Content uploaded by Qazi Zafar Iqbal

Author content

All content in this area was uploaded by Qazi Zafar Iqbal on Jan 09, 2018

Content may be subject to copyright.

Ant Colony Optimization based Energy

Management Controller for Smart Grid

Sahar Rahim1, Zafar Iqbal2, Nusrat Shaheen1, Zahoor Ali Khan3,4,

Umar Qasim5, Shahid Ahmed Khan1, Nadeem Javaid1,*

1COMSATS Institute of Information Technology, Islamabad, 44000, Pakistan

2UIIT, Pir Mehr Ali Shah Arid Agriculture University, Rawalpind 46000, Pakistan

3Faculty of Engg., Dalhousie University, Halifax, NS B3J 4R2, Canada

4CIS Higher Colleges of Technology, Fujairah 4114, United Arab Emirates

5Cameron Library, University of Alberta, Edmonton, AB, T6G 2J8 Canada

∗Corresponding author: www.njavaid.com, nadeemjavaidqau@gmail.com

Abstract—In this paper, we introduce a generic architecture

for demand side management (DSM) and use combined model

of time of use tariff and inclined block rates. The problem

formulation is carried via multiple knapsack and its solution is

obtained via ant colony optimization (ACO). Simulation results

show that the designed model for energy management achieves

our objectives; it is proven as a cost-effective solution to increase

sustainability of smart grid. The ACO based energy management

controller performs more efﬁciently than energy management

controller without ACO based scheduling in terms of electricity

bill reduction, peak to average ratio minimization and user

comfort level maximization.

Keywords: Smart Grid; Demand Side Management; Multiple

Knapsack Problems; Time Of Use Tariff; Inclined Block Rate;

Renewable Energy Sources.

I. INTRODUCTION

TRADITIONAL electrical power system is inadequate to

meet modern power grid challenges such as reliability,

stability, robustness, etc. [1]. Thus, a new infrastructure is

needed to smartly meet these challenges and reduce pressure

on global environment. In this regard, smart grid (SG) in-

tegrates communication technologies, computational abilities,

control systems and sensors with existing grid and enables two

way ﬂow of information between utility and end users. Main

aims of SG are to enhance efﬁciency, sustainability, capacity

and customer engagement [2].

One of the important aspects of SG is demand side man-

agement (DSM) which is the best way to maintain balance

between demand and supply. Two main functions of DSM are

load management and demand response (DR). Load manage-

ment focuses on the improvement of energy efﬁciency [3],

while DR is a responsive action taken by a customer against

dynamic price models [4]. The common objectives of SG are

electricity bill reduction, minimization of aggregated power

consumption and minimization of both electricity bill and

aggregated power. To achieve these objectives, many DSM

techniques and algorithms are proposed in the previous years;

integer linear programming [5], mixed integer linear program-

ming [6], mixed integer non-linear programming [7], convex

programming [8], etc. However, these techniques can not

tackle large number of different household appliances having

unpredictable, non-linear and complex energy consumption

patterns due to randomness in human behavior. Moreover, to

attain electricity cost minimization objective, they ignore user

comfort level and their electricity pricing model is also not

compatible with real scenarios.

In this paper, ant colony optimization (ACO) technique is

used for DR (in DSM) due to its exceptional characteristics;

ﬂexibility for speciﬁed constraints, ease of implementation,

low computational complexity and low computational time

[9]. We ﬁrst design an energy management controller (EMC)

model smart homes using multiple knapsack problem (MKP)

and then apply ACO to get feasible solution for designed

objective function. To calculate electricity bills, we use time

of use (TOU) tariff model with inclined block rate (IBR), so

that peak formation is avoided. Effectiveness of the designed

EMC model is shown via simulations where unscheduled and

ACO based schedules cases are compared in terms of energy

consumption pattern, electricity bill, peak to average ratio

(PAR), user comfort level, and execution time.

