INTELLIGENT HOUSEHOLD ENERGY MANAGEMENT

RECOMENDER SYSTEM

Nazaraf Shah, Chen-Fang Tsai

Department of Computing and the Digital Environment, Coventry University, Coventry, U.K.

Department of Industrial Management, Aletheia University, Taipei, Taiwan

Kuo-Ming Chao, Chi-Chun Lo

Department of Computing and the Digital Environment, Coventry University, Coventry, U.K.

Institute of Information Management, National Chiao-Tung University Hsin-Chu, Hsinchu, 300, Taiwan

Keywords: Rule Based System, Ontology, Energy Management, Wireless Network, Energy Consumption Monitoring.

Abstract: Recent years have seen extensive research in home energy management systems to address the issues of

rising energy prices and global warming. The focus of these research efforts is to create a smart

environment which integrates household energy consumption appliances and devices into a home area

network. This home area network collects energy consumption data constantly in real time in order support

data analysis, decision making and enable the householders to have a transparent view of their energy

consumption. The ultimate goal is to use Information and Communication Technologies (ICT) to help

householders to reduce their energy consumption while maintaining level of their comfort. The proposed

recommender system is a subsystem of an integrated energy management system which involves innovative

technologies to monitor and analyse energy consumption of households in real time and enables them to

have more detailed picture of their energy consumption and also provide them advice on efficient energy

usage. The recommender system is supported by the monitoring system which consists of a network of

energy consumption monitoring sensors. These sensors read energy consumption of household appliances in

real time and send the data to a central server for storage, analysis and query purposes. In this paper we

present a recommender system which provides advice to householders proactively by taking in account their

energy consumption patterns and also provides answers to their queries regarding efficient use of energy.

1 INTRODUCTION

The number of appliances, lightings and other

electronics equipments being used in household are

continue to rise. This increase is due to

householders’ continuous desire for increased

comfort, convenience and security. According to

International Energy Outlook the projected growth

of world energy demand is 44 % from 2006 to 2030

(IEO). It is a challenging issue to meet such demand

while minimising CO2 emission and optimising the

energy consumption.

UK domestic sector is responsible for 30% of the

total energy consumption [DTI]. In absence of

efficient energy consumption management we

expect a gradual increase in domestic energy

consumption due increased use of energy

consumption appliances and increase in population.

One of the effective ways to address this challenge is

to minimise the energy consumption by optimising

the energy consumption and influencing

householders’ behaviour towards their energy

consumption. In this paper we present an energy

consumption recommender system. The proposed

system aims at helping householders to take control

of their energy consumption by allowing them to

have appliance level view of energy consumption

and providing them advice on how to reduce energy

consumption without scarifying their comfort level.

The proposed system is a part of a larger project

“The Digital Environment Home Energy

Management System” (DEHEMS) (DEHEMS)

which is FP7 EU funded project. The objective of

DEHEMS is to investigate various ways of using

Shah N., Tsai C., Chao K. and Lo C. (2010).

INTELLIGENT HOUSEHOLD ENERGY MANAGEMENT RECOMENDER SYSTEM.

In Proceedings of the Multi-Conference on Innovative Developments in ICT, pages 51-56

DOI: 10.5220/0003046200510056

 SciTePress

ICT to improving household energy efficiency and

reduce CO2 emission and save money. The

DEHEMS aims at putting households in control of

their energy consumption by enabling them to view

and understand their energy consumption patterns

and helping them actively to reduce their energy

usage, costs and carbon emissions.

AIM (AIM) is another FP7 funded project for

design and implementation of a system that aims to

minimise energy waste in a domestic environment.

In contrast to DEHEM the focus of the AIM is to

exploit the use wireless sensor monitoring network

to control home appliances according to user profiles

(Barbato, 2009).

Rui et.al, proposed an architecture for home

energy appliances management and control (Rui ,

2005). Their proposed system is more focused on

use of hardware components such as sensors

actuators and communication network to manage

energy consumption in home environment. Another

strand of research focuses on providing intelligent

interfaces to increase awareness of energy usage and

hence influence the house holder’s behaviour (Jussi ,

2008, Wood 2007).

