A Distributed ICT Architecture for Continuous Frequency Control

Christian Giovanelli

, Olli Kilkki

, Antti Alah

aiv

, Ilkka Seilonen

Matti Lehtonen

and Valeriy Vyatkin

1,2

School of Electrical Engineering, Department of Electrical Engineering and Automation,

Aalto University, 02150 Espoo, Finland

SRT, Lule

a University of Technology, Lule

a, Sweden

Keywords:

Smart Grid, Demand Response, Distributed ICT Architecture, Continuous Frequency Control.

Abstract:

The active participation of consumers in frequency control can mitigate the negative effects of variable renew-

able generation in a power system. This study aims at designing a distributed information and communication

technology architecture for automated demand response. The distributed architecture enables a set of con-

sumers to perform frequency control while being coordinated by an aggregator. Moreover, decision-making

algorithms are designed to enable the demand response to participate in frequency control and to provide re-

quired reserves. An asynchronous message-oridented middleware is utilized to interface the consumers with

the aggregator. In addition, the communication logic between the actors is deﬁned. The distributed architec-

ture is then evaluated through the implementation of a prototype application. Simulated results show that the

designed architecture can be utilized for frequency control in automated demand response.

1 INTRODUCTION

In electrical power systems, the supply and demand

are required to be continuously in balance. Any im-

balance either increases or decreases the system fre-

quency which may endanger the system security and

the continuity of electricity supply. In order to co-

ordinate these frequency excursions, power systems

utilize different frequency control processes. In the

Nordic power system, the ﬁrst control process to react

to the frequency oscillations is called frequency con-

tainment process. Its tasks are to maintain the balance

in time frame of seconds and to stabilize the system

frequency after disturbances (Entsoe, 2015). Thus,

the process has a crucial role in the frequency coordi-

nation. The process activates two types of reserves

in the considered Nordic power system: frequency

containment reserves for normal operation (FCR-N),

and frequency containment reserves for disturbance

(FCR-D) (Fingrid, 2016).

Due to the expansion of variable renewable in the

power supply, the frequency containment is becom-

ing increasingly challenging. For this reason, it is

important to identify new solutions for the frequency

containment, which are not relying only on the sup-

ply. One such a solution is to employ electricity

consumption in the frequency containment process

(Siano, 2014). In particular, the Demand Response

(DR) could contribute to provide the required reserves

by shaping the electricity consumption (Short et al.,

2007). In fact, a large number of consumers provided

with intelligent electronic devices (IED), capable of

shaping and shifting their load, could take part in the

DR. However, the involvement of consumers in the

frequency control process would require an increas-

ing quantity of information exchange, which would

demand improvements to the current information and

communication technologies (ICT) infrastructure of

the power grid.

ICT systems will be among the key elements that

will drive the enhancement of the future power grid

(Gungor et al., 2013). ICT systems could provide

the necessary communication infrastructure, informa-

tion technologies, and applications for an improved

power delivery. Among others, ICT systems enabling

consumers to participate in DR would be needed to

ensure better coordination and decision-making for

the distributed resources. These architectures are re-

quired to be reliable, ﬂexible, and scalable. Thus,

internet-based architectures, such as service-oriented

(Grijalva and Tariq, 2011) and cloud-based (Kim

et al., 2010), have been identiﬁed as possible driver

technologies for modernizing the current power grid.

The aim of this paper is to design a distributed

ICT architecture and an application logic capable of

performing automated demand response (ADR) for

Giovanelli, C., Kilkki, O., Alahäivälä, A., Seilonen, I., Lehtonen, M. and Vyatkin, V.

A Distributed ICT Architecture for Continuous Frequency Control.

DOI: 10.5220/0006242201350142

In Proceedings of the 6th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS 2017), pages 135-142

ISBN: 978-989-758-241-7

135

continuous frequency control (CFC) of FCR-N. The

architecture consist of an aggregator and a set of con-

sumers that are communicating in order to perform

ADR for CFC. The communication is based on an

asynchronous Message-oriented Middleware (MOM)

that interfaces all the participants of the ADR. In ad-

dition, the aggregator utilizes a speciﬁcally designed

algorithm to allocate the DR instructions that con-

sumers have to adhere based on the frequency vari-

ations in the grid. To evaluate the ICT architecture

and the MOM, a set of scenarios are simulated. These

simulations verify the feasibility of the architecture

and validate that the MOM can be utilized for CFC.

