![Table I: Comparison of the existing energy harvesting techniques [17].](/figures/table-i-comparison-of-the-existing-energy-harvesting-1bn9hccs.webp)
1
Internet of Hybrid Energy Harvesting Things
Ozgur B. Akan, Fellow, IEEE, Oktay Cetinkaya, Student Member, IEEE, Caglar Koca, Student
Member, IEEE, Mustafa Ozger, Student Member, IEEE
Abstract—Internet of Things (IoT) is a perfect candidate to
realize efficient observation and management for Smart City
concept. This requires deployment of large number of wireless
devices. However, replenishing batteries of thousands, maybe
millions of devices may be hard or even impossible. In order
to solve this problem, Internet of Energy Harvesting Things
(IoEHT) is proposed. Although the first studies on IoEHT
focused on energy harvesting as an auxiliary power provision
method, now completely battery-free, self-sufficient systems are
envisioned. Taking advantage of diverse sources that the appli-
cation areas in Smart Cities offer helps us to fully appreciate
the capacity of energy harvesting. In this way, we address the
primary shortcomings of IoEHT; availability, unreliability and
insufficiency by the Internet of Hybrid Energy Harvesting Things
(IoHEHT). In this work, we survey the various energy harvesting
opportunities, propose an hybrid energy harvesting system and
discuss energy and data management issues for battery-free
operation. We also point out to hardware requirements and
present the open research directions for different network layers
specific to Internet of hybrid energy harvesting things for Smart
City concept.
Index Terms—Hybrid Energy Harvesting, Wireless Networks,
Internet of Things, Smart Cities.
I. INTRODUCTION
Enhanced management of cities brings a new paradigm,
named as Smart Cities [1], [2], which achieves environment
sensing and better utilization of city resources. Particular
application areas of Smart Cities are intelligent transport sys-
tems, smart grid, smart home, smart agriculture and structural
health [3]. The realization of them requires utilization of
cutting edge technologies such as the Internet of Things (IoT).
Sensing and controlling features of the IoT are keys to this
realization. Using IoT, the physical world can be observed,
and information related to the surroundings is gathered, such
that the physical world is digitized. Using IoT technology, we
can access to this digitized world via the Internet connection,
and move one step closer to the Smart City concept [4], [5].
In order to achieve continuous monitoring and control, an
auxiliary or even a completely distinct power source should
be equipped to the sensors. However, even this option may
or may not be applicable in some cases mostly due to size
constraints or design restrictions. Hence, energy harvesting
methods come into prominence to alleviate the problems of
energy-constrained wireless networks by exploiting a stray
source or converting energy from one form to another [6],
[7].
Ozgur B. Akan is with Internet of Everything (IoE) Group, Electrical
Engineering Division, Department of Engineering, University of Cambridge,
CB3 0FA Cambridge, UK (e-mail: oba21@cam.ac.uk).
O. Cetinkaya, C. Koca and M. Ozger are with the Next-generation and
Wireless Communications Laboratory (NWCL), Department of Electrical and
Electronics Engineering, Koc University, Istanbul, 34450, Turkey (e-mail:
{okcetinkaya13, cagkoca, mozger}@ku.edu.tr).
There are numerous potential alternatives to collect energy,
but their availability depends on the environmental variables,
ambient parameters, or other time-varying and highly random
external factors. The ongoing limits on the power extraction
capabilities force wireless devices for an energy trade-off
between proper system operation and the desired network
lifetime, whereby an upper bound is placed on the communi-
cation reliability. Due to this reason, hybrid energy scavenging
approaches possess a great potential to extend the lifetime of
wireless devices by operating in a complementary manner. A
power supply fed by multiple available sources will eventually
enhance the overall functionality, reliability, and efficiency of
both the system and communication [8]–[13].
The hybrid energy harvesting wireless smart nodes sense the
parameters of interest, process the collected data, and report
the resulting information to a base station/coordinator/gateway
over an Internet connection where the conditions of application
area are monitored, stored, and relevant authorities are alerted.
Energy modeling is crucial in any harvesting mechanism,
as optimal transmission policy directly depends on the energy
model. Hybrid energy harvesting enhances energy availability,
and therefore, improves the energy model of the system.
Moreover, in order to survive in the most dire circumstances of
Smart Cities, data management protocols, specific to IoHEHT
are needed. Furthermore, hybrid energy harvesting proposal
for IoT-enabled Smart Cities requires novel approaches in each
network layer to overcome the challenges posed by IoT and
Smart Cities to enable seamless operation. Hence, we lay the
foundations of battery-free IoHEHT networks.
