Research in Energy Harvesting Wireless Sensor

Networks and the Challenges Ahead

Winston K.G. Seah, Y.K. Tan, and Alvin T.S. Chan

Abstract Wireless sensor networks (WSNs) are set to form a signiﬁcant part of the

new pervasive Internet, often referred to as the Internet of Things. WSNs have

traditionally been powered by limited energy sources, viz. batteries, limiting their

operational lifetime. To ensure the sustainability of WSNs, researchers have turned

to alternative energy sources for power. Harvesting ambient energy from the

environment to power WSNs is a promising approach, but it is currently unable

to provide a sustained energy supply to support continuous operation. Sensor nodes

therefore need to exploit the sporadic availability of energy to quickly sense and

transmit the data. We ﬁrst review the recent developments in energy harvesting

technology and research on networking protocol design for WSNs powered by

ambient energy harvesting. Then, we discuss some of the challenges face d by

researchers in designing networking protocols and summarize the open research

problems.

Keywords Energy harvesting/scavenging, Protocol design, Wireless sensor

network

W.K.G. Seah (*)

School of Engineering and Computer Science, Victoria University of Wellington, P.O. Box 600,

Wellington 6140, New Zealand

e-mail: [email protected]

Y.K. Tan

Energy Research Institute @ Nanyang Technological University (ERI@N), Research Techno

Plaza, X-Frontier, Level 5, 50 Nanyang Drive, Singapore 637553, Singapore

e-mail: [email protected]

A.T.S. Chan

The Hong Kong Polytechnic University, Hung Hom, Hong Kong

e-mail: [email protected]

D. Filippini (ed.), Autonomous Sensor Networks: Collective Sensing Strategies for Analytical Purposes,

Springer Series on Chemical Sensors and Biosensors (2013) 13: 73–94

DOI 10.1007/5346_2012_27,

Springer-Verlag Berlin Heidelberg 2012,

Published online: 23 August 2012

Contents

1 Introduction ................................................................................... 74

2 Overview of WSN-HEAP Concept . . ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...................... 75

3 Energy Harvesting Technology . . . . . . . . . . . . . . . . . . . . . . . . ...................................... 77

3.1 Overview of Renewable Energy Harvesting .......................................... 78

3.2 Modelling of Energy Harvesting Sources . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

3.3 Design Principle of WSN with Energy Harvesting Technology ..................... 83

4 Networking Protocol Design ................................................................. 84

4.1 Power and Topology Management . . . . . . . . . . . . . . . . . . . . . . . ............................. 84

4.2 Data Delivery ........................................................................... 87

5 Challenges in WSN-HEAP ................................................................... 88

5.1 Correlation in Natural Phenomenon . . . . . . . . . . . . . . . .. . .. . . . . . . . . . . . . . . .. . .. . . . . . . . . . . . . 89

5.2 Modelling of Energy Harvesting Process ............................................. 89

5.3 Bio-Inspired and Learning Approaches ............................................... 90

5.4 Middleware . .. . .. . .. .................................................................... 90

6 Conclusion .................................................................................... 91

References . . . . . .. . . . . . . . . . . . .. . . . . . . . . .. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . .. . . . . . . . . . . . . . . . . . . 92

Abbreviations

CPU Central processing unit

CSMA Carrier sense multiple access

EH Energy harvesting

EHOR Energy harvesting opportunistic routing

ESC Energy synchronized communication

LZT Lead zirconate titanate

MAC Medium access control

MJ Mega joules

MTPP Multi-tier probabilistic polling

PRT Probabilistic retransmission protocol

PZ Piezoelectric

RL Reinforcement learning

WSN-HEAP Wireless sensor networks powered by ambient energy harvesting

WSNs Wireless sensor networks

1 Introduction

Wireless sensor networks (WSNs) are set to form a signiﬁcant portion of the smart

pervasive Internet of Things. Like the traditional Internet and many other

technologies, defence applications have motivated the research in WSNs [1]. The

often-cited deployment scenario for WSNs is that of a large number of small

wireless sensor nodes randomly scattered over the area of interest to collect

information on entities of interest. Research has predominantly assumed that sensor

nodes are powered by a portable and limited energy source, viz. batteries. Onc e a

74 W.K.G. Seah et al.

sensor node’s power supply is exhausted, it can no longer fulﬁl its role unless the

source of energy is replenished. Therefore, it is generally accepted that the useful-

ness of a wireless sensor expires when its battery runs out. Much of the research on

WSNs has therefore focused on efﬁcient methods to minimize energy usage in

order to extend the lifetime of the nodes that form the network.

