3.12: Correlation results for water boiler.

Override Option Temperature SCP

OFF No correlation No Correlation

ON No correlation No Correlation

Comparison of average

The y-axis and x-axis are showing P_ON respectively P_{OF F}. Correlation exists if the user is in the fourth quadrant, where P_{OF F} is positive and P_ON negative. It can be seen from figure 3.9 that the majority of the end-users were outside the fourth quadrant and correlation between price and overrides cannot be expected.

P OFF [öre/kWh] -20 -10 0 10 20 30 40 PON [öre/kWh] -15 -10 -5 0 5 10 15 20 25 30

Comparison of average price values for Water Boiler

(44)

A standard override week for the water boiler can be seen in figure 3.10. The stan-dard override week will illustrate how many end-users that will do an override for each hour of a week. Clear patterns can be seen. The peaks and valleys occurred approximately at the same time and the height of the peaks were about the same heights for all days for the week. No bigger difference can be seen from weekdays and weekends. Time [h] Mon-00:00 06:00 12:00 18:00 Tue - 00:00 06:00 12:00 18:00 Wed-00:00 06:00 12:00 18:00 Thu-00:00 06:00 12:00 18:00 Fri-00:00 06:00 12:00 18:00 Sat-00:00 06:00 12:00 18:00 Sun-00:00 06:00 12:00 18:00 Percentage [%] 0 5 10 15 20

25 Override Option OFF

Time [h] Mon-00:00 06:00 12:00 18:00 Tue - 00:00 06:00 12:00 18:00 Wed-00:00 06:00 12:00 18:00 Thu-00:00 06:00 12:00 18:00 Fri-00:00 06:00 12:00 18:00 Sat-00:00 06:00 12:00 18:00 Sun-00:00 06:00 12:00 18:00 Percentage [%] 0 5 10 15 20 Override Option ON

Override for Water Boiler

Figure 3.10: Standard override week for water boiler.

(45)

3.2. RESULTS OVERRIDE DATA ANALYSIS 43 Time [h] 00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 Percentage [%] 0 5 10 15 20 25 30 35 40

Override Option OFF

Time [h] 00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 Percentage [%] 0 5 10 15 20 25 30 35 Override Option ON

Average Day for Water Boiler

Figure 3.11: standard override day for water boiler.

Time [h] 00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 Percentage [%] 0 5 10 15 20

Time [h] 00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 Percentage [%] 0 2 4 6 8 10 12 14 Override Option ON Override Option Control schema: Device Switched On

Overrides with Generated Controll Schema for Water Boiler

(46)

Space Heaters

In this section the results of the combined space heating device will only be shown. For a deeper analysis for each space heating device see the appendix 8.1.3. For this part was 47 end-users used.

Correlation

The results of the correlation analysis are shown in table 3.13. No correlation be-tween the different override options and the temperature or the SCP could be found. In the appendix 8.1.2 can respective scatter plot be seen.

Table 3.13: Correlation results for space heater.

Override Option Temperature SCP

OFF No correlation No Correlation

ON No correlation No Correlation

Comparison of average

The y-axis and x-axis are showing P_ON respectively P_{OF F}. Correlation exists if the user is in the fourth quadrant, where P_{OF F} is positive and P_ON negative. It can be seen from figure 3.13 that only two end-users were in the fourth quadrant and correlation between price and overrides cannot be expected.

P OFF [öre/kWh] -15 -10 -5 0 5 10 15 20 P ON [öre/kWh] -10 -5 0 5 10 15 20 25 30

Comparison of average price values for Space Heater

Figure 3.13: Comparison of price average for space heater.

(47)

3.2. RESULTS OVERRIDE DATA ANALYSIS 45 Time [h] Mon-00:00 06:00 12:00 18:00 Tue - 00:00 06:00 12:00 18:00 Wed-00:00 06:00 12:00 18:00 Thu-00:00 06:00 12:00 18:00 Fri-00:00 06:00 12:00 18:00 Sat-00:00 06:00 12:00 18:00 Sun-00:00 06:00 12:00 18:00 Percentage [%] 0 2 4 6 8 10 12 14 16 18

Time [h] Mon-00:00 06:00 12:00 18:00 Tue - 00:00 06:00 12:00 18:00 Wed-00:00 06:00 12:00 18:00 Thu-00:00 06:00 12:00 18:00 Fri-00:00 06:00 12:00 18:00 Sat-00:00 06:00 12:00 18:00 Sun-00:00 06:00 12:00 18:00 Percentage [%] 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Override Option ON

Override for Space Heater

Figure 3.14: Standard override week for space heater.

(48)

46 CHAPTER 3. OVERRIDE DATA ANALYSIS Time [h] 00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 Percentage [%] 0 5 10 15 20 25 30 35 40 45

Override Option OFF

Time [h] 00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 Percentage [%] 0 5 10 15 20 25 Override Option ON

Average Day for Space Heater

Figure 3.15: Standard override day for space heater.

Time [h] 00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 Percentage [%] 0 2 4 6 8 10 12 14 16

Time [h] 00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 Percentage [%] 0 0.5 1 1.5 2 2.5 3 3.5 4 Override Option ON Override Option Control schedule: Device Switched On

Overrides with Generated Controll schedule for Space Heater

Modeling, Simulations and Analysis of a Demand-Side Management Pilot Project

Modeling, Simulations and

Analysis of a Demand-Side

Management Pilot Project

A collaboration together with Vattenfall R&D

ELIN VIBERG

Abstract

Referat

Acknowledgment

List of symbols

Contents

Chapter 1

Introduction

1.1

Background

1.2

Purpose of master thesis

1.3

Thesis goals and objectives

1.4

Delimitation of study

Chapter 2

Background Study

2.1

Demand Response

2.2

Smart Customer Gotland Project