You are on page 1of 10

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication.

<

Multi Agent based Distributed Energy Resource


Management for Intelligent Microgrids
H. S. V. S. Kumar Nunna, Student Member, IEEE, and Suryanarayana Doolla, Member, IEEE

Abstract Microgrid is a combination of distributed


generators, storage systems and controllable loads connected to
low voltage network that can operate either in grid connected or
in island mode. High penetration of power at distribution level
creates such multiple microgrids. This paper proposes a two level
architecture for distributed energy resource management for
multiple microgrids using multi agent systems (MAS). In order to
match the buyers and sellers in the energy market, symmetrical
assignment problem based on nave auction algorithm is used.
The developed mechanism allows the pool members such as
generation agents, load agents, auction agents, grid agents and
storage agents to participate in market. Three different scenarios
are identified based on the supply-demand mismatch among the
participating microgrids. At the end of this paper two case
studies are presented with two and four interconnected
microgrids participating in market. Simulation results clearly
indicate that the agent based management is effective in resource
management among multiple microgrids economically and
profitably.
Index Terms Microgrid, Energy markets, Multi Agent
Systems (MAS), Symmetrical Assignment problem, Auction
algorithms

I. INTRODUCTION

ISTRIBUTED generation (DG) is becoming as a new


paradigm to produce on-site highly reliable and good
quality electrical power [1]. The concept of DG is quite
interesting when different kinds of energy resources are
available, such as photovoltaic (PV) panels, fuel cells (FCs),
or wind turbines. The DG of different kinds of energy systems
allows for the integration of renewable and nonconventional
energy resources into distribution system. Optimally sizing
and placing DGs in distribution system brings significant
reduction in distribution loss, congestion relief etc [2].
Microgrid is a platform to integrate DERs into distribution
network. The DERs may include DGs and distributed storage
(DS). Managing microgrid having wide variety of DERs is a
challenging task and the complexity involved is much higher
if non-dispatchable DERs like wind and PV are involved. The
Manuscript received on April 30, 2011. Accepted for publication on March
19, 2012.
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.
H. S. V. S. Kumar Nunna and Suryanarayana Doolla are with Department
of Energy Science and Engineering, Indian Institute of Technology Bombay,
Powai, Mumbai-400076, India. (Phone: +91 22- 2576- 7872; Fax: +91 222576-4890; email: suryad@iitb.ac.in)

recent literature has shown that making the existing


distribution grids smarter is possible with customer driven
microgrids by paying special attention to distributed storage
[3].
In general microgrids are provided with the intelligence to
form intentional island within the distribution system [4].
Intentional islands are formed when faults continue to persist
on the utility side, quality of power is poor on grid side or grid
prices are high. Besides the benefits resulted by integrating
DERs into distribution system, there are numerous economic,
commercial, and technical challenges to be faced [4]. In [5]
some of the technical challenges like active power
management were addressed with viable solutions.
One of the common challenging tasks involved in microgrid
scenario is managing microgrids with non-dispatchable DERs.
Due to the presence of non dispatchable generation, there is
always an issue of demand and supply in a microgrid. One of
the solutions to such problem is to equip microgrids with
diesel generators. However the operating cost and emission
levels of such systems are significantly high compared to
green DERs and hence, to make microgrids smarter it is
important to embed necessary intelligence into existing control
strategies [5-7]. This paper addresses a solution to the
aforementioned problem by implementing agent based virtual
market. The key intention of this mechanism is to enable
microgrids to participate in the simulated market and thereby
effectively utilize the DERs. The developed intelligence routes
the power from surplus locations to deficient locations.
However the cost at which power will be traded is decided by
pool members following the auction protocol discussed in
section IV.
In this paper MAS are used to model market scenario with
energy buyers and energy sellers. JADE (Java Application
DEevelopment framework), is used to develop the proposed
MAS architecture. The proposed architecture consists of
several intelligent agents at two levels viz., Market level and
Field level. The field level agents are responsible for
balancing local load and generation. In case of mismatch of
demand and supply at field level, the microgrid intelligent
agent (MIA) takes appropriate decision to participate in the
market either as a generator agent or as a load agent. When
MIA participating as load, it starts bidding for the amount of
power required to supply local load. MIA continues to do this
bidding till bidding price reaches maximum affordable price
or till it finds alternate best choice (say self owned diesel
generator). When MIA participates in the bidding as a
generator agent, it accepts the bids whose weighted average is

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

<
more than or equal to the value called as ARL (Acceptable
Revenue Level). The proposed approach is tested on a multibus microgrid and is proven to be beneficial for all the
participants.
Agent oriented programming approach of MAS and its
engineering applications are discussed briefly in the next
section. The proposed architecture for microgrid market is
presented in section III. Section IV details the implementation
of the proposed architecture using JADE. In section V, a
microgrid test case with two and four microgrids forming pool
is presented. Representative market scenarios for microgrids
are highlighted in Appendix
II. MULTI AGENT SYSTEMS (MAS) - APPLICATIONS OF AGENT
ORIENTED PROGRAMMING (AOP)
Multi agent system is a system with two or more agents or
intelligent agents or even combination of both [8]. An agent is
a software (or hardware) entity that is placed in some
environment and is able to autonomously react to the changes
in that environment. This nature of agent is called reactivity
[8]. This definition of agent doesnt give the total flavor of the
agent technology. The notion of agent is meant to be a tool
for analyzing systems, but not an absolute characterization that
divides the world into agents and no-agents is another
definition of an agent which gives the major flavor of an agent
[9]. An intelligent agent is an agent which exhibits pro-activity
(goal-directed behaviour), social ability (able to interact with
other intelligent agents) and reactivity [10].
The key features of agent technology viz., data and method
encapsulation made the agent orient programming (AOP) as a
best suit to develop secure applications. These features of an
agent provide security against accessing data stored in the
agent intelligence and methods (actions) which can perform
built in or user developed functions. Object oriented
programming can provide data encapsulation but not method
encapsulation. More precisely, in object orient programming
case external objects can call functions which the object has to
execute but no choice to escape or postpone. In case of AOP
an agent will have intelligence to schedule the requests sent by
other agents to take certain actions. In AOP, agent can be
accessed by other agents using standard messaging interfaces
[11].
MAS have been applied to wide range of engineering
applications. The application of MAS is to construct robust,
flexible and extensible systems or as a modeling approach.
MAS applications to power engineering problems are detailed
in [10]. The authors have discussed in detail the possible areas
where MAS can be applied which include diagnostics,
condition monitoring, power system restoration, market
simulation, system control, protection and automation. Recent
publications [12-29] on multi agent systems applications to
power engineering are significantly focused on distributed
energy scheduling, smart grids, power system topology
verification, energy management in hybrid systems, power
system contingency analysis, controlling and scheduling of
energy resources in hybrid energy systems.