There are a number of freely available web

based tools for providing householders advice on

their energy consumption (PowerMeter, Hohm, i-

measure). There are also a number of commercial

ICT based energy management system available

(Plugwise, Plogg, Agilewaves). These tools and

systems broadly focus on issues of energy

consumption monitoring, displaying energy

consumption data and basic statistical analysis of the

data. The recommender system on the other hand

semantically encodes the energy consumption

activities of home environment and provides

intelligent and tailored energy saving advice to

householders using heuristic rules.

The paper is organised as follow. In section 2 we

briefly describe the DEHEMS system high level

architecture and its features. Section 3 discusses

the proposed energy saving recommender system

and its components and finally section 4 concludes

the paper.

2 DEHEMS SYSTEM

ARCHITECTURE

Energy consumption monitoring functionality is at

the heart of DEHEMS system, it provides essential

energy consumption information to be used by

different subsystems in DEHEMS. In this section we

briefly describe the DEHEMS high level

architecture as shown in figure 1.

The DEHEMS is based on a sensors network of

energy consumption measuring sensors. Zigbee

protocol is used for networking and data exchange in

DEMEMS system. The network has ability to

seamlessly integrate other Zigbee compliant sensors

into the system as required. The sensors collect

energy consumption data of electrical appliances

every three minute and send the data to the local

data collector which in turn forwards the data to the

central server. In the next stage of the DEHEMS

project gas consumption measuring sensors,

occupancy sensors and temperature sensors will be

incorporated into system in order to measure gas

consumption of space heating, water heating and

cooking.

The real time collection of data makes it possible

to understand correlation between appliances,

statistical analysis, intelligent advice generation and

various kind of query support. It also allows

householders to see the effect their energy

consumption activities in real time.

Figure 1: DEHEMS System Architecture.

The sensors attached to an electrical power

consuming appliances has sensory, limited

computation, and wireless communication

capabilities. These sensors form a Zigbgee mesh

network and a coordinator node coordinates the

communication between data collector and sensor

network.

Each sensor in the network has a unique identity

that is used to identify its associated appliance. This

identity is sent to a central server along with every

INNOV 2010 - International Multi-Conference on Innovative Developments in ICT

energy consumption reading to enable the server to

identify each appliance data uniquely.

The central server stores the energy consumption

data over a long period of time in order to enable the

households to view their history of appliances level

energy consumption and enables the recommender

subsystem to generate efficient usage advice based

on appliance level energy consumption data.

The operation of the DEHEMS system is

organised into three layers, know as service demand

layer service broker layer and service provider layer.

The service demand layer receives input from and

provides feedback to the householder via user

interface. The service broker receives the requests

from service demand layer and converts them into

service requests for the service provider layer. The

service provider layer comprises a semantic layer

which generates various applicable options to

address service requests from the service broker.

The proposed recommender system encodes the

energy consumption activities and their relationship

semantically making it a component of the semantic

layer.

3 ENERGY USUAGE

RECOMENDER SYSTTEM

The proposed recommender system defines a

knowledge base of energy saving tips. These tips are

classified based on their characteristics and the

energy consumption activities they belong to. Such

classification enables the recommender subsystem to

produce focused and intelligent energy saving

advice in response to the user’s queries and their

energy consumption behaviours.

Householders’ engagement with the system is an

essential factor to make them aware of the

consequences of their energy consumption

behaviour and hence influence their behaviour

towards efficient energy usage. For example a

when a householder asks the system to provide

energy saving advice on washing activity. The

system interactively asks user questions to acquire

essential data to produce a more accurate and

intelligent advice rather than providing random tips

on washing. The recommender system is also able to

acquire this data from the system once enough

statistically significant is available in the system.

Although the system provides user choices for

getting more specific advice, but it also allows the

users to get general advice on energy saving

regarding specific activity. For example in case of

washing activity the system may ask a householder

if he/she wants advice on washing temperature,

washing load, fabric types or overall advice on

washing and then generates advice based on his/her

response. In case the householder wants advice on

washing temperature the system then ask him/her

about their current temperature setting (eventually

these values will be acquired from appliance

ontology). The system then perform reasoning to

conclude applicable piece of advice and the amount

of energy that the householder would be able to save

by changing washing temperature to a suggested

temperature. This process is depicted in figure 2. A

simple formula below shows that one of three

advices will be picked based on value of the washing

temperature supplied.

output = advice (

(Tw> Ti) | (Tw < Ti) | (Tw = Ti))

Where Tw is current washing temperature and Ti is

ideal washing temperature.

