Rs

MRa

Sb

c

d

e

a

ARR1A

KHB(RES

1

wasafotd

h0

Energy and Buildings 138 (2017) 215–227

Contents lists available at ScienceDirect

Energy and Buildings

j ourna l ho me page: www.elsev ier .com/ locate /enbui ld

eal time optimal schedule controller for home energy managementystem using new binary backtracking search algorithm

aytham S. Ahmed a,b, Azah Mohamed a, Tamer Khatib c,∗, Hussain Shareef d,aad Z. Homod e, Jamal Abd Ali a

Department of Electrical, Electronic and System Engineering, Faculty of Engineering and Built Environments, Universiti Kebangsaan Malaysia, Bangi,elangor, MalaysiaGeneral Directorate of Electrical Energy Production- Basrah, Ministry of Electricity, IraqDepartment of Energy Engineering and Environment, An-Najah National University, 97300 Nablus, PalestineDepartment of Electrical Engineering, United Arab Emirates University, 15551 Al-Ain, UAEDepartment of Oil and Gas Engineering, Basrah University for Oil and Gas, 61004 Basrah, Iraq

r t i c l e i n f o

rticle history:eceived 31 October 2016eceived in revised form1 December 2016ccepted 16 December 2016

eywords:ome energy management systeminary backtracking search algorithmBBSA)esidential demand responsenergy efficiencychedule controller

a b s t r a c t

In the domestic sector, increased energy consumption of home appliances has become a growing issue.Thus, reducing and scheduling energy usage is the key for any home energy management system (HEMS).To better match demand and supply, many utilities offer residential demand response program to changethe pattern of power consumption of a residential customer by curtailing or shifting their energy useduring the peak time period. In the present study, real time optimal schedule controller for HEMS isproposed using a new binary backtracking search algorithm (BBSA) to manage the energy consumption.The BBSA gives optimal schedule for home devices in order to limit the demand of total load and schedulethe operation of home appliances at specific times during the day. Hardware prototype of smart socketsand graphical user interface software were designed to demonstrate the proposed HEMS and to providethe interface between loads and scheduler, respectively. A set of the most common home appliances,namely, air conditioner, water heater, refrigerator, and washing machine has been considered to becontrolled. The proposed scheduling algorithm is applied under two cases in which the first case considersoperation at weekday from 4 to 11 pm and the second case considers weekend at different time of the

day. Experimental results of the proposed BBSA schedule controller are compared with the binary particleswarm optimization (BPSO) schedule controller to verify the accuracy of the developed controller in theHEMS. The BBSA schedule controller provides better results compared to that of the BPSO schedulecontroller in reducing the energy consumption and the total electricity bill and save the energy at peakhours of certain loads.

© 2016 Elsevier B.V. All rights reserved.

. Introduction

Many critical shutdowns happened in the electric power net-orks in many countries such as in Italy in 2003, Indonesia in 2005

nd in both Canada and United States in 2003 [1]. Power outagesometimes occur due to increased number of residential appli-nces, increasing cooling and lighting demand and over-demandrom customers in the period of peak hours [2]. Peak time load

ccurs in the grid when most customers are using electricity athe same time in a day [3]. As a solution to meet the high poweremand, electricity suppliers are forced to increase generation

∗ Corresponding author.E-mail address: t.khatib@najah.edu (T. Khatib).

ttp://dx.doi.org/10.1016/j.enbuild.2016.12.052378-7788/© 2016 Elsevier B.V. All rights reserved.

by building additional conventional power plants [4]. However,this solution is unsustainable because of low plant utilization fac-tor, increased carbon dioxide emission that contributes to climatechange and increased the costs of investment [5].

Demand response (DR) program is one of the demand side man-agement (DSM) solutions that has been investigated in the UnitedKingdom and other countries since 1970s to reduce energy con-sumption during peak hours and to increase plant utilization [6].DR has been defined as changes in electricity use by demand-sideresources from their normal consumption patterns in response tochanges in the price of electricity or to incentive payments designed

to induce lower electricity use at times of high wholesale marketprices or when system reliability is jeopardized [7]. The DR schemecan help the participating customers to save on electricity billswhen they reduce their electricity consumptions during peak hour

2 nd Bui

ahbEbpuia[idgttupdicetswsitcsswiacwHa[s

tiscsaDatt[tdhtttsr[tAdHwc

16 M.S. Ahmed et al. / Energy a

nd shifting peak time load to off-peak time [8]. Several studiesave been focused on peak load energy reduction and DR systemsenefits and challenges [9–12]. According to the US Department ofnergy, the residential DR programs can be classified into incentive-ased programs and price based programs [13]. In incentive-basedrograms, the utility can access and control the appliances of endsers and provide financial incentives for demand reduction dur-

ng peak hours to participating customers and the consumers get discount rate for their participation such as direct load control14]. However, by direct access to control on/off home appliancesntrude occupant’s privacy and this option is regarded as one of therawbacks of incentive-based programs [15]. In price based pro-rams, consumers voluntarily modify the power consumption inheir houses based on time-based electricity and follow the realime cost of electricity by using tariff price program such as time ofse pricing, real time pricing and critical peak pricing [16]. Theserograms will offer different prices at different times of the day likeuring off-peak and on peak periods to reflect the ability of the util-

ty to produce the energy needed. Time-of-use pricing is the mostommon residential electricity tariff and currently considered forxecution in many utilities around the world which is divided intoime slots at various seasons of the year or hours of the day usingmart energy meters [17]. The smart energy meter can provide two-ay communication between customers and utilities and the DR

ignal can offer incentive to customers to make decisions accord-ngly on electricity consumption [18]. Moreover, the homes havehe option to send the status and information about their energyonsumption [19,20]. A smart home enabled with residential DRtrategies is usually equipped with a home energy managementystem (HEMS) that manages controllable appliances associatedith smart outlets and smart meter [21]. The technology for HEMS

s efficient with communication technology, which connects homeppliances for remote management based on the internet and aombination of the home network to reduce the peak demandithout affecting the comfort of the consumers [22,23]. By usingEMS, the overall electrical appliance loads can be fully automatednd controlled with enabled DR programs at the domestic sector24]. Various studies have discussed the concept and schedulingtrategies of smart HEMS [25–30].