2
In [14-15] MAS framework is used to enable mutual
coordination among various agents performing scheduling,
checking limit violations and clearing markets following
Contract Net Protocol (CNP). MAS were also applied in
simulation of energy markets using risk based continuous
double auction algorithm [16]. In [17] MAS were used to
coordinate microgrids to participate in ancillary service
markets (frequency control reserve markets). Ref [18] showed
the applicability of MAS to evaluate emission free electricity
markets by effectively coordinating green energy sources.
Apart from market simulation, MAS was also used to control
and incorporate necessary intelligence to detect faults on
utility feeders there by switching microgrid to island mode
[19]. In [20], MAS was used to implement CNP as mechanism
to control distribution feeder voltage by providing necessary
support from connected distributed generators. An attempt
was made to use MAS frame work to control and coordinate
storage systems of a microgrid operating in grid connected
mode [21]. Besides controlling storage systems MAS frame
work can also be extended to control hybrid renewable energy
generation by wind/PV [22-24].
MAS were also incorporated in time consuming tasks such
as contingency analysis, priority based distribution system
restoration [25-28]. In [29], MAS in connection with Petri
Nets were used to verify the power system topology during
real-time to avoid errors in the power network connectivity
models.
The market for power related commodities is usually
performed in intervals of 5, 15 and 60min [30].
Implementation of multi-agent systems for operation of a
microgrid is presented in [31-32]. The application confirms to
FIPA and is implemented using JADE software [33-34].
Auction algorithm to solve asymmetric assignment problem is
proposed in [35]. This uses idea of reverse auction where in
addition to agents bidding for objects by raising the prices new
agents offering discounts were also introduced which increase
competition and increase benefit to consumer. The work
reported in the literature so far focused on bidding process in
single microgrid while the present manuscript proposed
market scenario for load and generation agents multiple
microgrids with and without storage systems.
III. PROPOSED MICROGRID MARKET ARCHITECTURE
The proposed architecture has two layers/levels for
microgrid resource management viz., market level and field
level as shown in Fig. 1. The top level or market level is
designed to cater the trading among the microgrids and/or with
the grid while maintaining market fair and transparent and the
bottom level (field level) is responsible to manage individual
microgrids to match the supply and demand. The market layer
also detects and record successful negotiations among the
agents.
A. Resource Management at Field Level
A microgrid consisting of distributed generators (DG) and
loads constitute the field level of the architecture shown in
Fig. 1. The information regarding load requirement and

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

<

generation availability of an individual microgrid is available


with load agent (LA) and generation agent (GA) respectively.
This information is passed to these agents from field level load
and generation agents (FLA and FGA) of participating
generators and loads of the microgrids. FLA and FGA are
responsible for submission of this information to the microgrid
intelligent agent (MIA) via their respective brokers/trading
agents (LA/GA) during the discussion interval.
In addition to generation availability information, GA also
passes the Acceptable Revenue Level (ARL) for each
generator. ARL is defined as unit cost of energy that an
individual generator (or FGA) is willing to trade the power in
the market for.

Fig. 2: Gantt chart for market operation in microgrids.

MIA plays role of a manager when interacting with field


level agents and a trader when interacting with market level.
The mismatch between supply and demand is a common issue
encountered in a microgrid. It is more complicated in case of
systems with non-dispatchable generators. MIA shall execute
series of tasks to ensure proper trading and hence supplydemand balance in a microgrid.
MIA initiates market participation, when there is a supplydemand mismatch within the microgrid. The market level role
of MIA will be explained in the later section.
In any discussion period, upon receiving load requirement
and generation availability information from LA and GA of
the corresponding microgrid, MIA computes net power (NP)
available as
G

NP =

specified
i

i =1

Lspecified

(2)

Where, G is the total number of DGs connected to the


Fig.1: Proposed agent architecture for resource management in multiple
microgrids.

In a grid connected system with single microgrid, if the


generated power is higher than the load, the surplus power is
sold to the grid at buying price (BP) of grid. On the contrary,
if the load in the microgrid is higher than the generated power,
power is purchased from grid at selling price (SP). In the
present work, it is assumed that grid is always available to
export or import power and SP and BP are constant. However,
it is always possible to incorporate any constraints on
import/export of power in the proposed architecture at any
level. The relation between SP, BP and ARL in general is
given by (1)
(1)
BP < ARL SP
The market for power related commodities is usually
performed in intervals of 5, 15 and 60min [30]. In the present
work, a day (24 hrs) is divided into 96 blocks of 15min each
called as demand intervals/blocks and it is assumed that during
this interval the demand and supply remains constant. In the
proposed mechanism, commitment interval (B(n)) is preceded
by discussion interval (B(n-1)) by one block. Grid announces
the prices BP and SP at time just before B(n-1) and the MAA
(market administrator agent of market level) announces
bidding at B(n-1). The first three minutes of discussion
interval is known as discussion period/auction period for the
agents of microgrid. The loads and generators place their bids
through respective agents. Trading strategy implemented in
this paper is based on naive auction algorithm which is
discussed in section IV. The Gantt chart indicating discussion
interval, commitment interval and discussion period in the
trading process are given in Fig. 2.

microgrid, LSpecified is the aggregated demand of the microgrid


in kW, PiSpecified is the specified power availability from i

th

DG in kW. The immediate action of MIA depends on the sign


of the NP. If NP 0 then, MIA carries out optimization as per
(3).

