Sie sind auf Seite 1von 10

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO.

2, JUNE 2012

1029

Agent Based Restoration With Distributed Energy


Storage Support in Smart Grids
Cuong P. Nguyen, Member, IEEE, and Alexander J. Flueck, Senior Member, IEEE

AbstractThe goal of this paper is to present a new and


completely distributed algorithm for service restoration with distributed energy storage support following fault detection, location,
and isolation. The distributed algorithm makes use of intelligent
agents, which possess three key characteristics, namely autonomy,
local view, and decentralization. The switch agents will detect,
locate and isolate the fault, then restore the load. The distributed
energy storage agent will support the system in grid-connected as
well as islanded operation. Important restoration issues such as
load priority restoration and islanding coordination of multiple
distributed energy storage systems will be discussed. Two case
studies on the modified IEEE 34 node test feeder will be presented.
Index TermsActive network management, advanced distribution automation, distributed energy storage, islanded operation,
islanding coordination, multiagent systems, priority load restoration, smart grid.

I. INTRODUCTION

ESTORATION is an important part of Advanced Distribution Automation (ADA), which seeks to restore the
unfaulted and outaged parts of a system due to the isolation
of a fault. Mathematically, the restoration problem is a combinatorial problem with the objective of maximizing the supply
of power for as many customers as possible while satisfying
source, line/cable loading, and often radial network constraints.
Different reconfiguration techniques have been proposed to
solve this problem, such as deterministic mathematical programming, heuristic techniques and knowledge-based systems.
In deterministic mathematical programming, the restoration
problem is often formulated as a mixed integer programming
(MIP) problem. Then, any available MIP technique can be
applied to solve the restoration problem. Nagata et al. [1] proposed a two-stage algorithm which decomposes the restoration
problem into two subproblems (the maximization of available
power to the de-energized area, and the minimization of the
amount of unserved energy). The algorithm is limited to dc
models which do not take into account the reactive power as
well as voltage variations. In [2], the authors presented the
Manuscript received December 20, 2010; revised June 20, 2011; accepted
January 18, 2012. Date of current version May 21, 2012. This work was supported by the Perfect Power project at Illinois Institute of Technology, fully
funded by the Department of Energy, Illinois Institute of Technology, and S&C
Electric Company under Award DE-FC26-08NT02875. Paper no. TSG-003812010.
C. P. Nguyen is with the GE Energy, Schenectady, NY 12345 USA (e-mail:
cuong.p.nguyen@ge.com).
A. J. Flueck is with the Department of Electrical and Computer Engineering,
Illinois Institute of Technology, Chicago, IL 60616 USA (e-mail: flueck@iit.
edu).
Digital Object Identifier 10.1109/TSG.2012.2186833

network reconfiguration for service restoration in shipboard


power distribution systems. The problem is formulated as a
variation of the fixed charge network flow problem, and solved
by the CPLEX commercial package. Due to the linearization
of some of the problem constraints, the solution does not
guarantee a global optimum. Perez-Guerrero et al. [3] solved
optimal restoration of distribution systems using a dynamic
programming approach. During the solution process, many
states close to each other are grouped. Then, only the best states
are selected to reduce the problem size, so that solution speedup
is achieved. While mathematical programming might guarantee
a global optimal solution, its drawback is the computational
intensity, so it might not be practical for large distribution
systems.
On the other hand, heuristic techniques and knowledge based
systems utilize the operators knowledge and experience to
narrow down the search space for the combinatorial restoration
problem. Thus, the solution is achieved in a shorter time. In [4],
Liu et al. reported an expert system algorithm for restoration
and loss reduction of distribution systems. The authors constructed a knowledge base which contains rules that implement
a solution approach that system operators can use in order to
restore as many loads as possible. A recent work by Tsai [5] extended [4] by considering load variation so multiple restoration
plans are obtained. An early work on heuristic search approach
to distribution system restoration was introduced by Morelato
and Monticelli [6]. The proposed approach used a knowledge
guided search strategy to reach the solution. Artificial neural
network (ANN) [7], ant immune system-ant colony optimization (AIS-ACO) [8], Genetic Algorithm (GA) [9], among
others are also used extensively for restoration problems.
However, most of the restoration techniques surveyed above
solve the reconfiguration problem from a centralized point of
view, which in reality often requires a low-latency communication system transferring a potentially large amount of data between field devices and the control center. In addition, the control center requires expensive computing capability. As a result,
centralized techniques are subject to a single point of failure risk
and large capital costs.
Meanwhile, multiagent systems (MAS) have recently
emerged as a competitive technology for the advanced distribution automation requirements of smart grid. Essentially, a smart
grid is an advanced grid that makes use of distributed intelligence to fulfill its duties of self-healing, high reliability, high
quality, and demand response [10]. The MAS can overcome the
disadvantages of centralized control by distributing the control
at the component level, thus avoiding a single point of failure,
while utilizing peer-to-peer communication for collaboration to

