Demand-side Flexibility Management for Electric Vehicle Charging

(1)

Demand-side Flexibility Management for Electric Vehicle Charging Hafiz Muhammad Qasim

Master's Thesis

University of Eastern Finland School of Computing

Computer Science

29 April, 2020

(2)

UNIVERSITY OF EASTERN FINLAND, Faculty of Science and Forestry, Joensuu School of Computing

Hafiz Muhammad Qasim: Demand-side Flexibility Management for Electric Vehicle Charging

Master’s Thesis, 58 p.

Supervisors of the Master’s Thesis: Prof. Xiao-Zhi Gao and PhD Salmi Tuukka Reviewers: PhD Jussi Kiljander and MSc Janne Takalo-Mattila

29 April, 2020

Abstract: The sustainable development goals defined by the United Nations provide framework to make planet earth a safer place for humans. One of the goals is reduce greenhouse gas emissions substantially by 2030. To achieve these goals, all the countries are shifting from fuel based economy to green economy. The major transitions are undergoing in the domain of energy sector and automobile sector. The energy sector is shifting towards renewable energy and the automobile sector moving towards electric vehicles. These transitions require changes in the existing power transmission systems because the flexibility at generation side is getting challenged with each coming day. There are many demand-side flexibility management methods available based on electricity prices. The incentive based methods are mainly getting more traction to offer some kind of rewards to consumers if they enrol into these demand-side flexibility management programs. This thesis presents a proof-of-concept implementation of an EV controller which is the component of a flexibility management system, called energy management agent, for a shared electric vehicle. The energy management agent reads the latest state of charge of EV in order to plan the charging schedule and then represents possible changes to the plan as flexibilities. Control of charging is executed by interfacing with the EV charging station.

Keywords: demand-side flexibility, energy efficiency, energy management agent, controlled charging, electric vehicles

(3)

ii

Acknowledgements

I want to express my deepest gratitude to the University of Eastern Finland and the Department of Computing for giving me the opportunity to come and study at one of the finest educational institutions in Finland without worrying about paying tuition fees and managing my living expenses during the whole period of my Master's degree.

I am also thankful to my supervisors Prof. Xiao-Zhi Gao and Senior Research Scientist Salmi Tuukka, for all the guidance and valuable comments on my thesis. I want to extend my gratitude to Senior Research Scientist Jussi Kiljander and Research Scien- tist Janne Takalo-Mattila for their continuous support during my time at VTT. I owe special thanks to Research Team Leader Tuomas Paaso and Head of Data-driven So- lutions Janne Järvinen for allowing me to work on my thesis at VTT Technical Re- search Centre of Finland.

Lastly, I appreciate all the support and love of my friends and family, especially my mother.

Oulu, 29 April, 2020 Hafiz Muhammad Qasim

(4)

Abbreviations

BEV Battery electric vehicle CO2 Carbon dioxide

CPP Critical-peak pricing

DR Demand response

DSF Demand-side flexibility

DSM Demand-side flexibility management DSO Distribution system operator

EMA Energy management agent ESS Energy storage system EV Electric vehicle

HEV Hybrid electric vehicle ICE Internal combustion engine

IRENA International Renewable Energy Agency kWh kilo watt-hours

LED Light emitting diode

MQTT Message Queuing Telemetry Transport PHEV Plug-in hybrid electric vehicle

RE Renewable energy

RES Renewable energy sources RTP Real-time pricing

SDG Sustainable development goal SOC State of charge

ToUP Time of use pricing

TSO Transmission system operator U.S. United States

UK United Kingdom

UN United Nation V1G Vehicle to grid

V2B Vehicle to building and building to vehicle V2G Vehicle to grid and grid to vehicle

V2H Vehicle to home and home to vehicle

V2X Vehicle to home, grid or building and home, grid or building to vehicle VRE Variable renewable energy

Notations

𝑅_{𝑓𝑎𝑐𝑡𝑜𝑟} Reducing factor based on fixed values 𝑆𝑂𝐶_𝑓 Final state of charge of EV

𝑆𝑂𝐶_𝑖 Initial state of charge of EV

𝑊ℎ_{𝑏𝑎𝑡𝑡𝑒𝑟𝑦} Capacity of battery pack of EV in kilo watt-hours 𝑊_{𝑐ℎ𝑎𝑟𝑔𝑒𝑟} Power provided by EV charger in watts

𝜂_𝑏 Charging Efficiency

(5)

iv

∆𝑡 Time delta is the difference between two timestamps 𝐸 Energy required to charge battery to full capacity 𝑇 Time required to charge battery to full capacity

(6)

1 INTRODUCTION

The United Nations (UN) agenda 2030 includes 17 clear and targeted sustainable development goals (SDGs) for the world where people live prosperously and safely on a healthy planet [1]. Among all SDGs, the most important one addresses climate change which is also linked to another goal that provides framework for affordable and clean energy. The report published by the UN states that if the world does not cut the growing greenhouse gas emissions immediately, then over the next few decades, global warming is expected to hit 1.5°C. So, in order to curb the increase of global temperature levels, it is necessary to find out the sources of greenhouse gas emissions.

The worldwide CO2 emission level is shown in Figure 1 which represents the growth by different fuel types [2]. The data shows that coal, oil, and gas are the major fuel types responsible for CO2 emissions as the growth has been very significant starting at 5 billion ton per annum CO2 emissions in 1950 to 35 billion ton per annum in 2017.

Figure 1. Carbon dioxide (CO2) emission levels by fuel types [2]

In addition to this, Figure 2 shows worldwide CO2 emissions by different sectors that utilizes the traditional fuel types shown in Figure 1. The data shows that electricity, heat production and transport sectors are responsible for almost 70% of emissions [3].

(9)

The data also shows the steady growth in emissions by these three sectors starting from 50% in 1960 to around 70% in 2014.

Figure 2. Carbon dioxide (CO2) emissions by different sectors [3]

As it is evident by the data that the energy (electricity & heat) and transport sectors are mainly responsible for 70% CO2 emissions worldwide , now the energy sector is tran- sitioning from traditional fossil-fuel based energy production to renewable energy production. Similarly, the transport sector is experiencing major changes from oil based vehicles to electric vehicles. These transitions in energy and transport sector are undergoing throughout the world to achieve the SDGs targets set by the UN under the agenda 2030 [1]. Next, we will examine the transition in energy sector, and then we will explore the transformation of transport sector.

