Anda di halaman 1dari 5

GC'12 Workshop: Smart Grid Communications: Design for Performance

DRIFT: Differentiated RF Power Transmission for


Wireless Sensor Network Deployment in the
Smart Grid
Melike Erol-Kantarci, Hussein T. Mouftah
School of Electrical Engineering and Computer Science
University of Ottawa
Ottawa, ON, Canada
e-mails:{melike.erolkantarci, mouftah}@uottawa.ca

AbstractSmart grid calls for low-cost, ne-grained and longlasting monitoring solutions, to be able to provide reliable service
to customers, enhance situational awareness capabilities and
rene the operation of the grid and the microgrids, in addition to
enabling prompt utility reaction to emergencies. Wireless Sensor
Networks (WSNs) are promising candidates for monitoring the
smart grid, given their capability to cover a large geographic
region at low-cost. However, they do not provide long-lasting
operation capability due to the limited battery lifetime of the
sensors. Particularly, when sensors are deployed in hard-to-reach
or hazardous environments, replacing the batteries of the sensors
increase the cost of monitoring signicantly. In the literature,
energy-efcient protocols and ambient energy harvesting have
been proposed to extend the lifetime of the sensors while neither
of those offer a concrete solution for the smart grid. In this context, recent advances in Radio Frequency (RF) energy harvesting
offers a unique solution to make WSNs operationally ready for
smart grid monitoring tasks. Studies on RF energy harvesting
have focused on uniform power delivery to all sensors, however
it becomes essential to differentiate between critical zones and
less critical zones in smart grid monitoring tasks. In this paper,
we propose the Differentiated RF Power Transmission (DRIFT)
scheme which is based on an Integer Linear Programming (ILP)
model that maximizes the power received by the high priority
sensor nodes. We compare the performance of DRIFT with a
path length minimizing approach, namely Sustainable wireless
Rechargeable Sensor network (SuReSense). We show that DRIFT
is able to provide more power to high priority nodes than
SuReSense. We also show that there is a tradeoff between power
maximization and path length minimization.
Index TermsRF energy harvesting, smart grid, wireless
sensor networks.

I. I NTRODUCTION
Smart grid is the future electric power grid that aims to
enhance the reliability, security and efciency of the electrical services by integrating Information and Communication
Technologies (ICTs) to the power systems. Monitoring smart
grid equipment and assets is particularly important since the
healthy operation of the smart grid depends on the reliability of
the physical infrastructure. Wireless Sensor Networks (WSN)
have provided low-cost, ne-grained and robust monitoring
solutions for many critical domains such as military and health,
978-1-4673-4941-3/12/$31.00 2012 IEEE

which positions them as a strong candidate for smart grid


monitoring [1], [2].
One of the major challenges for WSNs in the smart grid
monitoring tasks is their relatively short lifetime due to limited
sensor batteries. In the smart grid, sensor nodes may not
be able to directly use mains power due to being placed
away from mains, or in the monitored environment, only
high/medium voltage may be available which needs to be
stepped down before use. Thus, energy harvesting emerges as
a signicant issue in long-term smart grid monitoring applications. In the literature, many studies have offered solutions
to extend the lifetime of WSNs. Those include duty-cycling,
energy-efcient protocols and ambient energy harvesting techniques. However, none of those solutions can support a WSN
for more than several years. For this reason, long-lasting smart
grid monitoring tasks call for novel solutions. RF energy
harvesting has recently emerged as a promising technology
to provide perpetual power to sensors. WSNs with RF energy
harvesting capability can be utilized in the smart grid as well
as building automation, industrial monitoring and so on [3].
RF-energy harvesting relies on wireless power transmission
where Electromagnetic (EM) signals from a transmitter can be
received by one or more sensors and stored in their capacitors.
The range of wireless power transmission is limited, hence
a WSN deployed in a large geographic region may require
several xed power transmitters to be mounted at xed locations or a mobile power transmitter may traverse the network.
Recently, these approaches have been considered in a few
studies. In [4], the authors have studied recharging of the
RFID tags by the RFID readers where they have minimized
the number of stationary readers in the network. In [5], the
authors have studied the optimal path of a mobile power
transmitter device when the charger visits each and every node
for power transfer. In [6], we have proposed the Sustainable
wireless Rechargeable Sensor network (SuReSense) scheme
which utilizes a mobile charger and selects the minimum
number of landmarks that are traversed by this mobile charger.
In this paper, we extend SuReSense considering different
priorities of the monitored equipment. In a smart grid asset,

