Abstract
—
The energy internet concept has been considered as a
new development stage of the Smart Grid, which aims to increase
the energy transmission efficiency and optimise the energy
dispatching in time and space. Energy router is a core device in the
energy internet and it connects all the devices together into a net
structure and manages power flows among them. The research work
presented in this paper described the energy router’s structure and
function expectations from the network perspective, and improved
the existing energy router design. Open-shortest-path first (OSPF)
protocol and virtual circuit switching mode are referenced from the
Internet in the energy local area network (e-LAN) design. This paper
proposed a design of an energy routing algorithm based on graph
theory in an e-LAN. A lowest-cost routing selection algorithm is
designed according to the features of power transmission, and a
source selection and routing design algorithm is proposed for very
heavy load conditions. Both algorithms have been verified by case
analyses.
Index Terms
—
energy internet, energy router, routing algorithm,
Smart Grid.
I. I
NTRODUCTION
In recent years, the increasing concern with the global energy
shortages and environment problems has stimulated worldwide
active research on the renewable energy sources, such as solar,
wind and tide [1]-[4]. Because of their distributed, intermittent
and fluctuated characteristics [5], [6], legacy grid cannot support
these increased renewable sources effectively. Smart Grid with
enhanced communication and sensing capabilities offers a
suitable platform for exploiting the use of distributed renewable
energy sources [7], [8]. In the Smart Grid, customers play both
roles of the energy consumers and the energy producers [7]-[9],
and power flows bidirectionally and flexibly end to end.
In the Smart Grid scenario, customers can exchange energy
equally and freely, which is similar to the information exchange
in the Internet. Based on this similarity, the concept ‘energy
internet’ was proposed [10]-[12], which has been considered as a
new development stage of the Smart Grid. It aims to increase the
energy transmission efficiency and optimise the energy
dispatching in time and space among renewable energy sources,
energy storage units (ESUs) and load [12]. In a typical energy
internet, all the devices are connected into a net structure by
Copyright (c) 2009 IEEE. Personal use of this material is permitted. However,
permission to use this material for any other purposes must be obtained from the
IEEE by sending a request to pubs-permissions@ieee.org.
This work is sponsored by the National Nature Science Foundation of China under
Grants 51577170, and the European Union’s Horizon 2020 research and
innovation programme under grant agreement No. 734769.
energy routers [13]. According to the scale division of the Internet,
a countrywide or worldwide large-scale energy wide area
network (e-WAN) is divided into thousands of small-scale energy
local area networks (e-LANs), such as a community or a campus,
as shown in Fig.1. The energy router is a technological
combination of power forwarding and information exchanges,
and is the core device in the energy internet [14], [15].
For the community with renewable energy sources connected,
the construction of e-LAN is especially valuable for the following
three reasons [16]. Firstly, it can decrease the dependence on the
main grid. Secondly, energy generated by renewable energy
sources can be traded with other customers instead of storing for
future use, which can reduce the start-up cost and energy storage
devices size [7]. Thirdly, with properly designed energy routing,
the optimised energy transmission and power dispatching can be
achieved, which will reduce the energy losses. In the e-LAN, the
design of the energy router and routing algorithm are two key
factors, which determines the performance of the network.
The design of energy routers and energy routing algorithms
have been proposed in some literatures. Future renewable electric
energy delivery and management (FREEDM) system centre
proposed the concept of energy router in [1], which can be
considered as an integrated microgrid. In [1], research work
mainly focused on the power electronic design of the energy
router, such as the SSTs and the internal power control of energy
routers. However, the output power of some ports is
uncontrollable, and it did not mention the network forming and
energy routing.
In [15] and [17]-[19], P. Yi et al. proposed an energy router
deployed in an electric vehicles (EV) based energy internet. In
this EV energy internet, EVs are responsible for transmitting,
distributing, and storing energy from renewable energy sources
to the places that need the energy. The energy router is defined as
the charging station with battery, which can receive energy from
one EV and forward energy to another. Based on this model, [15],
[17] and [19] proposed the algorithms about energy routers
placement, routing optimization especially in traffic jam, and
multi-source shortest energy route algorithm in both ideal and
realistic conditions. All these algorithms can be considered as
generalized routing algorithms. However, the EV energy internet
has some limitations. Firstly, it takes long time for energy
transmission compared with power grid. Secondly, the charging
stations and public buses need to add ESUs to satisfy the energy
transmission requirement, which increases the cost and power
consumption. Thirdly, all the generalized routing algorithms are
subjected to the fixed public bus lines and traffic conditions.