1949-3053/$31.00 2012 IEEE

1030

achieve near global objectives. The research and development


of MAS in power engineering have been highlighted in a recent
comprehensive review paper [11]. However, not much work
has been seen on multiagent approaches for service restoration,
especially with distributed energy storage support.
Distributed energy storage systems (DES) together with distributed intelligence and smart control (DISC) were envisioned
as two of the top five concerns for the future electric grid as early
as 2003 by the Department of Energy [12]. The two issues since
then have attracted many electric power system researchers, designers, and engineers.
Today, the technology of DES has advanced to the point that
successful cost-effective demonstrations in real systems have
been observed and therefore utilities have begun to consider it
as a crucial piece of technology in upgrading the aging grid. As
an example, the first megawatt scale deployment of the DES
technology was successfully carried out in the American electric power grid in 2006. The project aimed to reduce the peak
load on overloaded equipment in the distribution substation. Realizing multiple benefits that DES technology has to offer, the
utility continued to install three other DES units of larger capacity in their distribution system in 2008, providing not only
peak shaving but also dynamic islanding [13].
Distributed energy storage systems are so flexible that they
can be applied virtually any place in a modern electric power
system. In power generation, DES can serve as spinning reserve,
enabling the grid to transition in a smoother way to a new stable
operating point in the case of losing a generation unit [14]. More
recently, DES technology has been used to provide frequency
regulation in interconnected systems. The recent trend toward
harvesting intermittent renewable energy resourcessuch as
wind power and solar poweralso shows the strong need for
energy storage [15], [16]. The synergy of renewable energy and
energy storage will allow for flexible dispatch, so the energy can
be used when it is needed.
In transmission, DES can be used to support load carrying
capacity and stability by injecting power to damp out a system
disturbance when it happens [17], [18]. DES can also be used to
support a voltage dip in the transmission system when acting as
a reactive power source. To accommodate load growth, a DES
installation near the load center can be a very competitive alternative to a transmission upgrade investment by deferring the
expensive cost of upgrading or replacing the existing transmission facilities.
In distribution, DES can be located in various places from
the substation down to the end customers. During normal operation, DES can level load by storing energy during off-peak
hours and then releasing it during the peak period. If embedded
with advanced electronic components, DES also has the ability
to manage the real/reactive power, thus reduce overall system
losses [19] and improve the voltage profile [20]. In the event of
any fault or disturbance that causes a power outage in some part
of the grid, DES may act temporarily as a stand-alone source to
supply the load in its island while the system is in the process
of reconfiguration.
Due to multiple benefits listed above, distributed energy
storage has the potential of becoming a key component of smart
grids if barriers imposed by technology costs are overcome.

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO. 2, JUNE 2012

Fig. 1. Switching device and switching agent relationship.

This paper will explore the agent based restoration problem


with the support of distributed energy storage.
II. MULTIAGENT SYSTEM (MAS) AND SIMULATION TEST BED
The proposed MAS in this paper consists of two classes of
agents, namely switching agent (series agent), and distributed
energy storage agent (shunt agent). The MAS is designed based
on the team concept [21]. A team corresponds to a power line
segment or a group of segments bounded by agent-controlled
switching devices (series devices). In other words, the location
of the switching agents determines the number of teams in the
MAS. When an agent-controlled shunt device, such as a distributed energy storage system, is connected to a line segment,
it will increase the number of agents in the team but not the
number of teams in the MAS. The agents within a team can communicate with each other, while a team can communicate with
other teams next to it via common teammate(s). The teams
of agents usually communicate with each other to negotiate the
most efficient and expeditious reconfiguration of the system in
response to fault conditions and other circuit abnormalities.
The foundation of an unbalanced, three-phase distribution
Network Simulator (called dNetSim) and Java Agent DEvelopment Framework (called JADE) were presented in recent papers
by the authors [22], [23]. Basically, dNetSim is responsible for
solving the unbalanced power flow while the JADE platform
enables the autonomous agents to work out the best setting for
each agent-controlled device in the MAS at any given time.
III. SWITCHING AGENT
A. Power System Model
A switching device in branch i-j can be modeled as an ideal
branch from node X to node Y, see Fig. 1. The X and Y sides of
the device correspond to the X and Y teams, which are local to
the agent.
For example, switching device SWI_18 (Fig. 6) resides in
branch 816818. The X team of this agent is Team 2 and the
Y team of this agent is Team 3.
B. Agent Model
Fig. 2 shows a high level switching agent state diagram. A
description of the detailed diagram with 11 states is given in
[25]. Basically, there are four blocks.
1) NORMAL: The agent resides in this block under normal
conditions, i.e., all sensor measurements are within the
normal operating range.
2) ABNORMAL: The agent enters this block if there is any
problem in the network, such as an overload or short-circuit

NGUYEN AND FLUECK: AGENT BASED RESTORATION WITH DISTRIBUTED ENERGY STORAGE SUPPORT IN SMART GRIDS

1031

Fig. 4. Model of DES in different modes.


Fig. 2. High level switching agent state diagram.