In this paper, we first present existing energy harvesting
(EH) techniques, and then propose a new EH framework. It is
called hybrid energy harvesting, and copes with randomness
of harvestable resources by utilizing different EH methods
together. Furthermore, an applicable design for a hybrid EH
sensor system is presented. We also model energy and data,
and study mathematically the decrease of harvestable energy
variance by the hybrid approach. We test our new EH frame-
work with a simulation of a communication scenario, showing
that hybrid energy harvester can achieve lower drop rates for
the same reporting frequency. We propose a model for energy
and data queue management according to the proposed EH
method. Open issues and problems are discussed for each layer
in IoT networks utilizing hybrid EH.
The remainder of this paper is organized as follows. First,
we commence with a literature review of the existing energy
harvesting techniques. Then, we extend our study in Section
III to basic principles of hybrid energy harvesting systems
including their basis, main constraints and applicable proce-
dures in the IoT domain. This is followed by the performance
analysis of IoHEHT by investigating energy and data models
2
Table I: Comparison of the existing energy harvesting techniques [17].
Size
System
Complexity
Energy
Availability
Characteristics Harvester
Energy
Density
Advantages Disadvantages
Solar Good Medium Fair
Uncontrollable,
Predictable
PV Panel
15 − 100
mW/cm
2
Environmental,
Constant and consistent,
High output voltage
Not always available,
Sensitive structure,
Deployment constraints
Artificial
Light
Good Medium Fair
Partly-controllable,
Predictable
PV Cell
10 − 100
µW/cm
2
Abundant in indoor,
Easy to implement
Low power density,
Sensitive structure
Airflow Poor High Good
Uncontrollable,
Unpredictable
Piezo-turbines
Anemometers
100
mW/cm
2
Environmental,
Independent of grid,
Available day and night
Fluctuating density,
Hard to implement,
Requires construction
Motion Fair High Fair
Controllable,
Partly-predictable
Piezoelectrics
200
µW/cm
2
No ext. power source,
Compact configuration,
Light weight
Charge leakage,
Depolarization,
Highly variable output
Thermal Good Medium Poor
Uncontrollable,
Unpredictable
Thermocouple
' 50
µW/cm
2
Low-maintenance
Independent of grid,
Scalable
Not always available,
Requires efficient
heat sinking
RF Fair Medium Good
Partly-controllable,
Partly-predictable
Rectennas
1 − 10
µW/cm
2
Abundant in urban lands,
Allows mobility
Scarce in rural areas,
Low power density,
Distance dependent
M-Field Very good Low Good
Controllable,
Predictable
Current
transformers
150
µW/cm
3
No ext. power source,
Easy to implement,
Non-complex structure
Requires high and
perpetual current flow,
Safety vulnerabilities
E-Field Fair Very low Very good
Controllable,
Predictable
Metallic
plates
17
µW/cm
3
No need of current flow,
Easy to implement,
Always available
Being capacitive,
Mechanical constraints
and applicable transmission policies in Sections IV and V,
respectively. We address the applicable transmission policies
as well as open research directions for different network layers
specific to the IoHEHT procedures in Section VI. Finally, we
conclude our discussion in Section VII.
II. EXISTING ENERGY HARVESTING TECHNIQUES
When a small-scale industrial, medical, and/or educational
facility is envisioned, the continuity of communication is of
paramount importance. Any interruption or failure may not be
tolerated due to the vitality of the task that is being fulfilled.
This fact one again reveals the need for a complementary pro-
cedure, i.e., a hybrid energy harvesting architecture. Existing
energy sources can be broadly divided into four groups as
light, heat, motion, and electromagnetic (EM) radiation, in
which availability, controllability, and predictability of these
sources determine the models and specifications of the har-
vesting procedures that are going to be employed [6], [7].
By regarding this separation, the frequency of preference,
and the motivation of our proposal some leading energy
harvesting methods are discussed below, and a detailed com-
parison is illustrated in Table I.
A. Light Energy Harvesting
Energy harvesting form light sources is a well-established
method of power provision that gathers energy from ambient
lights, either from sun or artificial light sources, with respect
to a phenomena called as photo-voltaic (PV) effect [6], [7].
In outdoor, for the monitoring of overhead power lines, solar
cell inlaid photo-voltaic panels are used to convert solar
energy into electricity [14], [15]. For indoor applications,
specialized photo-voltaic materials, which are better suited
for diffused lights, are employed for taking advantage of
the light emitted from ambient elements. Even though the
PV modules are getting cheap, easy to use and efficient,
due to the dramatic fluctuations on the output power, large
surface area requirements, inoperability at night and ongo-
ing installation and maintenance costs, their use in mission
critical applications is limited [16]. However, for intermittent
reporting allowed ambient sensing and management services
of Smart Home/Building architectures IoT-capable light EH
sensor nodes are intensively preferred.