1491

some equipment need precise monitoring and report more


frequently, such as transformers or capacitor banks in the
substations while ambient measurements may be less critical
for healthy operation of the grid. Therefore, sensors deployed
in critical areas require a robust source of power. Our proposed
scheme; Differentiated RF Power Transmission (DRIFT), aims
to deliver more power to high priority nodes using a mobile
power transmitter. We formulate an Integer Linear Programming (ILP) model whose objective is to maximize the power
received by the high priority nodes from a landmark. A
landmark is a location determined by the ILP where the mobile
charger parks and emits RF signals in order to transmit power
to the sensor nodes in its vicinity. We show that power received
by high priority sensor nodes for DRIFT is signicantly higher
than the power received with SuReSense. On the other hand,
SuReSense is able to reduce the path length of the mobile
power transmitter. Hence, we show that there is a tradeoff
between the amount of received power in one operation cycle
and the duration spent while waiting for the mobile charger.
The rest of the paper is organized as follows. In Section
II, we present the related work. In Section III we explain the
principles of RF energy harvesting in detail. In Section IV
we present the ILP formulation of the DRIFT scheme and
in Section V we evaluate the performance of DRIFT. Finally,
Section VI concludes the paper.
II. R ELATED W ORK
Using WSNs for smart grid monitoring has been recently
studied in [1], [2], [7]. In [1], the authors have presented
wireless link quality measurements from IEEE 802.15.4-based
sensor networks deployed in several smart grid assets. In
[2], the authors have explored the utilization of multimedia
sensor networks and actor networks in power plants, renewable
energy generation sites, overhead transmission lines, transmission towers, distribution transformers, feeder lines and
consumer premises. In [7], the authors have presented results
from a dynamic link quality based routing algorithm from a
WSN that is utilized for collecting data from circuit breakers
and transformers, in addition to ambient temperature and gas
density measurements at a test substation in Kentucky. Furthermore, in [8], an IEEE 802.15.4-based WSN has been employed
as a part of a residential energy management application that
helps reducing the electricity expenses of the consumers.
Briey, WSNs are promising in many smart grid application domains. However, for equipment and asset monitoring
applications their limited lifetime appears as a drawback.
In the literature, extending the lifetime of WSNs has been
widely studied. Duty cycling is a commonly adopted technique
besides energy-efcient protocols. An extensive survey of
energy conservation techniques has been presented in [9],
[11]. Furthermore, [10] presents a detailed survey on ambient
energy harvesting options such as sun, vibration, body heat,
foot strike and so on. Meanwhile, coupling duty cycling
with the slow energy-harvesting process has been studied in
[12] where the authors schedule active and inactive periods
according to energy income. Although previously proposed

