Wireless sensor networks for cost efficient residential energy management in the smart grid

314 IEEE TRANSACTIONS ON SMART GRID, VOL. 2, NO. 2, JUNE 2011

Wireless Sensor Networks for Cost-EfficientResidential Energy Management in the Smart Grid

Melike Erol-Kantarci, Member, IEEE, and Hussein T. Mouftah, Fellow, IEEE

Abstract—Wireless sensor networks (WSNs) will play a key rolein the extension of the smart grid towards residential premises,and enable various demand and energy management applications.Efficient demand-supply balance and reducing electricity expensesand carbon emissions will be the immediate benefits of these appli-cations. In this paper, we evaluate the performance of an in-homeenergy management (iHEM) application. The performance ofiHEM is compared with an optimization-based residential energymanagement (OREM) scheme whose objective is to minimize theenergy expenses of the consumers. We show that iHEM decreasesenergy expenses, reduces the contribution of the consumers to thepeak load, reduces the carbon emissions of the household, and itssavings are close to OREM. On the other hand, iHEM applicationis more flexible as it allows communication between the controllerand the consumer utilizing the wireless sensor home area net-work (WSHAN). We evaluate the performance of iHEM underthe presence of local energy generation capability, prioritizedappliances, and for real-time pricing. We show that iHEM reducesthe expenses of the consumers for each case. Furthermore, weshow that packet delivery ratio, delay, and jitter of the WSHANimprove as the packet size of the monitoring applications, thatalso utilize the WSHAN, decreases.

Index Terms—Cost optimization, energy and demand manage-ment, home automation, smart grid, wireless sensor networks.

NOMENCLATURE

Energy consumption of appliance .

Length of the cycle of appliance .

Unit price for slot .

Timeslot of the arrival of request of applianceon day .

The ratio of time slot occupied by request ofappliance on day .

Length of timeslot .

Maximum allowable delay.

Delay of appliance .

Requested start time of appliance .

Manuscript received September 01, 2010; revised December 19, 2010; ac-cepted January 29, 2011. Date of publication March 17, 2011; date of currentversion May 25, 2011. This work was supported in part by an ORF-RE grantfrom the Ministry of Research and Innovation of the Government of Ontario,Canada. Paper no. TSG-00120-2010.

The authors are with the School of Information Technology and Engineering,University of Ottawa, K1N 6N5, Ottawa, ON, Canada. (e-mail: [email protected]; [email protected]).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TSG.2011.2114678

Set of appliances.

Set of days.

Set of time slots.

Set of requests for one day.

I. INTRODUCTION

A CCORDING TO the recent Annual Energy Outlookreport of the U.S. Energy Information Administration,

residential electricity demand is forecasted to increase by 24%within the following several decades [1], while the globalelectricity consumption trend is also reported to be increasingcontinuously [2]. The negative impacts of rising consumptionare becoming more evident with the diminishing fossil fuelsand accumulating greenhouse gases (GHG). Moreover, themismatch between demand and supply and lack of automationand monitoring tools have already caused major blackoutsworldwide. Apparently, the traditional power grid has shownsigns of inefficient operation and has been experiencing diffi-culties in meeting the requirements of the 21st century. As aresult, the Energy Independence and Security Act of 2007 gavea start for the smart grid implementation in the United States.

The milestone in the process of transition from the tradi-tional grid to the smart grid is the integration of information andcommunication technologies (ICT) to the power grid. The ad-vances in ICT can be employed to increase automation, integratedistributed renewable resources, secure the grid infrastructure,adopt electric vehicles (EVs), and enable efficient demand-sideenergy management. Within the concept of demand-side en-ergy management, residential energy management is recentlyattracting increasing interest from the research community. Thetraditional grid has demand response programs for large-scaleconsumers such as industrial plants or commercial buildings;however, it does not have a similar mechanism for the residen-tial consumers mostly due to two reasons. First, it has been hardto handle the large number of residential units without commu-nication, sensors, and efficient automation tools. Second, theimpact of demand response programs has been considered tobe relatively small when compared with their implementationcost. However in the smart grid, smart meters, low-cost sensors,smart appliances, and communications set the stage for novelresidential energy management techniques that involve commu-nications and interaction between consumers, devices, and thegrid.

In the state-of-the-art smart grid implementation, smartmeters have been installed in the majority of the consumerpremises, and time of use (TOU) tariffs will be activated shortly

1949-3053/$26.00 © 2011 IEEE

EROL-KANTARCI AND MOUFTAH: WIRELESS SENSOR NETWORKS FOR COST-EFFICIENT RESIDENTIAL ENERGY MANAGEMENT IN THE SMART GRID 315

by a number of utilities in North America. Consumers canalso generate renewable energy, consume some portion of itlocally, and sell the excess energy to the utility companies.For example, Ontario government’s micro feed-in tariff (FIT)program allows home owners to sell locally generated energy[3].

Recently, residential energy management has become an ac-tive topic and several appliance scheduling schemes have beenproposed. In [4], the authors design an automatic controllerto schedule appliances to provide an optimum cost. Similarly,in [5], the authors use the particle optimization techniqueto schedule demands in an automated way while in [6], theauthors optimize the schedule of a microCHP device using aneural network-based prediction approach. In [7], the authorsfocus on reducing the peak-to-average electricity usage ratio byfinding an optimal consumption schedule for the subscribers ina neighborhood where they employ a game theoretic approach.Although optimization-based demand scheduling schemes areable to provide energy savings, individual preferences of con-sumers are not considered in those schemes. However, by theuse of communication among the consumers and the devices, itis possible to design nonintrusive energy management schemes.

