ON THE APPLICATION OF AUTONOMIC

AND CONTEXT-AWARE COMPUTING TO SUPPORT HOME

ENERGY MANAGEMENT

Boris Shishkov

IICREST / TBM – Delft University of Technology, Jaffalaan 5, Delft, The Netherlands

Martijn Warnier

TBM – Delft University of Technology, Jaffalaan 5, Delft, The Netherlands

Marten van Sinderen

IS – University of Twente, Drienerlolaan 5, Enschede, The Netherlands

Keywords: Green Household, Home Energy Management, Autonomic Computing, Context-aware Computing.

Abstract: Conventional energy sources are becoming scarce and with no (eco-friendly) alternatives deployed at a

large scale, it is currently important finding ways to better manage energy consumption. We propose in this

paper ICT-related solution directions that concern the energy consumption management within a household.

In particular, we consider two underlying objectives, namely: (i) to minimize the energy consumption in

households; (ii) to avoid energy consumption peaks for larger residential areas. The proposed solution

directions envision a service-oriented approach that is used to integrate ideas from Autonomic Computing

and Context-aware Computing: the former influences our considering a selective on/off powering of

thermostatically controlled appliances, which allows for energy redistribution over time; the latter

influences our using context information to analyze the energy requirements of a household at a particular

moment and based on this information, appliances can be powered down. Household-internally, this can

help adjusting energy consumption as low as it can be with no violation of the preferences of residents.

Area-wise, this can help avoiding energy consumption peaks. It is expected thus that such an approach can

contribute to the reduction of home energy consumption in an effective and user-friendly way. Our

proposed solution directions are not only introduced and motivated but also partially elaborated through a

small illustrative example.

1 INTRODUCTION

Conventional energy sources are becoming scarce

and with no (eco-friendly) alternatives deployed at a

large scale, it is currently important finding ways to

better manage energy consumption. This is a key

challenge related to the societal goal of green

household and sustainable living (Carvalho, 2009).

Looking at this challenge from the perspective of a

single household (as a unit) seems a logical starting

point, before scaling up to collections of units, such

as neighborhood, city, or region. Hence, we propose

in this paper ICT-related solution directions that

concern the energy consumption management within

a household. In particular, we consider two

underlying objectives, namely: (i) to minimize the

energy consumption in households; (ii) to avoid

energy consumption peaks for larger residential

areas.

The proposed solution directions envision a

service-oriented approach that is used to integrate

ideas from Autonomic Computing and Context-

aware Computing.

Autonomic Computing (Kephart & Chess, 2003)

has been proposed as a way to empower systems

with self-management capabilities, in order to

increase availability and reduce time-consuming and

307

Shishkov B., Warnier M. and van Sinderen M. (2010).

ON THE APPLICATION OF AUTONOMIC AND CONTEXT-AWARE COMPUTING TO SUPPORT HOME ENERGY MANAGEMENT.

In Proceedings of the 12th International Conference on Enterprise Information Systems - Information Systems Analysis and Speciﬁcation, pages

307-313

 SciTePress

error-prone human management. Hence, Autonomic

Computing can endow systems with properties, such

as self-configuring, self-healing, and self-

optimizing. This influences our considering a

selective on/off powering of Thermostatically

Controlled Appliances (TCAs), which allows for

energy redistribution over time.

Context-aware Computing (Schilit et al., 1994)

has been proposed as a way to allow systems to

specialize their external behaviors depending on the

perceived needs of their users. Perceived needs are

deduced from context information which is captured

in turn from context sources, such as sensors in the

user environment. This influences our using context

information to analyze the energy requirements of a

household at a particular moment and based on this

information, appliances can be powered down

As for TCAs (such as fridges, water heaters, air

conditioners, and other energy-hungry appliances),

they are in our focus mainly because of their great

deal in the overall household energy consumption to

date – taking for example USA, 25% of all

household usage in the country points to TCAs.