Rapid technological progress has made available low-cost sensors and commu-

nication networks which led to the development of various other potential WSN

applications [1]. A salient feature of these new applications is the need for sensors

to operate for much longer durations (like years or even decades) after they are

deployed, for example, in in situ environmental/habitat monitoring and structural

health monitoring of critical infrastructures and buildings, where batteries are hard

(or even impossible) to replace after the sensors are deployed. Recently, alternative

energy sources for WSNs are actively investigated. Of particular interest is the

harvesting of ambient energy from the environment and converting it into electric-

ity to power the sensor nodes. While renewable energy technology is not new (e.g.,

solar and wind) the systems in use are far too large for WSNs. Those small enough

for use in wireless sensors are unlikely to provide a sustained supply to support

continuous operation. They are likely to be able to provide only enough energy to

power sensors sporadically and sensor nodes therefore need to exploit the sporadic

availability of energy to quickly sense and transmit the data.

In this chapter, we ﬁrst discuss the concept of a WSN in which sensor nodes rely

solely on harvested energy for power, referred to as WSNs powered by ambient

energy harvesting, or WSN-HEAP for short [2] and not hybrid systems that use

energy harvesting to supplement batteries [3, 4]. Since then, there have been active

research efforts by both the academia and the industry in this area; for earlier work

reported in the literature, the reader can refer to [2] and the references therein. We

then review the developments in energy harvesting systems and protocol design.

Before concluding, we discuss the open research problems and challenges ahead

that need to be addressed.

2 Overview of WSN-HEAP Concept

In WSN-HEAP, each sensor node uses one or more energy harvesting devices to

harvest ambient energy, such as light, vibration and heat, from the environment and

stores the harvested energy in a storage device. The differences in the system

architecture between a battery-powered wireless sensor node and WSN-HEAP

node are shown in Fig. 1. Harvesting energy from the environment is not new and

has been in use for decades. The most common forms of ambient energy include

water (hydro-electric power generation) , light (solar panels), wind (wind turbines)

and thermal (particularly in areas with volcanic activities). Harvesting energy for

low-power (and possibly embedded) devices like wireless sensors presents a new

challenge as the energy harvesting device has to be comparable in size (i.e. small

enough) with the sensors. Furthermore, the placement of the sensors may not be in

locations that will achieve optimal energy harvesting performance. There are

Research in Energy Harvesting Wireless Sensor Networks 75

complex trade-offs to be considered when designing energy harvesting systems for

WSNs arising from the interaction of various factors like the characteristics of the

energy sources, energy storage device(s) used, power management functionality of

the nodes and protocols and the applications’ requirements. Currently, the main

sources of ambient energy considered suitable for use with WSNs are solar,

mechanical (vibration or strain), thermal and electromagnetic energy [5].

Besides the energy harvesting component, another critical component of a

WSN-HEAP node is the energy storage device. Baring wear-and-tear and other

forms of physical damage, the goal is to substantially minimize, or ideally, totally

eliminate the need to physically replace the energy storage device or manually

replenish the energy. The primary candidates for energy storage in WSN-HEAP are

rechargeable battery and super-capacitor. The key advantage that the super-

capacitor has over the rechargeable battery is its virtually unlimited recharge

cycles—in the order of a million cycles as compared to less than 1,000 cycles for

rechargeable batteries. This makes super-capacitors the more viable energy storage

option for WSN-HEAP. Issues in storage devices and low-power electronics design

that are suitable for use in WSN-HEAP are discussed in [6].