In our previous work, we proposed an iHEM applica-tion which employs a wireless sensor home area network(WSHAN), and exploits communications among the appli-ances and an energy management unit (EMU) [8], [9]. In theiHEM application, EMU communicates with the appliances,smart meter, and storage units to determine a convenient timeto accommodate the consumer demands. In this paper, we aimto compare the performance of the iHEM application with anoptimization-based scheduling technique. For this purpose, wedeveloped the optimization-based residential energy manage-ment (OREM) scheme, which aims to minimize the energyexpenses of the consumers by scheduling appliances to lessexpensive hours according to the TOU tariff. We show that theiHEM application decreases the cost of energy consumptionand its savings are close to those of the optimal solution. Italso reduces the contribution of the residential consumers tothe peak load and the carbon emissions of the household. Inthis paper, we also elaborated on the use of real-time pricingand priority-based appliance scheduling. Furthermore, weevaluated the performance of the WSHAN in terms of deliveryratio, delay, and jitter. We showed that increasing the packetsize of the underlying applications degrades the performanceof the WSHAN.

Our contribution is primarily an energy and demand manage-ment application that: 1) is cost-effective; 2) is able to choosebetween grid-supplied or locally generated power; and 3)involves consumer participation in energy saving actions usingwireless communications. Our scheme achieves energy costsavings similar to the optimization-based automated controlschemes that are present in the literature. Additionally it givesflexibility in using local resources, and increases consumercomfort by interacting with the users. This is an initial steptowards the convergence of pervasive communications andenergy management where consumer needs can be perceivedby the cyberphysical infrastructures and nonintrusive actionscan be taken with less consumer interaction. Low cost, rapid

deployment together with flexibility of device locations makeWSHANs promising candidates for pervasive energy manage-ment applications.

The rest of the paper is organized as follows. In Section II wepresent the related work. Section III introduces the recent com-munication technologies that are available for the residential en-ergy management applications. In Section IV we introduce theOREM scheme. In Section V we explain our iHEM scheme indetail, and in Section VI we give our simulation results. Finally,Section VII concludes the paper.

II. RELATED WORK

In the literature, demand management have been studied inseveral works. In [10], the authors propose cycling on and offrefrigerators for frequency regulation services. The EuropeanSMART-A project discusses delaying the cycles of appliancesaccording to the local power generation capacity of a house[11]. Aggregated residential demand response programs havealso been considered in [12].

A residential load control (RLC) scheme that is suitable forgrids with real-time pricing is proposed in [4]. The authors focuson an automatic controller that is able to predict the price of elec-tricity during the scheduling horizon and schedule appliances toprovide an optimum cost and waiting time within that horizon.Our optimization based solution is different than [4], as in ourscheme, consumers can choose an upper limit for the waitingtime at the setup time and we make use of TOU rates and ex-ploit communications.

In [5], the authors propose a decision-support tool (DsT) forsmart homes. A PHEV, space heater, water heater, pool pump,and a PV system are scheduled based on various TOU tariffsby using the particle swarm optimization technique. In [5], thecommunication among the distributed resources and consumershas not been considered, whereas in our scheme, the controllerand the users communicate through appliance interfaces.

In [6], several management and control schemes are proposedfor microgrids and for single houses. The authors use a neuralnetwork-based prediction approach to predict the day-ahead de-mand. According to the predicted demand, the schedule of themicroCHP device in each house is optimized. In addition, localappliances are controlled to optimize electricity import/exportof the home. Our optimization-based residential energy man-agement is different than [6] since we aim to minimize the costof electricity based on TOU rates. Our work relies on demandshifting rather than scheduling generation and consumption toattain a balance. Moreover, in our paper, we assume that eachhouse makes independent decisions unlike a set of houses beingcontrolled by a steering signal from a global controller as de-scribed in [6].

In [7], the authors focus on reducing the peak-to-averageelectricity usage ratio by finding an optimal consumptionschedule (OCS) for the subscribers in a neighborhood. Theauthors employ a game theoretic approach. In [13], the authorspropose an energy management protocol which allows con-sumers to set a maximum consumption value and the residentialgateway is able to turn off the appliances that are in standbymode, or overwriting the user defined programmes with lessenergy consuming ones. However, defining a maximum value


for consumption is not practical and overwriting consumersettings may result in discomfort of the inhabitants. Our iHEMapplication interacts with the consumers via appliance inter-faces, and consumers negotiate with the controller using athree-way handshake protocol.

In [8] and [9], we proposed a TOU-aware appliance coor-dination scheme (ACS). Consumers are supplied with a sug-gested time when they wish to turn on their appliances. Thesuggested time is calculated based on TOU rates, generation ca-pacity, stored energy, and concurrent demands. The iHEM ap-plication used in this paper is based on ACS, and it is explainedin detail in Section V. A comparison of similar previous studieshas been given in Table I.

We believe that communication among smart devices, sensornodes, and controllers can be densely employed in residentialenergy management applications. Besides residential premises,smart grid also needs a communication infrastructure to connectthe utility assets [14]–[16]. Furthermore, several recent studiespropose using wireless sensor networks (WSN) and wirelessmultimedia sensor networks for monitoring and securing thepower transmission and distribution segment [17]–[20]. In [21],the authors investigate the use of IEEE 802.15.4 based WSNsin the substations where the time-critical applications have beenshown to suffer from delay. On the other hand WSNs are con-sidered to be convenient for residential non-time-critical appli-cations. Moreover, the experimental results of [22] imply thatusing Zigbee in the AMI and in the smart home increases inter-operability.

III. COMMUNICATION TECHNOLOGIES FOR HOME

AUTOMATION TOOLS

Residential energy management is possible with HANs. AHAN is a network of appliances, thermostats, pool pumps,lights, and other consumer devices. Although utility owned andoperated wide area networks (WAN) require high speed andreliable technologies, performance and reliability constraintsare expected to be more relaxed in the HAN, and the technologyoptions are more diverse than the utility backbones [23]. HANscan be implemented by wireline or wireless communications.There are a large number of wireline communication standardswhich have evolved in time. Currently, three standards havebeen selected by NIST as the possible power line carrier stan-dards in smart grids, which are IEEE P1901 (BPL), ITU-TG.Hn (HomeGrid), and ANSI/CEA 709.2 standards [24].Wireline technologies have the advantage of using the existingin-home wiring, and they generally have higher data ratesthan the wireless technologies. However, they are challengedby noisy channel conditions, channel characteristics that varydepending on the number and the type of appliances pluggedin, electromagnetic interference (EMI) due to unshielded powerlines and signal interference among the units of a building [25].