In recent years, many decision-support tools and optimizationechniques have been applied to help domestic customers in reduc-ng the energy consumption by creating an optimal home appliancecheduling of energy usage based on different pricing schemes andomfort settings. To achieve this goal, there is need to implementmart control in the domestic building. In [31], distributed controllgorithms have been applied to schedule home appliances withR based on the communication network architecture. The geneticlgorithm with artificial neural network (ANN) has been appliedo schedule home devices and optimize energy consumption inhe domestic sector by reducing the demand during peak periods32]. Hybrid lightning search algorithm based ANN has been usedo predict schedule controller in a HEMS [33]. In [34], autonomousemand-side management system based on game theory approachas been developed to optimize the appliances energy consump-ion and manage residential loads that help homeowners to selecthe priority of appliances considering either electricity cost reduc-ion or customer comfort. In [35], electrical appliances have beencheduled to reduce electricity cost based on dynamic pricing usingobust optimization and stochastic techniques. In a related work in36], hardware based HEMS incorporating DR has been developedo control the customers household by using a rule based technique.

prototype smart HEMS using machine learning algorithm with

ynamic price response has been developed in [36]. A rule basedEMS to control and schedule four home appliances that includeater heater, air-conditioner, clothes dryer and electric vehicle

onsidering residential demand response application and rules has

ldings 138 (2017) 215–227

been developed in [37]. In [39], a smart plug was developed withZigbee sensor for measuring and monitoring the power consump-tion of home appliances in the HEMS. A major issue that is facedin scheduling and shifting user loads is the minimization of theelectricity consumption during peak hours without affecting thecomfort of occupants. However, most previous researchers focusedon reducing electricity bills and saving energy without consideringuser comfort. Thus, to schedule and control home appliances thereis need to take into account all possible challenges, such as comfortlevel, demand limit and tariff.

This paper presents a new binary backtracking search algorithm(BBSA) based real time optimum schedule controller for HEMS toachieve energy savings and limit the household peak demand inthe home on the basis of the scheduled operation of several appli-ances according to a specific time, resident comfort constraintsand priority. Furthermore, hardware smart socket and graphicaluser interface (GUI) of HEMS are designed and build with Zigbeemodules and various sensors are used to control a certain load.In addition, this paper suggested two cases to schedule the homeappliances. The first case considers DR signal from 4 to 11 pm on aweek day and the second case considers different time of the dayat weekend. An average single-family home size 2300-square feetwith 5 persons have been considered as a case study in Malaysia.Four appliances, namely, air conditioner (AC), water heater (WH),refrigerator (REF), and washing machine (WM), are considered con-trollable. To validate the developed BBSA schedule controller theresults are compared with the binary PSO schedule controller.

2. Development and design of the home energymanagement system

The HEMS consists of four devices, which are AC, WH, WM,and REF. The overall system can help homeowners in controllingand monitoring the energy consumption of the appliances by gath-ering the data from selected loads by utilizing smart socket. Thesmart socket provides the interface between the developed HEMand the non-smart load appliances in real time. The smart socketis designed to provide remote control of non-smart loads thus pro-viding a practical solution to interface the loads with the developedsystem and schedule on/off status of selected loads [38]. In theoperation of the HEMS, it is assumed that the house is equippedwith smart meter which continuously communicates between util-ity and homeowner and can receive DR signals from the utility. Thissignal and other user preference conditions are used as inputs tothe system to perform local control.

The DR signal with information on the duration, situation andamount of load to be shed is assumed to be received directly froma smart meter in the HEMS. The following subsections describe therequired setup of the proposed HEMS (Fig. 1).

2.1. Graphical user interface design

The HEMS includes GUI and associated software to assist cus-tomers in monitoring the total and individual power consumption,monitor the on/off status of home appliances, the room tempera-ture, and water temperature for WH as shown in Fig. 2. As depictedin the figure, the dashboard can display the power consumptionin (Watt) for each load, the room and water heater temperaturein ◦C, the total power consumption and also the total energy in(kWh). Furthermore, it is possible to turn on/off each home appli-

ance manually from a push button switch display in the developeddashboard.

The proposed software includes the DR algorithm that repre-sents the main controller. In addition, this software is able to collect

M.S. Ahmed et al. / Energy and Buildings 138 (2017) 215–227 217

Fig. 1. Main elements of the developed HEMS.

erface

tt

2

stpcc

Fig. 2. Graphical user int

he power consumption and temperature data from the loads usinghe interface and store in a database.

.2. Smart socket hardware for individual electrical appliance

A prototype smart socket is developed as a stand-alone deviceimilar to a plug with ZigBee wireless communication to connect

he non-smart loads to the HEMS. The smart socket measures theower consumption by reading the current and voltage of theonnected electrical appliance and controls it with an electrome-hanical relay in the slave socket [39]. It also provides remote

of the proposed HEMS.

monitoring and off/on control of the appliances for implementingthe HEMS in smart homes. The slave circuit of the smart sockethardware is shown in Fig. 3. A node is connected to a home appli-ances and it is used for sending power consumption data of eachdevice by means of ZigBee wireless module to a data collectiondevice such as a personal computer (PC). This design is differentthan others because it can provide power supply to all sensors

and components in the circuit unlike the some devices with bat-teries to provide DC voltage to the rest of the circuit. In addition,the smart socket is equipped with temperature sensor to measurethe temperature inside the room.

218 M.S. Ahmed et al. / Energy and Buildings 138 (2017) 215–227

CH 2

LM 1117LM 7805

CH13.3V

CURRENT SENSOR712

Vo

Filter

GND5V

CH1

1 202 193 184 175 166 157 148 139 1210 11

XBEE PRO

CH 2 12 V

3.3 V5V

LM 324

TANSFORMER

220 V

L

N

5VVcc

C1

C2

C3

C4

C5

C6

R1

R3

R2

R4

R5R6

R7

R8

LED1

D1

RELAY

NPN

VR1

VR2

R9

1000µF

4.7µF

22µF

3KΩ

5KΩ

39KΩ

100KΩ

39KΩ

100KΩ 330KΩ 4.7µF5KΩ

10µF

470µF

IN4002

560Ω

12V

560Ω

L OA

D

CH 3 5 VCH 3

TEMPERATURESENSOR

Fig. 3. Slave circuit structure of smart socket node.