The remainder of the paper is organized as fol-

lows. Section 2 introduces the related work. Then,

Section 3 presents the requirements of the ADR sys-

tem, after which Section IV details the design of the

ADR architecture, focusing on both decision-making

algorithms and communication logic. In Section 5,

simulation results are presented, and in Section 6 the

discussion. The ﬁnal section then draws conclusions.

2 RELATED WORK

The control of FCR-N is conventionally implemented

in a decentralized manner by a droop control, i.e.,

generation plants adjust their output in proportion to

the system frequency. However, the growing pen-

etration of variable renewable generation, which is

replacing the conventional generation (Lalor et al.,

2005) and the increase of reserve requirements (Ha-

lamay et al., 2011), are threatening the stability of the

electrical grid. New strategies for CFC could consider

the demand-side for balancing the system frequency

(Short et al., 2007). Consumers could be involved

in ADR for CFC by controlling domestic appliances

based on the system frequency (Samarakoon et al.,

2012). Examples of such appliances are HVAC, re-

frigerators (Angeli and Kountouriotis, 2012), energy

storage devices (Megel et al., 2013), and electrical ve-

hicles (EVs) (Masuta and Yokoyama, 2012).

Control algorithms are required to provide

decision-making capabilities for enabling the partic-

ipation of the consumers in the DR. Various control

strategies have been designed for performing CFC

(Xu et al., 2011). Control strategies for CFC that

do not require a two-way communication and a cen-

tral coordination have been proposed (Molina-Garcia

et al., 2011). On the other hand, decision-making al-

gorithms have been designed in which an aggregator

coordinates the consumers during their participation

in the DR for CFC (Pourmousavi and Nehrir, 2012),

(Vrettos et al., 2014). The advantage of the latter solu-

tions consists of the possibility to coordinate the pro-

vided reserves of the ADR system. This advantage

becomes important in the case where the provided re-

serves have to be agreed beforehand, during DR plan-

ning phases such as day-ahead and intra-day market.

However, the aggregated coordination of consumers

require the development of new ICT technologies.

The future smart grid will be interdependent with

its ICT architecture (Yan et al., 2013). Currently, ICT

architectures for smart grids are in an initial stage of

development (Zaballos et al., 2011). In fact, vari-

ous technologies need to be developed in order to en-

hance the current ICT architecture (Fang et al., 2012).

Among others, service-oriented middleware for smart

grid have been proposed to have a central role in

the ICT architectures (Rodrigues, 2013). Service-

oriented middleware acts as a broker among many

heterogeneous entities which are communicating in

the smart grid. Considering ADR for CFC, communi-

cation represents an important component for mak-

ing various heterogeneous entities collaborating to-

gether (Kilkki et al., 2014). Nevertheless, ICT archi-

tectures enabling CFC should be further explored to

enhance the distributed control and the communica-

tion between consumers and service providers.

3 REQUIREMENTS

The design of the ADR system deﬁned in this study

needs to fulﬁll a set of requirements. The main re-

quirement is that the ADR architecture is capable of

performing CFC. The ADR architecture is required

to implement one aggregator and a set of consumers

that are collaborating together in order to execute

CFC. In addition, in order to deﬁne an ADR architec-

ture for CFC, several other requirements have to be

met. The following requirements can be categorized

in electrotechnical requirements, communication re-

quirements and system requirements.

To participate in CFC and provide FCR-N, the

ADR system should fulﬁll certain electrotechnical re-

quirements established by the transmission system

operator (TSO). This is to say that the ADR system

is expected to perform according to these electrotech-

nical requirements. In the Nordic power system, for

the case of FCR-N , the reserve should be fully avail-

able within three minutes after a step change of +/-

0.1 Hz from the nominal frequency, while a maxi-

mum dead-band of +/- 0.05 Hz is allowed (Fingrid,

2016). The reserves are not expected to react if the

frequency is within the dead-band. The amount of

provided reserve is the activated power after three

minutes against the step change. In this study, the

SMARTGREENS 2017 - 6th International Conference on Smart Cities and Green ICT Systems

136

minimum provided reserve size (0.1 MW) by a re-

source owner is not considered, nor any possible fre-

quency measurement errors.