B. Kinetic Energy Harvesting
Kinetic energy harvesting (KEH) is the conversion of am-
bient mechanical energy into electric power. Wind turbines,
anemometers and piezoelectric materials are being developed
to attain energy from highly random and unpredictable motion
variations driven by external factors [6], [7].
KEH is frequently preferred in indoor and outdoor domain,
as a variety of sources can be conveniently exploited to
drive low power consumptive wireless autonomous devices.
In outdoor, airflow operated IoT-capable sensor nodes are
satisfactorily utilized for remote monitoring of the spaced apart
grid assets. Similarly, for less power requiring wireless de-
vices, any source of motion variation offers sufficient solutions
for low duty-cycled communications. However, designing a
generalized harvesting system especially for vibrating sources
is an ongoing challenge. Since the conversion efficiency highly
varies with the resonant frequency of the vibration, a special-
ized design for each source may be necessary [14]–[17].
C. Thermal Energy Harvesting
Thermal energy harvesting, i.e., thermoelectric generation
(TEG), is simply based on converting temperature gradients
into utilizable electric power with respect to the Seeback Effect
3
occurred in semiconductor junctions [7]. TEG is an innate
power provision technique for Smart Grid communications,
in which temperature swings between the power line and
the environment is used to extract energy. In small scale,
peltier/thermoelectric coolers and thermocouples are widely
used for building delay-tolerant wireless indoor networks [17].
Although harnessing power from temperature gradients sounds
promising, there is a fundamental limit, namely Carnot limit,
to the maximum efficiency at which energy can be harvested
from a temperature difference [7], [14], [16].
D. Electromagnetic Energy Harvesting
EM energy harvesting includes collecting RF signals emit-
ted from base stations, network routers, smartphones, and
any other sources by using large aperture power receiving
antennae, and converting the attained waves into utilizable
DC power [6], [7]. Their performance depends strongly on
the RF to DC conversion efficiency and the amount of power
received by the antennae. Although this method is a reliable
solution unaffected by the environmental variables, providing
relatively low power densities, necessitating close deployment
to the network transmitters, and requiring additive components
such as filters and voltage multipliers can be counted as its
main shortcomings [14]–[17]. Moreover, in case the nodes are
sparsely deployed, available energy to be harvested may be
too low, which might limit the use of EM energy harvesting.
Due to the abundance of EM propagation in urban areas, RF
energy harvesting is mostly preferred to operate IoT-assisted
Smart City services.
E. Magnetic-field Energy Harvesting
M-field energy harvesting is based on coupling the field flow
around the AC current carrying conductors that is clamped by
current transformers (CT) [15], [18]. This technique is able
to provide an adequate rate of continuous power so long as
current flow in the line is sufficient. As the amount of current
on power distribution level is considered, M-field EH stands as
the best candidate for the energization of high power requiring
IoT networks. However, gathering energy from a high current
carrying asset in close proximity to the harvester in a safe way
is still a challenging issue. To mitigate the safety concerns,
M-field-based methods need to be equipped with advanced
protection and control mechanisms. This issue compels their
utilization in terms of circuit complexity and implementation
flexibility [17].
F. Electric-field Energy Harvesting
According to the basics of electrostatics, any conductive
material energized at some voltage level emits electric field.
In AC, time varying field results in a displacement current,
whereby the E-Field induced electric charges are dispatched
and collected in storing element. As the accumulated energy is
gathered from the surrounding field, this method is named E-
field energy harvesting (EFEH) [15], [17], [19], [20]. E-field
is the only source that is neither intermittent nor dependent
on the load [21]. As the voltage and the frequency are firmly
Energy
Source
Energy
Source
Energy
Source
H
i i+1 i+2
Energy
Buffer -C
Storage
Cap. -C
B
F
Energy
Buffer -C
B
Energy
Buffer -C
B
Storage
Cap. -C
F
Storage
Cap. -C
F
Energy Storage
Sensor Node
A
R
V
E
S
T
E
R
(s)
M
P
T
P
C
I
F
E
R
(s)
R
E
T
I
G
L
A
O
R
(s)
R
E
U
T
N
R
G
O
M
B
E
Y
E
C
I
N
R
E
PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT
PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT
PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT
PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT
Figure 1: An applicable design for a hybrid energy harvesting sensor
system. Note that, gray blocks represent sub-systems of modular
design.
regulated and exactingly maintained, the E-field is therefore
stable and predictable in its behavior. Thus, it can be referred
as the most promising way to compose long-term and self-
sustainable IoT networks notwithstanding the ambient factors.