Fig. 4. Prototype of smart socket; (a) PCB layout, (b) PCB printed, and (c) prototype hardware.

M.S. Ahmed et al. / Energy and Buildings 138 (2017) 215–227 219

g and design of smart socket.

fipcaBpcpcvivsat

co3

V

wfib

rb(

K

wc

iTs

I

wtt

I

wt

Table 1Configuration and specification of ZigBee.

Parameter Value and Unit

Power supply 2.8–3.7 VTransmit Power 1.3 mWOperating channel (CH) 2.415 GHzBaud rate (BD) 115200 bpsStop bits 1Data bits 8

Fig. 5. Laboratory testin

As shown in Fig. 3, the transformer steps down the mains voltagerom 220 V AC to 12 V DC to provide voltage to the electromechan-cal relay, LM7805, provide 5 V to the current sensor, LM 1117 androvide voltage to the ZigBee wireless module. Additionally, theurrent sensor signal at CH1 is connected to the ZigBee pin B20,nd the temperature sensor reading at CH3 is connected to pin17. The current sensor type ACS712 ±15 A is used and its out-ut is connected to an operational amplifier (op-amp) based signalonditioning circuit so as to read the required analog values to theins of the analog input of the Zigbee module. The output voltageonnected to voltage divide circuit with variable resistor and theoltage signal at CH2 is connected to pin B19. The proposed circuits used to record energy and power consumption by using current,oltage and also to turn on/off the loads. The ZigBee transceiverends digital data signals to other ZigBee transceiver which is usu-lly connected to the PC. Fig. 4 shows the PCB and the hardware ofhe implemented smart socket.

The output signal of the actual measured voltage value Vt is cal-ulated based on several signal conversion stages. The peak valuef Vt is proportional to the reference value of the ZigBee, which is.3 V. Thus,

t = Vout ∗ Kv (1)

here Voutis the scale down output voltage of the bridge recti-er, while Kvis the voltage scale down constant which is giveny(R1 + VR1)/ (VR1).

The conditioning circuit is used for the conversion from AC tooot mean square (RMS) signal. Since the ZigBee has a built in 10-its analog-to-digital converter ADC, therefore 10 bits = 1024 steps0–1023), so the step size is 3.3 V/1024 = 3.22 mV.

The voltage Vt is expressed as,

v ∗ ıvt ∗

(Vref /

(210 − 1

))(2)

here ıvt is the reading from channel CH2 of the analog-to-digital

onverter at time interval, t.Similar to the voltage signal measurements, the peak value of It

s proportional to the reference value of the ZigBee, which is 3.3 V.he conditioning circuit is used for the conversion from AC to RMSignal. Thus,

t = Viout ∗ Ki (3)

here Viout is the scale down output voltage which is equivalent tohe measured current which is limited by Vref < Vout < 0, while Ki ishe current scale down constant and can be given by (Imax/Vref):

The current, It is determined by using the following equation:

t = Ki ∗ ıit ∗(Vref /

(210 − 1

))(4)

here Vref is the supply voltage of the circuit which is 5 V and ıit ishe reading from channel CH1.

Data Process Power 64 mWPower level (PL) −10 dBm (0.1 mW)

The total power of the device is then calculated by using,

Pt = It ∗ Vt (5)

The performance of the prototype was tested in the laboratoryat Universiti Kebangsaan Malaysia (UKM) in terms of data trans-mission rate and accuracy of the received data as shown in Fig. 5.

The circuit board sizes are designed such that it can be placedinside the socket. The smart socket is also designed to satisfy thehigher and lower power ratings of the home appliances. To test theaccuracy of the received data, an energy analyzer is used to deter-mine the current and voltage readings. The received signals throughthe ZigBee wireless module are compared with that measured onthe oscilloscope. From the comparison, it is shown that the maxi-mum error of the received signals is small and found to be about5%.

2.3. ZigBee Wireless Protocol

The ZigBee wireless module is an independent network tech-nology module which is popular and low cost. It uses 2.4 GHz radiofrequency and work at the protocol of IEEE 802.15.4 with data trans-ferring between nodes up to 250 kbps [40]. The ZigBee has beendesigned in small size with various models and consumes verylow power when used for monitoring and control system. It trans-mits and receives data in the wireless sensor network [41]. Severalresearch works on energy consumption in smart home uses ZigBeewireless module due to the aforementioned advantages [42,43].Thus, in this work, five ZigBee type (XBee Series 2) as shown inFig. 6 are used. Four of the ZigBee wireless modules are connectedand configured with the sensors through signal conditioning cir-cuits to act as sensor nodes to transmit the data while the othernode acts as the coordinator node which is connected to a per-sonal computer (PC) with an appropriate MatLab software driver

to access the data and control the home appliances in real time. Thespecification and configuration of the ZigBee module is illustratedas shown in Table 1. The ZigBee configuration for both the coordi-nator and sensor are programmed based on the software X-CUT to

220 M.S. Ahmed et al. / Energy and Bui

sntcm

2

MtfltbT

3

toTsscs

3

poTT

ppts

Tito

Fig. 6. ZigBee type Xbee pro series 2.

et parameters by linking the module to the USB serial port. In theetwork, the baud rate (BD) and operating channel (CH) should behe same so as to ensure communication between the sender andoordinator. The analog to digital converter (ADC) of each ZigBeeodule has a resolution conversion of 10 bit.

.4. Experimental setup

The proposed system consists of PC as a control center with theatlab/software and Zigbee coordinator. The Zigbee is connected

o the PC which acts as a coordinator to receive and send data to theour home appliances and signals to switch on/off the controllableoads. The smart socket is connected to each home appliance andhe data will send to the control center by zigbee communicationy sending signals to turn on/off the home appliances in the HEMS.he experimental setup of the whole system is shown in Fig. 7.