Renewable energy (RE) is defined as the energy produced from continuously replen- ished natural processes such as solar, wind, hydroelectric, and geothermal. The energy produced from renewable energy sources is not constant and it is changing continuously because of unpredictable nature of such natural processes, and this is the reason that the energy produced from renewable energy sources (RESs) is commonly called variable renewable energy (VRE). In contrast to VRE, the energy produced from traditional fossil-fuels is categorized as the non-renewable energy.

(10)

Figure 3 shows the share of electricity produced using renewable and non-renewable energy sources. The left plot shows that non-renewable energy sources are shrinking rapidly as the share has been reduced from 80% to 70% in the last decade and it is projected that the share will further reduce to 50% by 2050. Also, it can be observed from the data that the use of coal is going to be reduced significantly as compared to other non-renewable energy sources. Moreover, the right plot shows the share of electricity produced by different VRE sources. The plot shows that the energy produced from solar and wind sources has been increased from 10% to around 35% in the last decade only and the share is projected to reach around 70% by 2050 [4].

Figure 3. Left side shows share (%) of electricity produced by different sources, right side shows share (%) of electricity produced by different renewable sources [4]

Similar to the transformation of energy sector, the automobile sector is also witnessing a major transition from fuel based vehicles to electric vehicles. The electric vehicles (EVs) are mainly of three types¹ known as hybrid electric vehicles (HEVs), plug-in hybrid electric vehicles (PHEVs), and battery electric vehicles (BEVs). Firstly, the HEVs use both fuel and battery to drive the car, and the battery of the car is charged by the car’s own mechanism known as regenerative braking. Secondly, the PHEVs also use gasoline and battery as a source of power. However, the difference is that in addition to the regenerative braking mechanism, the battery of the PHEVs can also be

(11)

charged by an external electric power source. Lastly, the BEVs are fully electric cars, and they do not have any gasoline engine. The BEVs are mainly charged by externals electric power sources. The BEVs are commonly also called as EVs, and the usage of such vehicles is growing very rapidly.

The long term projection of vehicle sales is shown in Figure 4. The forecast gives an overall picture of the transport sector from 2015 to 2040. The share of EVs is still quite small in 2020 as compared to internal combustion engine (ICE) vehicles. However, the EVs will outpace ICE vehicles by 2040 if the trend of increased sales of EVs con- tinues [5].

Figure 4. Long term projections of vehicle sales worldwide [5]

The International Renewable Energy Agency’s (IRENA) report [6] highlights the transition of energy sector and the transformation of transport sector which are causing stability issues on electric grid systems throughout the world. The usage of VRE sources including but not limited to solar, wind, hydroelectric, and geothermal is growing every year and the share of VRE sources is projected to be increased even in the coming years. The problem with VRE is that the uncertain variable production of energy makes the electric power transmission systems very unpredictable, and the out- come can be catastrophic if the supply and demand of energy are not balanced in the grid. Similarly, the decentralized deployment of VRE generation can affect the stability of grid system because of its intermittent nature. Moreover, the electrification of

(12)

transport and automobile sector is causing additional load on the grid system because the usage of fast chargers to charge the EVs is also growing. Therefore, because of such transitions and problems, the power system needs to be flexible to overcome the challenge of instability. The definition of flexibility is provided in [7]:

“Flexibility is the capability of a power system to cope with the variability and uncertainty that solar and wind energy introduce at different time scales, from the very short to the long term, minimising curtailment of power from these variable renewable energy (VRE) sources and reliably supplying all customer energy demand”

The definition above provides the basis for flexibility at the supply side. However, in order to mitigate the instability of the power system to reduce possible supply and demand differences caused by the transitions, the demand side also needs to have the flexibility the method often termed as demand-side flexibility (DSF) [8]. IRENA proposes following definition for demand side flexibility:

“Demand-side flexibility can be defined as a portion of the demand, in- cluding that coming from the electrification of other energy sectors (i.e., heat or transport via sector coupling), that could be reduced, increased or shifted in a specific period of time to: 1) facilitate the integration of VRE by reshaping load profiles to match VRE generation, 2) reduce peak load and seasonality and 3) reduce electricity generation costs by shifting load from periods with high price of supply to periods with lower prices.” [6]

The definition above provides the basis for the flexibility at the demand side. The definition of demand-side flexibility itself contains some terms like shifting, reducing, or increasing the demand for energy as shown in Figure 5, and it is necessary to describe these terms also. The term “reducing load” means the same as peak clipping, and it is the method in which the load is intentionally decreased during peak-hours. In contrast, the term “increasing load” resembles to valley filling and in this method, the load is increased at off-peak hours. Lastly, in the “load shifting” technique, the load profile is forecasted first and based on that, the load is shifted from peak-hours to off-peak hours.

However, these methods can only be applied to those electric appliances whose con-

(13)

Figure 5. Load shaping methods [9]

1.1 Objectives

The research presented in this thesis was carried out at VTT under the projects FLEX- IMAR and Smart Otaniemi. FLEXIMAR is a co-innovation project funded by Busi- ness Finland, and it has a close collaboration with Smart Otaniemi first pilots. The idea behind the development of the FLEXIMAR project is to create a new type of real-time energy market for demand-side flexibility management. The aim of the energy market is to enable consumers to have a more active role in the energy markets with the help of the energy management agent. The energy management agent will be responsible for automating the demand-side flexibility by controlling those resources which are flexible in terms of energy consumption.

One of the main objectives of this thesis was to evaluate the concept of energy management agent for EV charging. For the purpose of evaluation, another objective was to implement and test the proof-of-concept of an EV controller to verify the suitability of EVs for demand-side flexibility management. The aim of development of EV controller was to plan the load of EV charging consumption, forecast the flexibility potential of the load plan, and the most importantly send control commands to EV charger to follow the load plan if it has been modified by the energy management agent based on energy trade deals. Lastly, the third objective of the thesis was to analyze the effect of delayed charging on total energy consumed by EV. The third objective was important in the sense that if the EV consumes extra energy under the influence of delayed charging, then it is crucial to study the factors which have caused EV to consume

(14)

more energy. The author’s contribution to this work includes the development, integration, and testing of the reference implementation of EV controller for the energy management agent.