The rest of the paper is organized as follows. Section II

brieﬂy describes related work. Section III explains motivation

for proposed work. Section IV describes system model and

section V deals with problem formulation. Simulation results

are discussed in section VI. Finally, paper is concluded in

section VII by pointing out the future work.

II. RE LATE D WO RK

In [15], authors investigate the problem of household appli-

ance scheduling to enhance energy efﬁciency of electrical grid

and provide beneﬁts to end users. They proposed a solution

that optimally schedules a set of appliances. To minimize

customer electricity bills and maintain energy consumption

within a limit, they use day-ahead variable peak pricing model

and map their problem by using MKP. By limiting the energy

demand within certain capacity, problem of load shedding

can be removed. Results show that this model effectively

Fig. 1: SG architecture

reduces utility electricity bills while keeping power consump-

tion within pre-deﬁned limits. Another model of home energy

management controller for residential users is proposed in

[16]. Objective function is formulated by knapsack problem

and dynamic programming approach is used to solve problem

and to set consumer preferences for each appliance. These

priorities were the value of appliances that are used to schedule

the appliance to satisfy their operational time constraints to

avoid peak formation and to reduce electricity cost.

In [10], authors present an efﬁcient model of DSM that

reduces PAR and electricity bills for residential, industrial

and commercial users. Scheduling problem is formulated as

a minimization problem and then problem is evaluated by

using heuristic evolutionary approach. Heuristic algorithms

show better results because of their ﬂexible nature that allow

the implementation of individual load pattern in order to

minimize inconvenience. Proposed model is beneﬁcial for

both utilities and customers in a way that PAR reduction

causes minimization in the number of peak power plants

while incentive based model helps consumer to reduce their

electricity bills. Simulation results show that the proposed

DSM strategy achieves signiﬁcant savings, while reducing the

peak load demand of the smart grid.

In [11], authors discuss an efﬁcient architecture for energy

management system by using home area network (HAN) for

residential users. They combine real-time pricing (RTP) tariff

model with the IBR because when only the RTP is adopted,

there is a risk that most of the appliances operate during the

hours of lowest electricity price that cause peak formation. To

strengthen the stability of electricity system, peak formation

must be avoided. To solve these issues in an optimized way,

objective function is formulated. As this kind of optimization

problem is non-linear, therefore they use GA to optimize their

problem. Simulation results shows that proposed model is very

effective to reduce PAR and electricity cost.

Another DSM model is proposed in [12] for residential users

to reduce PAR and electricity bill minimization. GA is used

to get optimal start time of each appliance in each time slot

while satisfying its operational constraints. There is a tradeoff

between electricity cost and waiting time. When waiting time

of an appliance is zero, its electricity cost is increased and vice

versa. Combined model of RTP with IBR is used to avoid peak

formation. Simulations are carried out for single and multiple

users. Results show the effectiveness of proposed DSM model

for both single and multiple user scenarios.

An efﬁcient heuristic approach is presented in [17] for

scheduling of smart appliances in residential area. The pro-

posed algorithm is evaluated by comparing the electricity cost

and computational time with an exact algorithm. Variable

energy price model is used for scheduling of appliances.

Hourly prices for electricity, the operation start times of set of

appliances are optimized to reduce cost of energy consumption

while satisfying the operational and peak power constraints.

Results show that electricity cost obtained by heuristic algo-

rithm is within 5% of the optimal cost of exact algorithm

whereas computational time is reduced by exponential factor.

III. PROP OS ED S YS TE M MO DE L

In SG, DSM enables more efﬁcient and reliable grid op-

erations. Its two main functions are energy management and

demand side control activities for end users. In residential area,

every smart home is equipped with EMCs and smart meters

to make stable and reliable bi-directional communication be-

tween utilities and customers. All elements, such as electri-

cal appliances, sensors, local generation and energy storage

systems (ESSs) give their information to EMC through HAN

and EMC controls scheduling of appliances. After collecting

all information, EMC sends it to SG domain through WAN.