In [12] a secure energy routing mechanism was proposed. Each
house with its own renewable energy sources is considered as an
A Graph Theory Based Energy Routing Algorithm
in Energy Local Area Network (e-LAN)
Ruichi Wang, Student Member, IEEE, Jiande Wu, Member, IEEE, Zhongnan Qian, Zhengyu Lin,
Senior Member, IEEE, and Xiangning He, Fellow, IEEE
energy entity and connected with an energy router. Multiple
energy routers are connected into a net structure, which is the
energy internet. The mechanism is designed mainly for defending
against major internal attacks against routing protocols, while the
minimum energy losses based energy routing algorithm is
inaccurate. Many details were not discussed, such as the structure
and control of the energy router, power capacity of energy routers
and power links, dynamic regulations of the network topology,
routing information updating and exchange, etc.
According to the characteristics of power network, some
unique requirements should be considered in energy router and
routing algorithm design [20]. Firstly, the e-LAN contains
diverse distributed power generations and storage devices,
therefore the routing protocol should accommodate all of them.
Besides, in the e-LAN, energy transmission is demand dominated,
and for a certain load, the source is not specified. Also, to improve
public energy security, the number of sources for a specific load
should be as fewer as possible, to achieve higher reliability and
robustness of the system. To address these requirement, this paper
improves the existing energy router design [1] and proposes a
lowest-cost routing algorithm based on graph theory [21].
In this paper, the community based e-LAN is proposed and it
includes three types of devices: 1) renewable energy sources,
such as PV and wind turbines, 2) ESUs, such as batteries and
super capacitors, and 3) loads, such as lightings, household
electric appliances and EVs. All these devices are connected
together into a net structure by multiple energy routers, as shown
in Fig.2. In the proposed e-LAN, any two devices can exchange
energy via one or more energy routers and peer-to-peer energy
service is achievable. In this work, the transmission voltage
between energy routers is set dc 400V [22].
The energy router is responsible for energy forwarding and
information exchange. The proposed energy router design is able
to achieve controllable output power and power connections.
Routing selection is critical since it can make the system work
properly with high stability and efficiency. It also adopted
information routing protocols in the design of e-LAN, such as
routing information protocol (RIP) and open shortest path first
(OSPF). Common communication technologies will be used for
the communication between energy routers, such as optical,
power line communication (PLC), cellular communications,
Bluetooth and cognitive radio [23]-[25], it will not be discussed
in detail in this paper.
Because of its large power requirement, EV charging is a
challenge not only for the legacy grid [26] but also for the e-LAN.
According to the European EV charging standard IEC 61851, for
private EV, the medium power charging can be up to 22kW [27],
and thus it is likely that no single source in the e-LAN can power
the EV. A routing algorithm is specially designed for this
condition in this paper.
This paper is arranged as follows: Section II gives an overview
of the improved energy routers and routing mode from the
perspective of network. In Section III, the weighted energy
routing algorithm is proposed and verified by case analysis.
Finally, conclusions are given in Section IV.
Fig.1 Scale division of the energy internet
Fig.2 An example structure of the proposed e-LAN
e-WAN
e-LAN
e-LAN
Energy
router
Energy
router
Battery
PV
EV
Wind
turbine
PV
PV
Battery
EV
EV
Wind
turbine
230Vac
for household electric appliances
48Vdc
for LED lighting
EV
Wind
turbine
II. AN OVERVIEW OF ENERGY ROUTERS FROM THE
PERSPECTIVE OF NETWORK
The basic structure of the proposed energy router is shown in
Fig.3. It is composed of a power exchange structure, several
input/output ports and a controller, which is similar to the
information router shown in Fig.4.
The exchange structure consists of a common power bus, an
ESU and several converters. The 400V dc common power bus is
the intermediate link of the energy forwarding and the ESU is for
load balancing. It can absorb or compensate the power imbalance
during a short period. Compared with the main grid, the power
capacity of the ESU is much smaller, and thus the e-LAN has
higher requirement on the speed of power adjustment of sources.
It can be considered as a combination of memory-based and bus-
based exchange structures [28] in the information router. The
converters are the energy conversion unit. The input/output ports
allow energy sources, ESUs, load and other energy routers to be
connected.