1.2 Outline

In this thesis, we discuss different demand-side flexibility management terminologies in section 2.1. Section 2.2 illustrates EV charging as a flexibility resource and the benefits of its flexibility potential. Section 2.3 describes EV charging with demand-side flexibility management approaches.

In chapter 3, we will present VTT’s approach to handle demand-side flexibility. Sec- tion 3.4 deals with the internal architecture of the novel energy management agent which will offer residential consumers to have an active role in new energy markets.

In chapter 4, the reference implementation of the EV controller is given. Section 4.1 presents the experimental setup at VTT. The EV and the infrastructure for its controlled charging are discussed in section 4.1.1. Section 4.2 presents the methods for forecasting the load plan of EV, forecasting its flexibility potential, controlled charging, and an algorithm to monitor the consumption caused by EV charging.

Finally, in section 5.1 of chapter 5, the unit tests, integration tests, and validation tests are presented to verify the logic of the EV controller. Section 5.2 discusses the results obtained, the limitations, and future work to enhance the capability of the EV controller. Chapter 6 provides a summary of the thesis as a conclusion.

(15)

2 FLEXIBILITY MANAGEMENT APPROACHES

This Chapter mainly deals with background knowledge and previous work on demand- side flexibility management. Section 2.1 presents different categories of demand-side flexibility. Section 2.2 discusses EV charging as a flexibility resource, its flexibility potential, flexibility services provided by a single EV, and different smart charging techniques. Lastly, the different setup configurations and approaches for EV charging to support DSM are analysed in section 2.3.

2.1 Demand-side Flexibility Management

Demand-side flexibility management or more simply demand-side management (DSM) is a combination of activities and policies that encourages consumers to harness the potential of DSF in order to help decarbonize the energy production sector.

There are different types of consumers like commercial, industrial, and residential consumers. The scope of this thesis is limited to residential consumers who can utilize DSM. There are many definitions of DSM in the literature, but the most comprehen- sive definition of DSM is provided in [10] as following:

“DSM encompasses systematic activities at the interplay between grid op- erator and electricity consumer aiming at changing the amount and/or timing of the consumer’s use of electricity in order to increase grid per- formance and consumer benefits. DSM activities on the grid operator side involve the assessment of the need for load adjustment and the creation of financial incentives for the consumer, while the consumer reacts to these financial incentives and performs the actual physical load adjustment op- erations.”

There are number of objectives and benefits of DSM compiled by U.S. department of energy as following [11]:

 Reduction of greenhouse gas emissions by helping the transformation of energy sector

 Minimization of fuel imports that will help countries economically

(16)

 Stabilization of grid due to the integration of VRE sources

 Active role of consumers in electricity markets

 Reduction of electricity bills for consumers

 Reduction of investment by utility companies to maintain the stability of grid by reducing the peak load demand

 Increasing the efficiency of electric appliances

 Maintaining the reliability of power distribution system

The DSM is further classified into energy efficiency and demand response as shown in Figure 6.

Figure 6. Demand-side management categories and activities [12]

2.1.1 Types of loads

The prospective electric appliances (loads) intended to use in DSM programs can be classified into two types such as deferrable loads and adjustable loads [13] also known as flexible loads or flexible resources. The deferrable loads are those which can be rescheduled or those which provide the capability of shifting energy consumption to some other time. The examples of deferrable loads in a residential building includes washing machines and electric vehicles. However, the consumer needs to define some constraints or deadlines for deferrable loads in order to make sure that the DSM program is not affecting the daily lives of the consumer. On the other hand, the adjustable loads are those which have different energy consumption levels and the consumer can change the consumption behaviour according to the requirements. The electric appli-

(17)

ances that provides heating or cooling in the residential buildings are the classic examples of adjustable loads. However, the DSM programs should not influence on the comfort level of consumers. This is the reason that the consumers need to define the maximum and minimum consumption levels for adjustable loads to maintain the comfort level.

2.1.2 Energy Efficiency

One of the DSM programs’ objective is to transform the existing less energy efficient appliances present in a household to more energy efficient. Energy Efficiency (EE) is a way to manufacture electric appliances so that they waste less energy to do the same task². It involves the introduction of efficient electric appliances as well as modern building designs through which the usage of energy can be minimized without affecting the comfort levels. The energy efficient appliances includes, for example, LEDs for lighting in the house, electric vehicles to replace traditional fuel based vehicles for transportation, and efficient heat pumps to maintain the temperature inside the building. Similarly, in addition to energy efficient appliances, the architecture designs can include the inclusion of VRE sources to make buildings net zero-carbon emission sources [14].

2.1.3 Demand Response

Demand response (DR) is the class of DSM in which the consumers agree to intentionally change the consumption behaviour of the electric energy based on either real time pricing of electricity or incentives paid by the grid operator. The U.S. Department of Energy defines DR in the following way:

“Demand response is a tariff or program established to motivate changes in electric use by end-use customers in response to changes in the price of electricity over time, or to give incentive payments designed to induce

2 https://www.eesi.org/topics/energy-efficiency/description

(18)

lower electricity use at times of high market prices or when grid reliability is jeopardized.” [11]

DR programs are further categorized into two classes as shown in Figure 6:

 Price-based DR (Nondispatchable)

 Incentive-based DR (Dispatchable) 2.1.4 Price-based DR

The price-based DR programs depend on different methods that electricity utility companies utilize for billing purposes. The traditional electricity utility companies follow two methods to charge consumers for the electric energy they use [15]. One method is flat-rate billing in which the consumer is charged a fixed price per unit (kWh) energy and the utility companies usually charge consumers for the whole month. Another method is tiered-rate billing which has different threshold levels for different prices per unit energy. In the latter method, the rate of electricity can also be different according to the seasons, for example, in winters the prices of electricity can be different as compared to summer.

In new electricity markets [11], the time varying pricing of electricity usage enables the price-based DR programs. There are mainly three most popular time varying billing methods named as time-of-use pricing (ToUP), critical-peak pricing (CPP), and real-time pricing (RTP) [16]. In ToUP, the whole day is divided into different periods based on forecasted load on grid. The day is divided into peak hours when the gap between supply and demand is very less, mid-peak hours when load is relatively mod- erate, and off-peak hours when the supply is greater than demand. In CPP, the utility company decides the price of electricity based on forecasted critical future events which can jeopardize the grid stability, for example, the events like Christmas when the demand of electricity can increase or a time during the year when the demand of heating or cooling is increased. In RTP, the pricing is decided for very short intervals like hours based on real-time load on grid. These different dynamic or time varying pricing can be used to give benefits such as lower cost of electricity to consumers if they agree to participate in the DR programs.