TCAs work in a periodic fashion by turning their

thermostat ‘on’ and ‘off’. The thermostat keeps the

temperature of the device within the target range of

its operation, determined by a preference value pre-

set by a human being. There is some freedom to

time-shift the active/inactive state of these

appliances without exceeding their target range of

operation. Consequently, when considering a

collection of TCAs, we may manage them in such a

way that their active states have minimum overlap

thus reducing peaks in electricity usage (load

shifting). If we further could have real-time

information available on the working environment

of TCAs, such as the presence or activity mode of

persons, it may be possible to decide whether the

user set value for operation can be ignored. If such

conditions exist, respective TCAs may be

temporarily controlled on the basis of a lower set

value, thus reducing average electricity usage. All

those examples of intelligent management of TCAs,

as mentioned in this paragraph, illustrate the

potential for reducing home energy consumption,

provided nevertheless we have means to capture,

exchange and apply information on the status of

home appliances and their environment. This not

being the case with current energy management

systems, inspires us for proposing a solution

influenced by Autonomic Computing and Context-

aware Computing, as already mentioned.

Hence, household-internally, intelligently

controlling TCAs’ performance can help adjusting

energy consumption as low as it can be with no

violation of the preferences of residents. Area-wise,

this can help avoiding energy consumption peaks.

Finally, our envisioning a service-oriented

approach (used to integrate ideas from Autonomic

Computing and Context-aware Computing) is

motivated by the ease and flexibility of composition

and integration of application components,

associated with the Service-Oriented Architecture

(Erl, 2005; Leymann, 2005; Shishkov & Van

Sinderen, 2009).

It is expected thus that such a service-oriented

approach can contribute to the reduction of home

energy consumption in an effective and user-friendly

way. Our proposed solution directions are not only

introduced and motivated but also partially

elaborated through a small illustrative example.

The remaining of the paper is structured as

follows: Section 2 elaborates the computing

paradigms considered in this work, namely

Autonomic Computing and Context-aware

Computing. Section 3 introduces and motivates our

proposed solution directions inspired by the

mentioned computing paradigms. Section 4 partially

elaborates the introduced solution directions through

a small illustrative example. Section 5 discusses

related work and Section 6 presents the conclusions.

2 BACKGROUND

This section provides relevant background

information on Autonomic Computing and Context-

aware Computing.

2.1 Autonomic Computing

Autonomic Computing (Kephart & Chess, 2003) has

been proposed as a way to reduce the cost of

maintaining complex systems, and to increase the

human ability to manage these systems properly.

Autonomic Computing introduces a number of

autonomic properties, such as self-configuring, self-

healing, self-optimizing and self-protecting.

Extending and enhancing a system with these

properties is an important step towards a self-

management system.

In the context of knowledge based approaches

IBM has introduced an abstract architecture for

Autonomic Computing (Ganek & Corbi, 2003) that

identifies a number of fundamental concepts and

architectural building blocks for constructing self-

managed systems with autonomic properties. The

two main building blocks of the Autonomic

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

308

Computing architecture are autonomic managers and

managed resources.

Managed resources are hardware or software

components, for example a business application, a

router or a database. A managed resource is

managed by an autonomic manager. This autonomic

manager forms the central part of the autonomic

architecture. It collects data from managed

resources, which is used for diagnosing failures and

other unwanted behaviour. The autonomic manager

formalizes and executes remedy plans for the

managed resource which (should) correct the

unwanted behaviour. Internally the autonomic

manager implements a control loop that consist of

four components, the so called MAPE (IBM

Corporation 2005) functions: monitor, analyze,

plan, and execute. The MAPE functions share a

common knowledge base which is typically

predefined and domain-specific, i.e. new knowledge

is only added by system administrators and other

users, the system itself does not learn.

2.2 Context-aware Computing

Context-aware systems are primarily motivated by

their potential to increase user-perceived

effectiveness, i.e. to provide services that better suit

the end-user's needs, by taking account of the user

conditions. We refer to the collection of conditions

which characterize an end-user or his/her immediate

surroundings, and which are relevant for the system

in pursueing user-perceived effectiveness, as end-

user context, or context for short, in accordance to

definitions found in literature (Dey et al., 2001).

Context-awareness implies that context

information on the end-user must be captured, and

preferably so without conscious or active

involvement of the end-user. Context-aware systems

can be particularly useful if the end-user is mobile

and has a personal handheld device for the delivery

of services. Mobility implies dynamic context. For

example, different locations may have different

social environments and different network access

options, which offer opportunities for the provision

of adaptive or value-added services based on context

sensitivity. A context-aware system may provide

near real-time context-based adaptation during a

service delivery session with its mobile end-user.