There are various wireless solutions for communication links

between the smart meters and the EMCs such as ZigBee, Z-

Wave, Wi-Fi, or a wired (HomePlug) protocol [1]. Simple

architecture of DSM is shown in ﬁg. 2. In residential area

Sensors

Distributed RESs

Smart devices

Residential area

domain

ESSs

EMCs

HAN

SG domain

Distribution

Operation

Market

Service provider

Customers

WAN

Two way communication

One way communication

Fig. 2: Components of DSM

based DSM, we consider Nsmart homes and Msmart

appliances. In this model, all smart homes have smart metering

system and EMC. End users change their energy usage accord-

ing to incentive based schemes offered by utilities. In each

home, consumer inputs different parameters of appliances to

appliances scheduler and then appliance manager gives signal

to various appliances about their on/off status. For electricity

pricing model, TOU tariff is used to calculate electricity bill

against the energy consumption cost per day. In order to

design the optimization model for home energy management,

we have categorized the load according to the characteristic

of appliances and life style of end users as discussed in the

following section.

A. Load categorization

We classify appliances into three categories; ﬁxed, shiftable

and elastic appliances according to their power consumption

pattern and time of use [18]. Detail of all these categories is

given as follow:

1) Fixed appliances: These are also called regular appli-

ances because their usage or length of operation can not be

modiﬁed. For example, lights, fans, clothes iron, microwave

oven, toaster, tv, etc. We represent ﬁxed appliances by Fed and

its power consumption as ν.

2) Shiftable appliances: These are also called burst load

because these are manageable and can be shifted in time

without altering their load proﬁle. For example, washing

machine, dish washer, clothes dyer, etc. We denote shiftable

appliances by Sed and their power consumption by ∆. Each

shiftable appliance is characterized by its length of operation

which is denoted as τsed and it is pre-deﬁned by end users

each day.

3) Elastic appliances: These are also called interruptible

appliances because these are fully controllable in terms of both

usage time and power consumption proﬁle. For example, air

conditioner, refrigerator, water heater, space heater, etc. We

represent elastic appliances by Eed and its power consumption

is denoted by κ. Each elastic appliance eed ∈Eed has power

rating ρeed , power quantity factor λeed, length of operation

τeed , start time αeed and end time βeed . These attributes are

set by the consumer.

B. Energy consumption model

Let A={a1, a2, a3, . . . , am}be the set of appliances

such that a1,a2,a3,· · · ,amare number of appliances that

belong to each category. If t∈T={1,2,3,· · · ,24 }de-

notes the scheduling horizon, then hourly energy consumption

demand of a appliance is given as,

Ea(t) = {Ea

t1+Ea

t2+Ea

t3+. . . +Ea

t24 }(1)

where, Ea

t1,Ea

t2,Ea

t3,· · · ,Ea

t24 denotes energy consumption

demand of each appliance in the respective time slots. The

per day total energy consumption demand for all appliances

is calculated as follows,

ET=

24

X

t=1 A

X

a=1

E(i,t)(2)

C. Energy price model

A number of tariff models are available to deﬁne electric

energy prices for a day or for short time duration. Among

these, TOU tariff model is deﬁned for electricity prices depend

on the time of day and are pre-deﬁned in advance. Critical

peak pricing (CPP) is a variant of TOU in which price is

considerably raised in case of emergency situations (e.g. high

demand). RTP based electricity prices can change as often

as hourly, reﬂecting the utility cost of supplying energy to

customers at that speciﬁc time. In our model, we use TOU

with power dependent tariff known as inclined block tariff or

IBR. The energy price at time tis an increasing, piecewise,

linear function of the total energy demand. As E(t)is the total

power consumption of all appliances in a home at each time

slot tand it is calculated as,

E(t) =

24

X

t=1 ν(t) + ∆(t) + κ(t)(3)

To calculate electricity bills, energy price for each unit con-

sumed in each time slot is represented by C(t)and according

to IBR model, it is designed as,

C(t) =

C1(t) 0 ≤E(t)≤E1

th(t)

C2(t)E1

th(t)≤E(t)≤E2

th(t)

C3(t)E2

th(t)< E(t)

(4)

where, E1

th and E2

th are power consumption thresholds and

C1,C2and C3are costs for these particular cases.