(19)

2.1.5 Incentive-based DR

The incentive based DR programs rely on the consumption behaviour of consumers.

These type of DR programs encourage consumers to modify their load profiles volun- tarily but in some programs, the consumers may receive penalty if they do not follow the signals of utility company. The consumers usually receive some sort of incentives to participate in incentive-based DR programs [11].

There are different ways to participate in incentive-based DR programs generally known as direct load control, interruptible or curtailable, demand bidding, and ancillary services programs [16]. In direct load control DR programs, the utility companies has direct control over electric appliances. However, this method requires the installation of smart controllers that utility companies use to shift the load or turn on/off the appliance intended for participation in the DR program. In interruptible or curtailable program, the utility company cannot directly control the consumption of electric appliance, instead the utility company sends signals to consumers so that they can curtail the consumption between certain levels according to the agreement. In demand bidding programs, the consumers are allowed to bid the amount of load that they can curtail in the wholesale market and if the bid is approved, then the consumer has to follow that accordingly or otherwise the consumer might face penalties. Similarly, the ancillary services programs allow consumers to bid the amount of curtail-able load in a spot market and this load will be considered as operational reserve to help the utility company in the time of emergency or high demand. All these incentive-based DR programs encourages consumers to participate and in return they get some monetary benefits in the form of incentives [13].

2.1.6 Approaches to Change Consumption Pattern

The residential consumers should have the proper infrastructure to utilize one of the DSM programs. There are different approaches available that can facilitate consumers to change their intended consumption pattern according to the different DR methods.

Load curtailment is one of the approaches used for adjustable load types such as heat pumps [17]. Similarly, load shifting is another approach widely discussed in the literature. The load shifting approach is suitable for those type of loads which can be rescheduled or postponed, and electric vehicle is one of such load type that can be used

(20)

in this approach. Both load curtailment and load shifting may affect the comfort levels of a residential consumer which might cause problems in daily routines [18].

There are two possible solutions to address this issue such as integration of energy storage systems (ESSs) and installation of local VRE sources. The ESSs can be used to store electric energy from the grid when the price of electricity is low and this cheap energy can be used during times when there is a need to either shift load or curtail load at the time of peak hours [19]. In the same way, local VRE sources can be used to produce low cost energy causing zero-carbon emissions. In this way, the deferrable load types can be optimized in such a way that there consumption timing matches with the time when the production of VRE is high. The extra energy produced from VRE can be stored in ESSs to use it for adjustable loads [20].

2.2 EV Charging as Flexibility Resource

The growing demand of EVs is adding additional consumption load in a typical household as depicted in Figure 7. Among different electric appliances which are either controllable or uncontrollable in terms of consumption, EVs generally fall under controllable category [21]. In modern decentralized electric transmission systems, such a controlled load will help to make demand side more active in regards to market signals [22], [23]. However, in the absence of DSM and uncontrolled charging of EVs, the maintenance costs of transmission systems would rise to €41 billion till 2050 in UK only, and by implementing smart charging and DR at consumer side can help to save up to €11 to national exchequer [24].

(21)

2.2.1 Flexibility Potential of EVs

In [26], the author has conducted different surveys and based on the results we can conclude that the EVs are one of the potential candidates to be considered as a flexibility resource in a household because most of the cars including EVs are parked and in idle state around 16 hours out of 24 hours. The parking of a car is mainly of two types active-parking and inactive-parking. In the first type, it is the time the car is in the parked state during a journey e.g. if someone goes for shopping and park the car outside shopping mall then we call it as active-parking. For the latter case, if the person takes the car to home after completing a journey or returning from work at the end of the day, then we call it as inactive parking. In the case of EV, we can assume that it is connected to the charger when it is in the inactive-parking state.

Figure 8. Proportion of time EVs spend in active and inactive parking states daily [22]

In Figure 8, we can observe that there is a big window when the EV is in parked state mostly between 16:00 to 8:00 till the next morning. This trend is natural for private EV owners if they are taking cars to work place and then coming back home in the evening. In the default configuration, the charging of an EV starts automatically when someone plug-in the charging cable in the EV. Considering this big time window when EV is available for charging, we can simply modify the default charging behaviour of EV so that it does not start charging automatically and then we can shift the required electricity load from peak hours to off peak hours [22].

(22)

2.2.2 Factors Affecting Flexibility

In Figure 9, we can visualize the possible factors that might potentially affect the amount of flexibility offered by a single EV. First of all, if we consider the charging location, the maximum flexibility would be available if it is charged at home because the EV is parked and plugged-in most of the time when it is at home [26] as discussed in the previous section. Similarly, the amount of flexibility depends on the time (e.g.

day, evening or night) when the EV gets charged. If it is charged during a day time and peak hours, then the amount of flexibility would be more as compared to the scenario when it gets charged in off-peak hours as discussed in the previous section. In addition to this, the charging technology can affect the possible flexibility of an individual EV, the main reason is that the charging speed of home is always slow charger (e.g. 3.7-11 kW) as compared to the public charging stations where the charging speed is normally quite high (e.g. 22-140 kW) and such charging stations are growing at rapid pace [27] [28]. So, the fast charging can provide more flexibility as compared to the slow charging.

Moreover, the infrastructure V2X (like V2G or vehicle to grid etc.) can also harness the flexibility and the energy available in EV’s battery can be used for peak shaving which can eventually support the grid during peak load time [29]. Besides this, the individual driving behaviour can also affect the amount of flexibility available. For example, if the usage pattern of an EV is predictable and if its mobility time is the same every day, then this behaviour can possibly increase the flexibility. In the other case, if someone is using EV for commercial purposes e.g. as a taxi, then the usage behaviour is somehow unpredictable, so the flexibility would be limited in this case.

Lastly, those EVs which have bigger batteries can provide more flexibility [30]. The reason is that those larger battery packs would take more time to get charged as compared to smaller battery packs which take less time. So, having more charging time would provide a larger window and ultimately the charging load can be shifted to a later time. In a nutshell, the larger battery packs can provide more stability to grid in V2G model.