Fig. 3. Single line diagram of medium voltage application[24].

fault. In the latter case, the agent will detect the fault, then
locate and isolate the fault via MAS communication.
3) TRANSIENT: This contains the intermediate state where
the agent stays temporarily before transitioning to another
state.
4) RESTORATION: After moving to this block, the agent
will start the restoration process, including checking if the
team is ready for restoration, finding the restoration path,
and closing the switch to restore the team load.
During operation, the measurements including present status,
voltages
and
, and current
, are sampled regularly
and processed by the agent. Based on the measurements and
MAS communication, the agent will determine the new desired
status of the switching device. In addition, the switching agent
is also given the total load in each of the X and Y teams, the
capacity rating of the X and Y team line segments as well as
the capacity of the switching device, and the capacity rating
of a connected source, if any. These pieces of information are
used in the restoration process to describe source and network
constraints local to each agent.

beta alumina ceramic electrolyte. A NaS battery is characterized


by:
High operating temperatures of
to
.
High energy density: discharge duration can reach 6 hours
at rated output.
High efficiency of charge/discharge (82%92%).
A life cycle of 15 years.
Inexpensive material.
With these characteristics, a NaS battery is very attractive for
large-scale nonmobile applications such as distributed energy
storage for a distribution power system. Based on [26], the NaS
battery has been applied widely in the Japanese power grid. As
of 2009 the total installed capacity is about 270 MW. In the
United States, 9 MW of NaS battery system has been deployed
so far for peak shaving, backup power, firming wind capacity,
and other applications.
A. Power System Model
Depending on the condition of the grid, DES can work in
one of three modes, namely POWER CHARGE mode, POWER
DISCHARGE mode, or VOLTAGE CONTROL mode (Fig. 4).
First, in POWER DISCHARGE mode, the objective is to
discharge power into the grid at a user-defined complex power
setpoint. The DES can be modeled as a voltage dependent
current source. That is, given a total complex three-phase
power discharge setpoint
, if the terminal voltages
are
, then a set of three-phase balanced currents
can be injected into the system to meet the setpoint:
(1)
(2)
Therefore,
(3)

IV. DISTRIBUTED ENERGY STORAGE SYSTEM

(4)

This section presents the power system model and agent


model for the electronic converter storage system (ECS)a
popular type of distributed energy storage system.
ECS consists of a power electronics system and storage
system. The storage system, typically batteries, is of dc type. It
interfaces with an ac electric grid through an advanced power
electronics system. Fig. 3 shows an example of an ECS system.
Sodium sulfur battery (NaS) is a molten metal battery system
that is constructed from a molten sulfur (Na) positive electrode
and a molten sodium (S) negative electrode separated by a solid

(5)
It is seen from (3)(5) that both current magnitude and current
angle vary according to the three-phase terminal voltage.
Second, in POWER CHARGE mode, the objective is to
charge or draw the power from the grid at a user-defined
complex power setpoint. In this case, the DES can be modeled
as a voltage dependent current sink. That is, given the total
three-phase power charge setpoint
at the rated
voltage level
, if the terminal voltages are
,

1032

then a set of three-phase balanced currents


drawn from the system:

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO. 2, JUNE 2012

can be

TABLE I
POWER FLOW DATA OF THE MODIFIED IEEE 34 NODE TEST FEEDER

(6)
(7)
(8)
It is seen from (6)(8) that only the current angle varies according to the three-phase terminal voltage. The current magnitude is always constant in POWER CHARGE mode. Typically,
the charging power factor is maintained at 1.0, so
in (6) is
zeroed out.
Third, in VOLTAGE CONTROL mode, the objective is to
control the DES terminal voltage. The DES works like a generator that regulates its terminal voltage. That is, given the desired
voltage setpoint
, a set of balanced three-phase terminal voltages
is maintained as follows:
(9)
(10)
(11)
An example of DES power flow data is given in Table I with
the parameters explained as follows.

the rated power (kVA).

the rated line to line voltage (kV).

the maximum energy storage capacity


(kWh).

the real power setpoint (kW) in POWER DISCHARGE mode.

the power factor setpoint (%) in POWER DISCHARGE mode.

the available stored energy (kWh).

the real power setpoint (kW) in POWER


CHARGE mode.

the power factor setpoint (%) in POWER


CHARGE mode.

the minimum energy threshold


(%) before charging is required.

the high voltage limit (per unit).

the low voltage limit (per unit).

the maximum injected current


threshold (% of rated current) when a fault occurs while
working in POWER DISCHARGE mode.

the voltage setpoint (kV) in VOLTAGE


CONTROL mode.

the maximum injected current threshold


(%) in VOLTAGE CONTROL mode.

the time (seconds) before shutdown in


POWER DISCHARGE mode when the terminal voltage
is less than
.

the overload trip time (seconds) in


VOLTAGE CONTROL mode.

Fig. 5. High level DES agent state diagram.

the linear current increase time (seconds) in POWER DISCHARGE mode.