The energy characteristics of a WSN-HEAP node are distinctly different from

that of a battery-powered wireless sensor node, as illustrated in Fig. 2. In a battery-

powered node, the total energy reduces over time and the sensor remains opera-

tional until the energy level drops to an unusable level. At this stage, either the

battery needs to be replaced or the node is deemed to be unusable/dead. On the other

hand, the energy in a WSN-HEAP node is replenished with energy harvested from

the environment. The energy needs to be accumulated over time until a certain level

(e.g. E

min

in Fig. 2) before it can be used. Due to technology limitations and the

unpredictable nature of the environment, the rate of harvesting and charging is

unlikely to be able to support continuous sustained node operation. A WSN-HEAP

node is norm ally awake and operating for a short duration before it needs to shut

down to recharge. Similar cyclic trends (see Fig. 3) have also been observed in

experiments carried out with solar energy harvesting and reported in [ 7 ]. At the start

of their experiments, power was drawn from Battery 1 until its voltage falls below

4 V at sample 900; although Battery 2 was not used, it suffered from self-discharge

and its voltage level dropped slightly over time. The system then switched Battery 1

Fig. 1 Battery-operated wireless sensor vs WSN-HEAP node

76 W.K.G. Seah et al.

to energy harvesting mode and started drawing power from Battery 2. Therefore, the

voltage level of Battery 2 began falling at a faster rate until it dropped below 4 V at

2,600. Meanwhile, solar energy harvesting had recharged Battery 1 to its capacity

but while it was not used, it suffered from self-discharge, like Battery 2 previously.

At sample 2,600, the two batteries were switched over again. The voltage level of

Battery 1 dropped faster as power was drawn from it, while Battery 2 was recharged.

Hence, the study has validated the cyclic behaviour of the voltage levels shown in

the conceptual model presented in Fig. 2. A summary of key aspects and differences

between battery-powered WSN and WS N-HEAP is provided in Table 1.

3 Energy Harvesting Technology

To overcome the major hindrance of the “deploy and forget” nature of WSNs due to

the limit ation of available energy for the networ k constrained by the high power

consumption of the sensor nodes and the energy capacity and unpredictable lifeti me

performance of the battery, EH technology has emerged as a promising solution to

sustain the operation of WSN [8, 9].

3.5

4.5

5.5

1000 2000 3000 4000 5000 6000

7000

Voltage Battery (V)

Voltage Battery 1

Voltage Battery 2

Fig. 3 Energy characteristics of two-battery solar powered sensor system [7]

Fig. 2 Energy characteristics of WSN-HEAP node

Research in Energy Harvesting Wireless Sensor Networks 77

3.1 Overview of Renewable Energy Harvesting

Energy harvesting (EH) is a technique that captures, harvests or scavenges a variety

of unused ambient energy sources such as solar, thermal, vibration and wind, and

converts the harvested energy into electrical energy to recharge the batteries. The

harvested energy in WSNs is generally very small (of the order of mJ) as compared to

those large-scale EH applications using renewable energy sources such as solar farms

and wind farms of the order of several hundreds MJ. Unlike the large-scale power

stations that are ﬁxed at a given location, the small-scale energy sources are portable

and readily available for usage. Various energy harvesting sources excluding the

biological type, which can be converted into electrical energy, are shown in Fig. 4.

In our environment, there are full of wasted and unused ambient energy

generated from these energy sources seen in Fig. 4. These renewable energy sources

are ample and readily available in the environment and so it is not necessary to

deliberately expend efforts to create these energy sources like the example of

burning the non-renewable fossil fuels to create steam that in turn drives the

steam turbines to create electrical energy. Unlike exhaustible fossil fuels, the

majority of the environmental energy sources are renewable and sustainable for

almost inﬁnitely long periods. Numerous studies and experiments have been

conducted to investigate the levels of energy that could be harvested from the

ambient environment. A compilation list of various energy harvesting sources and

their power/energy densities is listed in Table 2.