. Proposed optimal schedule controller of HEMS

There are some difficulties to get the best optimal schedule forhe home appliances and switch customer loads to decrease the costf electric power consumption during DR event using simple rules.herefore, some advanced optimization techniques are required forcheduling home appliances. The optimal schedule controller con-ist of three sub components that include schedule conditions of theustomer, optimization technique and the design of the proposedchedule controller as explained in following sub-sections:

.1. Customer comfort preference and appliance priority

Each appliance is compared on several set points, including loadriority, power consumption, and customer preference, by settingsn the room temperature of AC and the water temperature of WH.he homeowner preference settings and load priority is shown inable 2.

Table 2 shows that the first priority is AC where the room tem-erature should be maintained in the range of 23–26 ◦C. The secondriority is WH with the hot water temperature maintained withinhe acceptable range of 40–60 ◦C. Fig. 8 shows the flowchart of thechedule controller conditions for home appliances.

At the initial, the data is read from all the mentioned appliances.

hen each appliance is compared with several set points indicated

n Table 2. If the total electrical power consumption is greater thanhe demand limit (DL), the algorithm will turn OFF the lowest pri-rity appliance starting with REF, and force the loads to shift their

ldings 138 (2017) 215–227

operating time after DR interval has elapsed so as to maintain thetotal power consumption below its DL.

3.2. Overview of back tracking search algorithm (BSA)

Optimization is a process that is used to find the best solutionof the objective function depending on the input information andoptimization limitations in order to improve performance of thesystem and to solve actual-valued numerical optimization prob-lem. The aim of the optimization techniques are to improve theapplications by finding minimal error, minimal cost, maximumperformance and efficiency [44]. BSA is an optimization algorithmproposed by Civicioglu in 2013 based on producing a trial pop-ulation which evolves using two new mutation and crossoveroperators [49]. BSA dominates in the search of the best value of thepopulations and searches in the space boundary to get the exploita-tion capabilities and very robust exploration [45]. It been utilizedin considerable research and is widely used to solve optimizationproblems [46,47]. Details of the BSA can be found in [48].

The process of initialization is the basic configuration of popu-lation to generate the value of initial population (Zij),

Zij = rand ∗(upj − lowj

)+ lowj (6)

where j = 1,2,3...M, M is the problem dimension and i=1,2,3...M, Mis the population size. The objective function (Obj) is calculated foreach population.

The historical population (oldzij) is used so as to calculate thesearch direction by using,

oldZij = rand ∗(upj − lowj

)+ lowj (7)

The historical population remembers a randomly chosen popu-lation for the previous generation to create a new trial populationby taking partial advantage of prior experiences and generate thesearch direction matrix. The following condition is demonstratedby comparing the two random variables:

ifa < bthenoldZij = Zij (8)

oldZij = permuting(oldZij

)(9)

The next process is mutation that generates theoldZijand thenew initial population. The BSA produces a trial population, Tij , andfrom previous generations takes its experiences of a partial featureby using,

mutant = Zij + rand ∗(oldZij − Zij

)(10)

The trial population, Tijis generated in the process named ascrossover. The crossover is utilized to update the binary matrix,BAij by generating the binary matrix and compare it between theTij andZ ij . The comparison is used to control the mechanism ofboundaries for trial population and to get the best objective value.

3.3. Design of the proposed BBSA schedule controller

The proposed binary BSA (BBSA) is used to control and schedulethe home appliance in the HEMS by creating the curtailment sched-ule at a specific time. The target of the proposed BBSA is obtainedby minimizing a predefined objective function. The implementa-tion of the proposed algorithm starts by resetting the parametersof the BBSA, namely, the number of population size (P), the max-

imum iterations (T), and the number of problem variables (N). Inthis work, 20 populations and 1000 iterations are used to get thebest results. The flow diagram of the proposed BBSA is shown inFig. 9. After generating the initial value of population (Zij), the sig-

M.S. Ahmed et al. / Energy and Buildings 138 (2017) 215–227 221

Fig. 7. Experimental setup of the HEMS.

Table 2Load priority and power ratings characteristics.

Appliance Priority of Loads Customer comfortable lifestyle. Power in (kW)

Air conditioner (AC) 1 Temperature of room 23–26 ◦C 1200Water heater (WH) 2 Temperature of water 40–60 ◦C 800Washing machine (WM) 3 Usage: Different intervals 500–600Refrigerator (REF) 4 Usage: 24 h 150–190

From Table 2 . read the appliances powerrating

START

Room temp, comfort level, load priorityand user demand limit (UDL).

Indicate the value of load to shed (LS)according to utility DR data signal

Calculate the new demand limit(NDL)=UDL-LS

Switch device status for all appliances andcalculate current demand (D)

WHWMREF AC

NDL< D

YES 1234

NO

ON

OFF OFFOFF OFF

ONONON

WH < priorityWM < priorityREF< priority AC < priority

YESNONONONO

YESYESYESOFF OFF OFF OFF

Fig. 8. Flowchart of the schedule controller conditions considering load priority.

222 M.S. Ahmed et al. / Energy and Bui

Yes

START

Calculated the objective function for eachpopulation (Zij) using Eq.(13)

Reset BSA parameters: maximum iteration (T),problem dimension (D), population size (N), F

t > TReach maximum population

No

i > NReach maximum population

Yes

n=n+1

No

Calculated the objective function of the eachparticles using Eq.(13)

Generate initial population (Zij) using Eq. (6)

Run the proposed schedule controller of thehome appliances for each population (Zij)

Run the proposed schedule controller of thehome appliances for each population (Tij)

i > NReach maximum population

i=i+1

Output the optimum schedule controller of thehome appliances

Generate initial historical population (old Zij)using Eq.(7)

Update historical population (old Zij) usingEq.s (8) and (9)

Generate the initial of the trail populationMutant using Eq. (10)

Calculate a binary integer valued martix (BAij)

Generate the final of the trail population (Tij)

i=i+1

Choose minimum of the objective function andfind the best schedule

Executed schedule controller conditions usingFig (8)

Executed schedule controller conditions usingFig (8)

Yes

No

Convert population value to binary value usingEq.s (11) and (12)

Convert population value to binary value usingEq.s (11) and (12)

mb

S

P

END

Fig. 9. Flowchart of the proposed BBSA based schedule controller.

oid equation is used to convert population to a real binary valueetween 0 and 1 by using the following equations:

i = 1(11)

1 + e−w

B ={

0, si < 0.5

1, si ≥ 0.5(12)

ldings 138 (2017) 215–227

whereSi is the sigmoid function,w is the population value and PB isits binary value.