techniques extend the lifetime of WSNs for several years, they


become inadequate in long-term smart grid monitoring tasks.
A recently emerging technology, i.e. RF energy harvesting
promises to solve this problem [3].
In [4], the authors have studied charging RFID tags by
RFID readers using wireless energy transfer. The authors have
rst considered stationary chargers and aimed to minimize the
number of chargers in the network. Then, considering mobile
tags, the authors have aimed at selecting reader locations that
provide adequate charging for the tags along their path. In [13],
the authors have proposed a charging-aware routing protocol
aiming to address simultaneous charging and packet transmission, particularly for the cases when power transmission and
packet transmission use the same frequency band. In [5], the
authors have aimed to maximize the docking time of a mobile
charger where the mobile charger is assumed to visit each
and every sensor node for replenishing their batteries. The
batteries are charged such that the minimum available energy
is higher than a threshold within one cycle of charging. In
[14], the authors have combined power transmission with data
collection using a mobile charger. In [15], the authors have
proposed the Mission-Aware Placement of Wireless Power
Transmitters (MAPIT) scheme that optimizes the placement
of RF power transmitters in the WSN while maximizing the
prot resulting from sensor-mission assignments.
In [6], we proposed the SuReSense scheme which uses
several mobile chargers to visit landmarks and replenish the
batteries of the sensor nodes in the vicinity of the landmarks. SuReSense consists of three steps. At the rst step,
minimum number of landmarks are selected according to
the locations and energy replenishment requirements of the
sensors. At the second step, landmarks are organized into
clusters where each cluster is serviced by a single mobile
charger. A mobile charger visits the assigned landmarks in
its cluster by following the shortest Hamiltonian cycle which
have been shown to be the optimal traveling path in [4].
SuReSense reduces the path length of the mobile charger,
allowing for a longer docking time to top up the battery of the
charger and reducing the waiting time of the sensors. However,
SureSense assumes that all of the sensors are identical and they
have the same priority. In this paper, we extend the wireless
power transmission idea of SuReSense to maximize the power
received by high priority sensor nodes.
III. W IRELESS P OWER T RANSMISSION BASICS
RF wireless energy harvesting can be realized by a transmitter Tpow and a receiver Rpow where dtr R. Here, dtr
denotes the distance between the transmitter and the receiver
and R denotes the power transmission range. According to the
free space model, the received power is inversely proportional
to d2tr and it is assumed to be 0 when dtr > R. Hence, the
received power is given by:

1492


Pr = Pt Gt Gr

4R

2
(1)

where is the wavelength, Gr is the linear gain and Pt Gt is


the transmitted power. For instance, for = 0.328 at 915MHz,
Gr = 3.98, transmitted power of 3W can be received as
0.325mW over a distance of 5m. Upon reception of RF signals,
the Rpow converts the received signal to DC voltage and stores
energy in a capacitor. Hence, a sensor equipped with a receiver,
converter and a capacitor can harvest energy and store it in its
battery for future use.
IV. D IFFERENTIATED RF P OWER T RANSMISSION FOR
W IRELESS S ENSOR N ETWORKS IN THE S MART G RID
RF energy harvesting is an opportunity to extend the lifetime
of the WSNs for long-term smart grid monitoring tasks. In the
smart grid environment, in addition to long-term monitoring
requirement, differentiating among the monitored equipment is
also essential. For instance, transformers and capacitor banks
in a substation are among the critical equipment that may
cause the failure of the distribution system. On the other
hand, ambient measurements from the same substation may
be less critical. In that case, smart grid applications will
demand more frequent status updates from a sensor collecting
data from a transformer than a sensor collecting ambient
data. Our aim is to provide more power and replenish the
batteries of the sensors that are collecting data from critical
equipment with high priority. Hence our Differentiated RF
Power Transmission (DRIFT) scheme maximizes the power
received by high priority nodes. DRIFT is based on our ILP
formulation whose objective is given by eq. (2):

i
M aximize
i Pxy
lxy
(2)
i

where i is a binary variable that is 1 if the sensor node


is collecting data from a critical equipment, hence has high
i
priority, and 0 otherwise. Pxy
is the power received by sensori from a landmark positioned at (x,y) coordinates. lxy is a
binary variable that is 1 if there is a landmark at (x,y), and
0 otherwise. Landmark is a location in the eld where it
is convenient for the mobile power transmitter to park and
emit signals to transmit power to the sensors in its vicinity.
In the end, optimal landmarks are selected and the mobile
power transmitter visits those landmarks using the shortest
Hamiltonian cycle which gives the optimum path length [5].
The objective function of eq. (2) is solved subject to several
constraints which are formulated below. We assume that the
total number of landmarks is limited by Nl , i.e., the maximum
number of landmarks, according to eq. (3):

lxy Nl ,
i
(3)
x

We avoid redundant landmarks where eq. (4) ensures that at


least one sensor is receiving power from (x,y) when there is
a landmark at (x,y).

i
zxy
lxy ,
x, y
(4)
i

In a WSN, the frequency of packet forwarding determines


the energy consumed by the sensor node mostly, since packet

transmission is one of the most energy consuming tasks.