Communication requirements have been identi-

ﬁed for the ADR system for CFC. The communica-

tion between the actors needs to allow the ADR sys-

tem to perform CFC. Due to the electrical requirement

of having the reserve active within three minutes, the

communication should be able to work under these

near real-time constraints. In addition, the communi-

cation should support interoperability between the ac-

tors, which should interact by means of a predeﬁned

format of data exchange. The communication should

also prove to be scalable up to several thousands of

consumers. Furthermore, the communication should

be reliable and secure.

The system requirements deﬁne the composition

of the system. The ADR system is composed of two

main actors: an aggregator and a set of consumers.

The aggregator acts as a mediator between consumers

and the utility operator (Gkatzikis et al., 2013). By

interfacing with the consumers participating in the

ADR, the aggregator has an obvious advantage to

have the complete overview of the entire system. For

this reason, the aggregator is required to implement

a decision-making algorithm capable of coordinating

the policies and allocating the reactions that each con-

sumer has to observe for performing CFC.

Consumers represent the second set of actors in

the ADR system. The ADR system is required to have

several thousands of consumers which take part in the

CFC by reacting to the frequency deviations from the

nominal frequency. Each consumer has to own IED

devices that can be used for CFC. Examples of such

IED devices are refrigerators, HVACs and EVs. In

addition, consumers are required to be equipped with

an energy management system (EMS) (Siano, 2014).

The EMS is a IED gateway device capable of inter-

facing with the various IED devices in the home area

network (HAN). Another functionality of the EMS is

to interface the household with the cloud-based sys-

tem for ADR via an IP-based protocol. Moreover, for

executing CFC, the EMS needs to be able to measure

the frequency of the power grid.

4 DESIGN

4.1 ADR System Design

As can be seen from the Figure 1, the ADR system is

composed by three main components: a software sys-

tem with a cloud-based architecture, an electrical grid,

and a set of consumers’ households. The cloud-based

system implements the necessary applications for ex-

ecuting the ADR for CFC. Moreover, the electri-

cal power grid provides electricity to the consumers’

households which can then measure the frequency

though the EMS. The households are equipped with

the EMS that is interfacing within the HAN with a set

of IED devices capable of CFC, and through the WAN

with the cloud-based system.

The cloud-based system for CFC implements two

main applications: the Aggregator and the DR Con-

sumer. The Aggregator application performs the re-

quired functionalities requested from the aggregator

(Gkatzikis et al., 2013). On the other hand, each con-

sumer has one instance of the DR Consumer appli-

cation. The DR Consumer application provides ser-

vices to the consumers for ADR which spans from

energy forecasting, optimization and CFC decision-

making. Moreover, the intra-cloud communication

between different applications is performed by an

asynchronous message oriented middleware (MOM).

The MOM is based on the Advanced Message

Queue Protocol (AMQP, 2016). AMQP provides

asynchronous message-oriented queuing communica-

tion, and it is capable of advanced routing messag-

ing. Among others, AMQP supports both publish-

subscribe and request-response communication. The

AMQP MOM provides several features which are of

fundamental importance to fulﬁll the communication

requirements of the ADR system. In fact, the AMQP

MOM supports the scalability of the communication

between applications. Moreover, the AMQP MOM

ensures that the various applications in the intra-cloud

system are decoupled. In addition, the AMQP MOM

offers a reliable and secure communication protocol.

Cloud

Aggregator

Consumer

Optimization

Service

AMQP

MOM

Electric Grid

(Frequency)

EMS

Power Plant

Figure 1: Overview of the ADR system for continuous fre-

quency control.

A Distributed ICT Architecture for Continuous Frequency Control

137

4.2 Decision-making Algorithms

To perform CFC, two main decision-making algo-

rithms have been implemented in the ADR system.

Firstly, a decision-making algorithm is executed by

each consumer. The algorithm establishes whether a

consumer participates in the CFC, and whether the

consumer reacts to over-frequency, under-frequency,

or both situations. This allows each consumer to be

independent in the decision of participating in the

CFC, based on their capabilities. The aggregator is

then executing the second decision-making algorithm,

in which an allocation algorithm schedules the CFC

reactions of the consumers, and provides the required

coordination for provisioning the reserves.