If the Si value is greater than 0.5 the PBis 1, otherwise the PBis 0. In addition, the schedule controller conditions are executedby following the flowchart shown in Fig. 9 and the load prior-ity given in Table 2. BSA and its binary version show significantpotential in locating near-optimal scheduling controller in highlydimension, and non-continuous optimization problems. After exe-cuting the conditions of the schedule controller, the proposed BBSArun schedule controller of the home appliances for each population(Zij). Thereafter, the objective function for each population is calcu-lated. The energy consumption is used as the objective function toimprove the performance of the HEMS by minimizing the energyconsumption for 24 h, which leads to reduce the electricity bill asshown in the following equation:

Obj =n∑i=0

Pt (i) ∗ T (13)

where Objis the objective function, Pt is the total power consump-tion in (kW) and T is the time in hour.

The scheduler controller is created by utilizing BBSA to turn thefour home appliances ON/OFF according to priority of appliances,and homeowner preferences, as shown in Eq. (14).

OSC =

⎡⎢⎢⎢⎢⎢⎣

AC1 WH1 REF1 WM1

AC2 WH2 REF2 WM2:::ACn WHn REFn WMn

⎤⎥⎥⎥⎥⎥⎦

(14)

where OSC is the optimal schedule controller, n is the dimensionwhich is equal to 42. The decision of proposed controller has beenmade for every 10 min for weekday and weekend. Every hour needs6 decisions as well as for 7 h there is need 42 decisions. Period ofweekday is 7 h from 4 to 11 pm as well as, weekend is 7 h dividedinto 9 am to 12 pm, 4–6 pm, and 9–11 pm. The dimension decisionvector of the population is 42 row and 4 columns.

4. Results and discussion

In this research, the DR event considered for two cases, namely,weekday and weekend. During weekday, the event starts from 4to 11 pm and at weekend the event starts at different time inter-vals from 9 am to 12 pm, 4–6 pm, and 9–11 pm respectively. Theproposed schedule controller is used to reduce the energy duringthe DR events considering four appliances per 7-h for weekday andweekend. The value of DL is considered to be 1300 kW. The totalpower consumption at weekday is found to be 23 kWh, whereasin the weekend it reaches 25 kWh per day. The behaviour of homeappliance for saving energy is important to be considered to givegood background to the researchers in the future. In addition, thebehaviour of each home appliance can be easily changed accordingto the customers setting, for example, the air conditioner and refrig-erator are working 24 h at weekday and weekend whereas washingmachine (WM) and water heater (WH) are operating according tothe home owner use. However, in the design of the experiment, theweather conditions do not change very much for Malaysia unlikecountries with four seasons [49].

The proposed BBSA schedule controller is compared with BPSOschedule controller to validate the developed controller. For fair

comparison, the population size and maximum iteration are 20and 1000 for the BBSA and BPSO controllers. Figs. 10 and 11 illus-trate convergence characteristics of the BBSA and BPSO schedulecontrollers. Fig. 10 shows the weekday results in which the BBSA

M.S. Ahmed et al. / Energy and Buildings 138 (2017) 215–227 223

0 20 0 400 600 800 10 001.82

1.83

1.84

1.85

1.86

1.87 x 104

Iterati on

Obj

ectiv

e fu

nctio

nBBSABPSO

0 200 400 600 800 1000

1.85

1.9

1.95

2 x 104

Iterati on

Obj

ectiv

e fu

nctio

n

BBSABPS O

Fig. 10. Performance comparison between BBSA and BPSO on a week day.

0 1 2 3 4 5 6 7 8 9 10 110

10002000

0 1 2 3 4 5 6 7 8 9 10 110

5001000

0 1 2 3 4 5 6 7 8 9 10 110

5001000

Pow

er (W

)

0 1 2 3 4 5 6 7 8 9 10 110

200400

0 1 2 3 4 5 6 7 8 9 10 110

1000

2000

3000

Tim

(d)

(e)

(c)

(b)

(a)

Fig. 12. Weekday power consumption without controller (a

0 1 2 3 4 5 6 7 8 9 10 110

10002000

0 1 2 3 4 5 6 7 8 9 10 110

5001000

0 1 2 3 4 5 6 7 8 9 10 110

5001000

Pow

er (W

)

0 1 2 3 4 5 6 7 8 9 10 110

200400

0 1 2 3 4 5 6 7 8 9 10 110

1000

2000

3000

Tim

(d)

(c)

(b)

(a)

(e)

Fig. 13. Real time power consumption with BBSA schedule controller at

Fig. 11. Performance comparisons between BBSA and BPSO at weekend day.

12 13 14 15 16 17 18 19 20 21 22 23 24

12 13 14 15 16 17 18 19 20 21 22 23 24

12 13 14 15 16 17 18 19 20 21 22 23 24

12 13 14 15 16 17 18 19 20 21 22 23 24

12 13 14 15 16 17 18 19 20 21 22 23 24e (hou r)

) AC, (b) WH, (c) WM (d) REF and (e) the total power.

12 13 14 15 16 17 18 19 20 21 22 23 24

12 13 14 15 16 17 18 19 20 21 22 23 24

12 13 14 15 16 17 18 19 20 21 22 23 24

12 13 14 15 16 17 18 19 20 21 22 23 24

12 13 14 15 16 17 18 19 20 21 22 23 24e (hour)

Deman d Period

weekday (a) AC, (b) WH, (c) WM (d), REF and (e) the total power.