B. Agent Model
The DES agent is represented by the SYSTEM CONTROL
block seen in Fig. 3. During operation, the agent receives the
measurements including interface switch status (not shown),
mode, terminal voltages, and injection currents. Then, the agent
decides the proper interface switch status, mode, and settings
for the DES.
Fig. 5 shows a high level DES agent state diagram. A description of the detailed diagram with 12 states is given in [25].
Basically, there are three blocks and two states.
1) POWER DISCHARGE: The DES injects power into the
grid at the user setpoint of
.
2) POWER CHARGE: The DES draws power from the
grid to charge its battery system at the user setpoint of
. This mode is not allowed
while the DES is operating in an island. Before the DES
enters this mode, MAS communication is required to determine if the distribution network and associated source
can supply the charging current.
3) VOLTAGE CONTROL: The DES works as a voltage
source in an island. A voltage setpoint of
is maintained at the DES terminals. Within an island,
only one DES is allowed to work in VOLTAGE CONTROL mode. That DES is also called the master DES.
The decision on which DES is the master depends on
the availability of the energy and size of the DES units
in the island, which is determined in real-time via MAS
communication.

NGUYEN AND FLUECK: AGENT BASED RESTORATION WITH DISTRIBUTED ENERGY STORAGE SUPPORT IN SMART GRIDS

1033

Fig. 6. Configuration of modified IEEE 34 node test feeder equipped with agent-controlled switches, and distributed energy storage.

4) SHUTDOWN: The DES has to be disconnected in this


state due to abnormalities in the network such as short-circuit or battery drain in the DES.
5) CLOSING: The DES is connected back to the network in
this state.
V. RESTORATION SIMULATION
The simulation on the IEEE 34-node distribution feeder [27]
is presented in this section, see Fig. 6. The 24.9 KV substation
resides at node 800. For the simulation, eight agent-controlled
switching devices [e.g., switch (SWI), breaker (BRK), and sectionalizer (SCT)] are added. In addition, three agent-controlled
distributed energy storage units DES_16, DES_58, and DES_60
are added for restoration support. The power system data of the
DES units are given in Table I. The 11 agents form a multiagent
system of nine teams. The naming convention for each agent is
AGT_ followed by the name of the device. This simulation
uses a timestep of 0.1 ms.
The agents communicate using Foundation for Intelligent
Physical Agents (FIPA) standard [28]. Some message performative examples include:
REQUEST: the sending agent requests the receiving agent
to perform some action.
QUERYIF: the sending agent asks the receiving agent
whether or not a given proposition is true.
INFORM: the sending agent informs the receiving agent
that a given proposition is true.
A. Case 1: Partial Restoration
The simulation starts at time 0. Then, a permanent unbalanced
bolted phase-A-to-phase-B fault (or A-B fault) occurs at node
828 at time 5.0 s. The simulation ends at 20.0 s. The Team 9 load
is critical while all other team loads are noncritical. The initial

energy capacity of DES_58 is decreased down to


, so it will need to charge during the simulation.
The switching sequence of switching devices and DES units
is given in Table II, while a short list of 184 messages being
exchanged in the MAS is shown in Fig. 7. The time line of the
simulation is explained as follows.
At 0.0000 s, the system operates in the normal state, i.e., all
switching devices are closed and work in NORMAL mode,
and all distributed energy storage system units operate in
POWER DISCHARGE mode at the desired setpoint.
At 0.7503 s, DES_58 available energy decreases
down to the minimum level
. AGT_DES_58 opens its
interface switch to isolate the unit, then sends out a
REQUEST cap=17.39 message to all Team 5 teammates for allocating a charge capacity of 17.39 amps.
The message gets propagated to all parts of the MAS to
search for the capacity (messages #2, #8, and #13). The
corresponding successful INFORM done messages are
#17, #19, and #20. Therefore, the charge path will be
.
At 1.7180 s, AGT_DES_58 closes its interface switch to
start charging the unit in POWER CHARGE mode. Fig. 9
shows that while charging, the real power output is negative. In other words, DES_58 draws power from the grid.
At 5.0000 s, a short-circuit fault occurs at node 828.
All the DES units sense the voltage dip at their terminals. AGT_DES_16 reacts to that by increasing the
output current linearly toward the
within
window to inject more power into the grid to meet the
power setpoint. Accordingly, the DES_16 current in Fig. 8
is seen to increase linearly from 5.0001 s to 5.0169 s.
Also due to the fault, AGT_BRK_02 and ACT_SCT_28
move to ABNORMAL mode (excessive current) while all

