Energy and Power Engineering, 2013, 5, 1330-1336
doi:10.4236/epe.2013.54B252 Published Online July 2013 (http://www.scirp.org/journal/epe)
A Multi-Agent Framework for Operation of a Smart Grid
Ruchi Gupta1, Deependra Kumar Jha1, Vinod Kumar Yadav1, Sanjeev Kumar2
1School of Electrical, Communication and Electronics Engineering, Galgotias University, Uttar Pradesh, India
2School of Basic and Applied Sciences, Department of Mathematics, Galgotias University, Uttar Pradesh, India
Received March, 2013
ABSTRACT
This paper presents the operation of a Multi-agent system (MAS) for the control of a smart grid. The proposed Mul-
ti-agent system consists of seven types of agents: Single Smart Grid Controller (SGC), Load Agents (LAGs), a Wind
Turbine Agent (WTAG), Photo-Voltaic Agents (PVAGs), a Micro-Hydro Turbine Agent (MHTAG), Diesel Agents
(DGAGs) and a Battery Agent (BAG). In a smart grid LAGs act as consumers or buyers, WTAG, PVAGs, MHTAG &
DGAGs acts as producers or sellers and BAG act as producer/consumer or seller/buyer. The paper demonstrates the use
of a Multi-agent system to control the smart grid in a simulated environment. In order to validate the performance of the
proposed system, it has been applied to a simple model system with different time zone i.e. day time and night time and
when power is available from the grid and when there is power shedding. Simulation results show that the proposed
Multi-agent system can perform the operation of the smart grid efficiently.
Keywords: Multi-agent System; Smart Grid; Micro-grid; Distributed Generation; Renewable Energy
1. Introduction
The security and resiliency of electric power supply to
serve critical facilities are of high importance in today’s
world. In India, with the increasing complexity of power
grids, growing demand and growing concerns for envi-
ronment accentuate the leap towards something ‘smarter’.
Instead of building large electric power grids and high
capacity transmission lines an intelligent approach is
essential for transforming the existing power grid to a
‘smarter grid’ widely referred as ‘smart grid’. Smart grid
technologies have the potential to transform the existing
grids to more efficient, self healing, reliable, safer and
less constrained grids [1].
A smart micro-grid can be defined as a low voltage
distribution network with distributed energy resource
(DER) units, such as the distributed generation (DG)
units and distributed storage (DS) units and loads. The
DG units utilize Diesel Engines, Micro turbines, Fuel Cells,
Photovoltaic (PV) panels, small wind turbines, and com-
bined heat and power (CHP) systems. The capacity of the
DG sources varies from few kW to 1-2 MW. The DS
units could be flywheels, energy capacitors and batteries.
The micro-grid can be made smarter by integrating
advanced sensing technologies, control methods and
communication techniques [1-3]. Smart grids can benefit
customers through providing uninterruptible power, en-
hancing local reliability, reducing transmission loss, and
supporting local voltage and frequency.
As the DER units typically operate at a distribution
voltage level and geographically close to loads, smart
micro-grids are developed to interconnect the energy
sources and loads in a relatively small area, such as a
suburban community, a university or school, a commer-
cial area, and an industrial site. Besides the environ-
mental benefit of using more renewable energy sources, a
smart grid can either be interconnected to the main grid
as a single aggregated load (or generator) or in case of
external faults or periods of emergency, it has the capa-
bility to separate, or island, from the main grid and oper-
ate independently, within limits.
Significant research works have been carried out on
operation and control of smart grid. Current trends to
control and monitor the operation of electrical power
systems are however moving towards the use of an
automated agent technology, known as multi-agent sys-
tem. In recent years, multi-agent based approach is pro-
posed/adopted to provide intelligent energy control and
management systems in smart grids. Multi-agent systems
have become the focus of intense research in European
countries [4, 5].
In India, a lot is yet to be done towards making power
grids smarter and reliable, capable of taking self correct-
ing decisions for optimal scheduling of generation re-
sources. In this paper the operation of a multi-agent sys-
tem for the control of a smart grid is described in an In-
dian scenario.
Present power scenario of India is described in Section
II. The rest of the paper is organized as follows: Section