Therefore, the frequency of forwarding events to the sink and
relaying the packets of the neighbors determine the amount
of energy consumed as well as the required amount of energy
for topping up the battery of a sensor node in each charging
cycle. Here, we denote the energy replenishment demand of
sensor-i by i . The supply of the mobile power transmitter, i.e.
l , should be greater than or equal to the energy requirement
of the sensors. This constraint is formulated by eq. (5).

i
zxy
i l ,
x, y
(5)
i

Equation (6) ensures that a sensor can receive power from


(x,y) if there is a landmark at (x,y), while eq. (7) guarantees
that each sensor is receiving power from at least one landmark.
i
zxy
lxy ,
i, x, y

i
zxy
1,
i
x

(6)
(7)

In wireless power transmission, the range is limited due to


attenuation as in any RF signal. The range limit is denoted by
R where dixy is the distance between sensor-i and the landmark
at (x,y).
i
zxy
dixy R,
i, x, y
(8)
Finally, eq. (9) ensures that high priority nodes receive more
power than low priority nodes:
i i
j j
Pxy
zxy Pxy
zxy 0,

si > sj

(9)

In the following section, we evaluate the performance of


DRIFT and compare it to SuReSense. The three-step algorithm
of SuReSense is briey explained in the previous section.
The rst step that minimizes the number of landmarks is
formulated with an ILP whose objective function is given by:

minimize
lxy
(10)
x

For further details of SuReSense, the reader is referred to [6].


V. P ERFORMANCE E VALUATION
We use the CPLEX optimization solver to determine the
landmark locations by solving the ILP formulation dened
in the previous section. We consider that sensor nodes are
randomly deployed in a rectangular eld of 100m X 100m.
The number of sensor nodes vary between 40 and 60. We
assume that the demand intensity of the nodes is identical and
is set to 2kJ. The battery capacity of the power transmitter is
assumed to be 10kJ. The number of landmarks is limited to
20, i.e. Nl = 20. We set the wireless energy transfer range to
R = 10m. We dene the ratio of high priority sensor nodes
to total number of sensor nodes as :
= Nh /Ns

(11)

where Nh denotes the number of high priority nodes and


Ns denotes the total number of sensor nodes. We initially
study DRIFT for different values of , i.e., = {15, 25} and

1493

3RZHUUHFHSWLRQHIILFLHQF\SHUKLJKSULRULW\QRGH

3RZHUUHFHLYHGSHUQRGH  P: 










'5,)7 
'5,)7 
6X5H6HQVH








1XPEHURIQRGHV













Fig. 1. Power received per node in mWs using SuReSense, DRIFT with
= 15 and = 25.

'5,)7 
'5,)7 



Fig. 2.



1XPEHURIQRGHV





Power reception efciency per high priority nodes.

3RZHUUHFHSWLRQHIILFLHQF\SHUORZSULRULW\QRGH



compare its performance to SuReSense. Then, we vary from


10 to 25 and investigate the impact of .
In Fig. 1, we present the total power received per node for
DRIFT when = 15 and = 25, in addition to SuReSense.
We dene the power received per node, , as:
h
l
= Paverage
+ Paverage