4.2.1 Consumer Participation Algorithm

Consumers participate in the CFC by either increase

or decrease their consumption. Each consumer can

decide whether or not to take part in the CFC through

the execution of the consumer participation algo-

rithm. This decision-making algorithm is imple-

mented in the DR Consumer application, which, by

knowing the status of the IED devices in the house-

hold, estimates two time based parameters called Par-

ticipation Time (PT) and Reaction Time (RT). These

two parameters are then used to decide whether the

consumer has the capability of providing reserves.

Figure 2 shows how PT and RT are calculated

from a thermodynamic load. Typically, in a thermo-

dynamic load the controlled variable has to stay be-

tween deﬁned upped (T

MAX

) and lower bound (T

MIN

)

limits. Given the current thermal status (T

CT S

) of the

controlled variable, it is possible to project the status

to the respective boundaries in order to calculate the

two parameters. As an example, assuming that Figure

2 represents the thermodynamic status of a refrigera-

tor as presented in (Stadler et al., 2009), the following

equations shows how PT and RT can be calculated for

such a case:

PT = −

ln(

MAX

− T

AMB

+ η

ptcl

CT S

− T

AMB

+ η

ptcl

) (1)

RT = −

ln(

MIN

− T

AMB

+ η

ptcl

CT S

− T

AMB

+ η

ptcl

) (2)

where τ is a time constant, T

AMB

represents the ex-

ternal ambient temperature, η is a coefﬁcient of per-

formance, ptcl consists in the rated power (W ) of the

appliance, and A represents the thermal conductance

(W /

◦

C). Besides the refrigerators, this example can

be applied to several types of thermal loads such as

HVAC and water boilers.

The parameters, estimated by the Consumer Par-

ticipation Algorithm, are of fundamental importance

for the aggregator which, through a decision-making

algorithm called Frequency Allocation Algorithm

(FAA), will use these parameters to decide the poli-

cies that each consumer will have to follow. In fact,

PT represents the time in which the consumer has the

capability of participating in the ADR, while RT de-

ﬁnes for how long the consumer can react to a fre-

quency deviation. The DR Consumer has to deﬁne

the participation and the reaction time for both: the

over-frequency and the under-frequency. Based on

the participation and the reaction time, each consumer

decides whether or not it is possible to take part in the

CFC, and then sends the decision to the aggregator.

Currently, the decision on the participation is based

on deﬁned thresholds for PT and RT, which are both

required to be more than a certain threshold value of

5 minutes.

Participation Time (PT)

Reaction Time (RT)

Time

Participation Time Projection

Reaction Time ProjectionCurrent Time (t)

Thermodynamic Load

Thermal Upper Bound Limit (T

MAX

)

Thermal Lower Bound Limit (T

MIN

)

Current Thermal

Status (T

CTS

)

Figure 2: An example of a thermodynamic load with ther-

mal boundary limits. The participation and reaction times

are estimated by projecting the current thermal status to the

upper and lower boundary.

4.2.2 Frequency Allocation Algorithm

The FAA is executed by the aggregator. The algo-

rithm aims to deﬁne the policies that each consumer

will have to follow to participate in the CFC and react

to the frequency deviation in the power grid. Based on

the latest updates received from the Consumer Partic-

ipation Algorithms of each consumer, the FAA elab-

orates the policies by allocating the CFC reactions in

the form of instruction messages for the consumers. If

FAA selects one consumer for the CFC, the respective

instruction message will contain the frequency thresh-

old from which the consumer should start reacting,

and provide accordingly the promised reserve.

Algorithm 1 provides the pseudocode of the FAA.

The algorithm starts with the procedure IsUpdate-

Needed. This procedure establishes whether a new

update of the consumer instructions is needed based

SMARTGREENS 2017 - 6th International Conference on Smart Cities and Green ICT Systems

138

on two criteria. The ﬁrst criteria is based on a timeout

of several minutes after which the CFC reactions are

required to be reallocated. In addition, the second cri-

teria decides whether a new update is needed before

the expiring of the aforementioned timeout. In fact, if

a quota of the allocated consumers updates the aggre-

gator declaring their inability to continue providing

the allocated reserves, then a new update of the con-

sumer instructions is executed.