D. Residential users

We design our model for three types of users in residential

area; passive, semi-active and active users.

1) Passive users: They only consume electrical energy of

the grid and does not generate or store electrical energy.

They can only shift there load from high peak to low peak

and reduce their electricity bills. The set of passive users is

represented by P.

2) Semi-active users: They have RESs such as solar panels

and wind turbines. They consume energy both from power grid

and RES to reduce their electricity bills. The set of semi-active

users is represented by S.

3) Active users: They take energy from RES and store it in

storage devices such as batteries as well as also take electrical

energy from grid to fulﬁll their need. The set of active users

is represented by A.

IV. PROBLEM FORMULATION

In this work, main objectives are to reduce consumer cost by

optimizing the energy consumption patterns of appliances to

maximize the comfort level of end user. Here, we formulate

our scheduling problem by using MKP. MKP is a resource

allocation problem that consists of “M” resources (capacities)

and set of “N” objects [19]. We take “j” number of knapsacks,

and map our scheduling problem in MKP as follows:

•We consider “j” number of knapsacks as power capacities

in each time slot.

•Number of appliances as number of objects.

•The weight of each object as the energy consumed by

appliances in each time slot. Note that it is independent

of “t”.

•The value of object in a speciﬁc time slot is the cost of

power consumption of the appliance in that time slot.

•The value of binary variable “χ” can be 0 or 1 depending

on the state of electrical appliance.

The total power consumption for all types of appliances

should not exceed maximum power capacity in each hour

denoted as γ(t), we introduce constraint which limits the

power consumption and depends on load proﬁle and its states.

Constraints show that power consumption is predeﬁned,

24

X

t=1 E(t)×χ(t)≤γ(t)(5)

Here, γ(t)is the power capacity in each hour that is available

from grid and χ(t)∈[0,1] denotes the states of appliances.

Total power consumption in each hour must be limited to this

capacity factor.

A. Objective function and its solution via ACO

The overall objective function of our scheduling problem is

to minimize electricity bill with optimal use of power from

grid and to minimize waiting time (to avoid frustration of

end users). Additionally, optimal integration of RESs is also

a key point to reduce green house gas (GHG) emission. We

formulate our objective function as an optimization function

and is modeled as,

min

24

X

t=1 a1·

A

X

a=1

(Ea(t)×Ca(t))+a2ϕa(t) (6)

where, Cais the electricity cost in each time slot that must be

minimized while keeping waiting time of shiftable appliances

minimized. a1and a2are weights of two parts of objective

function and their values are a1, a2∈[0,1] or a1+a2= 1. It

shows that either a1or a2would be 0 or 1. In this work,

our major concern is with electricity cost reduction with

maximizing comfort level of end users. For this purposed

model, we assume waiting time of each shiftable appliance

not greater than 5, if operation start time of an appliance is

greater than our assumption then utility pays penalty.