Table 1 Summary of key aspects and differences between WSN and WSN-HEAP

Battery-operated

WSNs

Battery-operated WSNs

supplemented with energy

harvesting WSN-HEAP

Goal Latency and

throughput is

usually traded off

for longer network

lifetime

Longer lifetime is achieved

since battery power is

supplemented by

harvested energy

Maximize throughput and

minimize delay since

energy is renewable and

the traditional concept of

network lifetime does not

apply

Protocol

design

Sleep-and-wakeup

schedules can be

determined

precisely

Sleep-and-wakeup schedules

can be determined if

predictions about future

energy availability are

correct

Sleep-and-wakeup schedules

cannot be predicted;

difﬁcult to know exactly

which is the awake next-

hop neighbour to forward

data to

Energy

model

Energy model is well

understood

Energy model can predicted

with high accuracy

Energy harvesting rate varies

across time, space as well

as the type of energy

harvesters used; energy

model is hard to predict

78 W.K.G. Seah et al.

Table 2 shows the performance of each EH source in terms of the power density

factor. It can be clearly observed that there is no unique solution suitable for all

environments and applications. According to Table 2, it can be observed that solar

energy source yields the highest power density. However, this may not be always

the case. Under illuminated indoor condition, the ambient light energy harvested by

the solar panel drops tremendously. The other energy harvesting sources could

provide higher power density depending on the renewable energy sources available

at the speciﬁ c application areas like outdoor bright sunny day with rich amount of

solar energy, along coastal area with a lot of wind energy, bridge structure with

vehicles travelling that has strong vibrations, etc. In addition, there could also be a

possibility of two or more energy sources available for harvesting at the same time.

As such, EH technology can provide numerous beneﬁt s to the end user and some

of the major beneﬁts about EH suitable for WSN are stated and elaborated in the

following list. Energy harvesting solutions can:

1. Reduce the dependency on battery power—with the advancement of microelec-

tronics technology, the power consumption of the sensor nodes are getting lesser

and lesser, hence harvested ambient/environmental energy may be sufﬁcient to

eliminate the need for batterie s comple tely.

2. Reduce installation cost—self-powered wireless sensor nodes do not require

power cables wiring and conduits, hence they are very easy to install and also

reduce the heavy installation cost.

Fig. 4 Energy harvesting sources and their energy harvesters, adapted from [10]

Research in Energy Harvesting Wireless Sensor Networks 79

3. Reduce maintenance cost—energy harvesting allows for the sensor nodes to

function unattended once deployed and eliminates service visits to replace

batteries.

4. Provide sensing and actuation capabilities—especially in hard-to-access haz-

ardous environments on a continuous basis.

5. Provide long-term solutions—a reliable self-powered sensor node will remain

functional virtually as long as the ambient energy is available. Self-powered

sensor nodes are perfectly suited for long-term applications looking at decades

of monitoring.

6. Reduce environmental impact—energy harvesting can eliminate the need for

millions on batteries and energy costs of battery replacements.

Clearly, it can be deduced from the list of beneﬁts that EH technology is a viable

solution to power WSNs and mobile devices for extended operation with the

supplement of the energy storage devices, if not completely eliminating the storage

devices such as batter ies.

Table 2 Energy harvesting opportunities and demonstrated capabilities adapted from [11]

Energy source Performance Notes

Ambient light 100 mW/cm

(direct sunlight) Common poly-crystalline solar cells are

16–17% efﬁcient, while standard

mono-crystalline cells approach 20%

100 mW/cm

(illuminated ofﬁce)

Thermal (a) 60 mW/cm

at 5k gradient Typical efﬁciency of thermoelectric

generators are 1% for DT < 313 K

(b) 135 mW/cm

at 10k gradient (a) Seiko Thermic wristwatch at 5 K body

heat

(b) Quoted for aThermolife

generator at

DT ¼ 10 K

Blood pressure 0.93 W at 100 mmHg When coupled with piezoelectric

generators, the power that can be

generated is order of mW when loaded

continuously and mW when loaded

intermittently

Vibration 4 mW/cm

(human motion-Hz)

800 mW/cm

(machines-kHz)

Predictions for 1 cm

generators. Highly

dependent on excitation (power tends

to be proportional to o, the driving

frequency and y

, the input

displacement

Hand linear

generator

2 mW/cm

Shake-driven ﬂashlight of 3 Hz

Push button 50 mJ/N Quoted at 3 V DC for the MIT Media Lab

Device

Heel strike 118 J/cm

Per walking step on piezoelectric insole

Ambient wind 1 mW/cm

Typical average wind speed of 3 m/s in the

ambient

Ambient radio

frequency

<1 mW/cm

Unless near an RF transmitter

Wireless energy

transfer

14 mW/cm

Separation distance of 2 m

80 W.K.G. Seah et al.

The latest trend in energy harvesting involves bioenergy, which is renewable

energy derived from biological sources. The energy in oxygen and glucose

molecules in blood has become the target of bioengineers looking for an energy

source to power implantable devices without having to use batteries [12]. Other

sources of bioe nergy include human and animal wastes, tree [13], etc.