1034

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO. 2, JUNE 2012

other switching agents move to TRANSIENT mode (low


voltage).
At 5.0168 s, AGT_SCT_28 opens its sectionalizer to isolate the fault after counting a given number of fault current
samples. Consequently, the voltage profile in the feeder upstream of the fault is recovered, causing AGT_DES_16 to
move back to normal discharge state at 5.0169 s.
Meanwhile, having isolated the fault, AGT_SCT_28 currently working in ABNORMAL mode, starts locating the
fault by sending out QUERYIF isFaulted messages to
Team 4 (message #23) and Team 2 (messages #22 and
#25) switching teammates. The respective replying messages #28 (INFORM TRUE), #27 (INFORM FALSE),
and #30 (INFORM FALSE) show that the fault current
is passing through Team 2, but not Team 4. Therefore, the
fault must be in Team 4. After fault isolation, the distribution system is split up into two parts. The part upstream of
Team 4 is still energized by the substation, while the part
downstream of Team 4 is islanded.
At 5.0169 s, AGT_DES_58 transitions from POWER
CHARGE mode to SHUTDOWN mode due to voltage loss,
turning off DES_58. Since there are two DES units in Team
5, the islanded operation starts by determining who will
operate in VOLTAGE CONTROL mode because only one
DES, also called the master DES, is allowed to control the
voltage. AGT_DES_58 sends a QUERYIF KVA=750
message #21 to AGT_DES_60 in Team 5 to query for
VOLTAGE CONTROL mode. The INFORM FALSE
reply #26 is received at 5.2838 s due to the smaller rated
kW capacity of DES_58. Therefore, AGT_DES_58 can not
work in VOLTAGE CONTROL mode.
At 5.0319 s, SCT_52 is opened due to extended loss
of voltage. The corresponding agent transitions to
RESTORATION mode.
At 5.0349 s, SWI_42, SCT_46, SWI_32, and SWI_62 are
also opened due to extended loss of voltage. The corresponding agents transition to RESTORATION mode.
At 5.1003 s, AGT_DES_60 transitions from POWER
DISCHARGE mode to SHUTDOWN mode due to the
extended loss of voltage, turning off DES_60. Then,
AGT_DES_60 sends a QUERYIF KVA=1500 message
#24 to AGT_DES_58 in team 5 to query for VOLTAGE
CONTROL mode. The replying INFORM TRUE message #29 is received at 5.4005 s due to the larger rated
kW capacity. Therefore, AGT_DES_60the master
DEScan work in VOLTAGE CONTROL mode.
At 5.4007 s, after transitioning from SHUTDOWN
mode to VOLTAGE CONTROL mode, AGT_DES_60
broadcasts QUERYIF isClear messages #31, #32, #33,
and #34 to all Team 5 switching members to query if
the team is clear for restoration. All replies (INFORM
TRUE messages #35, #36, #37, and #38) are received
by AGT_DES_60 by 5.7675 s, confirming Team 5 is ready
for restoration.
At 5.7677 s, AGT_DES_60 closes its interface switch to
regulate the voltage in Team 5, see Fig. 10. Then, it starts

searching for priority restoration path(s) by sending out


REQUEST priPath messages #40, #41, #42, and #52 to
Team 5 switching teammates. The request is forwarded to
other parts of the MAS, including messages #56 and #67.
The corresponding reply messages are #57, #58, #71, #72,
#78, and #80. Since only Team 9 has priority load, the priority restoration path is
as
seen in messages #72 and #80. Note that Team 6 does not
have priority load, but it still lies on the priority restoration
path. During the restoration process, team loads in a priority restoration path are restored first.
At 5.7679 s, after sensing good voltage at its terminal,
DES_58 decides to come back online to inject power into
the grid in POWER DISCHARGE mode.
At 10.5168 s, SWI_42 is closed after Team 6 is confirmed clear (messages #54 and #68) and the REQUEST
cap=10.3 message #109 is accepted (message #114) to
restore Team 6 load. The agent first transitions to TRANSIENT mode, then to NORMAL mode.
At 11.0942 s, DES_58 is shut down due to energy insufficiency. Note that in islanded operation, DES_58 is not
allowed to charge. Therefore, any charge request will be
rejected as seen in messages #129 and #138.
At 11.0992 s, SCT_46 is closed after the REQUEST
cap=9.82 amps (messages #115 and #116) is accepted
(messages #121 and #126) to restore Team 9 load. The
agent first transitions to TRANSIENT mode, then to
NORMAL mode.
After restoring the priority load, AGT_SCT_46 sends out
REQUEST updPath messages #130 and #133 to the
MAS to update the priority restoration path, the replying
INFORM done messages are #141, #144. As a result,
the priority restoration path is removed, allowing all other
nonpriority teams to start their restoration.
At 18.1165 s, SWI_62 is closed after the REQUEST
cap=2.09 amps (message #169) is accepted (message
#178) to restore Team 8 load. A similar attempt (message
#168) is also made by SWI_32 to restore its Team 7 load
of 12.7 amps in message #168, but due to the limited capacity of source DES_60, the attempt was not successful
(message #175).
This case study demonstrates partial service restoration following a serious fault. Due to the serious fault, the distribution
system is divided into two parts. The part upstream of the fault
is still energized by the substation while the part downstream
of the fault is supplied by the DES units. Load priority is taken
into account during the restoration process. The distributed energy storage system agents work in different modes based on the
different conditions of the distribution system. In particular, the
coordination of the two DES units in the same island is achieved.
If there are three or more DES units in the same team, then
a similar QUERYIF isMaster communication process can be
used for each DES. The DES that has available energy and the
largest capacity will be the master DES, which can work in
voltage control mode, while all other DES units will work in
power discharge mode.