Algorithm 1 : Improved Algorithm of ACO-EMC

1: Initialize all parameters (αa,βa,τa,ρa)

2: For all users n ∈N do

3: For all appliances a ∈A do

4: For all time slots t ∈T do

5: Randomly generate ant population

6: while Maximum number of iterations and min error not reached

do

7: For Each individual ant update pheromone refer [21]

8: For Each individual ant evaluate the objective function using

(29)

9: if Ea< E1then

10: calculate electricity bill using C1

11: else {E1< Ea< E2}

12: calculate electricity bill using C2

13: else {Ea> E2}

14: calculate electricity bill using C3

15: end if

16: if C(t)is high peak hour then

17: calculate ϕausing (28)

18: else

19: start an appliance

20: end if

21: local update pheromone for each ant refer [21]

22: choose best solution so far

23: global update pheromone for each ant refer [21]

24: repeat until iteration end

25: using Θ(t)when electricity bill is high

26: if E(t)is high then

27: Θ(t)energy

28: else

29: E(t)

30: end if

31: end while

ACO is a meta-heuristic optimization approach that is used

to solve discrete combinatorial optimization problems. It has

unique properties of self-healing, self-protection and self-

organization [13]. In literature, ACO is used for DSM in many

ways. For-example, authors in [14], investigate congestion

management and cost minimization problems. They formulate

their focused problem as a non-linear programming problem

and electricity bill minimization is achieved using ACO. To

our knowledge, ACO implementation in residential area is not

done before. In our work, we use ACO to evaluate the designed

optimization function to get optimized schedules for home

appliances. Our scheme gives novel idea to implement ACO as

optimization tool for DSM in residential area. In [20], linear

programming is used to designed the optimization function.

Refer to [21], we modiﬁed its algorithm for our designed

scenario. Algorithm. 3 gives detailed view of ACO based

EMC (ACO-EMC) model. ACO is used to evaluate objective

function (refer eq. 29) and its constraints (refer eq. 29a to eq.

29i) to get feasible operational time for all appliances. Our

proposed model is applicable for single and multiple homes

in residential areas. The improved ACO algorithm is shown

in algorithm 1.

V. SI MU LATI ON S AN D RE SU LTS

To evaluate different performance metrics of EMC, we

conduct extensive simulations in MATLAB. We use TOU tariff

model of Jemena Electricity Networks (VIC) Ltd [22], [23]

for residential area with IBR. For simulations, we design a

model for residential area in which each home is equipped

with 10 smart appliances and 4 end users. Appliances with

their parametric values that are used in simulations are shown

in table. I, table. II, and table. III, respectively. In table. I, ﬁxed

appliance has only ρaparameter measured in kWh because

these are non-manageable appliances and do not play any role

in load scheduling problem. Whereas, other two categories

TABLE I: Parameters of Fixed Appliances

Appliances ρa(kWh)

Lighting 0.6

Fans 0.75

Clothes iron 1.5

Microwave oven 1.18

Toaster 0.5

Coffee maker 0.8

of appliances; shiftable and elastic appliances are known as

schedulable appliances. As, in table. II, the parameters for

shiftable appliances are αa,βa,ξa,ϕaand ρaare kWh. ϕa

is the unique parameter in shiftable appliance because these

appliances can be interruptible during its length of use. For

elastic appliances, the parameters are αa,βaand ρain kWh

are shown in table. III.

TABLE II: Parameters of Shiftable Appliances

Appliances αa

(hours)

βa

(hours)

ϕa

(hours)

ρa

(kWh)

Washing machine 8 16 5 0.78

Dish washer 7 12 5 3.60

Clothes dyer 6 18 5 4.40

TABLE III: Parameters of Elastic Appliances

Appliances αa

(hours)

βa

(hours)

ρa

(kWh)

Air conditioner 6 24 1.44

Refrigerator 6 24 0.73

Water heater 6 24 4.45

Space heater 6 24 1.50

TABLE IV: ACO parametric list

Parameters Values

Ant quantity 10

Pheromone intensity factor 2

Visibility intensity factor 6

Evaporation rate 5

Trail decay factor 0.5

Stopping criteria Max. iteration

Max. iteration 600

Simulation parameters of ACO-EMC are given in table. IV,

respectively.