Currently, three wireless standards appear to be strong candi-dates for smart grid applications. These are Zigbee, Wi-Fi, andZ-wave standards. Wireless technologies use airborne signals,therefore they do not require cabling, and node locations can bemore flexible than wireline technologies. This advantage intro-duces the challenge of energy efficiency since the devices willbe operating on batteries in general. Zigbee uses a duty-cycling

principle to attain energy efficiency. However, it has limitedbandwidth since it has been initially designed for networks withlow communication intensity. On the other hand, recently intro-duced ultra-low power Wi-Fi chips position Wi-Fi as a strongalternative with its longer lifetime than the conventional Wi-Fiand with its higher bandwidth than Zigbee and Z-wave. Cur-rently, ZigBee Alliance [26] has developed Smart Energy Profile2.0 to meet the needs of smart metering and AMI, and providecommunication among utilities and household devices such assmart thermostats [27]. In the course of Zigbee Smart EnergyProfile, the Zigbee Alliance and the HomePlug Alliance is col-laborating to provide energy management in large apartmentsand commercial buildings.

In this paper, we utilize Zigbee since it is widely used forWSHANs and it is a strong candidate for smart grid energy man-agement applications. Zigbee is a short-range, low-data rate,energy-efficient wireless technology that is based on the IEEE802.15.4 standard. Zigbee utilizes 16 channels in the 2.4 GHzISM band worldwide, 13 channels in the 915 MHz band in NorthAmerica, and one channel in the 868 MHz band in Europe. Thesupported data rates are 250 kbps, 100 kbps (available in IEEE802.15.4-2006), 40 kbps, and 20 kbps, and its range varies be-tween 30–70 m indoors. Zigbee can support up to 64 000 nodes(devices). Zigbee certified devices can work for several yearswithout the need for battery replacement due to the low dutycycle mechanism. However, one of the drawbacks of Zigbee isthat it does not initially support IP addressing. Nevertheless, re-cently developed IPv6 over low-power wireless personal areanetworks (6LoWPAN) standard, which is defined in the IETFRFC 4944 [28], aims to integrate IPv6 addressing to LoWPANslike Zigbee. 6LoWPAN adds an adaptation layer to handle frag-mentation, reassembly, and header compression issues in orderto support IPv6 packets on the short packet structure of Zigbee.

IV. OPTIMIZATION-BASED RESIDENTIAL ENERGY

MANAGEMENT (OREM)

We propose an LP model to minimize the total cost of elec-tricity usage at home. Despite that home appliances consume thesame amount of energy regardless of the time they are switchedon, in the smart grid, as a result of the TOU tariffs, the hourswhen the appliances are used affect the cost of energy.

In the OREM scheme, we assume that one day is divided intoequal length consecutive timeslots which have varying pricesfor electricity consumption similar to TOU tariff. Our objectivefunction minimizes the total energy expenses by scheduling theappliances in the appropriate timeslots. In the LP model, con-sumer requests are given as an input and an optimum schedulingis achieved at the output.

Our objective function is defined in (1)

Minimize (1)

subject to the constraints given by (2)–(5). is the ratio of theduration that an appliance runs in a timeslot to the total length ofthe appliance cycle. is the average energy consumption of anappliance for a cycle. Appliances may have varying power con-sumption values within one cycle. For example, for a washer,


heating consumes the highest amount of energy. In the model,we assume an average consumption value for the whole cycle.The OREM scheme can be extended to schedule subcycles whennew appliances that can tolerate waiting times between cyclesbecome available. For the time being, we consider the conven-tional operating principles of the appliances where a cycle iscompleted without interruptions. is the length of the cycleof appliance and is the unit price for slot

(2)

In the OREM scheme, we ensure that the total duration of thecycles of the scheduled appliances does not exceed the lengthof the timeslot that is assigned for them in inequality (2)

(3)

A cycle may start at the end of one timeslot and it will naturallycontinue in the consecutive timeslot. Equation (3) ensures thatan appliance cycle is fully accommodated without experiencingany interruptions

(4)

(5)

OREM schedules the cycle of appliances to a convenienttimeslot. As a result, appliances may start later than the timethey are actually turned on, which creates a delay. To minimizethe cost of energy usage, appliances could be scheduled to lessexpensive timeslots; however, this generates bursts in thosetimeslots and it may increase the waiting time (i.e., delay).In our scheme, we bound the maximum delay, , to twotimeslots to reduce consumer discomfort and to avoid bursts ofrequest. Equations (4) and (5) ensure that the maximum delayis limited by an upper bound as the request is either accommo-dated in the present or the next timeslot. Hence, requests do notpile up in certain timeslots.

The OREM scheme assumes that the consumer requests aregiven in advance. Hence, we reach an optimal scheduling for theappliances. In practice, this information might not be availableand the appliances may have to be scheduled as the consumersturn them on. In this paper, our goal is to find a solution that givesa lower bound for the iHEM application, therefore we do not usea dynamic optimization scheme as it may yield to suboptimalsolutions.

V. IHEM APPLICATION

The aim of the iHEM application is decreasing the cost of en-ergy usage at home while causing the least comfort degradationfor the consumers. This scheme uses appliances with commu-nication capability, a WSHAN, and a central EMU. The iHEM

Fig. 1. iHEM packet formats. (a) START-REQ packet. (b) AVAIL-REQ packet.(c) UPDATE-AVAIL packet.

application is based on ACS of [8], which accommodates con-sumer demands at times when electricity usage is less expensiveaccording to the local TOU tariff.

In the iHEM application, consumers may turn on their ap-pliances at any time, regardless of peak time concern, and thescheme suggests start times to consumers. Consumer demandsare processed in near real time unlike the OREM scheme.