(12)

h
denotes the average power received per high
where Paverage
l
denotes the average power received
priority node and Paverage
per low priority node. As it is seen from Fig. 1, DRIFT is
able to deliver more power than SuReSense for both values.
Power received per node is slightly lower for = 25 when
compared with = 15, since the number of high priority
nodes is higher for = 25 and power received per node drops
as expected. Furthermore, as the number of nodes increases,
decreases due to more number of nodes sharing the battery
resource of a single mobile charger.
In Fig. 2 we present the power reception efciency per high
priority node (P REh ). We dene P REh as:

iH Pi
P REh = 
(13)
|H| 
iH Pi + |L|
jL Pj

where Pi is the power received by high priority node-i, Pj is


the power received by low priority node-j. Here, H denotes
the set of high priority nodes and L denotes the set of low
priority nodes.
Fig. 2 shows that P REh is higher for = 25 than = 15.
This means as the number of high priority nodes increases
power reception efciency of high priority nodes increases.
Furthermore, power reception efciency per high priority node
increases as the number of nodes increases.
In Fig. 3 we present the power reception efciency per low
priority node (P REl ). We dene P REl as:

jL Pj
P REl = |L| 
(14)

iH Pi +
jL Pj
|H|

'5,)7 
'5,)7 









Fig. 3.



1XPEHURIQRGHV





Power reception efciency per low priority nodes.

Fig. 3 shows that P REl decreases for higher value. This is


expected since P REl is the complementary of P REh .
In Fig. 4, we present the power reception efciency for
varying values where varies between 10 to 25. As
seen from the gure, power reception efciency for high
priority nodes slightly increases as increases, while power
reception efciency of low priority nodes slightly decreases as
increases. For all values, power received by high priority
nodes is higher than the power received by low priority nodes
which shows that DRIFT is able to deliver differentiated power
to sensor nodes.
In Fig. 5 we present the path traversal efciency of the
mobile charger for SuReSense and DRIFT with respect to
varying number of nodes. The path traversal efciency is
dened as:
LLandmark
path =
(15)
LSensors
where LLandmark denotes the length of the path when only
landmarks are visited while LSensors denotes the length of
the path when all of the sensors are visited for energy

1494

Transmission (DRIFT) scheme which delivers more power to


high priority nodes using a mobile power transmitter. The
mobile power transmitter parks at landmarks and replenishes
the batteries of the sensors in its vicinity. We have modeled
the problem using an Integer Linear Programming (ILP)
model whose objective is to maximize the power received by
high priority nodes from a landmark. We have compared the
performance of DRIFT with a previously proposed scheme,
namely Sustainable wireless Rechargeable Sensor network
(SuReSense). We have shown that the power received by
high priority sensor nodes using DRIFT is signicantly higher
than the case with SuReSense while SuReSense is able to
reduce the path length of the mobile power transmitter. Hence,
we show that there is a tradeoff between power reception
efciency and the length of the path traversed by the mobile
charger, hence the duration spent while waiting for a power
transmitter.
In our future work, we plan to address this tradeoff by
jointly maximizing the received power and minimizing the
waiting time.

3RZHUUHFHSWLRQHIILFLHQF\










Fig. 4.

3RZHUUHFHLYHGE\KLJKSULRULW\QRGHV
3RZHUUHFHLYHGE\ORZSULRULW\QRGHV


3HUFHQWDJHRIKLJKSULRULW\QRGHV



Power reception efciency under varying values.