224 M.S. Ahmed et al. / Energy and Buildings 138 (2017) 215–227

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

10002000

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

5001000

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

5001000

Pow

er (W

)

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

200400

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

1000

2000

3000

Time (hou r)

(c)

(b)

(a)

Demand Period(e)

(d)

roller at weekday AC, (b) WH, (c) WM (d) REF and (e) the total power.

shwsFtodBcc

4w

wtAcs

1ttWcas

aaF

feosctpp

16 17 18 19 20 21 22 230

500

1000

1500

2000

2500

3000

Time (hour)

Pow

er (W

)Before controll er BB SA sche dule controller

Fig. 15. The total power consumption at weekday before and after implementingthe BBSA based schedule controller.

16 17 18 19 20 21 22 230

500

1000

1500

2000

2500

3000

Time (hour)

Pow

er (W

)

Bef ore controll er BPSO schedule controll er

Fig. 14. Real time power consumption with BPSO schedule cont

chedule controller can reduce the energy consumption of theousehold appliances from 23 kWh per day to 18.1 kWh per dayhereas the BPSO schedule controller can reduce the energy con-

umption from 23 kWh per day to 18.4 kWh per day. Similarly,ig. 11 shows the results of a weekend day convergence characteris-ics. The BBSA schedule controller reduces the energy consumptionf the household appliances from 25 kWh per day to 18.3 kWh peray and BPSO from 25 kWh per day to 18.7 kWh. It can be noted thatBSA is better than BPSO schedule controller in reducing the energyonsumption at both weekday and weekend. Besides, the BBSAontroller achieves faster convergence than the BPSO controller.

.1. Proposed real time optimal scheduling controller result for aeek day

The power consumption data in real time is collected for 24 hithout controller at weekday in Malaysia. In the weekday case,

he power consumption for each home appliance which includesC, WH, WM, and REF and the total power consumption have beenollected from the HEMS without using the schedule controller ashown in Fig. 12.

The DR signal in the weekday is considered between 4pm to1pm. To determine the optimal schedule of the home appliances,he BBSA is used as a schedule controller to the HEMS for weekdayo achieve the best energy saving for each device, namely, AC, WH,

M, and REF as shown in Fig. 13. The BPSO schedule controller isompared with BBSA schedule controller to validate the developedlgorithm and the performances of AC, WH, WM, and REF are ashown in Fig. 14.

The total power consumption for home appliances before andfter using BBSA schedule controller at weekday is shown in Fig. 15,nd the comparison with BPSO schedule controller is shown inig. 16.

From Fig. 15, it can be noted that the BBSA schedule controlleror HEMS will turn off the lower priority device when the totallectrical power consumption exceeds the DL and schedule theperation to off peak hour so as to maintain the total power con-umption below its DL.The results of the proposed BBSA schedule

ontroller can significantly reduce the peak-hour energy consump-ion at weekday during the DR which is 4.8705 kWh per day (21.07%er day), whereas the energy saving by using the BPSO is 4.52 kwher day (20.55% per day) as shown in Fig. 16. A comparison of results

Fig. 16. The total power consumption at week day before and after implementingthe BPSO based schedule controller.

shows that the BBSA is better than the BPSO in terms of schedulinghousehold appliances and reducing the peak load at weekday whileguaranteeing homeowner comfort associated with the operation ofappliances.

4.2. Proposed real time optimal scheduling controller result for a

weekend day

In the weekend case, the DR signal in the weekend is representedby different times, which are between 9 am to 12 pm, 4 pm to 6 pm,

M.S. Ahmed et al. / Energy and Buildings 138 (2017) 215–227 225

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

10002000

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

5001000

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

5001000

Pow

er (W

)

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

200400

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

1000

2000

3000

Time (hour)

(a)

(c)

(b)

(d)

(e)

Fig. 17. Weekend power consumption for home appliances without controller.

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

10002000

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

5001000

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

5001000

Pow

er (W

)

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

200400

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

1000

2000

3000

Tim

(d)

(c)

(b)

(a)

(e)

roller

aap

astiw

bsc

ttta

Fig. 18. Real time power consumption with BBSA schedule cont

nd between 9 pm to 11 pm. The power consumption for each homeppliance at weekend is measured without controller by using theroposed experimental HEMS as shown in Fig. 17.

The result of the proposed BBSA schedule controller in the HEMSt weekend is shown in Fig. 18. The figure shows the power con-umption for each home appliance, which is WH, AC, WM, FEF andhe total power consumption. The DR for each time in the weekends presented by the dot green square. The same test is applied in the

eekend by optimal BPSO schedule controller as shown in Fig. 19.In addition, the total power consumption for home devices

efore and after using BBSA schedule controller at weekend ishown in Fig. 20, whereas, the comparison with BPSO scheduleontroller is shown in Fig. 21.

Fig. 20 shows the results of weekend day before and after using

he BBSA schedule controller at different times of the day. Theotal energy consumption is reduced during DR periods with theotal power consumption kept below 1300 W. The energy savingt weekend per 7 h for four home appliances is up to 6.6 kWh per

e (hour)

at weekend AC, (b) WH, (c) WM (d) REF and (e) the total power.

day (26.1% per day) by using BBSA, whereas, using BPSO schedulecontroller the saving energy is 6.3 kWh per day (25% per day).