As shown in Fig.3, the ports are divided into two categories:
devices connected ports (shown on the left side) and energy
routers connected ports (shown on the right side). On the left side,
the upper four ports are specially designed for renewable energy
sources, EV and ESU, while the two on the bottom are for normal
load: ac 230V (taking an example of the European standard, and
for other countries, the voltage level and frequency can be
designed according to their own standards) supplies household
electric appliances and dc 48V supplies LED lightings. On the
right side, the upper set are input ports, the middle set are the
input/output multiplexing ports and the bottom set are the output
ports. The DC-DC converters in the shadow are aimed at
controlling the output power, which is the main improvement
compared with the energy router proposed in [1]. Since all the
output power is controllable, routing algorithm can be adopted to
accurately control the power flow. For the input/output
multiplexing port, the DC-DC converter can be bypassed when
the port acts as an input port. The routing controller includes a
micro-processor and a communication module.
The proposed energy router has the following two functions.
Fig.3 Proposed structure of the energy router
Fig.4 Basic structure of the information router
A. Condition monitoring
During the power conversion and transmission process, over-
flow, voltage mismatch and current overshoot, etc. may cause
system crash and even safety accidents. Therefore, energy
exchange has much higher requirement on safety and reliability
than information exchange. Before the energy source is
connected to the system, strict input voltage check is carried out
to avoid damaging the energy router’s internal hardware. So high
voltage sensors and switches are put at each input port to monitor
and control the input voltage at all times. When there is a power
flow connection request, the controller checks the input voltage.
If the voltage matches the expected level and form, the high
voltage switch at the input port is switched on and the power is
allowed to flow in. Otherwise, the energy is blocked outside the
energy router. At the output ports, high voltage sensors and
switches are also needed for output voltage checking and
regulating. Power with voltage not matching the expected level
or form is not allowed to flow out.
The exchange structure is the core hardware of the energy
router and it is responsible for energy forwarding and conversion.
For information routers, over-flow may cause large time delay
and data packets lost. While for the energy routers, energy over-
flow may cause much more serious consequences, such as
overheating (lead to the efficiency decrease), device destroying,
or even the crash of the whole energy router. To avoid these
problems, the power values of converters and ports should be
real-time monitored.
B. Information exchange and routing design
In an e-LAN, dynamic routing algorithm should be employed
to adapt to the change of the network topology and frequent
connect/disconnect of devices. OSPF protocol, a widely used
information exchange method in the Internet [28], is referenced
to the e-LAN design in this paper. Each energy router floods the
connections and power status information of itself, directly
connected devices and power links to other energy routers [28],
so that all the energy routers acquire the whole network
connections and energy status information of the e-LAN. When
the connections or the power status have changed, the directly
connected energy router floods the update to all the others. The
power status information of all the devices, power links and
energy routers in the e-LAN is contained in three power
information tables and stored in every energy router. An example
is shown in Table I to Table Ⅲ. In these tables,
and
refer to the total power capacity and the left power capacity of the
corresponding ports and eff is the efficiency of renewable energy
sources. In Table Ⅱ, eff is the power conversion efficiency of the
power electronic converter in the corresponding port. In Table Ⅲ,
Routing
controller
Routing selection
& management
Forwarding
Energy routers
connected ports
AC/DC
Devices
connected ports
Common
power bus
400Vdc
Exchange structure
ESU
PV array
400Vdc
Wind turbine
230Vac
Load
230Vac
Load
48Vdc
EV charger
400Vdc
Battery
400Vdc
In put/outp ut
Output
AC/DC
DC/DC
DC/DC
DC/DC
DC/DC
DC/DC
Energy router
400Vdc
Energy router
400Vdc
Energy router
400Vdc
Energy router
400Vdc
Energy router
400Vdc
Energy router
400Vdc
DC/DC
……
……
……
In put
Exchange
structure
Routing
controll er
Routing selection
& management
Forwarding
…
…
Input ports Output ports
P
i/o
P
i/o
P
i/o
P
i/o
is the power link impedance and
is the transmission
voltage in the corresponding power link. Based on the equivalent
digraph of the e-LAN, each energy router executes the graph
traversal algorithm to find out all the possible power transmission
paths from other energy routers to itself for the following routing
calculation. This process repeats when the equivalent digraph
changes.