Although context-aware systems have received

much attention within the ICT research community,

they have not been fully successful so far from a

business point of view. This situation may change

rapidly however, due to the increased capabilities

and reduced prices of mobile devices, sensors, and

wireless networks, and due to the introduction of

new marketing strategies and service delivery

models (De Reuver & Haaker, 2009).

3 BASIC ARCHITECTURE

As mentioned in Section 1, we are interested in

pursuing two objectives with home energy

management: reduction of average energy

consumption and reduction of peak energy

consumption. Our proposed approach is to apply

results from Autonomic Computing and Context-

aware Computing in order to contribute to the

realization of these objectives. We introduce the

following terms:

 Actual Consumption: this is the actual

(measured) energy consumption of a

collection of home appliances of interest;

 Consumption Constraints: these are the

(maximum values) set for average and peak

energy consumption;

 Perceived Needs: these are the perceived

needs of the users regarding the operation

of appliances (should they be on or off, or

should their pre-set preference values be

kept or lowered);

 Supported Needs: this is an indication of

the needs that are supported by the current

status (on/off) and setting (preference

values) of appliances.

Analyze

Maintain

Consumption

Decrease

Consumption

Yes

Is Consumption

OK?

Figure 1: General plan for controlling home energy

consumption.

The general idea is that we can analyze actual

consumption with respect to consumption

constraints, and perceived needs with respect to

supported needs. If either consumption constraints

are (close to being) violated or supported needs are

unnecessarily high, then energy consumption should

be decreased and corresponding control actions need

ON THE APPLICATION OF AUTONOMIC AND CONTEXT-AWARE COMPUTING TO SUPPORT HOME ENERGY

MANAGEMENT

309

to be exercised on the appliances of interest. Figure

1 illustrates this general plan.

We will firstly address the problem of analyzing

consumption and then consider the analysis of

needs.

For analyzing consumption we adopt the MAPE

control loop (see Section 2), considering a pool of

appliances that are instrumented to allow monitoring

and control. The monitoring consists of measuring

the energy consumption of these appliances. The

measurements are fed to a control process, which

interprets these as the actual consumption, and

compares the consumption with the consumption

constraints. If, as a result of this analysis, it is

decided that control actions are needed, an action

plan is produced. The action plan is derived with an

algorithm that considers time-shifting of the active

state of appliances. Subsequently, the plan is

executed by performing the indicated control actions

on the selected appliances. Figure 2 illustrates the

application of the MAPE control loop.

Controlled domain

(e.g. household)

Analyzer

Planner

Appliances

Monitor actual consumption

Execute action plan: time-shift of active/inactive states

Control process

Consumption constraints

Figure 2: MAPE control loop applied to home energy

management.

With regard to analyzing consumer needs, we

adopt the event-control-action (ECA) pattern from

context-aware computing (Dockhorn Costa et al.,

2008). We consider the environment of the

appliances, and assume that this environment is

instrumented with sensors that are able to measure

relevant conditions. For example, measurements

may be used to determine context changes or

situations, such as the presence of one or more

persons in the house or in a particular room, the

activity mode (sitting, walking, sleeping) of a

person, or a person entering or leaving the house.

Context situations and changes can generally not be

directly or reliably measured by a single sensor. A

context management process is responsible for

producing events that indicate the occurrence of a

context situation or change, based on reasoning

which potentially involves sensor data from several

sources. Events are fed to a control process, which

applies them in rules to determine actions related to

perceived needs. For example, if nobody is in the

house, a rule may establish the action to set the

preferred value of the heating at 15 degrees Celsius.

Whether the actions are really needed depends on

the supported needs. For example, if the preferred

value of the heating is already set to 15 degrees

Celsius, no action is needed. The comparison of the

perceived and supported needs leads to an action

plan, which, if not empty, is subsequently executed

by performing the indicated control actions on the

selected appliances. Figure 3 illustrates the

application of the ECA pattern.

Controlled domain

(e.g. household)

Cxt mgr

Rule engine

Ctx mgr

Action

performer

Sensors

Appliances

Context events

Measure raw context data

Execute actions if perceived needs < supported needs

CMP

Rules

Figure 3: ECA pattern applied to home nergy management

(CP = Control Process, CMP = Context Management

Process.