When a consumer turns on an appliance, the appliance gener-ates a START-REQ packet and sends it to EMU. The format ofthe START-REQ packet is given in Fig. 1(a). The first field of thepacket is the Appliance ID. The sequence number field denotesthe sequence number of the request generated by the appliance,since the appliance may be turned on several times in one day.Start time is the timestamp given when the consumer turns onthe appliance. The duration field denotes the length of the appli-ance cycle. Each appliance has different cycle lengths. A cyclecould be a washing cycle for a washer or the time required forthe coffee maker to make the desired amount of coffee. This du-ration depends on the consumer preferences, i.e., the selectedappliance program.

Upon receiving a START-REQ packet, EMU communicateswith the storage unit of the local energy generator to retrievethe amount of the available energy. It also communicates withthe smart meter to receive updated price information from theutility. EMU and smart meter exchange packets periodicallywhile EMU communicates with the local storage unit when ademand arrives. It sends an availability request packet, namely,AVAIL-REQ. The AVAIL-REQ packet format is given inFig. 1(b). The storage ID field is the ID of the storage unitthat is attached to the local energy generation unit. When thehouse is equipped with multiple energy generation devicessuch as solar panels and small wind tribunes, the amount ofenergy stored in their local storage units may have to be inter-rogated separately. The packet sequence number is used forthe same purpose as described previously. Code field carriesthe controller command code. In our application this field isused for inquiring the amount of available energy, hence it is astatic value. However, other applications may also use this codefield, e.g., to send a command to the storage unit to dispatchenergy to the grid. Upon reception of AVAIL-REQ, the storageunit replies with a AVAIL-REP packet where the amount ofavailable energy is sent to the EMU.

After receiving the AVAIL-REP packet, EMU determines theconvenient starting time of the appliance by using Algorithm1. The algorithm first checks whether locally generated power


Fig. 2. Message flow for iHEM.

is adequate for accommodating the demand. If this is the case,the appliance starts operating, otherwise the algorithm checksif the demand has arrived at a peak hour, based on the requestedstart time, . If the demand corresponds to a peak hour, itis either shifted to off-peak hours or mid-peak hours as longas the waiting time does not exceed , i.e., maximumdelay. The computed delay is returned to the consumer asthe waiting time. prevents requests from piling up atcertain intervals and create new peak periods similar to OREM.EMU computes the waiting time as the difference betweenthe suggested and requested start time, and sends the waitingtime in the START-REP packet to the appliance. The consumerdecides whether to start the appliance right away (StartImme-diately()) or wait until the assigned timeslot depending on thewaiting time (StartDelayed()). The decision of the consumeris sent back to the EMU with a NOTIFICATION packet thathas the same format as the START-REQ packet. The start-timefield of the NOTIFICATION packet denotes the negotiatedrunning time of the appliance, i.e., it could be either the timethe appliance is turned on or the start time suggested by theEMU. This information is required to allocate energy on thelocal storage unit when it is used as the energy source. Since itis further possible to sell excess energy to the grid operators, theamount of energy that needs to be reserved for the appliancesthat will run with the local energy has to be known. EMUsends an UPDATE-AVAIL packet to the storage unit to updatethe amount of available energy (unallocated) on the unit. Theformat of the UPDATE-AVAIL packet is given in Fig. 1(c). Thetime diagram of the packet flow for the iHEM application isgiven in Fig. 2. The handshake protocol among the applianceand the EMU ensures that EMU does not force an automatedstart time. We avoid this approach to increase the comfort ofthe consumers and to provide more flexibility.

Algorithm 1—Scheduling at the EMU

1:2:3:4: if (stored energy available TRUE) then5:6: else7: if ( is in peak) then8:9: if then

10:11: if then12:13: else14:15: end if16: else17:18: end if19: else20: if ( is in mid-peak) then21:22: if then23:24: else25:26: end if27: else28:29: end if30: end if31: end if

We employ a WSHAN to relay the messages of iHEM ap-plication since EMU may be physically located far from the ap-pliances, or obstacles may prevent direct communication amongappliances and the EMU. Deploying a WSHAN only for energymanagement could be costly; however, we propose to use theexisting WSHAN that is already implemented for monitoringapplications in the smart home. The WSHAN can continue itsregular task, such as inhabitant health monitoring, and at thesame time, it can relay iHEM messages. In Section VI we showthe impact of these underlying applications on the performanceof the WSHAN.

We assume the sensor network implements the Zigbee pro-tocol and we give the network topology in Fig. 3. Zigbee allowstwo types of devices, which are full function device (FFD) andreduced function device (RFD). FFDs can be interconnected ina mesh topology, which means they can communicate with theirpeers while RFDs are simpler than FFDs, and they can be theedge nodes in a star topology. In our model home, the WSHANis organized in a cluster-tree topology where the nodes in thebedroom and bathrooms are RFDs and the nodes in the livingroom and smart meter are FFDs and EMU is the personal areanetwork (PAN) coordinator. It is possible to designate the smartmeter as the PAN coordinator; however, we prefer EMU dueto its central position. PAN coordinator may choose to operatein beacon-enabled mode or beaconless mode. In the beacon-en-abled mode, it defines the duty cycle with the superframe dura-tion (SD) of the superframe structure which is shown in Fig. 4.A superframe synchronizes the nodes in the network, and nodescommunicate only in the active period. In the contention accessperiod (CAP) of the superframe nodes compete to achieve ac-cess to transmit their data by using the slotted carrier sense mul-tiple access with collision avoidance (CSMA/CA) technique.Contention free period (CFP) provides guaranteed time slots(GTS) for the nodes that have previously reserved these slotsfor communication. One cycle of active and inactive periods can


TABLE ISUMMARY OF THE PREVIOUS WORKS

Fig. 3. WSHAN topology.

Fig. 4. 802.15.4/Zigbee superframe format.

occur within a beacon interval (BI), which starts at the beginningof a beacon frame and ends at the beginning of the next beaconframe. SD and BI are given as follows:

SD

(6)

BI

(7)

where SO is the superframe order and BO is the beacon order.In the standard, the range of the SO and BO is defined asSO BO .