3.2 Modelling of Energy Harvesting Sources

In order to design networking protocols for realistic WSN-HEAP applications, we

need to characterize the charging time of energy harvesters and radio transmission

behaviour of the EH system unit as shown in Fig. 5. Most manufacturer datasheets

of the energy harvesting devices only describe the average harvesting rates and not

the charging characteristics. As the source of harvested energy comes from the

environment, it is difﬁcult to predict the time when the energy is available.

Furthermore, the instantaneous energy level obtainable from the energy harvester

is variable and also inadequate to operate a sensor node. To mitigate the unpredict-

ability of the energy source, the harvested energy is stored in a buffer (e.g. super-

capacitor) until there is sufﬁcient to power the sensor node for the desired

operations like sensing, receiving and/or transmitting packets. Important energy

charging characteristics include charging time and number of packets that can be

transmitted per charge cycle (cf. Fig. 2).

A study to empirically characterize the charging times of commercially available

solar energy harvesting devices for WSN-HEAP is reported in [14]. To measure the

charging time, energy is accumulated by the harvester and stored in the storage

device. When the energy level has reached a sufﬁcient level (E

min

in Fig. 2), the

microcontroller and transceiver on the sensor node are switched on. The transmitter

then continuously broadcasts data packets until the energy is depl eted, after which

the microcontroller and transceiver are turned off. The energy storage device will

start to accumulate energy again and the process is repeated in the next cycle.

Figure 6 shows the probability density functions (pdf) of the charging times for

1,000 charge cycles for three deployment scenarios with the solar energy harvester

Environmental

Energy

Sources: Solar,

Vibration, Thermal

Energy

Harvester

1234

Energy

Storage

DC-DC

Converter

AC-DC

Converter

(optional)

Electrical

Load

Power

Management/

Conditioning

Controller

Fig. 5 General block diagram representation of energy harvesting system unit

Research in Energy Harvesting Wireless Sensor Networks 81

placed directly under, 1 and 2 m under a ﬂuorescent lamp. The results show that

there is greater variation (higher standard deviation) in the charging time required

for each charge cycle when the sensor node is further away from the light source.

Similar variations have also been observed with a thermal energy harvester

mounted on a central processing unit (CPU) heat sink.

The high variability in the energy harvesting process, in both time and space, has

also been observed by another study [15] using a time-slotted solar energy

harvesting node with different system and environmental parameters. Six different

statistical models, viz. uniform distribution, geometric distribution, transformed

geometric distribution, Poisson distribution, transformed Poisson distribution and a

Markovian model, were used to ﬁt the empirical datasets. The study concluded that

no single statistical model can ﬁt all the datasets, and empirical data is needed to

validate the theoretical analysis of time-slotted medium access control (MAC)

protocols for WSN-HEAP.

Analytical models supported with empirical data are derived in [16]. The study

involved harvesting solar as well as piezoelectric (PZ) energy. A solar harvesting

module was deployed next to a glass window in a typical ofﬁce building where it

harvested energy from sunlight and indoor ﬂuorescent light. Readings were

Directly under lamp. 1m under lamp.

0.045

0.04

0.035

0.03

0.025

Density

0.02

0.015

0.01

0.005

x 10

-3

1210 1220 1230 1240 1250

Charging Time (ms) Charging Time (ms)

Charging Time (ms)

1260 1270 1280 1290 4800 5000 5200 5400 5600 5800 6000 6200 6400 6600

2m under lamp

1.2

x 10

-3

0.8

0.6

0.4

0.2

Density

7500 8000 8500 9000 9500 10000 10500 11000 11500 12000

1.8

1.6

1.4

1.2

0.8

0.6

0.4

0.2

Fig. 6 Solar energy harvesting technology—charging times probability density functions [15]

82 W.K.G. Seah et al.