NGUYEN AND FLUECK: AGENT BASED RESTORATION WITH DISTRIBUTED ENERGY STORAGE SUPPORT IN SMART GRIDS

1035

Fig. 8. Case 1: DES_16 output.

Fig. 9. Case 1: DES_58 output.

Fig. 7. Case 1: Abbreviated list of agent communication messages.

B. Case 2: Full Restoration


The rated capacity of DES_60 is increased to 2250 KVA
for the full restoration simulation. The simulation starts at time
0. Then, a temporary breaker failure occurs at 5.0 s, causing
BRK_02 to open. The simulation ends at 20 s. Also, at 5.0 s,
the communication link from AGT_BRK_02 to AGT_SWI_18
fails permanently. Team 9 load is critical while all other team
loads are noncritical.

Fig. 10. Case 1: DES_60 output.

This restoration case will employ the break-before-make


transfer scheme to guarantee the radial configuration of the distribution network at all time. The scheme requires that the energized line segment has to be de-energized from the existing
source or break before it can be re-energized to an alternate

1036

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO. 2, JUNE 2012

TABLE II
CASE 2: FULL RESTORATION SWITCHING SEQUENCE

TABLE III
CASE 2: FULL RESTORATION SWITCHING SEQUENCE

source or make. The advantage of this common practice is to


avoid switching surges and loop currents. But the disadvantage
is that the line segment is subject to a momentary outage. On
the other hand, if make-before-break were to be used, then
the synchronization of the sources would be carried out before
the make action occurs. This may lead to longer restoration
times, but there would be no outage during the transfer.
The switching sequence of switching devices and DES units
is given in Table III. There are totally 230 messages being exchanged in the multiagent system during the simulation. The
time line of the simulation is explained as follows.
At 0.0000 s, all switching devices are closed and all
DES units work in POWER DISCHARGE mode in the
PQ_NORMAL state.
At 5.0000 s, BRK_02 opens due to temporary breaker
failure. The communication link from AGT_BRK_02 to
AGT_SWI_18 is also unavailable from this time onward.
Therefore, the distribution system can not be restored

back to the substation because BRK_02 fails to get the


QUERYIF isClear replies from all Team 2 teammates.
The distribution system can only be restored by DES units.
At 5.0151 s and 5.0181 s, after sensing the extended
voltage loss, all closed switching devices are opened by
the corresponding switching agents.
At 5.0835 s, all DES units are also opened due to the
extended voltage loss. After this time, similar to case 1,
AGT_DES_60 in Team 5 will work in VOLTAGE CONTROL mode while AGT_DES_58 in the same team will
operate in POWER DISCHARGE mode. AGT_DES_16 is
the only DES agent in Team 2, so it will work in VOLTAGE
CONTROL mode by default.
At 5.4570 s, AGT_DES_16 closes its interface switch
to work in VOLTAGE CONTROL mode after confirming Team 2 to be clear by MAS communication.
DES_16 acts as a new source to regulate the voltage in
Team 2. The part of the distribution system energized
by this source will be referred to as the upstream island. After closing its interface switch, AGT_DES_16
starts searching for priority load using MAS communication. The priority restoration path is found to be
.
At 5.8176 s, AGT_DES_60 (master DES) also closes
its device to work in VOLTAGE CONTROL mode after
Team 5 is clear. DES_60 acts as a new source to regulate the voltage in Team 5. The part of the distribution
system energized by this source will be referred to as
the downstream island. After closing, AGT_DES_60
starts searching for priority load using MAS communication. The priority restoration path is found to be
.
At 5.8178 s, after sensing good voltage at its terminal,
AGT_DES_58 also closes its interface switch to discharge
power into Team 5.
At 6.1728 s, SCT_28 in the upstream priority restoration
path is closed to restore Team 4 load of 1.71 amps.
At 6.6001 s, SWI_42 in the downstream priority restoration path is closed to restore Team 6 load of 10.3 amps.
At 7.2158 s, SCT_46 in the downstream priority restoration path is closed to restore Team 9 load of 9.82 amps.
After restoring the priority load, AGT_SCT_46 starts a request for updating the priority restoration path. As a result, the upstream and the downstream priority restoration
paths are removed, allowing other nonpriority loads to be
restored.
At 10.1331 s, SWI_62 is closed to restore Team 8 load of
2.09 amps.
At 10.1997 s, SWI_32 is closed to restore Team 7 load of
12.7 amps.
At 18.3741 s, the request for allocating Team 3 load of
13.06 amps from AGT_SWI_18 is successfully authorized by the downstream island. The upstream island can
not satisfy Team 3 capacity request due to the limited
capacity of the source DES_16. The successful restoration
path is
. This
is a special restoration path because SCT_52 is open.
Therefore, the break-before-make transfer scheme must