VI. SIMULATION RESULTS

We use ILOG CPLEX optimization suite to solve OREM ofSection IV. For iHEM simulations, we implemented a discrete

event simulator in C++. In the first part of this section, we an-alyze the contribution of the appliances to the energy bill andpeak load for the OREM scheme, for the iHEM applicationand for the case when there are no energy management mech-anisms. We evaluate the performance of the WSHAN in termsof delivery ratio, delay, and jitter considering iHEM applicationwith fixed packet size and other smart home applications withvarying packet sizes. In the following subsections, we analyzethe carbon emission reductions achieved by iHEM. We also ex-tend our simulations to three case studies, first, iHEM with localenergy generation, then iHEM with prioritized appliances, andfinally, iHEM for real-time pricing.

Residential energy consumption may vary depending on anumber of factors such as the size of a house, the number of oc-cupants, the location, and the season. These parameters impactheating, cooling, lighting, and similar loads of the household. In[29], the authors experimentally show that consumption transi-tion can be modeled by a Poisson process which correspondsto switching on an appliance in our case. In our simulations,to model the increasing demand during the peak hours we uti-lize a Poisson process with increasing arrival rate at peak hours.The interarrival times between two requests is negative expo-nentially distributed with a mean of 12 h. During morning peakperiods and evening peak periods the interarrival time is nega-tive exponentially distributed with a mean of 2 h.

We analyze the use of four appliances; washer, dryer, dish-washer, and coffee maker. The duration and energy consump-tion of these appliances are vendor specific; however, we usethe reference values for average load per cycle given in [11].The washer, dryer, dishwasher, and coffer maker is assumed toconsume 0.89 kWh, 2.46 kWh, 1.19 kWh, and 0.4 kWh whilethe duration of the appliance cycles are given as 30, 60, 90, and10 min, respectively.

We set the TOU rates as given in Table II while weekendsare off-peak periods. The rates are taken as in a typical winterTOU tariff [30]. In OREM scheme, the length of one timeslot isassumed to be 6 h, hence the maximum delay for OREM is setto hours. We simulate the schemes from 10 days to210 days (approximately seven months). We present results asthe average of 10 simulation runs.

In Fig. 5, we compare the savings of the iHEM application,the optimal solution provided by OREM, and the case with noenergy management. Note that total contribution of the appli-ances to the energy bill increases with increasing days because


TABLE IIWEEKDAY TOU RATES FOR WINTER

Fig. 5. Total contribution of appliances to the energy bill.

the bill is calculated cumulatively. As seen in Fig. 5, the iHEMapplication decreases the contribution of the appliances to theenergy bill and the savings of the iHEM application is close tothe optimal solution. After 210 days, iHEM scheme results inalmost 30% reduction in the energy bill while the optimal solu-tion reduces the bill by around 35%.

In Fig. 6, we give the deviation of the two cases: iHEM and noenergy management, from the optimal solution. The deviationfrom optimal solution when energy management is not used is0.5, while the deviation of iHEM from the optimal solution isaround 0.1. This shows that the iHEM application approachesthe savings provided by the OREM scheme. As more appliancesopt in the iHEM application, savings are expected to increaseas [8] shows that increasing the number of consumer requestsincrease savings.

In Fig. 7, we present the maximum delay experienced by theconsumer. Note that when energy management is not used, i.e.,the conventional use of the appliances, an appliance starts assoon as the consumer turns it on and the delay is almost zero.Therefore, we do not include this case in the plot. For the iHEMapplication, consumers experience a maximum delay around 6h. In the OREM scheme, demands can be delayed for two times-lots, i.e., 12 h. For several appliances such as washer or dish-washer, this much delay can be tolerated; however, for some ap-pliances, such as the coffee maker, the consumer will be likely

Fig. 6. Percentage of deviation from the optimal solution.

Fig. 7. Maximum delay observed by a consumer.

to choose to start the appliance immediately. The iHEM appli-cation allows consumers to negotiate or decline to negotiate onthe suggested time.

We showed that, on the demand side, residential energy man-agement schemes are useful for decreasing energy bills. On thesupplier side, reduction in peak load would be another benefitof these schemes. In Fig. 8, we show the contribution of the ap-pliances on the average demand. When energy management isnot employed, 0.3 of the load generated by the appliances takesplace during peak periods while the iHEM application shiftsthose requests from peak times and only 0.05 of the total house-hold load is accommodated during peak hours. Therefore, theiHEM application is also capable of reducing the peak demandin the smart grid. We compare the performance of our scheme


Fig. 8. Percentage of the contribution of the appliances to the total load on peakhours.

with the previously proposed schemes that aim to reduce themonthly energy expanses of the consumers. We compare iHEMwith RLC [4], DsT [5], and OCS [7] in terms of savings andpeak load reduction in Table III. iHEM provides similar savingswith DsT and RLC while OCS is able to provide more savings.On the other hand, iHEM outperforms OCS in peak load reduc-tion proportional to average load. In iHEM higher amount of thepeak load is shifted to off-peak and mid-peak periods.

In our simulations, we evaluate the performance of theWSHAN in terms of packet delivery ratio, end-to-end delayand jitter. Delivery ratio is the ratio of the number of suc-cessfully received packets to the number of sent packets.End-to-end delay is the interval between sending a packet fromthe application layer of the source and receiving the packet atthe application layer of the destination. Jitter is the differencebetween the delays experienced by the packets.

We consider a network topology as described in Fig. 3 whereone PAN coordinator is engaged for residential energy manage-ment and other types of WSHAN applications. The nodes area mixture of RFD and FFD devices, i.e., five FFD devices areused for routing packets and 14 RFD devices, four of which areconnected to the appliances are utilized. We use Zigbee protocolutilizing the 2.4 GHz ISM band and the bandwidth is 250 kb/s.Deploying a dedicated WSHAN for relaying iHEM packets iscostly; therefore, as we mentioned before, WSHAN relays thepackets of iHEM as well as a monitoring application. We showthe impact of the varying packet size of the monitoring appli-cation on the overall performance of the network. We vary thepacket sizes of this application between 32B and 128B. Weassume the nodes generate packets at 10 min intervals. Notethat when the packet size exceeds the maximum physical layerpacket size defined in IEEE 802.15.4 specifications (128B), it isfragmented to smaller packets.