In the Internet, information is actively sent by source hosts and
transmits to the destination hosts via the lowest-cost path which
is stored in the routing table. However, in the energy internet,
energy transmission is demand dominated and the source is not
specified. Also, the cost is disproportional to the added power
value (it will be discussed in detail in next section). Therefore,
every time a load is connected, a message about its power demand
is sent to the directly connected energy router. Besides, table
checking routing selection mode is not suitable in the e-LAN. The
routing selection algorithm proposed in this paper is carried out
in every energy router based on the local power information
tables and then one or more sources with related paths are
arranged to power the load.
TABLE I
DEVICES RELATED POWER INFORMATION TABLE
Device
/kW
/kW
D
1
6.0
5.5
0.96
D
2
12.0
9.2
0.95
D
3
5.0
4.0
---
D
4
14.0
10.9
0.96
D
5
22.0
22.0
---
TABLE
Ⅱ
ENERGY ROUTERS RELATED POWER INFORMATION TABLE
Router
Port
/kW
/kW
R
1
R
1a
20.0
19.5
1
R
1b
15.0
9.0
0.98
R
2
R
2a
15.0
12.2
0.97
R
2b
15.0
14.0
0.98
R
2c
15.0
13.0
0.98
R
3
R
3a
18.0
12.0
1
R
3b
15.0
14.0
0.98
R
3c
15.0
15.0
1
R
4
R
4a
15.0
9.0
0.98
R
4b
10.0
9.0
1
R
5
R
5a
15.0
13.0
1
R
5b
18.0
12.0
0.98
R
5c
15.0
15.0
1
R
6
R
6a
20.0
20.0
0.98
R
6b
20.0
20.0
1
P
6c
18.0
18.0
1
R
7
R
7a
25.0
25.0
1
R
7b
20.0
20.0
1
R
7c
25.0
23.0
1
R
7d
25.0
25.0
1
R
8
R
8a
20.0
18.0
0.98
R
8b
20.0
20.0
1
R
8c
18.0
12.0
0.98
R
9
R
9a
20.0
14.0
0.98
R
9b
20.0
16.9
1
TABLE
Ⅲ
POWER LINKS RELATED POWER INFORMATION TABLE
Power Link
/kW
/kW
/Ω
/V
L
1_3
30
24
0.6
400dc
L
2_3
20
19
0.64
400dc
L
2_5
20
20
0.51
400dc
L
3_7
45
45
0.94
400dc
L
4_5
24
18
0.19
400dc
L
5_6
20
20
0.45
400dc
L
6_7
40
40
0.24
400dc
L
8_9
32
26
0.6
400dc
L
6_8
30
30
0.21
400dc
L
7_8
30
30
0.21
400dc
On the network layer of the Internet, two packet switching
modes can be employed: virtual circuit and datagram [28]. Virtual
circuit provides connected and reliable service. In this mode, a
virtual circuit is set up from source to destination since the start
of the communication and lasts until the end of the
communication. For an end-to-end packet delivery, routing
selection is carried out only once before data delivery and the path
cannot be further optimised during data delivery process even if
the network state is changed. Datagram network provides
disconnected service. Sending data is divided into multiple data
packets and each packet chooses its own paths separately
according to current state of the network. It is more flexible,
efficient and reliable. In the Internet, usually datagram network is
adopted. However, in the e-LAN, virtual circuit is a better choice
for the following two reasons: 1) For datagram mode, every time
the path is changed, the related converters along the old path will
shut down and the converters along the new path will start up.
Such frequent startup and shutdown will cause additional power
losses and deteriorate the reliability and robustness of the e-LAN.
2) Compared with datagram mode, a great disadvantage of virtual
circuit mode is lower utilizations of data links. However, during
energy transmission, when a load is connected, the energy
transmission is usually continuous, until it is disconnected. In
another word, this virtual circuit is always engaged during energy
transmission process, which promises high utilization efficiency.
As a result, virtual circuit mode is employed in the e-LAN.
III. P
ROPOSED ROUTING ALGORITHM
Graph theory [21] is employed to describe the energy routing
selection problem. The proposed e-LAN shown in Fig.5(a) is
equivalent to the digraph as shown in Fig.5(b). All the energy
routers constitute the set of nodes =
{
,
, … …
}
and all
the power links constitute the set of edges =
_
, … (
_
is
the power link connecting energy router
and
). The direction
of the edge represents the energy flowing direction, and it is from
an energy router’s output port to an input port. The possible
transmission path from source A to destination B is expressed as:
(
)
=
,
, …
(1)
,
, …
are the energy routers along this path.