(23)

Time when Charged

Charging Technology

Desired and Current SOC Driving Pattern Charging Location

Factors Affecting Flexibility

Home Office

Public Charging Station

Day Night Evening

Slow Fast

V2X Enabled

Personal vehicles

Taxis

Battery Capacity

Figure 9. Factors influencing amount of flexibility from a single EV

2.2.3 Flexibility Services

Figure 10 illustrates the flexibility services that an EV can provide to a power system.

First of all, the flexibility offered by an EV can help to integrate renewable energy produced locally because with controlled system we can stop energy flow from grid and divert the energy produced by renewables to the EV charging system. Moreover, in V2B/H (i.e. vehicle to building or home) [31], the energy stored in the EV battery pack can provide backup power to the home.

Also, the available flexibility can help stabilize the frequency and control the voltage at DSO (distribution system operator) and TSO (transmission system operator) levels [30]. Additionally, the flexibility service provided by EV can help for peak-shaving and valley filling (as shown in Figure 11) of demand curve by adjusting the charging speed to slow and shifting the charging to night time and in the meantime the renewables energy can help to charge the car with the energy produced locally that will ultimately lower the cost of electricity.

(24)

Local Flexibility

System Flexibility Flexibility Services

Peak Shaving Portfolio Management Behind the Meter

DSO

Wholesale Market

TSO

Ancillary Services

Frequency Control Voltage Control

Capacity Management Renewable Integration

Backup Power

Figure 10. Flexibility services provided by EV to power system

Figure 11. Example of controlled and uncontrolled EV charging [32]

2.2.4 EV Charging Types

Figure 12 illustrates different smart charging schemes and their effect on flexibility offered by EVs. Firstly, the time of use policy is the most basic uncontrolled charging policy and it provides the least flexibility, because in this scheme the tariff of electricity depends on the peak and off-peak hours and these hours can be different for different demographics. The time of use charging policy can be used for peak-shaving during high load at the grid [33]. As compared to time of use policy, the basic on and off control with scheduled charging can provide better flexibility level and this policy can mainly help for congestion management at grid level. In this scheme, the users know

(25)

mobile apps provided by car manufactures, they can simply schedule the charging time.

Additionally, in unidirectional controlled policy like V1G, the charging is controlled at user side but it also depends on the real time price of energy as well as on market signals. This type of charging policy can be used to for frequency and voltage control and also for ancillary services as discussed in section 5.1.3. The V2X policies like vehicle to grid, vehicle to home or vehicle to building are quite new and still in the testing phase in most of the countries. The flexibility offered by V2X policies is high as compared to V1G, basic on/off control and time of use charging policy as demon- strated in Figure 12. In V2X control, the EV is charged at off-peak hours and then can support the power system at local level and grid level by discharging the battery to provide backup power [29]. Similarly, if V2X is combined with dynamic pricing, then this policy would provide the highest flexibility possible from a single EV. The bidi- rectional control provided by this policy can help for congestion management, as a backup power, load curtailment and peak-shaving [31].

Figure 12. Effect of different smart charging schemes on flexibility [28]

2.3 EV Charging with DSM

The EV charging as a potential flexible resource is the main focus of this thesis. So, it is important to highlight different setup configurations discussed in the literature.

There are mainly three setup configurations to support DSM. The Figure 13 shows the most appropriate configuration for DSM. In Table 1, different methods are compared that support EV charging with DSM.

(26)

Type 1. EV charging in the absence of ESSs and VREs Type 2. EV charging with VREs

Type 3. EV charging in the presence of both ESSs and VREs

Figure 13. EV charging system configuration to support DSM

Table 1. Comparison of different DSM methods to support EV charging in different setups Method

Reference Objective Local VRE

Generation ESS Pricing Grid

Interaction

[34] Load balancing   Incentive-based V1G

[35]

Load balancing + Min cost of

electricity

  RTP V1G

[36] Min cost of elec-

tricity   RTP V1G

tricity   ToUP V1G

[38] Load balancing + Min cost of

electricity

  RTP V2G

tricity   RTP V2G

[40]

Load balancing + Min cost of

electricity

  RTP V2G

(27)

In type 1, the electric energy to charge EV’s battery comes completely from the grid without the support of ESSs and local VREs. This type of charging can cause grid instability if the combined irregular and uncontrolled charging of thousands of EVs is considered in a demographic area. The irregular charging behaviour of EVs can be avoided with DSM if the consumer has the infrastructure to automate the charging with DR methods. Johal et. al. has analysed the effect of large proportion of EVs pen- etrated in the grid systems [34]. Based on the analysis, a load shaping strategy has been proposed to defer the deferrable loads in a household including EV. The proposed strategy works when a transformer exceeds its load capacity and then sends a signal to nearby house which has enrolled in the DR program. The simulation results show that the strategy is useful for controllable loads to defer the energy consumption while satisfying the constraints set by the consumer. Chen et. al. has proposed a method for residential consumers to automatically control different load types in a household [35].

The proposed method uses stochastic optimization and robust optimization to plan and control both deferrable and interruptible loads and at the same time protecting consumers against the financial risks that might occur because of RTP. The results prove stochastic optimization more promising, even though its computational complexity is higher as compared to robust optimization in terms of reducing total electricity cost for a consumer. Yao et. al. proposes a real-time charging mechanism for controlled EV charging by satisfying the constraints, such as deadline to charge the battery to a specific level, set by the consumer [36]. The proposed strategy is based on dynamic pricing of electricity and it controls the charging, i.e. turn on/off with constant and maximum current, according to the DR signals received from the utility company. The results obtained from a charging management system shows EV as a potential flexible resource to participate in DSM programs.

In type 2, there are local VRE sources, for example solar or wind, installed to support the charging of EV. Wi et al. devised an algorithm based on ToUP for smart charging of EVs in a residential building in the presence of VRE specially PV. The method works by first predicting the PV energy production and EV charging demand based on time series data, and then optimize the charging of EV for the time when there is more VRE production. The proposed method is effective to integrated locally produced VRE by planning the EV charging based on its SOC, charging speed, electricity prices, and end-user’s constraints. Chen et. al. proposes a strategy based on dynamic pricing to

(28)

automate the charging of EVs based on DR signals received from the grid operator [38]. The presented strategy schedules EV charging in such a manner that a major part of EV battery is charged with locally produced solar energy while some part of the battery is charged with the energy purchased from grid. The results obtained with simulations show that the strategy is useful to integrate local VRE and at the same time it reduces load on grid while the cost of EV charging is also drastically reduced.