In Fig. 9(a), we show that the packet delivery ratio of theWSHAN decreases as the packet size of the monitoring ap-plication increases. For packet size of 32B, the delivery ratiois almost 90%. The end-to-end delay is around 0.75 s as seenin Fig. 9(b). Shorter packets decrease contention period there-fore delivery ratio is high and delay is less for those packetsthan longer packets. For iHEM performance jitter is anotherimportant parameter. Jitter is the variation in delay and highjitter values mean that consumers may experience variableend-to-end delays when they are communicating with theEMU. In Fig. 9(c), we show that jitter is less than 0.06 s, whichis negligible considering the human response times.

A. Carbon Emissions for iHEM

Climate change and global warming are considered to be dueto the amount of accumulating GHG in the atmosphere; there-fore, reducing carbon emissions is very significant for a sustain-able habitat. Smart grid aims to reduce the CO emissions of thepower sector which currently contributes almost 40% of the totalemissions [31]. The most effective way of reducing emissions isto mitigate from fossil-fuel based energy generation to renew-able and clean energy resources such as hydro, solar, or windpower. However, the renewable resources are not integrated wellto the power grid yet, because of their intermittent nature, lackof storage technology to balance their output, and finally, lackof efficient transportation technologies. Moreover, during peakhours utilities bring peaker plants online which use resourcessuch as diesel or heavy oil, which have high emissions. This im-plies that the time of consumption affects the carbon footprintof the consumers [32].

In this section, we focus on the carbon emissions resultingfrom the electricity consumption of the appliances during peakhours, considering two different regional grids that have dif-ferent energy generation mixes [33]. We assume that Region 1is geographically rich in renewable resources. We assume thebase generation mix is as follows: 50% nuclear, 25% coal andnatural gas, 25% hydro, wind, and solar. For peak generationmix we assume: 40% nuclear, 40% diesel and heavy oil, 20%hydro, wind, and solar. For Region 2 we consider a grid wheregeneration mostly depends on fossil fuels. We assume the basegeneration mix is 30% nuclear, 60% coal and natural gas, 10%hydro, wind, and solar, and peak generation mix is 25% nuclear,70% diesel and heavy oil, 5% hydro, wind, and solar. We givethe carbon equivalent emissions of the generation resources inTable IV, which are taken from [34]. The emissions related withRegion 1 (R1) and Region 2 (R2) are calculated by using theabove mixture ratios. In Fig. 10 we show the carbon equiva-lent emissions for R1 and R2 for two cases, namely no energymanagement (NEM) and iHEM. iHEM can provide 10% to 20%lower emissions depending on the regional characteristics whereR1 represents a relatively optimistic scenario with higher pene-tration of renewable resources, and R2 represents a pessimisticscenario with less renewable energy generation penetration.

B. iHEM With Local Power Generation

In the smart grid, energy management for the consumers ex-tends beyond consumption control, and it includes the controlof the energy generation of the home. Energy generated at home


Fig. 9. WSHAN performance in terms of delivery ratio, delay, and jitter. (a) Delivery ratio. (b) Delay (s). (c) Jitter (s).

can be consumed by the appliances and electronic devices, andthe excess energy can be sold to the grid. In this section, we as-sume that the model smart home has three PV panels that areable to generate 350 W per day considering a climate with sev-eral hours of sun light. The rate of electricity sold to the griddepends on the regional grid operator and it is called feed-intariff (FIT). In Ontario 80.2 cents/kWh is the flat rate for solargeneration, and we use this value in our simulations. In Fig. 11,we show the savings introduced by the regular iHEM scheme(iHEM w/o feed in) and iHEM with local energy generation(iHEM with feed in). Naturally, iHEM with feed in leads to lessexpenses because some of the appliance requests can be sup-plied by local resources, furthermore excess energy is sold tothe grid. In Fig. 12 we show the daily utilization of local energyresources, the grid, and the amount of energy sold to the grid.Almost 12 kWh of energy is used from the grid, almost 1 kWhis used from the local resources and the remaining is sold to thegrid. Note that these results may vary depending on the season,

weather conditions, location of the house, and the number andthe efficiency of the panels used.

C. iHEM With Priority-Based Scheduling

In this set of simulations, we give priority to a subset of theappliances. High priority appliances are turned on immediatelyregardless of the peak hours. This scenario corresponds to a casewhere users have either preconfigured a subset of their appli-ances as high priority appliances or several appliances are notable to communicate with the EMU. The latter case may be morecommon until smart appliances are widely adopted. In Fig. 13,iHEM w/o priority is the regular iHEM scheme while “iHEMwith priority” includes appliances with different priorities. Nat-urally, savings of the consumer reduce when appliances havepriorities because this reduces the flexibility of scheduling theappliances in the off-peak hours. Similarly, the peak load in-creases for “iHEM with priority” as seen from Fig. 14.


TABLE IIICOMPARISON OF COST AND PEAK LOAD REDUCTION

Fig. 10. Carbon emissions in two regions with different energy generation mix.

TABLE IVCARBON EMISSIONS OF ENERGY GENERATION RESOURCES AND TWO

REGIONAL GRIDS

D. iHEM With Real-Time Pricing

Throughout the previous sections, we assumed TOU pricingis used in the grid where TOU defines a fixed rate for a cer-tain amount of time. In the smart grid, it is also possible to havereal-time (dynamic) pricing. Dynamic pricing reflects the ac-tual price of the electricity in the market to the consumer bills.The market price of electricity is generally determined by the in-dependent system operator where the day-ahead or hour-aheadprices are announced to the consumers. Raw market price ofthe electricity depends on several factors such as the load fore-casts, supplier bids, and importer bids. The final price is deter-mined after taxes, regulatory charges, transmission and distribu-tion fees, and other service charges are added to the raw marketprice. In this subsection, we analyze the performance of iHEMfor real-time pricing. In Fig. 15, we present the contribution of

Fig. 11. Appliance electricity usage expenses with local power generation.