Minimize C i Pi T

i =1
Subjected to the constraints

i =1

Specified
0 Pi Pi

P = L

Specified

(3)

Ci is the cost of generating power from i th DG in


cents/kWh, T is the duration of the commitment interval in
th
hours, and Pi is the amount of load optimally allotted to i
Where,

DG in kW. Based on (3), MIA optimizes the local distribution


of loads for available local generation. The next action of MIA
depends on the value of the NP computed using (2). If NP >0,
then MIA enters market as a generator otherwise it remains in
the same level. On the other hand, if NP is <0 then, MIA
enters the market as an aggregate load.
B. Trading Strategy at Market Level without Storage
The market level of the proposed architecture is responsible
for maintaining fair market and power management among
multiple microgrids. The agents involved in this layer are
market level load agent (MLLA), market level generation
agent (MLGA), auction agent (AA) and market administrator
agent (MAA). The hierarchy of these agents is shown in Fig. 1

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

<

As discussed before, MIA represents itself as load or


generator to the market level architecture. Bases on supply and
demand among the microgrids, the tasks performed by MIA
are narrated in the following three scenarios.
Scenario 1: When MIA represents itself as load to market
level.
MIA submits bid in the market with an initial value of
starting bid (SB). In the present developed market model, it is
considered that the starting bid by the MIA is always greater
than or equal to buying price of the grid (BP). If the bid value
is less than BP, the generators would be interested in trading
with main grid rather than participating in local market. The
step wise bidding process between MIA, AA and other market
agents is explained in section IV(C).
This bidding process continues progressively till bid reaches
a value called maximum affordable price or maximum bid
(bidmax), also bidmax SP. When MIA reaches bidmax, it
stops increasing bid and continues to bid with the same value.
However a bid reaches to bidmax only at the end of the
discussion period. At the end of the discussion period, MIA
remains either with a set of accepted bids or with a single
accepted bid. If the bid gets accepted, then MIA stores the
information and calculates unit price of electricity and
communicates to FLAs of the microgrid.
Scenario 2: When MIA represents itself as generator to
market level.
When NP > 0, MIA represents itself as a generator to the
market level. MIA submits the available generation
information to AA. The step wise bidding process among
MIA, AA and market agents is discussed in detail in section
IV(C).
Upon receiving offer from AA, MIA compares offer with
ARL calculated using (4).
G

ARL =

(ARL
i =1

MIAs, MLLA divides the consolidated load requirement into


packets of equal size (say 10 kW) and creates an agents called
market load agents (MLA) for each energy packet. The role of
MLA is to purchase energy packet from the market on behalf
of MIA participating as a load. The process is being used by
MLGA to create market generation agents. However, care
should be taken while selecting size of the each packet as each
packet represents an agent in the market and negotiation
involving large number of agents will introduce significant
delay [31]. MLLA and MLGA submit the information about
the newly created agents to AA. The number of market
generation and load agents reflects the supply demand relation
in the market. Upon receiving this information, AA starts
appending market grid agents (MGrdAs) in accordance to the
set of newly created MLAs and MGAs. This task of AA is
essential in order to apply solution of the symmetrical
assignment problem to the market simulation. The developed
market model follows solution of the symmetrical assignment
problem using naive auction algorithm.
C. Trading Strategy at Market Level with Storage
The proposed architecture can be extended to mange power
among microgrids having storage systems. In such scenario, a
new agent called buffer agent (BA) starts participating in filed
level as shown in Fig. 3. The trading mechanism of the agents
of the architecture shown in Fig. 3 is similar to that of Fig. 1
except in AA role. AA makes BAs to address the mismatch
between supply and demand primarily. If mismatch still exists
then AA communicates with grid.

(PiSpecified - Pi ))

(4)

(P

Specified
i

- Pi )

Fig.3: Agent architecture for resource management in multiple microgrids


with storage.

i =1

th

Where, ARL i is the ARL specified by i DG. If the offer


is greater than or equal to ARL, MIA accepts the offer.
Otherwise, it rejects and continues in the waiting state. MIA
follows this strategy till the end of the discussion period is
reached. If by any chance MIA has not received any offers
(just before end of auction period) which are acceptable then,
the most recently received offer is compared with BP of the
grid. If the offer is greater than BP then the best choice for
MIA is to accept instead of selling the power to grid at a lower
price.
Scenario 3: If the load and generation in an individual
microgrid are matched, MIA closes the discussion and stores
the set points of generators. These set points are released to
the generators during the commitment interval B(n). It is very
clear that MIA is acting as a bridge between the market and
field levels of the proposed architecture.
Next to MIA, market level contains MLLA and MLGA.
These are responsible for creating market load and generation
agents. After receiving market participation information from

The role of BA is to update AA with the status of the


storage system. BA is also responsible in maintaining storage
systems within the range (Smin, Smax). For example, if a battery
storage system is used then BA shall maintain state of charge
within the range (SOCmin, SOCmax).
BAs participation status is mainly decided by the supply
demand relation in market and filling/draining ability of the
storage systems. BAs calculate the ability (Pin) of the storage
systems to get filled as,
S(i 1)

S
, PMax
Pin (i) = min max
(5)
T

And draining ability (Pout) as,

S(i 1) Smin
, PMax
Pout (i) = min
(6)
T

Where,
Pin, Pout are the abilities of a storage system to get filled and
drained in ith commitment interval respectively, kW.

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

<

PMax is the rated capacity of the storage system, kW.


T is the duration of the commitment interval, hours.
S(i 1) = S (1 E ) is the status of a storage system during
(i-1)th commitment interval, kWh.
S is the status of the system at the end of the most recent
filling interval, kWh.
, is the self draining factor of the storage system, kWh per
month.
E is the elapsed time from the most recent filling interval to
(i-1), months.
If supply is more than demand then BAs participate in
market as loads and place bids in the market as per the
strategy given in section IV (C). If the demand is more than
the supply then BAs participate in market as generators by
submitting cost of stored energy (CSE) after accounting self
draining as ARL and is given by,
CSE Re
(7)
ARL (i ) =
(1 E )
Where, CSERe is the cost of the stored energy at the end of
the most recent filling interval in cents/kWh. The outcome of
the participation is a contract for charging or discharging in
the next commitment interval. If the contract is to fill then, the
cost of the stored energy (CSE) will be updated at the end of
the ith interval as follows,
CSE (i) =

(ARL(i) S(i 1)) + (CP(i) T CS(i))


S(i 1) + (CS(i) T)

(8)

accordance to the set of newly created MLAs, MGAs and


MSAs.

(a)

(b)
Fig. 4: (a) AA role in forming symmetrical assignment problem without
storage and (b) with storage

Fig. 4 shows the role of AA in forming equal number of


suppliers and buyers in the market. This task of AA is
essential in order to apply solution of the symmetrical
assignment problem to the market simulation. In the
developed market model naive auction algorithm is used.
Pools of agents which are actually participating in the trade
are shown in Fig. 5. Another important role of AA is to
negotiate with all the members of the pool.