NGUYEN AND FLUECK: AGENT BASED RESTORATION WITH DISTRIBUTED ENERGY STORAGE SUPPORT IN SMART GRIDS

be activated to reconfigure the network such that the upstream island can be merged with the downstream island.
Initially, SCT_28 is opened to break Team 4 from the
upstream island.
At 18.3743 s, SCT_52 is closed to make Team 4.
At 18.6077 s, DES_16 is shut down to break Team 2.
At 18.6079 s, SCT_28 is closed to make Team 2.
At 18.6081 s, after sensing good voltage at its terminal,
DES_16 is reconnected to operate in POWER DISCHARGE mode.
At 18.7745 s, SWI_18 is closed to restore Team 3.
This case study shows the full service restoration following
breaker failure and communication failure events. The failures
cause the system to operate in two islands initially. Due to the
limited capacity of one of the DES units, merging is necessary
to achieve full service restoration. Priority load is also considered during the restoration process. The break-before-make
transfer scheme is employed for reconfiguring the network
during the restoration.
C. Loss Reduction and Voltage Improvement in Normal
Operation
Loss reduction is observed in the distribution system under
normal operation. With the presence of the DES units, the total
system loss is 120.056 kW as compared to 274.673 kW without
the DES units. The loss is reduced mainly because part of the
load is locally supplied by the DES units rather than the substation, which is electrically farther away.
In terms of voltage profile, most of the node voltages are increased slightly with the presence of the DES units. In particular,
the maximum voltage increase of 7.4% occurs at node 814.
VI. CONCLUSION AND FUTURE WORK
As the communication network penetrates deeper into the
distribution system and as distributed computing power improves in both speed and cost, the implementation of agent
based advanced distribution automation is becoming more
practical. The dNetSim-JADE simulation package provides a
realistic testing environment for advanced distribution automation via autonomous agents.
A new model and distributed control strategy for DES were
presented. Multiple benefits of DES were demonstrated through
two test cases: 1) In normal operation, the total system loss
was cut by more than a half and the system voltage profile was
slightly improved; 2) in abnormal operation, dynamic islanding
was achieved to restore the islanded portion of the grid; 3) Coordination of multiple DES units in the same island as well as in
different islands was achieved; 4) Load priority was considered
during restoration.
Future work will explore the optimal dispatch problem of
multiple distributed energy storage system units from the distributed perspective.
ACKNOWLEDGMENT
The authors would like to thank the sponsors for their financial support.

1037

Department of Energy Disclaimer: Neither makes any warranty, express or implied, or assumes any legal liability of responsibility for the accuracy, completeness, or usefulness of any
information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights.
Reference herein to any specific commercial product, process,
or service by trade name, trademark, manufacturer, or otherwise
does not necessarily constitute or imply its endorsement, recommendation, or favoring by the U.S. Government or any agency
thereof. The views and opinions of authors expressed herein do
not necessarily state or reflect those of the U.S. Government or
any agency thereof.

REFERENCES
[1] T. Nagata, S. Hatakeyama, M. Yasouka, and H. Sasaki, An efficient
method for power distribution system restoration based on mathematical programming and operation strategy, in Proc. Int. Conf. Power
Syst. Technol., 2000.
[2] K. Butler, N. Sarma, and V. R. Prasad, Network reconfiguration for
service restoration in shipboard power distribution systems, IEEE
Trans. Power. Syst., vol. 16, no. 4, pp. 653661, Nov. 2001.
[3] R. Perez-Guerrero, G. Heydt, N. Jack, B. Keel, and A. Castelhano, Optimal restoration of distribution systems using dynamic programming,
IEEE Trans. Power. Del., vol. 23, no. 3, pp. 15891596, 2008.
[4] C.-C. Liu, S. Lee, and S. Venkata, An expert system operational aid
for restoration and loss reduction of distribution systems, IEEE Trans.
Power. Syst., vol. 3, no. 2, pp. 619626, May 1988.
[5] M. Tsai, Development of an object-oriented service restoration expert
system with load variations, IEEE Trans. Power. Syst., vol. 23, no. 1,
pp. 219225, 2008.
[6] A. Morelato and A. Monticelli, Heuristic search approach to distribution system restoration, IEEE Trans. Power. Del., vol. 4, no. 4, pp.
22352241, Oct. 1989.
[7] Y. Hsu and H. Huang, Distribution system service restoration using
the artificial neural network approach and pattern recognition method,
IEE Proc. Gener., Transm., Distrib., vol. 142, no. 3, pp. 251256, May
1995.
[8] A. Ahuja, S. Das, and A. Pahwa, An ais-aco hybrid approach for
multi-objective distribution system reconfiguration, IEEE Trans.
Power. Syst., vol. 22, no. 3, pp. 11011111, 2007.
[9] Y. Kumar, B. Das, and J. Sharma, Multiobjective, multiconstraint service restoration of electric power distribution system with priority customers, IEEE Trans. Power. Del., vol. 23, no. 1, pp. 261270, 2008.
[10] A. Flueck and Z. Li, The journey to perfect power at Illinois Institute
of Technology, IEEE Power Energy Mag., vol. 6, no. 6, pp. 3647,
Nov.Dec. 2008.
[11] S. McArthur, E. Davidson, V. Catterson, A. Dimeas, N. Hatziargyriou, F. Ponci, and T. Funabashi, Multi-agent systems for power
engineering applicationsPart I: concepts, approaches, and technical
challenges, IEEE Trans. Power. Syst., vol. 22, no. 4, pp. 17431752,
2007.
[12] U. Doe, Grid 2030: A national vision for electricitys second 100
years, 2003 [Online]. Available: http://www.climatevision.gov/sectors/electricpower/pdfs/electric_vision.pdf
[13] A. Nourai and C. Schafer, Changing the electricity game, IEEE
Power Energy Mag., vol. 7, no. 4, pp. 4247, 2009.
[14] P. Mercier, R. Cherkaoui, and A. Oudalov, Optimizing a battery
energy storage system for frequency control application in an isolated power system, IEEE Trans. Power. Syst., vol. 24, no. 3, pp.
14691477, 2009.
[15] D. Lee and L. Wang, Small-signal stability analysis of an autonomous
hybrid renewable energy power generation/energy storage system part
i: time-domain simulations, IEEE Trans. Energy Convers., vol. 23, no.
1, pp. 311320, 2008.
[16] K. Qian, Z. Li, C. Zhou, and Y. Yuan, Benefits of energy storage in
power systems with high level of intermittent generation, in Proc. 20th
Int. Conf. Exhib. Electr. Distrib.Part 1, 2009, pp. 14.
[17] Z. Yang, C. Shen, L. Zhang, M. Crow, and S. Atcitty, Integration of
a statcom and battery energy storage, IEEE Trans. Power. Syst., vol.
16, no. 2, pp. 254260, May 2001.