In order to combine the previously mentioned

solution approaches, we need ease and flexibility in

composing and integrating application components.

We adopt a service-orientated architecture (SOA)

because of its strengths in this direction (Leymann,

2005). In general, SOA brings flexibility to (re-)use

applications and develop new applications and

systems. Development may be fast and cost-

effective with services of 3rd party applications.

Maintenance is easy – implementations of services

can be replaced without affecting functionality and

functionality might be changed or extended

according to new requirements by changing or

extending the composition of services. These

strengths are relevant especially in our case of

considering an ICT system for home energy

management. For the sake of brevity, we will not

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

310

elaborate further on SOA concepts and related

standards, but instead refer interested readers to

(Papazoglou, 2007). Services encapsulate ICT

functionality and externalize their public functions

through well-defined interfaces. By appropriately

composing services, it is possible to support a

desired business process. For achieving this, good

coordination is required as well as proper

information exchange.

We propose a service model that is driven by a

coordination service and an information service.

These services are essential since they support the

deliveries of all other services. The coordination

service orchestrates the overall work of the system,

invoking other services at the right moment and

offering them also the right input. This service is

supported from the information service on most of

the service invocations because in invoking a

service, specific and actual information inputs would

be needed. This vision is reflected in Figure 4.

system

system components

coordination service

inf. service

service

…..

resident

TCA

adjust

preferences

info

control

environment

control

Figure 4: Facilitating residents, appliances and their

environment in a service-oriented way.

As shown in Figure 4, all services, including the

coordination service and information service, have

their underlying software/hardware components

which implement required ICT functionality.

Our service model helps to bridge between user

needs and system capabilities, by considering the

entities involved: (i) residents who have their own

preferences; (ii) Home appliances which have their

own capabilities; (iii) the living environment having

its own characteristics. What the system would

typically do is.

- Let a resident adjust personal preferences and

if possible adjust appliances in such a way

that those are met;

- Inform a resident about the household energy

consumption;

- Monitor appliances, by properly capturing

information that reflects their energy

consumption;

- Exercise control over appliances driven by

external request(s) and/or residents’

preferences;

- Allow external control that is in line with

relevant public regulations

4 EXAMPLE

To illustrate our basic architecture we return to the

TCA example mentioned in Section 1. Specifically,

we consider a 2-person household with a fridge, a

freezer and a central heating system. All TCAs are

equipped with sensors to measure performance of

the appliances. Additional sensors are present

throughout the house. These sensors are used to

collect relevant context information, e.g. to

determine if a person is in the house, in which room

(s)he is, and what (s)he is doing (active, sitting,

sleeping). Both context information and TCA-

specific information is sent to an analysis service.

This analysis service keeps track of the current

state of energy consumption and predicts what

energy usage is required in the (near) future. For

example, if a resident enters his/her empty house it

can be expected that energy usage as a whole will

rise. Behavioral patterns of the residents of a house

can be learned: does a person usually start with

preparing a meal within 30 minutes after arriving in

the evening? This information can gradually be

acquired by analyzing the energy usage of the

residents over a longer period of time. The analysis

service should learn from the residents and be

adaptive to their behavior. It should also take factors

into account that are external to the household. If

peak usage in a neighborhood should be avoided,

control systems of different households should

coordinate their behavior to jointly prevent this.

The information that the analysis service

acquires is necessary for the planning service. This

service needs to carefully plan how the TCAs in the

household should function for a given time period. It

is essential that this service can coordinate the

energy usage of the different appliances. TCAs can

all be put on and off in sync or, in contrast, be put on

and off alternating over different time periods. The

planning service will select the appropriate energy

consumption plan for each appliance, based on the

requirements of the analysis service.

The execution plans for the individual devices

are then forwarded by a control service to the

fridge, the freezer and the central heating system,

where they are executed.

ON THE APPLICATION OF AUTONOMIC AND CONTEXT-AWARE COMPUTING TO SUPPORT HOME ENERGY

MANAGEMENT

311

The residents can give their own requirements to

the system. For example, if the house is empty or the

residents are sleeping, the preferred room

temperature can be decreased to 15 °C.