Where,
S(i 1) = S(i 1) S min
CP is the contract price in cents per kWh.
CS is the contract size in kW.
CSE (i-1) is the cost of stored energy in cents per kWh.
The microgrids considered in this study are community level
microgrids and hence all the storage agents work under
nonprofit mode. In case of customer driven microgrids, the
storage systems are profit motivated and hence sell energy at a
price greater than CSE. The roles of intelligent agents other
than AA of market level are same as before. If the number of
market generation agents is not equal to the number of market
load agents then AA invites BAs to participate in the market.
BAs acknowledge with the status of the storage systems. AA
decides the number of market storage agents (NB) to be added
in the market as,
B

P i=1SB(i)
,
NB = min

AS AS

(9)

Where,
P is supply-demand mismatch in the market, kW.
AS is Agent size, kW.
B is the total number of storage systems
SB(i) is the ability of the ith storage system in kW, and is
equal to Pin for charging and Pout for discharging.
Upon adding NB number of market storage agents (MSAs),
AA rechecks the mismatch, and if the mismatch is non zero
then AA starts appending market grid agents (MGrdAs) in

Fig. 5: Pool of trading agents.

The role of Market Administrator Agent (MAA) is to


announce start of auction, discussion intervals, prepare
monthly reports, negotiate with utility grids, recording
successful bids, and passing on stored set points to each MIA.
IV. IMPLEMENTATION OF MICROGRID BIDDING STRATEGY
USING JADE
JADE (Java Agent Development framework) is a software
development framework for developing MAS and applications
conforming to FIPA (Foundation for Intelligent Physical
Agents) standards for intelligent agents [33-34]. The
developed MAS based architecture enables individual
microgrids to practice power trading which is based on Naive
Auction Algorithm applied as a solution for symmetrical
assignment problem [35-36]. However to adopt this solution,
there should be equal number of buyers and sellers in the
market.

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

<

A. Naive Auction Algorithm applied as a solution for


symmetrical assignment problem [31], [35-36]
The naive auction algorithm progresses in rounds and in
each round only one buyer (also called a person) bids on the
desired object. The mechanism involved in naive auction
algorithm is given below:
In the classical example (assignment problem) there are X
objects and Y persons which are to be matched on a one-toone basis. The benefit associated with matching a person to an
object is defined as utility, u(i, j). The objective of assignment
problem is to find and assignment (a set of person-object
pairs) that maximizes total benefit given by (10):

Total Benefit =

i =1

u(i, j)

(10)

In real life auction, there is always a price pj associated with


each object, where j is the object index. However, in the
current application, price is an algorithmic variable which has
no physical significance. The price pj helps to progress and
terminate the naive auction algorithm. The actual value gained
by a person allotted with an object is the difference between
utility and the price of the object.
As discussed earlier, naive auction algorithm proceeds in
rounds or iterations and terminates when all objects and
persons are matched. In any round only one person (or the
bidder) can submit bid on the desired object. In each round,
the bidder increases the price of the desired object by a value
called bid increment. Bid increment b(i,j) is the increment in
bid by person i for the object j and hence price indicates
global desire. The bid increment is calculated by using (11).

b(i, j) = {u(i, j) p j} {u(i, j* ) p j*}

(11)

Where, j is the next best desired object for a person i.


This process resembles the real auction as bidding increments
and price increases aid competition by making the bidders
desired object less attractive to other competitors.
However, a buyer can have more than one desired object
which will yield same maximum profit. In such case, the
auction algorithm results a non terminating quarrel among
bidders as no bid increment is possible. To avoid such infinite
auction rounds, a perturbation mechanism was introduced.
This mechanism demands that in each round, a bidder should
increase the bid at least by a non negative value . In such
case, just as in real auction objects become less attractive after
finite number of rounds resulting a solution. The value of
influences the number of steps required to converge. A more
detailed explanation of naive auction algorithm and its
application for microgrid control is given in [31].
B. Designing of Agents
In order to simulate the behaviours of the various entities of
the market, a total of twelve agents are created and are broadly
categorized as,
Trading Agents: FLA, FGA, BA, and GRID
Trading brokers: LA, GA, MLLA, and MLGA
Auctioneer Agents: AA
Administrative Agents: MIA and MAA
Marketing Agent: MGA, MLA, MSA, and
MGrdAs.

The intelligence developed for each of the above agents is an


extension of the behaviours available in JADE package like
OneShotBbehaviour, SequentialBehaviour, ParallelBehaviour,
and TickerBehaviour. The following are some of the
developed behaviours:
StartAuction
MarketParticipation
FormSymAssignement
StartNegotaiation
StartBidding
UpdateStatus
Reception
In this application content of the messages exchanged between agents
is set using setContentObject() method. However, the message
communication can also be done by developing ontology as
discussed in [31].

C. Implementation of proposed architecture on multiple


microgrids
The steps involved in overall market simulation are given
below:
Step 1: Announcement of discussion period:
MAA announces start of the discussion period, SP, and BP
Step 2: Load and generation data
FLAs and FGAs of each microgrid submits the information
about load requirement and generation availability for the
commitment period to LA and GA which in turn consolidate
perceived information and submit to corresponding MIA.
Step 3: Decision for role of MIA
MIA traces out the cheapest way of supplying received load
demand by availing local generation using (2) and (3). The
local generation can be more or less than the local load
requirement. If surplus power/load is available in a microgrid
then corresponding MIA initiates market participation by
sending request either MLGA/MLLA.
Step 4: Creating market agents by AA
MLLA and MLGA consolidate the received data and create
MLAs and MGAs, and submit this information to AA.
AA checks the number of MLAs and MGAs and if they are
not equal, then AA appends suitable number of market grid
agents (MGrdAs) to make the market symmetrical. If storage
systems are present then AA appends suitable number of
MSAs as per (8). AA announces start of negotiation to all pool
members and conducts the auction as per naive auction
protocol. The outcome of auction algorithm is an assignment
between buyers and sellers.
Step 5: Weighted average for the bids
AA calculates the offer to be informed to MIA participating
as generator (MIAGenerator) by taking weighted average of the
bids submitted by the agents which are assigned to
corresponding MIAGenerator.
Step 6: Market negotiation by AA
AA negotiates the offer with all MIAs acting as generators.
Based on their ARL they may accept or reject the offer. The
ARL is computed as per (4). If the offer is acceptable to
MIAGenerator, then the algorithm terminates and Step 8 follows.
If the offer is not accepted by the MIAGenerator, then AA sends
message to MIA acting as load (MIALoad) to submit new bids
as per (12) and algorithm directs to Step 4. The bidding curve
shown in Fig. 6 is a time based bidding curve which can
accurately model the behavior of true buyer in the market.

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

<

Buyer in the market starts bidding on the desired object with a


value greater than the existing offer made by other
competitors, in order to win the bid.
bidmax SB
(12)
bid(t) = (
) t + SB
Td

20kW which it had purchased from MG-1 at 12.37 cents/kWh.