Fig. 12. Local energy and grid usage profile.

the appliances for the regular iHEM scheme (iHEM-TOU) andiHEM for real-time pricing. iHEM for real-time pricing still in-troduces savings when compared to the case without any energymanagement however iHEM performs better with TOU pricingbecause the scheduling can be coordinated better when the off-peak price stays fixed for a certain amount of time. Schedulingunder real-time pricing may require demand prediction in orderto increase the performance of scheduling.

VII. CONCLUSION

Residential energy management, smart appliances,WSHANs, and their integration to smart grid applications


Fig. 13. Total contribution of appliances to the energy bill with priority-basedscheduling.

Fig. 14. Contribution of the appliances to the total load on peak hours withpriority-based scheduling.

are becoming popular topics as the governments and the util-ities are urging for migration to the smart grid. In this paper,we introduce the OREM and the iHEM schemes to reducethe share of the appliances in the energy bills and to reducetheir contribution to the peak load. We show that the iHEMapplication decreases the contribution of the appliances to theenergy bill, significantly. Meanwhile, the savings of the iHEMscheme is close to the savings of the optimal solution providedby the OREM scheme. From the utility perspective, reducingthe peak load is an important issue. The iHEM application is

Fig. 15. Total contribution of appliances to the energy bill with real-timepricing.

shown to decrease the load on the peak hours and the power-re-lated carbon emissions, as well. In this paper, we also evaluatethe performance of iHEM under various scenarios, which areiHEM with the presence of local energy generation, iHEM withprioritized appliances (or appliances without communicationcapabilities), and iHEM with real-time pricing. In each case,we showed that iHEM reduces the expenses of the consumerscompared to the case without energy management. Further-more, we elaborated on the performance of the WSHAN interms of packet delivery ratio, delay, and jitter. Since the solepurpose of the WSHAN is not relaying iHEM messages but it isalso responsible for the smart home monitoring application, theperformance of WSHAN depends on the packet sizes generatedby the monitoring application. Through simulations, we showthat as the packet size of the underlying monitoring applicationdecreases, the delivery ratio increases and the delay decreases,which translates into improved network performance.

As a future work, we are planning to include learning tech-niques from the artificial intelligence (AI) field to increase con-sumer comfort and pervasiveness of our application. Further-more, our schemes can be extended for a new class of appli-ances that allow subcycle scheduling. The availability of suchappliances will enrich the opportunities of residential demandmanagement applications.

REFERENCES

[1] J. J. Conti, P. D. Holtberg, J. A. Beamon, A. M. Schaal, G. E.Sweetnam, and A. S. Kydes, Annual energy outlook with projectionsto 2035, report of U.S. Energy Information Administration (EIA),Apr. 2010 [Online]. Available: http://www.eia.doe.gov

[2] T. F. Garrity, “Getting smart,” IEEE Power Energy Mag., vol. 6, no. 2,pp. 38–45, Apr.–Mar. 2008.

[3] Ontario Power Authority FIT Program [Online]. Available: http://fit.powerauthority.on.ca

[4] A.-H. Mohsenian-Rad and A. Leon-Garcia, “Optimal residential loadcontrol with price prediction in real-time electricity pricing environ-ments,” IEEE Trans. Smart Grid, vol. 1, no. 2, pp. 120–133, 2010.


[5] M. A. A. Pedrasa, T. D. Spooner, and I. F. MacGill, “Coordinatedscheduling of residential distributed energy resources to optimize smarthome energy services,” IEEE Trans. Smart Grid, vol. 1, no. 2, pp.134–143, 2010.

[6] A. Molderink, V. Bakker, M. Bosman Johann, L. Hurink, and G. J. M.Smit, “Management and control of domestic smart grid technology,”IEEE Trans. Smart Grid, vol. 1, no. 2, pp. 109–119, 2010.

[7] A.-H. Mohsenian-Rad, V. W. S. Wong, J. Jatskevich, and R. Schober,“Optimal and autonomous incentive-based energy consumption sched-uling algorithm for smart grid,” presented at the IEEE PES Innov.Smart Grid Technol. Conf., Gaithersburg, MD, Jan. 2010.

[8] M. Erol-Kantarci and H. T. Mouftah, “Using wireless sensor networksfor energy-aware homes in smart grids,” in Proc. IEEE Symp. Comput.Commun. (ISCC), Jun. 2010, pp. 456–458.

[9] M. Erol-Kantarci and H. T. Mouftah, “TOU-aware energy manage-ment and wireless sensor networks for reducing peak load in smartgrids,” presented at the Green Wireless Commun. Networks Workshop(GreeNet), IEEE VTC Fall, Ottawa, ON, Canada, Sep. 2010.

[10] D. G. Infield, J. Short, C. Horne, and L. L. Freris, “Potential for do-mestic dynamic demand-side management in the UK,” presented at theIEEE Power Eng. Soc. Gen. Meet., Tampa, FL, Jun. 2007.

[11] R. Stamminger, Synergy potential of smart appliances Univ. Bonn,Bonn, Germany, Mar. 2009 [Online]. Available: http://www.smart-a.org, Deliverable 2.3 of work package 2 from the Smart-A project

[12] F. Rahimi and A. Ipakchi, “Demand response as a market resourceunder the smart grid paradigm,” IEEE Trans. Smart Grid, vol. 1, no.1, pp. 82–88, Jun. 2010.

[13] S. Tompros, N. Mouratidis, M. Draaijer, A. Foglar, and H. Hrasnica,“Enabling applicability of energy saving applications on the appliancesof the home environment,” IEEE Netw., vol. 23, no. 6, pp. 8–16, Nov.2009.

[14] V. C. Gungor and F. C. Lambert, “A survey on communication net-works for electric system automation,” Comput. Netw., vol. 50, pp.877–897, 2006.