A coordination service is needed to orchestrate

the overall work of the system, invoking other

services at the right moment and offering them also

the right input, which service. This service in turn is

supported by an information service.

Based on this example, we can thus identify 5

services. These services can be included as building

blocks in a SOA-based platform in the area of home

energy management, in addition to middleware

functions that support the realization on particular

platforms and technologies.

5 RELATED WORK

In the domain of energy management, research

mainly focuses on either energy management on the

provider side or on management at the consumer

side. The latter is often referred to as 'demand side

management' and forms the main focus of this paper.

To the best of our knowledge, little work has

been done on demand side energy management

systems that consider context information, such as if

a person is present in a house or not. As mentioned

in Section 5, an important challenge is the

development of reasoning algorithms that can defer

relevant and reliable context information from raw

context data produced by low cost sensors. Roy et al.

(2006) propose an algorithm that makes location and

activity tracking in multi-inhabitant homes possible,

enabling an adaptive energy management scheme.

Our approach is more general, targeted to

architectural solutions, and could benefit from the

incorporation of such algorithms.

Most related approaches on demand side

management however do not consider context, but

monitor actual consumption and apply some form of

load-shifting.

Dynamic energy markets have been used to

stimulate the energy usage of users. These markets

allow users to actively buy energy for a certain time

slot, usually a couple of hours, while the price

changes dynamically with demand. This can be

exploited for load shifting, as has been shown by the

work of Faruqui and George (2005), McDonough

and Kraus (2007) and Hopper et al. (2006). In

contrast to our work, an open energy market that

allows energy contracts for short periods of time is

required for these approaches.

Stadler et al. (2009) consider cooling devices

which are switched off during peak usage and

switched back on when energy consumption usage

of the grid is low. The grid itself signals the devices

when either condition is met. In contrast to our

work, here all devices are either on or off, no mixing

of on and off devises is allowed.

Pournaras, Warnier and Brazier (2009) propose

EPOS, the Energy Plan Overlay Self-stabilisation

system. In this work TCAs are controlled by

software agent and organized in a tree overlay. The

global goal of stabilization emerges through local

knowledge, local decisions and local interactions

among the software agents. EPOS mainly deals with

the scalability issues that arise when thousands of

TCAs have to communicate with each other. Our

approach starts with energy management in

individual households, and scalability over more

devices should be more clustered, going from

households, to city blocks, to neighborhoods to

towns and regions.

Energy management has further mainly been

deployed in industrial settings. Middelberg, Zhang

and Xia (2009) propose an approach based on a

binary integer programming problem that addresses

the energy management of a colliery. A similar

integer programming model is proposed by Asok

(2006) for the energy management of steel plants.

In both cases the environment is relatively static and

no context information is required. Both approaches

lack adaptive behaviour and seem to be less

applicable to demand side energy management in

households.

6 CONCLUSIONS

In this paper, we have presented solution directions

that concern the application of two computing

paradigms to support home energy management,

namely Autonomic Computing and Context-aware

Computing. In particular, we have proposed an

energy management approach, based on ideas from

the mentioned paradigms, and using a service-

oriented architecture. We believe that this represents

a modest but useful contribution to finding technical

solutions for more advanced home energy

management visions, taking the following into

account as distinctive features of our solution

directions:

- well-established monitoring and control

mechanisms concerning energy-consuming

appliances;

- near real-time context-based adaptation during

a service delivery that allows for adjusting energy-

consuming appliances in such a way that adequately

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

312

suits the residents’ needs, by taking account of their

conditions;

- ease and flexibility in composing and

integrating application components.

Inspired by this work, we envision furthering our

research in the following directions: (i) continuing

with the development of the proposed approach at a

lower level, considering particular platforms and

technologies through which the approach could be

realized; (ii) exploratory case-study research that

would help considering our approach in real-life

context, which would be of great importance for

adding more practical insight as enrichment to our

ideas; (iii) re-visiting our peak-prediction vision, by

considering probabilities and statistics, for

supporting our system in a sound and reliable way;

(iv) acquiring more specific domain knowledge from

environmental organizations and energy companies

with the purpose of tuning our approach in such a

way that it is maximum useful in supporting real-life

problems; (v) analyzing this usefulness in different

ways, including through simulation, that would

provide valuable feedback for us as architects but

would also facilitate our discussions with domain

experts who would better understand our nicely

visualized ideas.

