Beruflich Dokumente
Kultur Dokumente
I. I NTRODUCTION
A lot of effort has been invested when it comes to
managing and distributing resources among large distributed
systems. Together with the distribution of such resources,
improving the energy efficiency of a system can be done at
the demand-management level [1]. A common resource to
be shared is electrical power, powering all kind of devices
around us, such as households appliances (fridge, computer,
toaster etc.). An Internet of Things (IoT) [2] is a term to describe a network of thingsobjects not traditionally thought
of as computers (e.g. cars, household appliances), which
may nonetheless be connected using Internet protocols and
technologies (TCP/IP, etc.). Assuming that such a network of
objects is connected to a shared, variable power supply, it is
fruitful to consider protocols by which they may function
effectively in a decentralized manner, efficiently sharing
electrical power. Here is presented a novel approach where
a network of connected, autonomous IoT objects efficiently
share electrical power from a common, limited power supply.
Every object within the IoT has an individual power requirement. The power rating of each object is assumed to be
constant (i.e., not changing over time), such that objects can
either be powered (i.e., consuming their full power demand)
Zorzi et al. [12] point out that there is presently a firsttime opportunity to use the IoT approach to interact with
surrounding environments and to exchange information that
previously was not available. They address a series of issues
when comes to the IoT paradigm, including connectivity,
scalability, self-management, capabilityall within the context of energy management. Monnier [13] discusses how
objects around us need to be more and more connected and
suggest that connectivity is the key to automation. Perera
et al. [6] survey context-aware computing by an IoT.
Niyato et al. [14] present a system that uses machineto-machine (M2M) communication to reduce the costs of
a home energy management system. They describe a smart
grid system as having three major parts: power generation,
power distribution and power consumption. Their M2M
communication takes place between home appliances and
smart meters, and the system is dependent on a central
node or control center [14]. Our approach overcomes the
limitation inherent in using a central controller of any sort.
Objects in our model use a sort of M2M communication, but
in a different context; objects communicate among themselves and decide their own actions (e.g. to get powered
if suitable conditions exist). Karnouskos [15] supports such
an approach, and points out that communication between
objects and other entities, with alternative energy resources
which are smaller and decentralized, is one way to achieve
common goals such as energy efciency. Walczak et al. [16]
also describe M2M communication as part of the IoT and
the Future Internet Engineering.
The combination of intelligent consumers and providers of
sustainable energy has been recently combined conceptually
as the Internet of Energy [17], [18]. Bui et al. [17] present
a set of advantages of combining the smart grid and IoT,
such as decentralization of the control procedures. They
mention that from a communication perspective, the smart
grid distribution network must also be scalable, given the
high number of devices that are presumed to take part in the
future Internet of Energy [19].
Siano [20] presents ways in which systems can lower
peak demand, and Gelazanskas and Gamage [21] propose a
novel electricity demand control technique using real-time
pricing. Such methods often require coercive stopping of
consumption at a system-wide level. Our work overcomes
the limitations of stopping objects of the IoT/smart grid
consumers from consuming power at peak times, instead
making the objects responsible for their own usage times
(objects decide themselves when it is a suitable time to get
powered) but subject to global availability limits.
The works discussed above consider how to reduce energy
consumption in networked systems, but there is also a
substantial literature (though not directly pertinent to our
problem domain) on energy efficiency in networking itself
as may be seen in the survey by Bolla et al. [22].
III. P ROBLEM S TATEMENT
In this section, an abstract model of the IoT within a
demand-management problem is discussed. The definitions
n1
P
f (i )
i=0
Algorithm 1: Flood Algorithm with object j flooding message over all other objects.
1
broadcast[j , ]
for i 0 to n 1 do
if i 6= j then
Send msg[j , i , ]
end
end
Upon Receiving msg[]
store(i , )
ack [i , j , ]
3
4
5
6
for j 0 to n 1 do
3
4
5
6
7
8
9
for j 0 to n 1 do
Upon Receiving msg[]
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
11
12
13
14
15
if (i ) = (j ) then
Remove i from Qj []
end
else
if sizeof (Pj []) = 1 then
Remove (i ) from Pj []
end
end
if powered = TRUE then
w w + f (i )
end
end
if (i ) = (j ) then
Insert i in Qj []
ack [j , i , ((i ) = (j ))]
end
else
if (i )
/ Pj [] then
Insert (i ) in Pj []
ack [j , i , (i ) 6= (j ))]
end
end
end
for j 0 to n 1 do
Sort Qj [] in increasing order of f ()
Sort Pj [] in decreasing order
end
10
In the case that the leaving object i has the same priority
18
19
20
21
22
23
24
25
26
27
28
29
30
Object Name
Electric Shower
Dishwasher
Washing Machine
Tumble Dryer
Iron
Toaster
Oven
Grill
Fridge-Freezer
Vacuum Cleaner
Hairdryer
Plasma TV
Desktop Computer
Broadband Router
Smart Phone
Abbreviation
ES
Dw
WM
TD
Ir
Ts
Ov
Gr
FF
VC
Hd
TV
PC
BR
SP
PowerDemand
7kW
1.5kW
1.5kW
3kW
1.6kW
1kW
2kW
1.8kW
0.4kW
1kw
1kW
0.4kW
0.15kW
0.01kW
0.005kW
Priority
6
1
2
2
2
8
8
8
10
3
9
5
9
10
10
TABLE I
( SOURCE : [24])
Object
SP
BR
FF
PC
Hd
Ts
Gr
Ov
ES
TV
VC
WM
Ir
TD
Dw
Case 1 (10kW)
on
on
on
on
on
on
on
on
off
on
on
on
off
off
off
Case 2 (10kW)
on
on
on
on
on
N/A
on
N/A
off
on
on
on
on
off
on
Case 3 (11kW)
on
on
on
on
on
N/A
on
N/A
on
on
off
off
off
off
off
TABLE II
L EGEND : ON = OBJECT IS POWERED ; OFF = OBJECT IS NOT POWERED ;
N/A = OBJECT DOES NOT REQUIRE POWER CURRENTLY.
Fig. 1.
IV. R ESULTS
In this section, the results of our work are presented with
a simple simulation example. A sample IoT consisting of
objects from a hypothetical smart house as consumers of
power is considered with realistic data from a published
source (see Table I). Given limited proposed power budgets,
we simulated which objects are powered and in which order
in various scenarios (3 cases). Each object has a power
demand (usage setting), and a priority level (from 1 to 10)
denoting its importance relative to the other objects. The
power resource budget is an amount of power in kilowatts
(kW) available to share among the objects.
After running the Exploration Algorithm, each object
knows where it stands in the system relative to other objects.
Further, according to the Power Algorithm, the highest priority object with the smallest power consumption is powered
first. Figure 1 shows that object SP should be powered first,
and also describes the order in which power flows among
the IoT objects.
For a power availability of 23 kW, assuming that all