The FAA begins by retrieving the latest update

messages of each consumer through the GetLates-

tUserUpdates procedure. Then, the algorithm pro-

ceeds by allocating in sequence the under-frequencies

and the over-frequencies to the consumers based on

the updates received. The algorithm ranks the best

consumer for CFC through the OrderByUnderFre-

quency and the OrderByOverFrequency procedures,

which order the consumers respectively for the under-

frequencies and the over-frequencies. These two pro-

cedures order the consumers according to their ability

to perform CFC. The ordering is performed based on

the PT, the RT and an aging parameter, which are sent

from each consumer to the aggregator. While the PT

and the RT are estimated by the DR Consumers, the

aging parameter represents the number of times that

each consumer have acted to the CFC, and it is used

during the ordering to avoid overloading only a small

set of consumers with many CFC reactions. The or-

dering, for both under-frequency and over-frequency,

gives the priority to the consumers with higher PT and

RT, and a smaller aging parameter.

Following the procedures of ordering the con-

sumers, the AllocateUnderFrequency and Alloca-

teOverFrequency procedures are used to allocate the

reaction frequencies to the consumers according to

the given ranking in the ordering. The FAA allo-

cates the frequencies for CFC reaction until the CFC

reaction of the ADR system can cover the required

target reserves. Finally, the algorithm proceeds to

send the new instructions to the consumers through

the SendInstructionsToConsumers procedure.

4.3 Communication Logic

The communication logic of the ADR system can be

divided in two phases: the status update phase and

the instruction phase. These phases are asynchronous

between each other. The status update phase is used

by the consumers to provide the last updates to the

aggregator, while the instruction phase is used by the

aggregator to update the consumers with new policies

for the CFC.

Algorithm 1: Frequency Allocation Algorithm.

1: function FREQUENCYALLOCATION

2: loop

3: . // Checks if a new update is needed

4: if IsUpdateNeeded() then

5: GetLatestUserUpdates()

6: . // Under Frequency Allocation

7: OrderByUnderFrequency()

8: AllocateUnderFrequency()

9: . // Over Frequency Allocation

10: OrderByOverFrequency()

11: AllocateOverFrequency()

12: . // Send instructions to Consumers

13: SendInstructionsToConsumers()

4.3.1 Status Update Phase

The status update phase consists of all the necessary

communication steps needed by the consumers to no-

tify the aggregator about their possibility to partici-

pate in the ADR for CFC. Figure 3 shows the se-

quence diagram of the update phase. Before detailing

the communication logic, it is important to note that

the messages exchanged between each actor during

the update phase are asynchronous.

The update phase starts with each IED device that

sends status update messages to the respective EMS.

The EMS operates as a network gateway by interfac-

ing with various IED devices using different protocols

within the HAN. After processing the data of the IED

devices, the EMS sends a message containing the in-

formation of the IED devices that can be used for CFC

to the DR Consumer. Then, the DR Consumer aggre-

gates the provided data and establishes whether or not

the consumer could participate in the CFC, either to

the over-frequency, the under-frequency or both the

controls. In the last step, the DR Consumer updates

the aggregator, with a Frequency Control S-U mes-

sage, about its capability in participating to the CFC.

4.3.2 Instruction Phase

The second phase of the communication logic is

called instruction phase. During this phase, the ag-

gregator sends the necessary instructions to the con-

sumers, enabling their participation to the CFC. In the

FAA, as shown in Algorithm 1, the aggregator starts

by calling the SendInstructionsToConsumers proce-

dure. Figure 4 presents the sequence diagram of the

communication logic for the instruction phase.

The aggregator starts the instruction phase as the

last task in the execution of the FAA. The algorithm

determines the set of consumers that will be involved

A Distributed ICT Architecture for Continuous Frequency Control

139

icDraw UML, 1-1 \\home.or

.aalto.fi\

iovanc1\data\Documents\ADR

EM\ADR

EM.mdzi

FinalC

Academic Version for Teaching Only, Commercial Development is strictly Prohibited

FinalConsToAgg FinalConsToAgginteraction [ ]

: DR Consumer (n) : IED Device (n m ) : Aggregator : EMS (n)

Consumer Participation Algorithm4:

Frequency Control S-U5:

IED Device S-U (Status-Update)1:

IED Gateway Data Processing2:

IED Devices S-U3:

Figure 3: A sequence diagram showing the update phase of

the CFC communication logic.