'5,)7 
6X5H6HQVH

3DWKWUDYHUVDOHIILFLHQF\ SDWK




R EFERENCES



[1] V.C. Gungor, B. Lu, G.P. Hancke, Opportunities and challenges of


wireless sensor networks in smart grid, IEEE Transactions on Industrial
Electronics, vol. 57, no. 10, Oct 2011, pp. 3557-3564.
[2] M. Erol-Kantarci, H. T. Mouftah, Wireless Multimedia Sensor and
Actor Networks for the Next-Generation Power Grid, Elsevier Ad Hoc
Networks Journal, vol.9 no.4, 2011, pp. 542-511.
[3] Powercast Corporation, [Online] http://www.powercastco.com/
[4] S. He, J. Chen, F. Jiang, D.K.Y. Yau, G. Xing, Y. Sun, Energy
Provisioning in Wireless Rechargeable Sensor Networks, in Proc. of
IEEE INFOCOM, Shanghai, China, April 10-15, 2011, pp. 2006-2014.
[5] L. Shi, L. Xie, Y.T. Hou, H.D. Sherali, On Renewable Sensor Networks
with Wireless Energy Transfer, in Proc. of IEEE INFOCOM, Shanghai,
China, April 10-15, 2011, pp. 1350-1358.
[6] M. Erol-Kantarci, H.T. Mouftah, SuReSense: Sustainable Wireless
Rechargeable Sensor Networks for the Smart Grid, IEEE Wireless
Communications, June 2012.
[7] A. Nasipuri, R. Cox, J. Conrad, L. Van der Zel, B. Rodriguez, R.
McKosky, Design considerations for a large-scale wireless sensor
network for substation monitoring, in Proc. of IEEE 35th Conference on
Local Computer Networks (LCN), Denver, CO, Oct. 2010, pp.866-873.
[8] M. Erol-Kantarci, H. T. Mouftah, Wireless Sensor Networks for CostEfcient Residential Energy Management in the Smart Grid, IEEE
Transactions on Smart Grid, vol.2, no.2, June 2011, pp.314-325.
[9] G. Anastasi, M. Conti, M. Di Francesco, A. Passarella, Energy conservation in wireless sensor networks: A survey, Ad Hoc Networks, vol.
7, no. 3, May 2009, pp. 537-568.
[10] S. Sudevalayam, P. Kulkarni, Energy Harvesting Sensor Nodes: Survey
and Implications, IEEE Communications Surveys & Tutorials, vol.13,
no.3, pp.443-461, Third Quarter 2011.
[11] S. Sendra, J. Lloret, M. Garcia, J. F. Toledo, Power saving and
energy optimization techniques for Wireless Sensor Networks, Journal
of Communications, vol. 6, no 6, pp. 439-459, August 2011.
[12] V. Pryyma, D. Turgut, L. Boloni, Active time scheduling for rechargeable sensor networks, Computer Nets, vol. 54, no. 4, 2010, pp.631-640.
[13] R. D. Mohammady, K. Chowdhury, M. Di Felice, Routing and Link
Layer Protocol Design for Sensor Networks with Wireless Energy
Transfer, IEEE GLOBECOM, Miami, December 2010.
[14] M. Zhao, J. Li, Y. Yang, Joint Mobile Energy Replenishment and Data
Gathering in Wireless Rechargeable Sensor Networks in Proc. of Int.
Teletrafc Congress, Sept. 6-8, 2011, San Francisco, USA.
[15] M. Erol-Kantarci, H. T. Mouftah, Mission-Aware Placement of RFbased Power Transmitters in Wireless Sensor Networks, IEEE Symposium on Computers and Communications (ISCC), June 30-July 5,
Cappadocia, Turkey.












1XPEHURIQRGHV





Fig. 5. Path traversal efciency of the mobile charger for SuReSense and
DRIFT.

replenishment. As seen from the gure, SuReSense incurs


lower path length than DRIFT. Thus, our results show that
there is a tradeoff between power reception efciency and path
traversal efciency.
VI. C ONCLUSION
Smart grid monitoring tasks demand long-term operation
capability where the monitored equipment may also have different levels of priority in terms energy replenishment needs.
RF energy harvesting technology provides the capability to
charge the batteries of the sensors and signicantly extends the
lifetime of the Wireless Sensor Networks (WSN), while novel
solutions are required for providing differentiated power delivery to sensor nodes. For instance, in the smart grid, sensors
collecting data from critical equipment such as transformers
and capacitor banks transmit data more frequently than the
sensors collecting ambient measurements, hence they need to
have higher priority in terms of delivered power.
In this paper, we have proposed Differentiated RF Power

1495