III. HYBRID ENERGY HARVESTING
All the energy harvesting methods discussed above are
used in such applications like wireless networking and remote
monitoring. However, availability of natural sources affects the
power density, and durability of their operation dramatically.
To exemplify, solar energy is extremely sensitive to the envi-
ronment, i.e., it is only exploitable during daytime. The very
same problem is also seen in non-environmental sources, in
which the harvesting performance is highly threatened by the
randomness of the ambient variables, although the sources are
partly-controllable in general. As all available techniques of
EH depend strongly on environmental conditions, grid-based
variables or any other uncontrollable parameters, hybrid solu-
tions become even more important for sustaining information
and/or time critical communications [8]–[11].
In order to obtain the best performance achievable, a two-
staged performance maximization process is recommended for
hybrid energy harvesting systems [12]. Fig. 1 depicts such
a possible architecture. In the first stage, the harvesters are
required to maintain their operation as collecting maximum
energy possible from the available sources. For that pur-
pose, such approaches like maximum power point tracking
(MPPT) are developed for compensating inconsistencies and
accordingly maximizing the scavenging efficiencies. As each
harvesting method has an optimal operation point that varies
with the amount of harvastable energy, MPPT procedures
should be capable of real-time tracking, and highly responsive
to any change in sources’ conditions. In addition to this power
extraction related approach, further effort should be focused on
how to convert, and transfer the gathered energy as efficiently
as possible, since the scavenged energy is still quite low and
highly time-varying.
The second stage includes efficient combination and man-
agement of the exploited sources. As energy is gathered
simultaneously from distinct harvesters, an energy combiner
is required to accumulate the individual contributions of each
4
Harvesting
(copper) Plate
3~ 1600kVA
Dry-type TR
Fluorescent
Reflector
Light Bulb
1
D
D
2
C
B
I
D
High current carrying
seconder side conductor
Current
Transformer
Primer Side
Conductors
C
B
C
B
Photovoltaic
Cell (PV Cell)
R
SH
D
1
I
PV
R
S
I
D
Piezoelectric
Cooling Vent
Piezo-turbine
E-field Energy Harvesting
M-field Energy Harvesting
Artificial Light Energy Harvesting
Airflow Energy Harvesting
Energy Harvesting from Vibrations
I
PV
C
F
C
F
C
F
C
B
C
F
Rectifier
Regulator
C
F
Regulator
Regulator
Regulator
Energy
Combiner
Energy
Storage
Sensor
Node
C
F
Array
PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT
PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT
PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT
PRODUCED BY AN AUTODESK EDUCATIONAL PRODUCT
Figure 2: Representative drawing of the proposed hybrid energy harvesting architecture for IoT.
system in a storage whereby the overall energy is delivered
to a wireless device, i.e., sensor node, autonomously. The
combiner needs a modular design that supports a variety of
EHs and their corresponding circuitries to be attached as sub-
systems. Note that, the eventual standardization of IoT ease the
modular design of such systems. In this way, the connection
of complementary sources is ensured in a very straightforward
manner at the expense of few components. For such an
architecture, an adaptive connection mechanism is needed to
isolate the harvesters from each other, such that undesired
interferences are prevented, i.e., charging each other instead
of the storing element. [12], [13]. In addition, combining
sources in close proximity with each other using this circuitry
autonomously allows charge conveyance when the collected
energy is high enough for transmission, and switch off the
sensory circuit when the voltage of the storage drops beyond
a certain threshold [8], [22]. This operation not only prevents
redundant and undesired discharge of the storage to 0 V, but
also allows more frequent data transmission by shortening
the charging time [17]. As the energy collected by different
sources can be combined in a universal depository, i.e., energy
storage in Fig. 1, it can also be kept separately in sub-level
5
buffers to supply different loads or sensors. This operation,
i.e., supporting various energy harvesting techniques as well as
energy storing systems points out to a newly-emerging topic,
namely multi-input multi-output (MIMO) energy harvesting
(MIMO-EH). However, MIMO-EH is still at its infancy.
Fig. 2 illustrates the physical model depiction of a repre-
sentative IoT scenario for a transformer, a pillar of the Smart
Grid infrastructure, powered by hybrid energy harvesting.
The hybrid energy harvesting node equipped with specialized
sensors such as; light, temperature, humidity, and presence, is
envisioned to observe the parameters of both the room and
transformer, process the extracted data, and notify upper level
authorities over the Internet for decision-making procedures.