The excess power of 10 kW of MG-1 is sold to the main grid 9
cents /kWh. In block five and six, power is imported and
exported from the grid respectively by the MG-1 and MG-2.
In block seven and eight, power is exported and imported from
the grid by both the microgrids respectively in the absence of
storage systems.
TABLE 1(A). CASE STUDY SYSTEM DATA FOR TWO MICROGRIDS AND
MARKET SIMULATION WITHOUT STORAGE

Blk

Fig. 6. Bidding curve

Step 7: Communicating Success


AA communicates the success of the bid to corresponding
members who submitted the bid i.e., MIAs acting as load.
Step 8: End of discussion period
At the end of the discussion period MAA announces stop
discussion and records all successful bids.
In order to evaluate the effectiveness of the proposed MASbased resource management, a system with two interconnected
microgrids shown in Fig. 7 is simulated. In order to show the
applicability of the proposed architecture, the system shown in
Fig. 7 is simulated by introducing storage system of capacity
40 kWh in each microgrid. The initial status of these systems
is 50% of their rated capacity at a CSE of 11.25 cents/kWh.
The maximum allowable depth of draining and self draining
factors of each storage system are 5 kWh and 25% of the rated
capacity per month respectively. Details of loads and
generations in the microgrids for various intervals (blocks) are
given in Table. 1(A).

MG

7
8

Blk
Fig. 7. System with two interconnected microgrids.

In Table 1(A), an export of 20 kW at 9 cents/kWh (buying


price of the grid, BP) to the main grid by the MG1 is observed
in the first block. It is also observed that MG2 has purchased
30 kW at 12.76 cents/kWh from MG1 during this period based
on naive auction algorithm. Detailed process involved in
arriving at the final bid value by MG-2 is explained in
Appendix. In the second block, a surplus power of 30 kW
available from the MG1 which is purchased by MG2 (after the
bidding process) at 11.25 cents/kWh. In the case of third
interval, MG-2 requires additional 70 kW power and hence it
has purchased 50 kW from MG-1 at 11.25 cents/kWh and
remaining 20 kW from main grid at 13.5 cents/kWh which is
selling price (SP) of the grid.
It is observed that in block four, the generation in MG-1 is
reduced by 20 kW and MG-2 requires additional power of

L2

DG1*
(ARL)

DG2*
(ARL)

1
2
3
4
5
6
7
8

100

MG
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2

50

100
(10.13)
100
(10.13)
100
(10.13)
100
(10.13)
100
(10.13)
100
(10.13)
100
(10.13)
100
(10.13)
100
(10.13)
100
(10.13)
100
(10.13)

Export/
Import
(Cost)#
+50
(11.25)

Grid
Power
(Cost)#
-20
(9)

100
(11.25)
100
2
100
130
-30 (12.76)
0
(10.13)
100
+30
1
100
70
0
(11.25)
(11.25)
100
2
100
130
-30 (11.25)
0
(10.13)
100
+50
1
100
50
0
(11.25)
(11.25)
100
-70
+20
2
100
170
(10.13)
(11.89)
(13.5)
80
+30
-10
1
100
50
(11.25)
(11.25)
(9)
100
2
100
120
-20 (12.37)
0
(10.13)
100
+50
1
100
50
0
(11.25)
(9)
100
+10
2
100
90
0
(10.13)
(9)
80
1
100
100
0
-20 (13.5)
(11.25)
100
2
100
120 70 (10.13)
0
-50 (13.5)
(10.13)
100
100
+50
1
100
50
0
(10.13)
(11.25)
(9.0)
100
100
+50
2
100
50
0
(10.13)
(10.13)
(9.0)
100
40
-60
1
100
100
0
(10.13)
(11.25)
(13.5)
50
100
-50
2
100
100
0
(10.13)
(10.13)
(13.5)
*: Units for load and generation are in kW, ARL in cents/kWh
#: - sign indicates exporting to grid and + sign indicate importing
from grid
TABLE 1(B). MARKET SIMULATION WITH STORAGE
1

L1

Surplus
(Cost)$
+50/(11.25)
-30/(11.25)
+30/(11.25)
-30/(11.25)
+50/(11.25)
-70/(11.25)
+30/(11.25)
-20/(11.25)
+50/(11.25)
+10/(10.13)
-20/(11.16)
-50/(11.21)
+50/(11.25)
+50/(10.13)
-60/(11.95)
-50/(11.27)

Storage
(Cost)*$
0
-20/(11.25)
0
0
0
+20/(11.25)
0
-10/(11.25)
-40/(10.97)
-20/(11.25)
+40/(11.16)
+30/(11.25)
-40/(11.11)
-40/(11.11)
+40/(11.14)
+40/(11.20)

Grid Power
(Cost)#$
0
0
0
0
0
0
0
0
0
0
0
0
10/(9.0)
10/(9.0)
+20/(13.5)
+10/(13.5)

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

<

8
#: - sign indicates exporting to grid and + sign indicate importing
from grid;
*: + sign indicates draining and - sign indicate filling
$: Units for cost are in cents/kWh

In Table 1(B), no export to or import from the main grid is


observed in the first six blocks. In blocks seven and eight,
available power is surplus/deficit when compared to
filling/draining capacity of the storage systems and hence an
export/import of 20kW/30kW is observed. The advantage of
using storage (based bidding) can be clearly understood by
comparing block 1 of Table 1(A) and 1(B). The total cost of
purchase by MG2 is reduced to 11.25 cents/kWh from 12.76
cents/kWh when local community based storage is introduced.
Fig. 8 depicts snapshot of the actual communication taking
place between the agents for one transaction period (block) of
15 minutes. This snapshot was obtained using a sniffer agent
in JADE software.

(a)

Fig. 8. (a) Communication and (b) Sniffer diagrams for market negotiation for
two inter connected microgrids.