in the CFC. Whereupon, the aggregator sends an In-

struction Message to each DR Consumers which con-

tains the instructions needed to react to the CFC. If the

consumer has to use multiple IED devices for CFC,

the DR Consumer disaggregates the instructions of

the aggregator, and allocates them to each different

IED device. Then, the DR Consumer sends an IED

Control Instruction message to the EMS which con-

tains the control instructions scheduled for the IED

devices. The EMS, when instructed to apply CFC,

measures the frequency of the electric grid, and it re-

acts following the provided instructions. When the

frequency exceeds the given limits, the EMS sends a

DR Reaction Control message to the IED devices in-

volved, which are then participating to the CFC.

icDraw UML, 1-1 \\home.or

.aalto.fi\

iovanc1\data\Documents\ADR

EM\ADR

EM.mdzi

FinalA

Academic Version for Teaching Only, Commercial Development is strictly Prohibited

FinalAggToCons FinalAggToConsinteraction [ ]

: DR Consumer (n) : IED Device (n m ) : Aggregator : EMS (n)

IED Control Policies4:

Household Data Disaggregation3:

Frequency Allocation Algorithm1:

Instruction Message2:

Measure Frequency5:

DR Control Reaction6:

Figure 4: A sequence diagram that shows the communica-

tion logic of the instruction phase of the ADR system for

CFC.

5 EXPERIMENTATION

The ADR system was evaluated through simulations

based on a prototype application. The prototype ap-

plication has been developed using Java. The proto-

type application is composed of different entities: one

Aggregator application, a set of DR Consumer appli-

cations, and a RabbitMQ MOM (RabbitMQ, 2016)

which relies on the AMQP protocol. For each con-

sumer, the DR Consumer application is interfaced

with the respective EMS. The EMS model provides

the interconnection with the IED devices of the house-

hold. The IED devices of each consumer are modeled

as refrigerators, which will be utilized by the con-

sumers to participate in the ADR for CFC. The re-

frigerator population is based on the models presented

in (Stadler et al., 2009). Moreover, end-to-end com-

munication delays were injected into the ADR system

and uniformly distributed between 1 and 3 seconds.

For the following simulations, various types of

data were utilized. The frequency data were taken

from different sources. The ﬁrst set of the frequency

data was deﬁned manually with some ad hoc patterns,

while the second set consists of real frequency data

(Mainsfrequency, 2012). The number of consumers

utilized was 5000. The aggregator target reserves

was 40kW for both the over-frequency and the under-

frequency (de la Torre Rodriguez et al., 2014). The

maximum deviation from the nominal frequency in

which the reserves were fully activated was +/- 0.1

Hz, while the dead-band speciﬁed was +/- 0.02 Hz.

Moreover, the system frequency dynamics were not

affected by the simulated system, since the controlled

power was relatively low.

In order to verify that the ADR system is capa-

ble of performing CFC, simulations were executed.

Figure 5 presents one hour simulation of the devel-

oped prototype application for the ADR system in

which CFC is performed. Based on the frequency

deviation from the nominal value (50 Hz), the ADR

system was reacting by providing the necessary re-

serves. The aggregated consumption shows how the

consumers of the ADR system were reacting to the

frequency changes of the electrical grid. Again, the

estimated nominal consumption shows instead the es-

timated consumption of the system without being in-

volved in the CFC. The deviation between the aggre-

gated consumption and the estimated nominal con-

sumption represent the amount of reserves that the

ADR system for CFC was able to provide. Lastly,

the vertical dashed lines represent the times in which

the aggregator starts updating the consumers after ex-

ecuting the FAA through the instruction phase.

SMARTGREENS 2017 - 6th International Conference on Smart Cities and Green ICT Systems

140

DEAD BAND

Time (min)

Consumption (kW)

Frequency (Hz)

ADR for CFC

Aggregated Consumption with ADR

Estimated Nominal Consumption without ADR

System Frequency

100

120

140

160

180

200

49.85

49.9

49.95

50.05

50.1

50.15

Highcharts.com

Figure 5: One hour simulation of the ADR system for CFC. The aggregated consumption of the ADR system reacts to the

frequency changes of the system by providing the allocated reserves by the frequency allocation algorithm.