With Internet connectivity, preclusive actions can be simulta-
neously fulfilled against any intruder and/or unexpected varia-
tions in medium parameters. The lifetime of the IoT network
can be further prolonged by harvesting multiple-sources in
the vicinity of the environment, which guarantees interruption-
free operation of the transformer. In this figure, there are five
distinct sources of interest for energy provision. The nature
of these sources do differ immensely which inevitably affects
the characteristics of the energy gathered. In other words,
certain harvesting methods require rectification, regulation
and/or conversion processes due to their high voltage low
current AC output, while some others need only one or two of
these procedures. However, in general, the circuits employed
after power acquisition stage can be referred as roughly similar
to each other. From Fig. 1 and Fig. 2, the diodes, i.e., rectifiers,
are for both rectifying the alternating current, and preventing
the harnessed energy from back feeding. The converted energy
is first stored in an energy buffer C
B
before regulation. As
the name suggests, regulators ensure delivering suitable and
stable voltage supply to the other parts of the circuit. They can
also be supported by additive smoothing and charge control
circuits for enhanced performance. The regulated energy is
then accumulated in a storage capacitor C
F
, to be combined
with the output of other distinct sources/harvesters. The energy
combined is stored in a quick-charged, long-lasting, and high
power-condensed super capacitor to boost longevity. DC-to-
DC converters, which are not shown in figures, can also be
employed to adjust the voltage output of the energy storage
to ensure proper operation of the attached load, i.e., sensor
node. Overall performance of such a system depends on
the efficiency of the equipped components and employed
procedures, as well as duty cycle of the sensor node; and the
protocol stack.
Harvesting energy from several sources simultaneously acts
as an insurance in case of energy scarcity. In other words,
each harvester mechanism is partly responsible for energy
acquisition, and they complement each other when any of
them fail to provide enough power in the absence and/or in-
sufficiency of the exploited source. As this operation increases
the overall system reliability, it becomes possible to run the
wireless devices as if they have a constant energy source like
batteries. By using hybrid energy harvesting-enabled Internet-
capable sensors, sensory data can be remotely observed by
a network coordinator, and necessary actions can be directed
over Internet. This better supported operation will eventually
help to achieve more reliable, responsive, and inter-operable
IoT networks for advanced Smart City services. Following
sections are investigating this proposition and questioning the
availability of transmit power maximization with respect to
hybrid energy profile of the universal harvesting system.
IV. ENERGY AND DATA MODELING
A crucial aspect of energy harvesting is profiling the energy.
Any EH system should be designed specific to the energy
profile of the resource to be exploited. The most common
assumption in energy profiling is offline profiling, where it
is assumed that the energy availability and data transmission
requirements are known beforehand. In this case, network
design should be optimized to the expected harvestable energy,
i.e., any design powered by solar energy harvesting should
keep in mind that there will be no harvestable energy at
night. In case such an information does not exist, harvesters
should adjust to the energy and data arrivals, i.e., online profile.
Such designs need to handle more uncertainties in the energy
arrivals.
Whether an offline energy profile exists or not, harvester
design must consider two principles: Energy causality and
Data causality. Energy causality implies energy cannot be
used before it is harvested. Similarly, data causality implies
any data that has not arrived cannot be transmitted.
In order to model the harvested energy and energy required
for transmission, we use energy line and data line, respectively.
Energy line, e(t) is the total amount of energy harvested until
time t, while data line, d(t) is the total amount of energy
required to process all arrived data packages until time t, i.e.,
e(t) =
Z
t
0
P
harvested
(t
0
)dt
0
, (1)
d(t) =
Z
t
0
P
used
(t
0
)dt
0
(2)
where P
harvested
(t) and P
used
(t) are the harvested and ex-
hausted power between t and t +∆t, respectively. Note that if
the system is supplied with a battery, e(t) should be a constant
line. Also note that, any packet arriving to the queue causes
an increase in the data line and any packet dropped from the
queue causes a drop in the data line. Total available energy in
the system is
E
av ailable
(t) = min(C, e(t) − d(t)) (3)
where C is the total energy storage capacity. Since E
av ailable
cannot be negative, if minimum amount of energy required to
process the arrived data, exceeds total harvested energy, i.e.,
some data must be dropped. The area lying between energy
line and data line is called feasible energy tunnel. Energy line,
data line, feasible energy tunnel and storage element size for
a generic harvesting scenario is shown in Fig. 3(a).
Examining Fig. 3(a), we realize that any energy alloca-
tion policy must lie between the total harvested energy and
minimum amount of energy required to process all data.
Furthermore, an optimal policy should minimize the storage
overflow while maximizing the transmission rate. Such an
optimal energy allocation policy is proven to be the shortest