The communication shown between MIAs and BAs and the


dotted lines in Fig. 8 correspond to storage systems. The six
blocks (15 min each) illustrate the performance of the
proposed scheme under different load, generation profiles and
market condition. Market simulation is also performed on
system with four inter connected microgrids with storage
systems shown in Fig. 9 and the results are tabulated in Table
2. The proposed architecture and market mechanism can easily
be extended for multiple microgrids.
Note that, the microgrids shown in Fig. 9 except microgrid 4
are having battery energy storage systems (BESSs). BESS is
assumed to be storage system for this case study. These
systems are rated at a capacity of 40kWh and are initially
charged to 50% of their rated capacity at a CES of 11.25
cents/kWh. The maximum allowable depth of draining and
self draining factors of each BESS are 5 kWh and 25% of the
rated capacity per month respectively.
In Table 2, some of the important scenarios of the system
shown in Fig. 9 are presented. It is observed that, during first,
second and fourth blocks, storage systems are filled by the
surplus power available in the market.
In block three, five and six storage systems are supplying
deficit power. In block four, deficit is more than the capacities
of the storage systems and hence an import of 10 kW from
main grid at SP (13.5 cents/kWh) is observed. In block five,
the surplus power is more than the capacities of the storage
systems and hence an export of 30 kW to main grid at BP (9.0
cents/kWh) is observed. The simulation setup for
interconnected microgrid with the central market agents is
shown in Fig. 10.
The results show that the proposed architecture based on
naive auction protocol performs the resource management as
desired under all conditions.
Utility Substation
DG1
(100 kW)

Microgrid-2

Large Load

L1

Microgrid-4

Z2
Z1

DG1
(100 kW)

DG2
(100 kW)
DG1
(100 kW)

L2

Microgrid-3
L2

L1
DG1
(100 kW)

BESS

BESS

L1

L2

DG2
(100 kW)

BESS

L2
Z3

DG2
(100 kW)
Microgrid-1

L1
DG2
(100 kW)

Fig. 9. System with four interconnected microgrids with BESS.

DG2a/ARL

100/10.13
100/11.25
100/10.13
100/10.13

80/10.13
100/11.25
100/11.25
80/11.25

20/11.25
0
+30/11.25 30/11.25
+20/11.25
0
0

Grid
Powera,b
(Cost)

DG1a/ARL

100
70
80
80

Storagea,*
(Cost)

L2

100
100
100
100

(Cost)a,b

L1

1
2
3
4

Export/Import

MG

(b)

Block

TABLE 2. CASE STUDY SYSTEM DATA FOR FOUR MICROGRID SYSTEM

0
0
0
0

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

<

9
0/10.13 +20/10.13 10/11.25
0
1
100
800 100/10.13 100
2
100
700 100/11.25 100
0/11.25 +30/11.25 40/11.25
0
3
100
800 100/10.13 100
0/11.25 +20/11.25 40/10.69
0
4
100
600 100/10.13 80
0/11.25 +20/11.25

0
1
100
1000 100/10.13 80
0/10.13 20/11.25
0
0
2
100
800 100/11.25 100
0
0/11.25 +20/11.25 +10/11.25
3
3
100
1000 100/10.13 70
0/11.25 30/11.25 +40/11.06
0
4
100
1000 100/10.13 80
0/11.25 20/11.25

0
1
100
1000 100/10.13 50
0/10.13 50/11.25 +40/11.25
0
2
100
1000 80/11.25
0/11.25 40/11.11 +40/11.25
0
60
4
3
100
1000 100/10.13 60
+100
0/11.25 40/11.77 +40/11.06
4
100
1000 100/10.13 100
0/11.25
0

0
1
100
500 100/10.13 100
0/10.13 +50/10.13 40/11.25
0
2
100
600 100/11.25 100
0/11.25 +40/11.25 40/10.41
0
5
3
100
600 100/10.13 100
0/11.25 +40/11.25 40/10.69
100
4
100
800 100/10.13 100
0/11.25
+20/9.0

200
1
100
1000 100/10.13 100
0/10.13
0
0
0
2
100
1000 70/11.25 100
0/11.25
30
+10/11.01
0
6
3
100
800 100/10.13 80
0/11.25
0
+40/10.88
0
4
100
1000 100/10.13 80
0/11.25
20

0
a: Units for load, genneration, storage, and
d export/import are inn kW, ARL in
c
cents/kWh;
b sign indicates export
b:
e
to grid at BP and
a + sign indicates import
i
from grid at SP;
S
*: + sign indicates draining
d
and sign in
ndicates filling
2

F 10. Simulation setup for interco


Fig.
onnected microgridds with Central Aggent
Maarket

V. CON
NCLUSION
In this paper a tw
wo level archiitecture for disstributed resouurce
maanagement in multiple inteer connected microgrids
m
using
MA
AS is presenteed. The markeet agents parrticipate in real
r
tim
me bidding based
b
on data from agentss at field levvel.
Biddding action iss simulated usin
ng naive auctioon algorithm. The
T
inttelligent agentt of each micrrogrid following the proposed
moodel behaves like a true buyer or truue seller in the
maarket so as to
t fulfill the needs of thhe microgrid. In
ordder to even the
t
number of
o load and generation
g
ageents
parrticipating in the trading the conceptt of symmetriical
asssignment is ussed. The resultts of case studyy for two and four
fo
microgrids clearrly indicate that
t
proposedd architecture is
efffective in resouurce managemeent among mulltiple microgridds.
APPEENDIX
In order to give
g
full insig
ght of the maarket mechaniism
prooposed, a systtem with two microgrids is considered. The
T
sysstem data conssidered for a bllock (period off 15 min) is givven
in Table A1. In the
t beginning of an auction period local looad
his informationn to LA and GA
G
andd generator aggents submit th
resspectively. As per Table A1,
A MG-1 is having
h
a surpplus
pow
wer of 50 kW
W, while MG-2 is having deficit
d
of 30 kW.
k
Heence the intelliggent agent of MG-1(MIA
M
epresents itselff as
1) re
gennerator and inntelligent agen
nt of MG-2 (M
MIA2) represeents
itseelf as load in the market. The
T ARL for MIA
M 1 as a sinngle
enttity is calculateed from (4).

ARL
L=

0 10.13 + 50
5 11.25
50

= 11.25 cents/kWhh

TABLE AI
SYSTEM
Y
DATA FOR MICROGRIDS UNDEER CONSIDERATIO
ON BLOCK1

MG-11
(ARL))*
MG-22
(ARL))*

Load 1
(kW)
100

Load 2
L
(kW)
50

100

130

DG-1
(kW)
100
(10.13)
100
(10.13)

DG-2
(kW)
100
(11.25)
100
(10.13)