As for the realization of the proposed solution

directions, we envision several main challenges: (i)

scaling up to collections of households, taking into

account that our solutions would have to be repeated

at different granularity levels (e.g. household,

residential area, city); (ii) peak prediction indicators

should be identified, for reliably predicting

consumption peaks; (iii) algorithms are needed to

support the schemas for consumption decrease that

may be enforced; (iv) needs of the residents should

be considered carefully, in order to avoid irritations

that result from enforced consumption decreases; (v)

intelligent consumption scheduling and/or other

alternatives to consumption decrease should be

considered in this is better for the comfort of

residents.

REFERENCES

Ashok, S., 2006. Peak-load management in steel plants,

In: Applied Energy 83(5), 413 – 424.

Carvalho, Maria da Graca, 2009. Building a low carbon

society. In: 5th Dubrovnik Conf. on Sustainable Dev.

of Energy Water and Environm. Systems.

De Reuver, M., Haaker, T., 2009. Designing viable

business models for context-aware services,

Telematics and Informatics 26(3), 240-248.

Dey, A., Abowd, G.D., Salber, D., 2001. A conceptual

framework and toolkit for supporting rapid

prototyping of context-aware applications, HCI 16(2),

97-166.

Dockhorn Costa, P. and Ferreira Pires, L. and van

Sinderen, M.J., 2008. Concepts and architectures for

mobile context-aware applications. In: Research on

mobile multimedia. Inf. Science Ref., Hershey, NY.

Erl, T., 2005. Service-oriented architecture: concepts,

technology, and design, Prentice Hall PTR, NJ.

Faruqui, A. & George, S., 2005. Quantifying customer

response to dynamic pricing. In: The Electricity

Journal 18(4), 53–63.

Ganek, A.G. and Corbi, T.A., 2003. The dawning of the

Autonomic Computing era. IBM Systems Journal 42-

Hopper, N., Goldman, C., Bharvirkar, R., Neenan, B.,

2006. Customer response to day- ahead market hourly

pricing: Choices and performance, Util. Policy 14(2),

126–134.

IBM Corporation, 2005. An architectural blueprint for

Autonomic Computing. White Paper.

Kephart, J.O. and Chess, D.M., 2003. The vision of

Autonomic Computing. IEEE Computer Society.

Leymann, L., 2005. Combining web services and the grid:

Towards adaptive enterprise applications. CAiSE

Workshops (2), 9-21

Mazza, P., 2002. The smart energy network: Electrical

power for the 21st century. Climate Solutions.

McDonough, C. & Kraus, R., 2007. Does dynamic pricing

make sense for mass market customers? In: The

Electricity Journal 20(7), 26–37.

Middelberg, A., Zhang, J. & Xia, X., 2009. An optimal

control model for load shifting - With application in

the energy management of a colliery. In: Applied

Energy 86(7-8), 1266 – 1273.

Papazoglou, M., 2007. Web services: principles and

technology. Boston: Pearson Prentice Hall.

Pournaras, E., Warnier, M. and Brazier, F. M. T., 2009. A

Distributed Agent-based Approach to Stabilization of

Global Resource Utilization. In: Int. Conf. on

Complex, Intelligent and Software Intensive Systems

(CISIS'09).

Roy, N., Roy, A., Das, S.K., 2006. Context-aware

resource management in multi-inhabitant smart

homes: a Nash H-learning based approach, In:

PERCOM 2006), IEEE.

Schilit, B., Adams, N., Want, R., 1994. Context-aware

computing applications, In: WMCSA 1994), IEEE

Computer Society, 85-90.

Shishkov, B. and Van Sinderen, M.J., 2009. Service-

oriented coordination platform for technology-

enhanced learning. In I-WEST’09, 3rd Int. Workshop

on Enterprise Systems and Technology. INSTICC

Press.

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

2009. Modelling and evaluation of control schemes for

enhancing load shift of electricity demand for cooling

devices. In: Env. Modelling & Software 24(2), 285 –

295.

ON THE APPLICATION OF AUTONOMIC AND CONTEXT-AWARE COMPUTING TO SUPPORT HOME ENERGY

MANAGEMENT

313