The recommender system also informs the

household about the effect of one energy

consumption activity to another energy consumption

activity. For example if the tumble dryer is left to

over dry the clothes this will have effect on ironing

activity as the ironing of over dry clothes causes

increased consumption of energy in ironing

activity.

A proactive function of recommender system is

to display the energy saving advice concerning

current activities being performed by the household.

When a household log into DEHEMS system the

recommender detect his/her current energy

consumption activities and displays the energy

saving advice in a context sensitive way.

The recommender system has access to an

ontology which defines the various energy

consumption activities in a home environment and

their associated energy saving tips. The main

objective of the recommender subsystem is to

enhance the household engagement with the system

by providing them customised and context specific

advice on their energy consumption there by

influencing their energy consumption behaviour.

The energy consuming appliances in domestic

environment vary greatly in terms of their efficiency

size and operating characteristics. Such variations

make it difficult to produce one-size fit all energy

saving advice. In order to address this issue the

recommender system also makes use of energy

consuming appliances ontology. The energy

consumption appliances ontology enables the

recommender system to take into account various

INTELLIGENT HOUSEHOLD ENERGY MANAGEMENT RECOMENDER SYSTEM

characteristic of the appliances while generation

energy saving advice for householders.

Figure 2: Advice generation process.

The declarative knowledge of the energy

consumption activities in domestic environment is

encoded using protégé [protégé] knowledge-

framework. The encoded knowledge is made of

energy consumption related concepts and

relationship between these concepts. Jess [Jess] rule

engine is used to encode reasoning rules which

operate on knowledge encoded in the ontology.

The problem solving method used by the

recommender system employs heuristic

classification approach (Clancy, 1985).

The following two subsections briefly describe

the ontology of energy consumption activities of and

heuristic classification approach used in

recommender system.

3.1 Classification of Home Energy

Consumption Activities

There are various ways in which energy

consumption activities of household appliances can

be classified. Figure 4 represents an activity based

appliance classification in which appliances energy

consumption activities are grouped together based

on a high level energy consumption activity.

These activities are treated as domain’s concepts

and represented by frames in ontology. We have

used following three types of links to create

relationships amongst the concepts.

a. Instance: An instance represents is-a

relationship between a concrete instance and

its associated concept. For example

relationship between “clothes washing”

concepts and its all concrete instances.

b. Sub-class: A sub-class relationship

represents the child parent relationship.

c. Property: It is used to represent

characteristics of a concept.

Figure shows 3 a frame and its relation to properties

and other frames. Each energy consumption activity

has its start time end time and its energy

consumption during this period. As shown in figure

3 the clothes washing activity has

hasClothWashingTip relationship to

clothWahingTips frame, which in turn has slots

relating washing temperature and load for which this

tip is appropriate to. Every energy consumption

activity concept has at least one energy saving tip

associated with it and there is no upper limit on

number of energy saving tips that an energy

consumption activity can have.

The hasEffect slot represents the other energy

consumption activities that could be affected by this

activity, i.e. clothes not properly spun by clothes

washing activity have effect on energy consumption

of the tumble dryers.

Frame:

{

Clothes

Washin

hasEffectOn: washing

washin

Temp: inte

startTime: Strin

endTime: String

consumption: inte

standb

: inte

hasWashingTips: Frame

Frame:

{

ClothWashin

Tips

}

hasLoad: in

hasTemperature: in

description: strin

Priority: in

Figure 3: Ontology Frames and Slots Relationships.

3.2 Heuristic Classification Approach

The reasoning process in recommender system is

based on heuristic classification (HC) problem

solving method. HC is a well understood and a

widely used problem solving method. A number of

well-known expert system are based HC (Shortliffe,

1976, Bennett, 1978).

In HC approach, programs employ an inference

structure that systematically relates data to a pre-

enumerated set of solutions by abstraction, heuristic

association and refinement (Clancy, 1985). A

heuristic classification approach includes four main

components in its knowledge base : data, data

INNOV 2010 - International Multi-Conference on Innovative Developments in ICT

Figure 4: Activity based classification of energy consumption appliances.

abstractions, solution abstractions and solutions.

When data/symptoms are observed, the system

populates data/symptoms to the data abstraction; the

data abstraction then matches the solution

abstraction; and refines the solution. For example in

energy consumption advice regarding temperature of

clothes washing activity, the temperature of washing

machine provides the data to be used in HC

inference process. The temperature value is then

abstracted to normal, abnormal or below normal data

abstractions. This data abstraction is then matched to

solution abstraction which is the hierarchy below

washing activity. Then solution refinement is

applied considering all tips related to washing

temperature and advice is presented to household

based on best fit.