NP
(kW)
+50
-30

**Units for load andd generation are inn kW, ARL in centts/kWh
# + sign indicates export
e
and sign indicate
i
import

However, it is obsserved that AR


RL for DG-1 < DG-2 and
t power gennerated by DG
G with low ARL shall be
hence the
utilizedd locally and rest
r
shall be made
m
available for trading.
For thiss reason, the suurplus power of
o 50 kW is avvailable from
MIA1 at
a 11.25 cents//kWh for the next
n
commitm
ment interval.
The NP
P in MG-2 is -30 kW and heence MIA2 reprresents itself
as an agggregate load to the market. The initial biid submitted
by MIA
A2 is 9.1 centts/kWh. The information
i
received from
MIA1 and
a
MIA2 by MLLA and MLGA is useed to create
market load and generration agents (M
MLAs and MG
GAs).
In thhe present studdy, the size of agent is consiidered as 10
kW andd hence MLLA
A creates three load agents and MLGA
creates five generatinng agents and passes the infformation to
mmetric assignnment problem discussed in
AA. Baased on the sym
section IV, AA creeates two new
w marketing grid agents
A) and initiates negotiation to
t all market agents
a
as per
(MGrdA
naive auction
a
algorithhm. The pricee submitted byy MGrdA is
equal too buying price (BP) of the griid which is fixxed at 9 cents
/kWh.
Afterr negotiation, AA calculatess the weightedd average of
the bidss and communnicates to MIA
A1. The computted weighted
averagee in the first iteeration is 9.06 cents/kWh. Thhis offer may
be acceepted or rejectted by MIA1 based
b
on its ARL.
A
For the
present block 9.06 cents/kWh iss less than ARL
A
(11.25
Wh) and hencce MIA1 rejeccts the offer. AA informs
cents/kW
MGrdA
As and MIA2 to
t submit new bid. As discuussed before,
the buyying price (BP)) of grid remaiins constant (99 cents/kWh)
in the present
p
contextt and there is no
n change in bid submitted
by MG
GrdAs. MIA2 shall
s
submit a new bid and the bidding
process repeats and iss continued till the offer is acceptable
a
to
MIA1. In the presentt example, whhen the bid submitted by
MIA2 is
i 12.76 cents//kWh, the weighted averagee of bids as
computted by AA is obtained as 11.25
1
cents/kW
Wh which is
equal too ARL of MIA
A1 and hence thhe new bid is acceptable
a
to
MIA1. It
I is to be noteed that the bid submitted by MIA2 is less
than thee bidmax or sellling price (SP)) of the grid.
MAA
A records this event and passes on seet points to
particippating MIAs. During
D
the coommitment intterval MIAs
commuunicate the obbtained set pooints to local load agents
throughh traders.
REFERENCEES
[1] J. M.
M Guerrero, F. Blaabjerg, T. Zheleev, K. Hemmes, E. Monmasson,
S. Jemei,
J
M. P. Com
mech, R. Granaddino, and J.I. Fraau, Distributed
geneeration: Toward a new energy paraddigm , IEEE Ind. Electron. Mag.,
pp. 52-64, Mar. 2010..

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

<

10

[2] Y. M. Atwa, E.F.


E
El-Saadany, M.M.A. Salama, and R. Seethapaathy,

[23] X. Zhou,
Z
Y. Gu, Y. Ma,
M L. Cui, and S.
S Liu, Hybrid opperation control

Optimal renew
wable resources mix
m for distributionn system energy loss
minimization, IEEE Trans. Pow
wer Syst., vol. 25, no. 1, pp. 360-3370,
Feb. 2010.
S. Suryanarayannan, F. Mancilla-D
David, J. Mitra, Y.
Y Li, Achieving the
smart grid throough customer-drriven microgrids supported by eneergy
storage, in Prroc. IEEE Int. Co
onf. Industrial Technology,
T
2010, pp.
884-890.
N. D. Hatziaargyriou, H. Assano, R. Iravanii, and C. Marnnay,
Microgrids, IEEE
I
Power & En
nergy Mag., vol. 5,, no. 4, 2007.
A. Timbus, M.
M Larsson, and C. Yuen, Acttive Managementt of
Distributed Eneergy Resources Ussing Standardized Communications and
Modern Inform
mation Technologiees, IEEE Trans. Ind.
In Electron., vol.. 56,
no. 10, pp. 40299-4037, Oct. 2009.
J. Vasquez, J. M. Guerrero, J. Miret, M. Castillla, and L. Garcia De
o intelligent miccrogrids, IEEE Ind.
I
Vicuna, Hieraarchical control of
Electron. Mag.,, pp. 23-29, Dec. 2010.
2
S. Chakraborty,, M. D. Weiss, and
d M. G. Simes, D
Distributed intelliggent
energy manageement system fo
or a single-phasee highfrequency AC
microgrid, IEE
EE Trans. Ind. Eleectron., vol. 54, no.
n 1, pp.97-109, Feb.
F
2007.
M. Wooldridgee, , G. Weiss, Ed
d., Intelligent Aggents, in Multi-aggent
Systems. Cambrridge, MA: MIT Press,
P
Apr. 1999, pp. 351.
S. Russell and P. Norvig, Artificcial Intelligence: A Modern Approaach.
Hall, 1995.
Englewood Clifffs, NJ: Prentice-H
S. D. J. McArthhur, E. M. Davidsson, V. M. Cattersson, A. L. Dimeass, N.
D. Hatziargyrioou, F. Ponci, and T.
T Funabashi, Muulti-Agent Systemss for
Power Engineeering ApplicationssPart I: Conceppts, Approaches, and
Technical Challlenges, IEEE Trrans. Power Syst.., vol. 22, no. 4, pp.
1743- 1752, Noov. 2007.
S. D. J. McArthhur, E. M. Davidson, V. M. Cattersson, A. L. Dimeass, N.
D. Hatziargyrioou, F. Ponci, and T.
T Funabashi, Muulti-Agent Systemss for
Power Engineerring Applications
Part II: Technollogies, Standards, and
Tools for Buildding Multi-agent Systems, IEEE Trrans. Power Syst., vol.
22, no. 4, pp. 17753- 1759, Nov. 2007.
Li, G. Poulton, G. James, A. Zem
man, P. Wang, M. Chadwick, and M.
M R.
Performance of Multi-agent
M
Coordiination of Distribuuted
Piraveenan, P
Energy Resourcces, in Proc. WSE
EAS Int. Conf. Computer
C
Engineerring
and Applicationns, Gold Coast, Au
ustralia, Jan. 17-199, 2007, pp. 568-5774.
Z. Jiang, Aggent based contro
ol framework foor distributed eneergy
resources microogrids in Proc. IEEE/WIC/ACM Int. Conf. Intelliggent
Agent Technoloogy, Hong Kong, Dec.
D 18-22, 2006, pp.646-652.
p
T. Logenthirann, D. Srinivasan
n, A.M. Khambaadkone, Multi-aggent
system for eneergy resource sch
heduling of integrrated microgrids in
i a
distributed systtem, Electric Pow
wer Systems Research, vol. 81, noo. 1,
pp. 138-148, 20011.
T. Logenthiran,, D. Srinivasan, D.
D Wong, Multi-aagent coordinationn for
DER in microgrid, in Proc. 200
08 IEEE Int. Conff. Sustainable Eneergy
I
2008, pp. 77
7-82.
Technologies, ICSET
B. Ramachandrran, S. K. Srivastaava, C. S. Edringtoon, and D. A. Carrtes,
An intelligentt auction scheme for smart grid market
m
using a hybbrid
immune algoritthm, IEEE Trans. Ind. Electron., vol. 58, no. 10, pp.
46034612,Octt. 2011.
C. Yuen, A. Oudalov,
O
and A. Timbus, "The prrovision of frequeency
control reservess from multiple miicro-grids, IEEE Trans. Ind. Electrron.,
vol. 58, no. 1, pp.
p 173-183, Jan. 2011.
2
K. Sichao, H. Yamamoto, and K. Yamaji, Evaaluation of CO2 free
electricity tradinng market in Japaan by multi-agent simulations, Eneergy
Policy, vol.38, no.7,
n
pp. 3309-331
19, Jul. 2010.
M. Pipattanasom
mporn, H. Feroze, and S. Rahman, Multi-Agent