[15] M. Qureshi, A. Raza, D. Kumar, S. Kim, U. Song, M. Park, H. Jang, H.Yang, and B. Park, “A survey of communication network paradigms forsubstation automation,” in Proc. IEEE Int. Symp. Power Line Commun.Its Appl., Jeju Island, South Korea, 2008, pp. 310–315.

[16] V. K. Sood, D. Fischer, J. M. Eklund, and T. Brown, “Developing acommunication infrastructure for the smart grid,” presented at the IEEEElect. Power Energy Conf. (EPEC), Montreal, QC, Canada, Oct. 2009.

[17] Y. Yang, F. Lambert, and D. Divan, “A survey on technologies forimplementing sensor networks for power delivery systems,” presentedat the Power Eng. Soc. Gen. Meet., Tampa, FL, 2007.

[18] V. C. Gungor, B. Lu, and G. P. Hancke, “Opportunities and challengesof wireless sensor networks in smart grid,” IEEE Trans. Ind. Electron.,vol. 57, no. 10, Oct. 2010.

[19] R. A. Leon, V. Vittal, and G. Manimaran, “Application of sensornetwork for secure electric energy infrastructure,” IEEE Trans. PowerDel., vol. 22, no. 2, pp. 1021–1028, Apr. 2007.

[20] M. Erol-Kantarci and H. T. Mouftah, “Wireless multimedia sensor andactor networks for the next-generation power grid,” J. Ad Hoc Net-works, 10.1016/j.adhoc.2010.08.005, to be published.

[21] S. Ullo, A. Vaccaro, and G. Velotto, “The role of pervasive and coop-erative sensor networks in smart grids communication,” in Proc. 15thIEEE Mediterranean Electrotech. Conf. (MELECON), Apr. 2010, pp.443–447.

[22] S. Luan, J. Teng, S. Chan, and L. Hwang, “Development of a smartpower meter for AMI based on ZigBee communication,” in Proc. Int.Conf. Power Electron. Drive Syst., Nov. 2–5, 2009, pp. 661–665.

[23] J. G. Cupp and M. E. Beehler, “Implementing smart grid communica-tions,” Burns & McDonnell Marketing Communications, 2008.

[24] NIST Framework and Roadmap for Smart Grid Interoperability Stan-dards, [Online]. Available: http://www.nist.gov/smartgrid

[25] S. Galli and O. Logvinov, “Recent developments in the standardizationof power line communications within the IEEE,” IEEE Commun. Mag.,vol. 46, no. 7, pp. 64–71, Jul. 2008.

[26] ZigBee Alliance, [Online]. Available: http://www.zigbee.org/[27] B. Heile, “Smart grids for green communications,” IEEE Wireless

Commun., vol. 17, no. 3, pp. 4–6, Jun. 2010.[28] G. Montenegro, N. Kushalnagar, J. Hui, and D. Culler, “Transmission

of IPv6 packets over IEEE 802.15.4 networks,” Request for Comments4944 [Online]. Available: http://www.ietf.org

[29] S. W. Lai, G. G. Messier, H. Zareipour, and C. H. Wai, “Wirelessnetwork performance for residential demand-side participation,” pre-sented at the IEEE PES Innov. Smart Grid Technol. Conf. Eur. (ISGTEur.), Gothenburg, Sweden, 2010.

[30] TOU Rates, [Online]. Available: http://www.hydroottawa.com[31] International Energy Agency Report on Energy Technology Perspec-

tives 2010, [Online]. Available: http://www.iea.org[32] M. Erol-Kantarci and H. T. Mouftah, “The impact of smart grid

residential energy management schemes on the carbon footprint ofthe household electricity consumption,” presented at the IEEE Elect.Power Energy Conf. (EPEC) Halifax, NS, Canada, Aug. 25–27, 2010.

[33] B. Kantarci and H. T. Mouftah, “Greening the availability design of op-tical WDM networks,” in Proc. IEEE Globecom 2010 Workshop GreenCommun., Dec. 2010, pp. 1447–1451.

[34] Hydro Quebec [Online]. Available: http://www.hydroquebec.com

Melike Erol-Kantarci (S’08–M’10) received theB.S., M.Sc., and Ph.D. degrees from Computer Engi-neering Department, Istanbul Technical University,Turkey, in 2001, 2004, and 2009, respectively.

She was a Fulbright Visiting Researcher at theComputer Science Department, University of Cali-fornia, Los Angeles, from September 2006 to August2007. She has been a Postdoctoral Fellow at theSchool of Information Technology and Engineering,University of Ottawa, ON, Canada, since October2009. Her main research interests are communication

protocols, underwater sensor networks, and wireless sensor networks.

Hussein Mouftah (S’74–M’76–SM’80–F’90)received the B.Sc. degree in electrical engineeringand the M.Sc. degree in computer science fromthe University of Alexandria, Alexandria, Egypt, in1969 and 1972, respectively, and the Ph.D. degreein electrical engineering from Laval University,Quebec City, Canada, in 1975.

He has three years of industrial experiencemainly at BNR of Ottawa, now Nortel Networks(1977–1979). He was with the ECE Departmentat Queen’s University (1979–2002), where he was

prior to his departure a Full Professor and the Department Associate Head. Hejoined the School of Information Technology and Engineering, University ofOttawa in September 2002 as a Canada Research Chair Professor. He is theauthor or coauthor of six books, 40 book chapters and more than 1000 technicalpapers and 10 patents in this area.

Dr. Mouftah is a Fellow of the Canadian Academy of Engineering, the En-gineering Institute of Canadam, and the Royal Society of Canada RSC: TheAcademy of Science. He served as Editor-in-Chief of the IEEE Communica-tions Magazine and IEEE ComSoc Director of Magazines, Chair of the AwardsCommittee and Director of Education. He has been a Distinguished Speaker ofthe IEEE ComSoc (2000–2007).

Business

Wireless sensor networks for cost efficient residential energy management in the smart grid