1038

[18] L. Zhang, F. Wang, Y. Liu, M. Ingram, S. Eckroad, and M. Crow,


Facts/ess allocation research for damping bulk power system low frequency oscillation, in Proc. IEEE 36th Power Electron. Specialists
Conf., 2005, pp. 24942500.
[19] A. Nourai, V. Kogan, and C. Schafer, Load leveling reduces t&d line
losses, IEEE Trans. Power. Del., vol. 23, no. 4, pp. 21682173, 2008.
[20] G. Koeppel, M. Geidl, and G. Andersson, Value of storage devices
in congestion constrained distribution networks, in Proc. Int. Conf.
Power Syst. Technol., 2004, vol. 1, pp. 624629.
[21] D. Staszesky, D. Craig, and C. Befus, Advanced feeder automation is
here, IEEE Power Energy Mag., vol. 3, no. 5, pp. 5663, 2005.
[22] A. Flueck and C. Nguyen, Integrating renewable and distributed
resourcesIIT perfect power smart grid prototype, in Proc. IEEE
Power Energy Soc. Gen. Meet., 2010, pp. 14.
[23] C. Nguyen and A. Flueck, Modeling of communication latency in
smart grid, in Proc. IEEE Power Energy Soc. Gen. Meet., 2011, pp.
16.
[24] A. Jones, White paper: The potential for energy storage in transmisson
and distribution systems, S&C Electr., pp. 16, May 2009.
[25] C. Nguyen, Power system voltage stability and agent based distribution automation in smart grid, Ph.D. dissertation, Illinois Institute of
Technology, Chicago, IL, 2011.
[26] B. Roberts, Capturing grid power, IEEE Power Energy Mag., vol. 7,
no. 4, pp. 3241, 2009.
[27] W. Kersting, Radial distribution test feeders, in Proc. IEEE Power
Eng. Soc. Winter Meet., 2001.
[28] FIPA communicative act library specification, 2002.

IEEE TRANSACTIONS ON SMART GRID, VOL. 3, NO. 2, JUNE 2012

Cuong P. Nguyen (S06M11) received the Ph.D.


degree in electrical engineering from the Illinois Institute of Technology, Chicago, in 2011.
He is currently an Application Engineer within
the Power Systems Software Group at GE Energy,
Schenectady, NY. His research interests include
agent-based distribution automation in smart grids,
modeling of smart grid components, power system
voltage stability and contingency ranking using
continuation methods, and large scale production
simulation software.

Alexander J. Flueck (S90M96SM09) received


the B.S., M.Eng., and the Ph.D. degrees in electrical
engineering from Cornell University, Ithaca, NY, in
1991, 1992, and 1996, respectively.
He is currently an Associate Professor at Illinois
Institute of Technology, Chicago. His research interests include autonomous agent applications to power
systems, transfer capability of large-scale electric
power systems, contingency screening of multiple
branch or generator outages with respect to voltage
collapse, transient stability, and parallel simulation
of power systems via message-passing techniques on distributed-memory
computer clusters.

Das könnte Ihnen auch gefallen