5. Conclusions

The application of the BBSA to determine the optimum schedulehas been presented for use in the HEMS in real time. The compar-ison of results showed that the BBSA schedule controller is betterthan the BPSO schedule controller in terms of reducing the energyconsumption of home appliances at weekday and weekend with-out affecting the comfort level of the customers associated with theoperation of the appliances. The BBSA schedule controller achievedin saving energy by 4.87 kWh per day (21.07% per day) at weekdayand 6.6 kWh per day (26.1% per day) at weekend. The energy saving

results of the BPSO schedule controller is 4.52 kWh per day (20.55%per day) at week day and 6.3 kWh/day (25% per day) in the DR peri-ods. The results proved that the performance of the BBSA schedulecontroller is better in terms of saving energy compared to the BPSO

226 M.S. Ahmed et al. / Energy and Buildings 138 (2017) 215–227

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

10002000

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

5001000

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

5001000

Pow

er (W

)

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

100200300

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

1000

2000

3000

Tim

(d)

(c)

(b)

(a)

(e)

Fig. 19. Real time power consumption with BPSO schedule controller at

9 10 11 120

1000

2000

3000

Pow

er (W

)

16 17 180

1000

2000

3000

Time (hour)21 22 230

1000

2000

3000(a) (b) (c)

Fig. 20. The total power consumption before and after implementing the BBSAbased controller in the weekend at (a) 9–12, (b) 16–18, and (c) 21–23.

9 10 11 120

1000

2000

3000

Pow

er (W

)

16 17 180

1000

2000

3000

Time (hour)21 22 230

1000

2000

3000(b) (c)(a)

Fc

smi

A

Mg

R

[

[

[

[

[

[

[

[

[

[

[Awareness on energy management in residential buildings: a case study in

ig. 21. The total power consumption before and after implementing the BPSOontoller in the weekend at (a) 9–12, (b) 16–18, and (c) 21–23.

chedule controller as well as actively controlling the loads whileaintaining the power demand of household appliances within the

mposed demand limit.

cknowledgment

The authors gratefully acknowledge the University Kebangsaanalaysia for the financial support for the project under research

rant DIP-2014-028.

eferences

[1] T. Xu, A.J. Masys, Critical infrastructure vulnerabilities: embracing a networkmindset, in: J.A. Masys (Ed.), Exploring the Security Landscape:

e (hour)

weekend (a) AC, (b) WH, (c) WM (d) REF and (e) the total power.

Non-Traditional Security Challenges, Springer International Publishing, Cham,2016, pp. 177–193.

[2] R.V. Jones, K.J. Lomas, Determinants of high electrical energy demand in UKhomes: socio-economic and dwelling characteristics, Energy Build. 101(2015) 24–34.

[3] Y. Liu, C. Yuen, R. Yu, Y. Zhang, S. Xie, Queuing-based energy consumptionmanagement for heterogeneous residential demands in smart grid, IEEETrans. Smart Grid 7 (3) (2016) 1650–1659.

[4] G. Papachristos, Household electricity consumption and CO2 emissions in theNetherlands: a model-based analysis, Energy Build. 86 (2015) 403–414.

[5] S. Agarwal, R. Satyanarain, T.F. Sing, D. Vollmer, Effects of constructionactivities on residential electricity consumption: evidence from Singapore’spublic housing estates, Energy Econ. 55 (2016) 101–111.

[6] M.M. Jalali, A. Kazemi, Demand side management in a smart grid withmultiple electricity suppliers, Energy 81 (2015) 766–776.

[7] A. Safdarian, M. Fotuhi-Firuzabad, M. Lehtonen, Benefits of demand responseon operation of distribution networks: a case study, IEEE Syst. J. 10 (1) (2016)189–197.

[8] F. Shariatzadeh, P. Mandal, A.K. Srivastava, Demand response for sustainableenergy systems: a review, application and implementation strategy, Renew.Sustain. Energy Rev. 45 (2015) 343–350.

[9] P. Bradley, M. Leach, J. Torriti, A review of the costs and benefits of demandresponse for electricity in the UK, Energy Policy 52 (2013) 312–327.

10] H.T. Haider, O.H. See, W. Elmenreich, A review of residential demand responseof smart grid, Renew. Sustain. Energy Rev. 59 (2016) 166–178.

11] A. Arteconi, D. Patteeuw, K. Bruninx, E. Delarue, W. D’haeseleer, L. Helsen,Active demand response with electric heating systems: impact of marketpenetration, Appl. Energy 177 (2016) 636–648.

12] M.H. Albadi, E.F. El-Saadany, A summary of demand response in electricitymarkets, Electr. Power Syst. Res. 78 (11) (2008) 1989–1996.

13] M. Lee, O. Aslam, B. Foster, D. Kathan, J. Kwok, L. Medearis, R. Palmer, P.Sporborg, M. Tita, Assessment of demand response and advanced metering,Federal Energy Regulatory Commission, Tech. Rep. (2013).

14] M.A.F. Ghazvini, P. Faria, S. Ramos, H. Morais, Z. Vale, Incentive-based demandresponse programs designed by asset-light retail electricity providers for theday-ahead market, Energy 82 (2015) 786–799.

15] C. Lang, E. Okwelum, The mitigating effect of strategic behavior on the netbenefits of a direct load control program, Energy Econ. 49 (2015) 141–148.

16] V. Hamidi, F. Li, F. Robinson, Demand response in the UK’s domestic sector,Electr. Power Syst. Res. 79 (12) (2009) 1722–1726.

17] K.S. Cetin, L. Manuel, A. Novoselac, Effect of technology-enabled time-of-useenergy pricing on thermal comfort and energy use inmechanically-conditioned residential buildings in cooling dominatedclimates, Build. Environ. 96 (2016) 118–130.

18] F. Clarizia, D. Gallo, C. Landi, M. Luiso, R. Rinaldi, Smart meter systems forsmart grid management, in: 2016 IEEE International Instrumentation andMeasurement Technology Conference Proceedings, IEEE, 2016, pp. 1–6.

19] D. Alahakoon, X. Yu, Smart electricity meter data intelligence for futureenergy systems: a survey, IEEE Trans. Ind. Inf. 12 (1) (2016) 425–436.

20] M.S. Ahmed, A. Mohamed, R.Z. Homod, H. Shareef, K. Khalid, (in press)

kajang and putrajaya, J. Eng. Sci. Technol. (2016).

nd Bui

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

algorithm, Measurement 78 (2016) 49–62.

M.S. Ahmed et al. / Energy a

21] S. Maharjan, Q. Zhu, Y. Zhang, S. Gjessing, T. Bas ar, Demand responsemanagement in the smart grid in a large population regime, IEEE Trans. SmartGrid 7 (1) (2016) 189–199.