6 DISCUSSION

As shown in Figure 5, the ADR system performs CFC

by reacting to the frequency changes in the electrical

grid and providing the necessary reserves. Currently,

consumers are using a heterogeneous population of

fridges to apply CFC. Adding more heterogeneity to

the system (i.e. by adding populations of different

IED devices for CFC as HVAC or EVs) as well as in-

creasing the number of consumers could provide both

a more realistic scenario and improve the quality of

the simulation results. This addition would proba-

bly not have a big impact in the communication of

the ADR system. On the other hand, it would affect

both the decision-making algorithms, which would

increase in complexity.

The decision-making algorithms deﬁned in this

study have proved to have a key role in the ADR

architecture. Undoubtedly, the FAA has resulted to

have a fundamental role in the decision-making pro-

cess which enables the ADR system to perform CFC.

The current disadvantage of this algorithm consists in

the fact of being centralized and executed by a single

entity as the aggregator. Thus, a possible improve-

ment would be to redeﬁne the algorithm in a fully

distributed way, in which the aggregator could spec-

ify the objectives (e.g. in term of reserves) that the

ADR system should achieve, while the consumers al-

together could negotiate the CFC reactions needed to

fulﬁll the speciﬁed objectives. In addition, the partici-

pation of the consumers in the CFC is currently based

only on the two parameters PT and RT. An improve-

ment of the consumer participation algorithm could

include user preferences as a key factor for deciding

whether a consumer can participate to the CFC.

Compared to the previous study (Giovanelli et al.,

2016), in which the communication was used for an

energy consumption planning phase with no time con-

straints, the current communication had to deal with

near real-time constraints. This constraint was due to

the electrical requirement of having the reserves fully

available within three minutes. This additional con-

straint has allowed to further stretch the asynchronous

MOM, which has resulted to well perform also in the

presented new scenario.

7 CONCLUSIONS

This paper presented a prototype of ADR system ca-

pable of performing CFC. A set of consumers was

communicating with an aggregator in order to pro-

vide the required reserves during events of frequency

deviations. For the intra-cloud communication, an

asynchronous MOM was utilized to interface the con-

sumers with the aggregator. Moreover, two decision-

making algorithms were deﬁned to enable the con-

sumers to participate in the CFC. Finally, the capabil-

ity of the ADR system to provide CFC was evaluated

based on simulations with real data.

Future research might explore the possibility of

improving the decision-making algorithm for fre-

quency allocation. A further study could enhance the

FAA by supporting different IEDs and exploiting their

different properties. Moreover, a distributed solution

for the FAA could further enhance the ADR system.

A Distributed ICT Architecture for Continuous Frequency Control

141

In fact, a distributed algorithm could provide a more

robust and reliable solution.

ACKNOWLEDGEMENTS

This research has been funded by the SAGA project

of the Aalto Energy Efﬁciency (AEF) Research Pro-

gram and the Flexible Customer project funded by the

Academy of Finland.

REFERENCES

AMQP (2016). Amqp web page, https://www.amqp.org/.

Angeli, D. and Kountouriotis, P.-A. (2012). A Stochastic

Approach to Dynamic-Demand Refrigerator Control.

IEEE Transactions on Control Systems Technology,

20(3):581–592.

de la Torre Rodriguez, M., Scherer, M., Whitley, D., and

Reyer, F. (2014). Frequency containment reserves di-

mensioning and target performance in the European

power system. In 2014 IEEE PES General Meeting —

Conference & Exposition, pages 1–5. IEEE.

Entsoe (2015). Supporting document for the network code

on load-frequency control and reserves.

Fang, X., Misra, S., Xue, G., and Yang, D. (2012). Smart

gridthe new and improved power grid: A survey. IEEE

communications surveys & tutorials, 14(4):944–980.

Fingrid (2016). Fingrid web page,

http://www.ﬁngrid.ﬁ/en/pages/default.aspx.

Giovanelli, C., Kilkki, O., Seilonen, I., and Vyatkin, V.

(2016). Distributed ict architecture and an application

for optimized automated demand response. In IEEE

PES ISGT Europe 2016. IEEE.

Gkatzikis, L., Koutsopoulos, I., and Salonidis, T. (2013).

The role of aggregators in smart grid demand response

markets. IEEE Journal on Selected Areas in Commu-

nications, 31(7):1247–1257.

Grijalva, S. and Tariq, M. U. (2011). Prosumer-based smart

grid architecture enables a ﬂat, sustainable electricity

industry. In ISGT 2011, pages 1–6. IEEE.