In type 3, the EV charging is supported by both local VRE sources and ESSs. A home energy management system is proposed in [39] to control and curtail different flexible resources in a household including EV in the presence of both VRE and ESS. The proposed system supports two way transmission of electric energy i.e. from grid to home and vice versa. The energy in ESS is stored from two sources i.e. energy produced from VRE as well as from the grid when the price of electricity is cheap. The energy stored in ESS is then sold to grid operator based on DR signals. The consumption profiles of flexible resources are forecasted with the help of a stochastic optimization approach. The flexible loads consume energy according to their schedules. The results show that this approach is useful to decrease cost of electricity. Erdinc et. al.

has developed a home energy management system with mixed-integer linear program- ming to automate the DR in both grid to vehicle and vehicle to grid i.e. V2G transmission of electricity [19]. The proposed strategy shows significant results in simulations for two way bidding and selling of electric energy. The RTP was used to purchase energy from the grid and the energy was sold back on flat prices of electricity. The results show that by utilizing this home energy management system completely, the electricity prices for residential consumers may be decreased significantly by 65%

overall. Atzeni et. al. proposes a non-cooperative and cooperative optimization model to integrate distributed generation of VRE and ESS to store the locally produce energy to effectively minimize the cost of electricity for EV charging and lower the demand curve on grid [40]. The model supports two way transmission of electricity in DR methods by storing the energy in ESS when the price of electricity is lower in the market. The proposed scheme is applicable to both residential and commercial consumers including small businesses.

(29)

3 VTT ENERGY MANAGEMENT AGENT

In this chapter, we discuss the architecture of the energy management agent (EMA).

Starting with Section 3.1, we describe the purpose of the EMA and discuss the problem we are trying to solve with it. Then, in Section 3.2, we describe the context view of the EMA, as illustrated in Figure 14. Subsequently, in Section 3.3, we discuss the terms or definitions used in the EMA architecture. Lastly, in Section 3.4, we discuss the internal architecture of the EMA, as illustrated in Figure 15.

3.1 Introduction

The EMA is an entity or more simply a software that automates DSM of electric energy on behalf of a consumer. The main objective of the EMA is to maximize the reward by shifting the energy consumption of a flexible resource while satisfying the consumer’s constraints. In the most simple case, the reward is money, but it can also include environmental aspects depending on what a specific consumer wants. The EMA is mainly responsible for controlling and monitoring a single metering point.

The EMA inherits its novelty from FLEXIMAR which is a new type of energy market place that makes it feasible for small electricity operators including individual consumers to participate in DSM programs. The existing markets have been developed to support only large players in the energy markets who have more flexibility potential to trade and it is very costly for individual consumers who have small amount of flexibilities to participate in such markets. The EMA is responsible to send load plans of flexible resources and trade deals in the form of flexibilities on behalf of consumers to this type of new energy markets.

Moreover, the existing energy management systems use top-down approach and flexible resources are directly controlled based on DR signals received from utility companies. However, the approach presented in this thesis makes use of bottom-up approach and each flexible resource has a separate controller which is responsible for planning load plans, representing possible changes to the load plan in the form of flexibilities, and then controlling the resource accordingly if EMA trades any flexibility in

(30)

the market. In other words, this approach does not let utility companies to have direct control on consumption profiles of flexible resources.

By default, any flexible resource starts consuming electric energy as soon as it is plugged-in to a power source. This kind of consumption behaviour is not ideal, and it can cause grid imbalance at peak-hours if we consider the combined behaviour of all such flexible resources in a demographic area This type of consumption behaviour is one of the reasons for grid imbalance; other reasons include the integration of renewable energy sources in the grid, the decentralized deployment of renewable energy generation, and the electrification of transportation and the automobile sector as discussed in the first chapter of this thesis. So, to tackle this problem, the EMA will offer DSM and will enable the consumer to have an active role in the energy markets by providing the opportunity to shift the load of flexible resources to off-peak hours while satisfying the constraints provided by the consumer to the EMA. At the same time, the consumer will get some reward from the energy market. Lastly, the EMA will enable the integration of locally produced renewable energy to use it for consumption of flexible resources.

3.2 Context View of EMA

Contextual view of the EMA is shown in Figure 14. The diagram represents the EMA as a black-box in which it is getting inputs and sending outputs to different external components. However, we are not concerned with the internal working of the EMA in this context.

(31)

In Figure 14, it is shown that the end-user is providing constraints regarding the flexible resources, and the EMA is sending feedback to the end-user as described in Section 3.3. Similarly, the EMA is asking the load plans from the flexible resources and then sending control commands to follow the modified load plan based on the trades made in the energy market. Likewise, the EMA is sending load plans and trade offers to the energy market according to the market window and the market resolution, and in return, the market is sending confirmations on trade offers to the EMA. Furthermore, the EMA is getting the base-load of a metering point, local energy production from renewables, environmental data such as temperature and humidity levels from sensors installed in the apartment, and the weather forecast from meteorological services. The weather forecast and the data from environmental sensors are useful to predict renewable energy production. It is important to note here that the type of energy market discussed exists only at the research phase, and it is not a mandatory entity in the EMA.

Also, the experimental setup, which is described in chapter 4, does not utilize such energy markets.

3.3 Definitions Used in EMA

Metering Point

A metering point is generally a point used for billing of electric energy and it can be a whole building, apartment or even a single electric appliance. The EMA can monitor the base consumption of a metering point.

Flexible Resource

An electric appliance is a flexible resource if it can change its energy consumption. An electric vehicle and a heat pump in a household are potential electric appliances, among many others, to be examined as flexible resources [30].

End-User

The end-user is basically a consumer who has installed the EMA in the house. The end-user is giving input to the EMA as constraints and getting some feedback as discussed next.

(32)

Constraints

The constraints are inputs given to the EMA according to a consumer’s preferences.

In the case of an electric vehicle, the constraint can be a scheduled time when the consumer expects the EV to be fully charged, and it is ready for a drive. In the case of a heat pump, the constraints are maximum and minimum temperatures that a consumer expects to experience in the house.