A. Electricity bill reduction

The maximum value of electricity bill in unscheduled model

is 266.3492 cent as shown in ﬁg. 3. It is reduced to 114.2536

cent in ACO-EMC. During peak hours (16-22), sufﬁcient

electricity cost reduction is shown for the designed ACO-EMC

model. ACO-EMC acts more effectively than the unscheduled-

EMC BPSO-EMC in achieving our designed objective of

electricity cost reduction due to its characteristics of local and

global exploration.

Fig. 3: Electricity bill (cent)

B. PAR

Performance of the designed model (ACO-EMC) with re-

spect to PAR reduction is shown in ﬁg. 4. It shows that PAR

is signiﬁcantly reduced for ACO-EMC as compared to the

unscheduled case because these are designed to avoid peak

formation in any hour of a day. Results prove that our proposed

model effectively tackle the peak formation problem. PAR

curves for ACO-EMC describe that power consumption of

appliances is optimally distributed in 24 hours without creating

peak in peak hours (16-22) of a day. We have used combined

model of TOU and IBR for electricity billing to avoid peak

formation via giving information to consumers.

C. Waiting time

User comfort is related to both electricity bill and waiting

time of an appliance. In order to achieve lower electricity

bills, smart users must operate their appliances according

to optimal schedule of EMC. During scheduling horizon of

shiftable appliances, operational time is not ﬁxed due to price

variation in dynamic pricing models. Generally, it is observed

that electricity cost reduction and waiting time show inverse

relationship. By applying waiting time constraints on the

objective function, we have enhanced the performance of EMC

in terms of user comfort and electricity bill reduction. In ﬁg. 5,

it is shown that electricity bill is high if rate of waiting time

is zero and it is low with increase in rate of waiting time for

the proposed model.

Fig. 4: PAR curve

Fig. 5: Possible trade off between electricity cost and waiting

time

VI. CONCLUSION AND FUTURE WO RK

In this paper, we have presented an efﬁcient DSM model

for residential energy management system in order to avoid

peak formations while decreasing the utilities electricity bill

by preserving user comfort level within acceptable limits. We

used ACO to solve our objective function and used combined

pricing models, TOU tariff and IBR model for electricity bill

calculation. From the results, it is clearly justiﬁed that our

proposed model works efﬁciently in terms of electricity bill

reduction, and minimization of PAR while considering user

satisfaction.

In future, we will focus on human behavior to achieve

comfort level of consumer and to minimize frustration cost

and improve security and privacy issues between end user and

utility.

REFERENCES

[1] V. Gungor, D. Sahin, T. Kocak, S. Ergut, C. buccella, C. Cecati and G.

Hancke, “Smart Grid Technologies: Communication Technologies and

Standards”, Industrial Informatics, IEEE Transaction, Vol. 7, No. 4, pp.

529-539, Nov. 2011.

[2] M. Hashmi, S. Hanninen, and K. Maki, “Survey of Smart Grid Concepts,

Architectures and Technological Demonstrations Worldwide”, IEEE PES

Conference, Medellin, Oct. 2011.

[3] L. Gelazanskas, and K. A. A. Gamage, “Demand Side Management in

Smart Grid: A Eeview and Proposals Forfuture Direction”, Sustainable

Cities and Society, Vol. 11, pp. 2230, Feb. 2014.

[4] P. Siano, “Demand Response and Smart Grids A Survey”, Renewable

and Sustainable Energy Reviews, Vol. 30, pp. 461478, Feb. 2014.

[5] A. Molderink, V. Bakker, M. G. Bosman, J. L. Hurink, and G. J. Smit,

“Domestic Energy Management Methodology for Optimizing Efﬁciency

in Smart Grids”, IEEE PowerTech, Bucharest, pp. 1-7, 28 Jun(2008) - 2

Jul (2009)

[6] T. Sousa, H. Morais, Z. Vale, P. Faria, and J. Soares, “Intelligent

Energy Resource Management Considering Vehicle-to-Grid: A Simulated

Annealing Approach”, IEEE Transaction Smart Grid, Vol. 3, No. 1, pp.