Gungor, V. C., Sahin, D., Kocak, T., Ergut, S., Buccella, C.,

Cecati, C., and Hancke, G. P. (2013). A Survey on

Smart Grid Potential Applications and Communica-

tion Requirements. IEEE Transactions on Industrial

Informatics, 9(1):28–42.

Halamay, D. A., Brekken, T. K. A., Simmons, A., and

McArthur, S. (2011). Reserve Requirement Impacts

of Large-Scale Integration of Wind, Solar, and Ocean

Wave Power Generation. IEEE Transactions on Sus-

tainable Energy, 2(3):321–328.

Kilkki, O., Kangasr

asi

o, A., Nikkil

a, R., Alah

aiv

a, A.,

and Seilonen, I. (2014). Agent-based modeling and

simulation of a smart grid: A case study of communi-

cation effects on frequency control. Engineering Ap-

plications of Artiﬁcial Intelligence, 33:91–98.

Kim, Y.-J., Thottan, M., Kolesnikov, V., and Lee, W. (2010).

A secure decentralized data-centric information in-

frastructure for smart grid. IEEE Communications

Magazine, 48(11):58–65.

Lalor, G., Mullane, A., and O’Malley, M. (2005). Fre-

quency Control and Wind Turbine Technologies.

IEEE Transactions on Power Systems, 20(4):1905–

1913.

Mainsfrequency (2012). Mainsfrequency web page,

http://www.mainsfrequency.com/services.htm.

Masuta, T. and Yokoyama, A. (2012). Supplementary load

frequency control by use of a number of both electric

vehicles and heat pump water heaters. IEEE Transac-

tions on Smart Grid, 3(3):1253–1262.

Megel, O., Mathieu, J. L., and Andersson, G. (2013). Max-

imizing the potential of energy storage to provide fast

frequency control. In IEEE PES ISGT Europe 2013,

pages 1–5. IEEE.

Molina-Garcia, A., Bouffard, F., and Kirschen, D. S.

(2011). Decentralized demand-side contribution to

primary frequency control. IEEE Transactions on

Power Systems, 26(1):411–419.

Pourmousavi, S. A. and Nehrir, M. H. (2012). Real-

Time Central Demand Response for Primary Fre-

quency Regulation in Microgrids. IEEE Transactions

on Smart Grid, 3(4):1988–1996.

RabbitMQ (2016). Rabbimq web page,

https://www.rabbitmq.com/.

Rodrigues, J. (2013). Service-oriented middleware for

smart grid: Principle, infrastructure, and application.

IEEE Communications Magazine, 51(1):84–89.

Samarakoon, K., Ekanayake, J., and Jenkins, N. (2012). In-

vestigation of Domestic Load Control to Provide Pri-

mary Frequency Response Using Smart Meters. IEEE

Transactions on Smart Grid, 3(1):282–292.

Short, J. A., Inﬁeld, D. G., and Freris, L. L. (2007). Sta-

bilization of Grid Frequency Through Dynamic De-

mand Control. IEEE Transactions on Power Systems,

22(3):1284–1293.

Siano, P. (2014). Demand response and smart gridsA sur-

vey. Renewable and Sustainable Energy Reviews,

30:461–478.

Stadler, M., Krause, W., Sonnenschein, M., and Vogel, U.

(2009). Modelling and evaluation of control schemes

for enhancing load shift of electricity demand for

cooling devices. Environmental Modelling & Soft-

ware, 24(2):285–295.

Vrettos, E., Oldewurtel, F., Zhu, F., and Andersson, G.

(2014). Robust Provision of Frequency Reserves by

Ofﬁce Building Aggregations. IFAC Proceedings Vol-

umes, 47(3):12068–12073.

Xu, Z., Ostergaard, J., and Togeby, M. (2011). Demand as

Frequency Controlled Reserve. IEEE Transactions on

Power Systems, 26(3):1062–1071.

Yan, Y., Qian, Y., Sharif, H., and Tipper, D. (2013). A

survey on smart grid communication infrastructures:

Motivations, requirements and challenges. IEEE com-

munications surveys & tutorials, 15(1):5–20.

Zaballos, A., Vallejo, A., and Selga, J. M. (2011). Heteroge-

neous communication architecture for the smart grid.

IEEE Network, 25(5):30–37.

SMARTGREENS 2017 - 6th International Conference on Smart Cities and Green ICT Systems

142