(a) Part of an e-LAN
R
9
R
1
R
2
R
3
R
4
R
5
R
6
R
8
R
7
L
1_3
L
2_3
L
2_5
L
3_7
L
4_5
L
5_6
L
6_7
L
8_9
L
6_8
L
7_8
a
b
c
a
b
a
b
c
a
b
c
a
b
c
d
a
b
c
a
b
c
a
b
D
1
D
2
D
3
D
4
D
5
a
b
(b) The equivalent digraph of the e-LAN
Fig.5 An example of an e-LAN and its equivalent digraph
A. Weighted routing algorithm
In practical applications, the total cost of transmission and
conversion involves multiple factors, such as power losses and
prices, therefore, it is necessary to assign weight to every edge
and node according to its cost. In this paper, we mainly focus on
the power losses and other factors will be considered in the future
study of this project. Define
as the weight of energy router
and
_
as the weight of the link
_
. To give priority to those
energy routers and power links with smaller losses, the weight
should be proportional to the added power losses , as shown in
(2) and (3).
=
(2)
_
=
_
(3)
and
are two adjustable coefficients and in this paper
they are defined as
=
= 1.
The energy losses in an energy router are composed of the
power conversion losses and power cable transmission losses
inside the energy router. The later can be ignored since the power
cables inside the energy routers are very short. The conversion
losses of an energy router are not a fixed value, but decided by
the converters at the ports through which the power flows. The
weight of port x of energy router
is expressed as:
=
(1
)
,
(4)
where
is the added power value through port x of energy
router
, and
is the efficiency of the electronic power
converter between port x and the common power bus (if there is
no converter, then
=1). For simplicity, the efficiency of a
converter is considered as a constant measured under the rated
power, and thus the conversion losses are linearly proportional to
the converted power approximately. The total power losses of an
energy router are expressed as (port x is the input port and port y
is the output port):
=
(
(
1
)
+
1
)
(5)
The total power losses in a power link between energy router
and
are calculated as:
_
=
_
_
_
.
(6)
_
is the power transmitting in the link. For added power
_
,
the link weight is:
_
=
_
_
(
_
+
_
_
)
,
(7)
where
_
is the link resistance,
_
is the link voltage,
_
is the
already existing power in the link, which can be found in the
power links related power information table, and
_
is the
added power. It can be seen from (7) that for power transmission,
the power losses are not a linear superposition. As shown in Fig.6,
the added power losses caused by the same added power
are increasing with the increase of the existing power in the power
link. As a result, the weight of power links is variable according
to the existing power and the added power. The total cost is the
sum of all the nodes’ and edges’ weights along the power
transmission path, as shown in (8):
=
+
_
,
(8)
Fig.6 Function graph of power link losses with respect to the transmission power
A Matlab based case analysis is carried out to verify the
proposed weighted routing algorithm and a comparison is made
with the existing energy routing algorithm. Part of an e-LAN
topology is shown in Fig.7 and Table Ⅳ shows the parameters. It
is assumed that the left power capacities of each device, power
link and energy router’s ports are large enough. Two optional
paths from source
to load
are found out after graph
traversal:
1
(
)
,
,
,
,
(9)
2
(
)
,
,
,
(10)
Load
400Vdc
400Vdc
400Vdc
R
1
R
2
R
3
R
4
R
5
L
1_2
L
2_3
L
3_5
L
1_4
L
4_5
D
1
D
2
a
b
c
b
a
a
b
a
b
a
b
c
400Vdc
400Vdc
Fig.7 Part of an e-LAN
TABLE
Ⅳ
PARAMETERS OF THE E-LAN TOPOLOGY
Device
D
1
0.95
D
2
---
Router
Port
R
1
P
a
0.97
P
b
0.98
P
c
0.98
R
2
P
a
1
P
b
0.98
R
3
P
a
1
P
b
0.98
R
4
P
a
1
R
2
R
1
R
3
R
4
R
5
R
6
R
7
R
8
R
9
L
1_3
L
2_3
L
2_5
L
3_7
L
4_5
L
5_6
L
6_7
L
6_8
L
7_8
L
8_9
Transmission
power P
Power losses p
△
p1
△
p2
0
△
P
△
P