Feedback

The feedback is output returned by the EMA to a consumer. It can be some information about the reward earned so far. Also, it can be a graph representing some statistical data to show when and how the EMA is shifting the consumption of the flexible resources connected to it.

Load Plan

The load plan is a flexible resource’s intended energy consumption behaviour divided according to the market resolution as intervals of fixed duration.

Environmental Sensors

The house or apartment building have different sensors installed to monitor the temperature, humidity, air pressure, CO2 levels and pressure difference between outer and inner environments. The sensors are sending each minute environmental measure- ments to the EMA. The data collected through these sensors is required to maintain the apartment heating. This data, combined with the data provided by meteorological services, is also being used to forecast renewable energy production.

Meteorological Services

The meteorological services is providing weather forecast which is useful to predict renewable energy production as well as outdoor temperature of the apartment because the outdoor temperature can influence the apartment heating that the EMA is maintaining.

Flexibility Potential

The flexibility potential is the ability to change the consumption behaviour of a flexible resource. The flexibility potential is of two types named as down flexibility and up

(33)

flexibility. Down (resp. up) flexibility means the ability to decrease (resp. increase) energy consumption for a specific time period.

Market Window

A market window is a timeframe or a period between two different times during which it is possible to trade flexibility offered by a flexible resource in an energy market. The preferred period is 24 hours to align with the current markets because the day ahead forecasting of electricity prices is common in current markets. However, the market window is configurable, and it is possible to configure the period accordingly.

Market Resolution

The whole market window is further resampled as shorter intervals, which can be of 15 or 60 minutes of duration according to the market conditions. Similar to the market window, the market resolution is also configurable and it is possible to change the sampling frequency of intervals.

Renewable Energy Sources

The EMA can also forecast the on-site production of energy from renewable energy sources (RESs) such as wind, solar, and hydrogen fuel cells. The renewable energy produced locally can be stored in batteries, and the EMA can supply this energy to flexible resources to decrease the billing of energy further.

Energy Market

The energy market, for example FLEXIMAR, is a platform where the EMA can trade flexibilities offered by flexible resources. The trading agent, described in Section 3.4.3, connects external market places to the EMA, as shown in Figure 15. The trading agent can send trade offers to the market, and then the market place can confirm if the trade offer gets accepted. This kind of energy market does not exist as of now and it is still at research level [41].

(34)

3.4 Internal Architecture of EMA

Internal architecture of the EMA is shown in Figure 15. Internally, it is composed of three components named as energy planner, trading agent and a separate resource controller for a different flexible resource. The trading agent is responsible to interact with energy markets and retailers, and the resource controller is responsible to interact with flexible resources. Additionally, the energy planner has the major responsibility as all the components are collaborating with it.

3.4.1 Internal Interfaces Trading agent – Energy planner

The interface between the trading agent and the energy planner is MQTT interface.

The energy planner sends information to the trading agent about flexibility potential, energy consumption, and load plans through the topics ‘flexibilities’, ‘consumption’, and ‘plan’ respectively. The energy consumption and load plan data is used for visualization purposes. Whereas, the trading agent sends information about confirmed trades to the energy planner through the topic ‘trade’.

Flexibility Topic

The energy planner sends new flexibilities to the trading agent using topic “[METER- ING_POINT]/[ID]/flexibilities”. The example of the message is as follows:

{

"1573641083000": {

"flexibilities": [50, 20, -40, -50], "minimum_profit": [0.0, 340, 278],

"expiration_time": [1573641083000, 1573641083000, 1573641083000]

} }

Here the key "1573641083000" is epoch time in milliseconds, "flexibilities" are the energy values in Wh, "minimum_profit" is the minimum price that we should receive when a trade-deal is made, and "expiration_time" is the time when the deal should be closed, or otherwise it will be discarded.

(35)

Consumption Topic

The energy planner sends the realized energy consumption data to the trading agent using topic “[METERING_POINT]/[ID]/consumption”. The example of the message is as follows:

{

"1573641083000": { "load": 100 },

"1573641183000": { "load": 200 }

}

Here the key "1573641083000" is epoch time in milliseconds, and the "load" represents amount of energy consumed in Wh.

Plan Topic

The energy planner sends the load plan to the trading agent using topic “[METER- ING_POINT]/[ID]/plan”. The format of the message is as follows:

{

"<timestamp1>": {"load" : <value1>},

"<timestamp2>": {"load" : <value2>}

}

Here the key "<timestamp1>" is epoch time in milliseconds, and the "load" represents amount of energy going to be consumed represented in Wh.

Trade Topic

The trading agent can send information about confirmed trades to the energy planner using topic “[METERING_POINT]/[ID]/trade” through the MQTT, and here ME- TERING_POINT is the metering point that the EMA is monitoring and ID is the instance of that metering point. In reality, there should be only one ID for a single metering point but we can create more IDs for testing and simulation purposes. The format of the message is as follows:

{

"<timestamp>": <value>

}

(36)

Here the key <timestamp> is epoch time in milliseconds and value is the amount of energy measured in Wh (Watt-hours). The value can also be negative depending on up or down flexibilities.

Energy Planner – Controller

The energy planner and resource controllers have a unidirectional association relation- ship and it means that both the energy planner and resource controller are related to each other but only the energy planner knows about the existence of the relation. The energy planner requires four necessary interfaces for load forecasting, flexibility forecasting, controlling and monitoring energy consumption from each resource controller.

In chapter 4, we will discuss these four interfaces in detail.

Figure 15. Architecture view of energy management agent

(37)

3.4.2 Energy Planner

The energy planner is the essential component of energy management agent as the other components of the whole system are collaborating with it as illustrated in Figure 15. Initially, before the start of the next market window, it asks load plans for the next market window from all the flexible resources linked to the EMA. The aggregate of load plans received from each flexible resource is the base-load plan of a metering point for the given market window. Secondly, it asks the up and down flexibility potential compared to the current load plan from all the flexible resources for the given market window. Similarly, the aggregate of flexibility potential from all the flexible resources is the flexibility potential of a metering point. Thirdly, the energy planner publishes both the aggregated load plan and the flexibility potential to the trading agent so that the trading agent can make trade offers in the energy markets. After that, the planner receives information about proposed trades from the trading agent and decides either to accept or to discard the proposed trades. In the first case, if the proposed trade offers are accepted, then the energy planner forwards the modified load plan to the flexible resources to follow it accordingly. In the meanwhile, the planner is forecasting the on-site renewable energy production, the total energy consumption of the metering point as well as the consumption of individual flexible resource.