535-542, Mar. 2012.

[7] J. Soares, T. Sousa, H. Morais, Z. Vale, and P. Faria, “An Optimal

Scheduling Problem in Distribution Networks Considering V2G”, in Proc.

IEEE CIASG, Paris, 11-15 Apr. 2011.

[8] K. M. Tsui and S. C. Chan, “Demand Response Optimization for Smart

Home Scheduling Under Real-Time Pricing”, IEEE Transaction Smart

Grid, Vol. 3, No. 4, pp. 1812-1821, Dec. 2012.

[9] D. Maringer, “Portfolio Management with Heuristic Optimization”,

Springer, 2004.

[10] T. Logenthiran, D. Srinivasan, and T. Z. Shun, “Demand Side Man-

agement in Smart Grid using Heuristic Optimization”, IEEE Transaction

Smart Grid, Vol. 3, No. 3, pp. 1244-1252, Sept. 2012.

[11] Z. Zhao, W. C. Lee, Y. Shin and K. Song, “An Optimal Power

Scheduling Method for Demand Response in Home Energy Management

System”, IEEE Transaction Smart Grid, Vol. 4, No. 3, pp. 1390-1400,

Sept 2013.

[12] M. A. Khan, N. Javaid, A. Mahmood, Z. A. Khan and N. Alrajeh, “A

Generic Demand-Side Management Model for Smart Grid”, International

Journal of Energy Research, Vol. 39, No. 7, pp. 954964, Jun. 2015.

[13] T. Dethlefs, T. Preisler and W. Renz, “Ant-Colony based Self-

Optimization for Demand-Side-Management”, Conference: SmartER Eu-

rope, Essen, Feb. 2015.

[14] B. Liu, J. Kang, N. Jiang and Y. Jing, “Cost Control of the Transmission

Congestion Management in Electricity Systems based on Ant Colony

Algorithm”, Energy and Power Engineering, Vol. 3, No. 1, pp. 17 - 23,

Feb. 2011.

[15] N. Kumaraguruparan, H. Sivaramakrishnan and S. Sapatnekar, “Residen-

tial task scheduling under dynamic pricing using the multiple knapsack

method”, IEEE Conference, Washington, DC, pp. 1-6, Jan. 2012.

[16] O. A. Sianaki, O. Hussian and A.R. Tabesh, “A Knapsack Problem

Approach for Achieving Efﬁcient Energy Consumption in Smart Grid

for End-user Life Style”,IEEE conference, Waltham, MA, Sept. 2010.

[17] C. Ogwumike, M. Short and M. Denai, “Near-Optimal Scheduling of

Residential Smart Home Appliances using Heuristic Approach”, IEEE

International Conference on Industrial Technology, Seville, Spain, 2015.

[18] A. Barbato and A. Capone, “Optimization Models and Methods for

Demand-Side Management of Residential Users: A Survey”, Sept. 2014

[19] H. Kellerer, U. Pferschy, and D. Pisinger, “Knapsack Problems”.

Springer, 2004.

[20] I. K. Yu and Y. H song, “A Novel Short-term Generation Scheduling

Technique of Thermal Units using Ant Colony Search Algorithms”,

International Journal of Electrical Power and Energy Systems, Vol. 23,

No. 6, pp. 471479, Aug. 2001.

[21] S. Rahim, S. A. Khan, N. Javaid, N. Shaheen, Z. Iqbal and G. Rehman,

“Towards Multiple Knapsack Problem Approach for Home Energy Man-

agement in Smart Grid”, 8th IEEE International Conference on Network-

Based Information Systems, Taipei, Taiwan, 2015.

[22] http://jemena.com.au/supply-interruptions/electricity(Last accessed: sep

20, 2015)

[23] “Jemena Electricity Networks (VIC) Ltd - Network Tariffs For The 2015

Calendar Year (Exclusive of GST)”