Systtems
in a Distributeed Smart Grid: Design
D
and Impleementation, in Proc.
P
IEEE PES 2009 Power System
ms Conf. and Exxposition (PSCE09),
Seattle, Washinngton. Mar. 2009, pp. 1-8.
M.E. Baran, I.M
M. El-Markabi, A
A multiagent-based dispatching scheeme
for distributed generators for vo
oltage support on distribution feedeers,
IEEE Trans. Poower Syst., vol.22,, no. 1, pp. 52-59, Feb. 2007.
J. Lagorse, D. Paire, and A. Mirraoui, Multi-agennt system for eneergy
wable Energy, vol. 35
management off distributed poweer sources, Renew
no. 1, pp. 174-1182, April 2010.
R. Roche, B. Bllunier , A. Miraouii, V. Hilaire, and A. Koukam, Muultiagent systems for
f grid energy management:
m
A shhort review, in Proc.
P
Ind. Electron. Conf.
C
, 2010, pp. 3341-3346.
3

metthod for micro-grid based on MAS, in Proc. 2010 IEEE


IE
Int. Conf.
Proggress in Informatiics and Computingg, PIC-2010, vol.1, pp. 72-75.
Z. Jun,
J
L. Junfeng, W.
W Jie, and H.W. Ngan,
N
A multi-aggent solution to
enerrgy management in hybrid renewaable energy generation system,
Rennewable Energy, vool. 36, no.5, pp. 13352-1363, May 20011.
A. Oonsivilai,
O
and K.. A. Greyson, Poower system continngency analysis
usinng multiagent syystems, in Procc. World Academ
my of Science,
Enggineering and Techhnology , 2009,pp.. 355-360.
P. Li,
L B. Song, W. Wang, and T. Wang,
W
Multi-agennt approach for
servvice restoration of microgrid, in Prooc. 5th IEEE Conff. Ind. Electron.
andd Applications, ICIE
IEA 2010 , pp. 9622-966.
C.-H
H. Lin, C.-S. Chenn, T.-T. Ku, C.-T. Tsai, C.-Y. Ho, Fault

detection,
isolaation and restooration using a multiagent-baseed distribution
autoomation system, 2009 4th IEE
EE Conf. Ind. Electron.
E
and
Appplications, ICIEA 2009
2
,vol.21, no.1,, pp. 2528-2533
P. Li,
L B. Song, W. Wang, and T. Wang,
W
Multi-agennt approach for
servvice restoration of microgrid, in Prroc. 5th IEEE Conff. Ind. Electron.
andd Applications, ICIE
IEA 2010 , pp. 9622-966.
K. Wilkosz,
W
A multii-agent system appproach to power system
s
topology
veriification with use of Petri Nets, in
i Proc. Modern Electric Power
Systtems 2010, Wroclaaw, Poland.
Elecctricity market [Onnline]. Available: http://en.wikipedia
h
a.org/wiki/
A. L.
L Dimeas, and N.
N D. Hatziargyriiou, Operation of
o a multiagent
systtem for microgrid control, IEEE Trans.
T
Power Syst., vol. 20, no. 3,
pp. 1447-1455, Aug. 2005.
2
H.S.V.S.K. Nunna, and S. Doolla Demand respoonse in smart
microgrids, in Prooc. 2011 IEEE PES Innovativee Smart Grid
(
India), pp. 131-136,
1
1-3 Dec. 2011.
Techhnologies - India (ISGT
Javaa Agent DEvvelopment fram
mework [Onlinee]. Available:
httpp://jade.tilab.com/
Fouundation of Intelliggent Physical Agennts (FIPA) [Onlinee]. Available:
httpp://www.fipa.org
D. P. Bertsekas, annd D. A. Castaoon, A forward/rreverse auction
mmetric assignmennt problems,
algoorithm for asym
Computational
Opttimization and Appplications, vol. 1 , no. 3, pp. 277-2977C, 1992.
Y. Shoham, and K. Leyton-Brown. (2011, April 288). Mutiagent
Systtems Apllications, Game-Theoreetic, and Logicaal Foundations
[Onnline]. Available: http://masfoundati
h
ions.org

[3]

[4]
[5]

[6]
[7]

[8]
[9]
[100]

[111]

[122]

[133]
[144]

[155]
[166]

[177]
[188]
[199]

[200]
[211]
[222]

[24]
[25]
[26]
[27]

[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]

V S. Kumar Nun
nna (S12) receivved the M.Tech
H. S. V.
degree in
i electrical enginneering from Natioonal Institute of
Technology Calicut, Keraala, India in 2010. He is currently
ment of Energy
pursuingg the PhD degreee in the Departm
Science and Engineering,, Indian Institute of Technology
Bombayy, India. His reesearch interests are in smart
a demand side management.
m
microgridds, multi agent system applications, and

Suryanarrayana Doolla (S04M07) receivved PhD degree


from Indian Institute of Technology
T
Delhii in 2007, and
d
in Energyy systems and enngineering from
M.Tech degree
India andd Indian Institute of
o Technology Boombay, India in
2002. He is currently workking as an Assistaant Professor in
Departmeent of Energy Sccience and Enginneering, Indian
Institute of Technology Bombay, India. His research
interests are
a in smart microogrids, communicaation protocol for power systems,
and gridd integration of distributed energgy resources andd demand side
managem
ment.

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

You might also like