Pseudo code of one of a Jess heuristic rule about

advice regarding washing machine temperature is

shown below.

Condition (LHS)

(temp← WashingTemperature)

( []applicableAdvice ← hasTips )

( energyConsumption ← Consumption)

Action (RHS)

(for each

advice ← [] applicableAdvice

if ( temperature(advice) > temp)

[] solutions ← advice

else if ( temperature(advice) < temp)

[]solutions← advice

else

solutions ← advice

)

selected_solution ← (selection (solutions,

energyConsumption, washingMachineModel) )

4 DISCUSSION

AND CONCLUSIONS

In this paper we have presented energy consumption

recommender system. The recommender system is a

part of DEHEM project which aims to influence

householder energy consumption in order enable

householder to make efficient use of energy

consumption and reduce CO2 emission. The

novelty of the recommender system lies in its ability

to generate customised and focused energy

consumption advice using HC approach. We have

implemented ontology of energy consumption

activities in home environment which separates

energy consumption activities from appliances

ontology. Since the proposed system is a part of

ongoing research, the system will be refined using

statistical machine learning approaches using data

collected over period of more than six months.

INTELLIGENT HOUSEHOLD ENERGY MANAGEMENT RECOMENDER SYSTEM

ACKNOWLEDGEMENTS

This research is carried out as a part of DEHEMS

which is funded from the European Community’s

Seventh Framework Programme FP7/2007-2013

under grant agreement No.224609.

REFERENCES

IEO International Energy Outlook, 2009,

http://www.eia.doe.gov/oiaf/ieo/pdf/0484(2009).pdf

Protégé, http://protege.stanford.edu/

Jess, http://www.jessrules.com/

Clancy W. J., 1985, Heuristic Classification, Artificial

Intelligence 27 Elsevier Science Publishers, 1985, pp.

289-350.

Shortliffe E. H., 1976, Computer Based Medical

Consultations: MYCIN, New York, Elsevier.

Bennett J., Creary L., Englemore R., Melosh R., 1978,

SACON: A Knowledge-Based Consultant for

Structural Analysis”, STAN-CS-78-699, Stanford

University, CA.

PowerMeter, Google PowerMeter, http://www.google.

com/powermeter/about/about.htmlHohm, Microsoft,

http://www.microsoft-hohm.com/

i-Measure, Oxford University, http://www.imeasure.

org.uk/index.php

Plugwise, http://www.plugwise.com/en/domestic/home-use

Plogg, http://www.plogginternational.com/ploggproducts

.html

Agilewaves, http://www.agilewaves.com/resource-monitor-

multi-unit-residential/

AIM, http://www.ict-aim.eu/home.html

Barbato A., Luca Borsani, Antonio Capone, Stefano

Melzi, 2009, Home Energy Saving through a User

Profiling System based on Wireless Sensors, First

ACM Workshop On Embedded Sensing Systems For

Energy-Efficiency In Buildings, USA

Rui SarnadasPaul Fonseca, J. Paulo Teixeira, Isabel

Teixeira, Antonio Macedo Silva, Alexandre Correia,

João Correia, Henrique Serra, Antonio Gano, A.

Miguel Campos, 2005, Intelligent Architecture for

Home Appliances and Energy Management Control,

Conference on Design of Integrated Circuits and

Systems, Lisbon.

Jussi Karlgren, Lennart E. Fahlén, Anders Wallberg, Pär

Hansson, Olov Ståhl, Jonas Söderberg, Karl-Petter

Åkesson, 2008, Socially Intelligent Interfaces for

Increased Energy Awareness in the Home, Internet of

Things, Springer-Verlag Berlin.

G. Wood, M. Newborough,2007, Energy-use information

transfer for intelligent homes: Enabling energy

conservation with central and local displays, Energy

and Buildings Volume 39, Issue 4,pp. 495-503

DTI. UK, Energy consumption in the UK http://

www.bis.gov.uk/files/file11250.pdf

DEHEMS, Digital Environment Home Energy

Management System, http://www.dehems.eu/.

INNOV 2010 - International Multi-Conference on Innovative Developments in ICT