3.4.3 Trading Agent

The trading agent has three main responsibilities described below:

1. Providing interface to connect with external market places 2. Communicate with energy planner internally in EMA 3. Provide a REST API to visualize the data

The trading agent is the component that connects external markets to the EMA. The main goal of the trading agent is to optimize reward given flexibility potential received from the energy planner and market conditions such as electricity pricing received from the market. Furthermore, the trading agent can send and receive trade offers to and from the market, and when the deal on a certain trade-offer finalizes, the trading agent receives confirmation from the market and then sends it to the energy planner so

(38)

that the energy planner modify the base load plan and then forwards the modified load plan to the flexible resource to follow it.

Moreover, the trading agent is providing a REST API to visualize the data in the user- interface. Table 2 gives information about different end-points of the API that the UI of the EMA can utilize. The trading agent uses energy consumption and load plan data, provided by the energy planner, for visualization purposes only.

Table 2. REST API end-points provided by trading agent for EMA UI

Type Path Params Description

GET /trading_place?instance_id=[id] None

Returns trading_place settings ({enabled: [True/False]}) for

given building instance.

POST /trading_place?instance_id=[id] {enabled: [True/False]}

Sets trading place configurations for given building instance. Instance id is required.)

GET /trader?instance_id=[id] None

Returns automatic trader settings ({enabled: [True/False]})

for given building instance.

POST /trader?instance_id=[id] {enabled: [True/False]} Sets automatic trader configurations. Instance id is required.)

GET /plans?instance_id=[id] None Returns plans for the given

building instance.

GET /flexibilities?instance_id=[id] None Returns flexibilities for the given building instance.

GET /consumption/?in-

stance_id=[id] None Returns consumptions for the

given building instance.

GET /virtual_apartments None Returns all the virtual instances

of the building.

GET /deals?instance_id=[id] None

Returns all the deals of the given instance id of the build-

ing.

GET /asks?instance_id=[id] None Returns all the ask orders.

POST /asks?instance_id=[id] {starttime, expirationtime,

load, price} Posts a new ask order.

GET /bids?instance_id=[id] None Returns all the bid orders.

POST /bids?instance_id=[id] {starttime, expirationtime,

load, price} Posts a new bid order.

3.4.4 Controller

The resource controller connects a flexible resource to EMA. Initially, the resource

(39)

The forecasting of load plan depends on each flexible resource. Then, the resource controller forecasts the flexibility potential of a flexible resource compared to the base load plan forecasted in the first step. After that, the resource controller sends control commands to flexible resource so that it can follow its intended load plan provided by the energy planner of the EMA. Finally, the resource controller is also responsible to monitor the actual energy consumption and provide this information to energy planner.

(40)

4 IMPLEMENTATION OF EV CONTROLLER

This chapter presents the proof-of-concept implementation of EV controller. Section 4.1 illustrates the experimental setup and its components at VTT. Section 4.1.1 provides the EV charging infrastructure. Lastly, section 4.2 discusses the implementation details of EV controller including charging profile of EV.

4.1 VTT Experimental Setup

The components of the experimental setup are shown in Figure 16 that gives high level representation of the whole setup. The setup consists of an apartment with different sensors installed in it. The setup also contains an electric vehicle and a charging post.

Apart from the apartment and electric vehicle, the setup consists of a windmill and solar panels to produce renewable energy locally. The renewable energy produced locally can be stored in a battery so that it can be used later-on to supply energy to the apartment and the electric vehicle. It is important to note here that the scope of this thesis is limited to the electric vehicle as a flexible resource. However, it is necessary to describe the whole experimental setup. The given setup resembles to type 3 configuration which is described in section 2.3. Although, the setup is capable of only V1G i.e. one way transmission from grid to vehicle yet.

Figure 16. Experimental Setup at VTT

(41)

The characteristics of the whole experimental setup are given in Tables 3-10.

Table 3. Solar panel characteristics

Solar panels Attributes Power output 7.2 kW

Panel area ~20 m²

Measurement Accurate production measurement

Table 4. Smart grid server at VTT

Smart grid server Attributes

Data storage Measurement data from solar panels, windmill, charging post, EV, apartment sensors,

Server For EMA

Table 5. Windmill

Windmill Attributes

Generator type Asynchronous wind turbine

Diameter 6 meter

Power output Nominal output 5.5 kW at 9 m/s Minimum wind speed 4 m/s

Maximum wind speed 25 m/s

Measurement Accurate production measurement

Table 6. Energy storage system for storing locally produced energy

Energy storage system Attributes

Capacity 58 kWh

Charging From grid or local renewable production

Supply To components of the setup

Table 7. VTT sensor equipped apartment

Apartment Attributes

Sensors

Temperature, humidity, air pressure, pressure difference between outer and inner environments, and CO2 levels

Controllable appliances Heat pump, lighting, sauna, oven

Table 8. Shared EV at V`TT

EV Attributes

Model Nissan Leaf 2016

Battery capacity 30 kWh

Range ~100 km

Power 47 kW or 63 hp

Torque 180 Nm

Weight 1080 kg

Slow charging ~11 hours Fast charging ~3.5 hours

State of charge Remotely collected in every 10 minutes

(42)

Table 9. Charging post for EV charging

Charging station Attributes

Type Ensto wall charger

Max power output 22 kW

Max current 32 A

Min current 6 A

Control Remotely controllable

Charging Possible with RFID identification 4.1.1 EV Charging Infrastructure

A more detailed experimental setup for EV charging is shown in Figure 17. The EV controller is the component of EMA (described in previous chapter), and this controller is linked to three cloud services represented in Figure 17 as Nissan server, Ensto charger server, and smart grid server.

The EV controller is utilizing the Nissan server to fetch the latest state of charge (SOC) of EV. In this scenario, the SOC is useful to forecast the load plan of EV. Also, the flexibility potential (down and up flexibility) is forecasted based on this load plan. The load plan and flexibility potential are required by the energy planner and these are published through the trading agent into the market.

Figure 17. EV charging infrastructure

Demand-side Flexibility Management for Electric Vehicle Charging