22] Y.-T. Lee, W.-H. Hsiao, C.-M. Huang, T.C. Seng-cho, An integrated cloud-basedsmart home management system with community hierarchy, IEEE Trans.Consum. Electron. 62 (1) (2016) 1–9.

23] D. Egarter, A. Monacchi, T. Khatib, W. Elmenreich, Integration of legacyappliances into home energy management systems, J. Ambient Intell. Hum.Comput. 7 (2) (2016) 171–185.

24] S. Gyamfi, S. Krumdieck, Scenario analysis of residential demand response atnetwork peak periods, Electr. Power Syst. Res. 93 (2012) 32–38.

25] B. Zhou, W. Li, K.W. Chan, Y. Cao, Y. Kuang, X. Liu, X. Wang, Smart homeenergy management systems: concept, configurations, and schedulingstrategies, Renew. Sustain. Energy Rev. 61 (2016) 30–40.

26] Y.-H. Lin, M.-S. Tsai, An advanced home energy management systemfacilitated by nonintrusive load monitoring with automated multiobjectivepower scheduling, IEEE Trans. Smart Grid 6 (4) (2015) 1839–1851.

27] M. Pipattanasomporn, M. Kuzlu, S. Rahman, An algorithm for intelligent homeenergy management and demand response analysis, IEEE Trans. Smart Grid 3(4) (2012) 2166–2173.

28] K.X. Perez, M. Baldea, T.F. Edgar, Integrated HVAC management and optimalscheduling of smart appliances for community peak load reduction, EnergyBuild. 123 (2016) 34–40.

29] D. Mahmood, N. Javaid, N. Alrajeh, Z.A. Khan, U. Qasim, I. Ahmed, M. Ilahi,Realistic scheduling mechanism for smart homes, Energies 9 (3) (2016) 202.

30] H.A. Özkan, Appliance based control for home power management systems,Energy 114 (2016) 693–707.

31] Z. Zhao, W.C. Lee, Y. Shin, K.-B. Song, An optimal power scheduling method fordemand response in home energy management system, IEEE Trans. SmartGrid 4 (3) (2013) 1391–1400.

32] B. Yuce, Y. Rezgui, M. Mourshed, ANN–GA smart appliance scheduling foroptimised energy management in the domestic sector, Energy Build. 111(2016) 311–325.

33] M. Ahmed, A. Mohamed, R. Homod, H. Shareef, Hybrid LSA-ANN based homeenergy management scheduling controller for residential demand responsestrategy, Energies 9 (9) (2016) 716.

34] N. Yaagoubi, H.T. Mouftah, User-aware game theoretic approach for demandmanagement, IEEE Trans. Smart Grid 6 (2) (2015) 716–725.

35] Z. Chen, L. Wu, Y. Fu, Real-time price-based demand response managementfor residential appliances via stochastic optimization and robustoptimization, IEEE Trans. Smart Grid 3 (4) (2012) 1822–1831.

[

[

ldings 138 (2017) 215–227 227

36] M. Kuzlu, M. Pipattanasomporn, S. Rahman, Hardware demonstration of ahome energy management system for demand response applications, IEEETrans. Smart Grid 3 (4) (2012) 1704–1711.

37] T. Kawakami, T. Yoshihisa, N. Fujita, M. Tsukamoto, A rule-based home energymanagement system using the Rete algorithm, 2013 IEEE 2nd GlobalConference on Consumer Electronics (GCCE) (2013).

38] M.S. Ahmed, A. Mohamed, R.Z. Homod, H. Shareef, A.H. Sabry, K.B. Khalid,Smart plug prototype for monitoring electrical appliances in Home EnergyManagement System, 2015 IEEE Student Conference on Research andDevelopment (SCOReD), IEEE (2015) 32–36.

39] L.-C. Huang, H.-C. Chang, C.-C. Chen, C.-C. Kuo, A ZigBee-based monitoring andprotection system for building electrical safety, Energy Build. 43 (6) (2011)1418–1426.

40] H.-W. Tseng, Y.-H. Lee, L.-Y. Yen, S.-Y. Yu, Y.-L. Chen, ZigBee (2.4 G) wirelesssensor network application on indoor intrusion detection, ConsumerElectronics-Taiwan (ICCE-TW) 2015 IEEE International Conference On, IEEE(2015) 434–435.

41] S.K. Gharghan, R. Nordin, M. Ismail, Energy-efficient ZigBee-based wirelesssensor network for track bicycle performance monitoring, Sensors 14 (8)(2014) 15573–15592.

42] K. Gill, S.-H. Yang, F. Yao, X. Lu, A zigbee-based home automation system, IEEETrans. Consum. Electron. 55 (2) (2009) 422–430.

43] A. Mahmood, I. Khan, S. Razzaq, Z. Najam, N. Khan, M. Rehman, N. Javaid,Home appliances coordination scheme for energy management (HACS4EM)using wireless sensor networks in smart grids, Procedia Comput. Sci. 32(2014) 469–476.

44] D. Setlhaolo, X. Xia, Optimal scheduling of household appliances with abattery storage system and coordination, Energy Build. 94 (2015) 61–70.

45] Q. Lin, L. Gao, X. Li, C. Zhang, A hybrid backtracking search algorithm forpermutation flow-shop scheduling problem, Comput. Ind. Eng. 85 (2015)437–446.

46] X. Song, X. Zhang, S. Zhao, L. Li, Backtracking search algorithm for effectiveand efficient surface wave analysis, J. Appl. Geophys. 114 (2015) 19–31.

47] J.A. Ali, M. Hannan, A. Mohamed, M.G. Abdolrasol, Fuzzy logic speed controlleroptimization approach for induction motor drive using backtracking search

48] P. Civicioglu, Backtracking search optimization algorithm for numericaloptimization problems, Appl. Math. Comput. 219 (15) (2013) 8121–8144.

49] A. Salahuddin, S. Curtis, Climate extremes in Malaysia and the equatorialsouth China sea, Global Planet. Change 78 (3–4) (2011) 83–91.