R. GUPTA ET AL. 1331
III gives the Multi-agent approach to smart grid opera-
tion. Multi-agent technology is shortly introduced. An
overview of the related work, general management and
control concept based on the related work is presented
with the help of a process flow-chart. In Section IV
Simulation results are presented based on the proposed
approach. Section V concludes the paper.
2. Present Power Scenario of India
The relationship between power consumption and na-
tional economic development has a great significance.
Indian power sector is facing challenges and despite sig-
nificant growth in generation over the years, it has been
suffering from shortages and supply constraints. All In-
dia region-wise generating installed capacity (MW) of
power utilities is given in Table 1 [6] and the corre-
sponding pie chart is shown in Figure 1.
India’s electricity sector is amongst the world’s most
active players in renewable energy utilization, especially
wind energy. As of December 2011, India had an in-
stalled capacity of about 22.4 GW of renewal technolo-
gies-based electricity. In1990, the capacity of renewable
energy sources (RES) was 18 MW whereas the genera-
tion during the year 1989-1990 was 6 MU in India. Ini-
tially, the annual capacity addition was very slow, but
from 2008 onwards the contribution from RES is consid-
erable. In 2012, the RES capacity was 24503. 45 MW
and the percentage share of RES in total generation ca-
pacity was 12.26% which is expected to increase to
17.12% by 2017. The percentage share of RES in total
generation in India during 2011-12 was around 5.5 %.
The category-wise details of electricity generation in
the country during August 2012 and during April 2012 to
August 2012 are given below in Table 2 [7].
Figure 1. All India Generating Installed Capacity (MW).
Table 1. Region –wise Electricity Generating Capacity of India.
Thermal
S.N. Region Coal Gas Diesel Total Nuclear Hydro RES Total
1 Northern 31623.50 4671.26 12.99 36307.75 1620.00 15467.75 4623.24 58018.74
2 Western 43537.00 8254.81 17.48 51809.29 1840.00 7447.50 8450.04 69546.83
3 Southern 23782.50 4962.78 939.32 29684.60 1320.00 11353.03 12096.78 54454.41
4 Eastern 22607.88 190.00 17.20 22815.08 0.00 3948.12 436.71 27199.91
5 North Eastern 60.00 824.20 142.74 1026.94 0.00 1200.00 243.28 2470.22
6 Islands 0.00 0.00 70.02 70.02 0.00 0.00 6.10 76.12
7 Total 121610.88 18903.00 1199.75 141713.68 4780.00 39416.40 25856.14 211766.22
Capacitive Generation Capacity in Industries having demand of 1 MW and above, Grid interactive (as on 31-03-2011) = 34444.12 MW
Table 2. Electricity generation in India (category-wise).
Category Monitored capacity
(MW) Actual Generation (MU) during August2012 Actual Gen (MU) during April2012 to August201 2
RES
Wind 14870.955 4508.687 18909.007
Solar 976.904 83.483 539.205
Biomass 707.250 105.995 781.449
Biogases 2881.210 243.555 2184.995
Small Hydro 1290.293 330.083 1059.842
Others 131.760 7.900 83.246
Total RES 20858.372 5279.702 23557.744
Total (conventional) 181710.520 74498.360 382466.050
Total 202568.892 79778.062 406023.794

R. GUPTA ET AL.
1332
3. Multi-Agent Based Operation of Smart
Grid
3.1. Multi-Agent System
A multi-agent system is a combination of several agents
working in collaboration pursuing assigned tasks to
achieve the overall goal of the system. The multi-agent
system has become an increasingly powerful tool in de-
veloping complex systems that take advantages of agent
properties: autonomy, sociality, reactivity and pro-activ-
ity. The multi-agent system is social-able in that they
interact with other agents via some kind of agent com-
munication language. The agents also perceive and react
to their environment. Lastly, the multi-agent system is
proactive in that they are able to exhibit goal-oriented
behavior by taking initiatives [1-4, 8-12].
In the context of power systems, multi-agent technolo-
gies can be applied in a variety of applications, such as to
perform power system disturbance diagnosis, power sys-
tem restoration, power system secondary voltage control
and power system visualization. Some of the recent re-
searchers have implemented multi-agent system to con-
trol the operation of a smart grid.
3.2. Agent Model of the Proposed Smart Grid
The proposed model of the smart grid consists of several
agents that communicate and coordinate with each other.
The smart grid model that is used in the proposed appli-
cation is simple, since we focus mainly on the coopera-
tion of agents. Figure 2 shows the architecture of the
smart grid. Smart grid is connected to the main grid. The
Smart Grid Controller announces the time zone (day/
night) and availability of grid. The production units that
belong to the smart grid adjust their set points based on
their availability, the time zone, availability of the grid,
grid price, their operational cost and the load demand [1].
a) Load Agent (LAG)
LAG is installed at each load centre. Main functions of
LAG are:
i. Load Forecasting: LAG makes a prediction about
the demand of next day. The forecast demand for
electricity is determined from the historical data and
the weather forecast for tomorrow.
ii. Sending Message: LAG sends a request message to
SGC for the power purchase. The message includes
the amount of power and its price.
iii. Receiving Message: LAG receives from SGC a
message that tells from where to purchase the elec-
tric power.
b) Battery Agent (BAG)
BAG, installed at each battery centre, has attributes
such as the amount of charge, the capacity and the status
(charge/discharge). The following describes the main
functions of BAG.
i. Measurement of the state of charge: BAG conducts
to measure the amount of battery charge. Battery is
charged when Grid is available and discharged when
Grid is not available.
ii. Sending Message: BAG sends a request message to
SGC for the power purchase or sale. The message
includes the amount of power and its price.
iii. Receiving Message: BAG receives from SGC a
message that tells when to purchase or sale the elec-
tric power.
c) Smart-Grid-Controller (SGC)
SGC, a special purpose agent facilitates the negotia-
tion process of the multi-agent system. The following
describes the main functions of the SGC.
i. Controlling the timing of the negotiation: SGC sug-
gests the timing of the negotiation by sending a
message to open the market to all agents in the smart
grid.
ii. Decision making of the operation: SGC makes a
decision of the operation of producer or pro-
ducer/consumer agents.
d) Wind Turbine Agent (WTAG)
WTAG, installed at each wind turbine generator, has
attributes such as the generator available power. The fol-
lowing describes the main functions of WTAG.
i. Sending Message: WTAG sends a request message
to SGC for the power sale. The message includes the
amount of power and its price.
ii. Receiving Message: WTAG receives from SGC a
message that tells where to sale the electric power.
e) Photo-Voltaic Agent (PVAG)
PVAG, installed at each photo-voltaic generator, has
attributes such as the generator available power. The fol-
lowing describes the main functions of PVAG.
i. Sending Message: PVAG sends a request message
to SGC for the power sale. The message includes
availability, the amount of power and its price.
ii. Receiving Message: PVAG receives from SGC a
message that tells where to sale the electric power.
f) Micro-Hydro Turbine Agent (MHTAG)
Figure 2. Agent model of the proposed smart grid

R. GUPTA ET AL. 1333
MHTAG, installed at each micro-hydro turbine gen-
erator, has attributes such as the generator available
power. The following describes the main functions of
MHTAG.
i. Sending Message: MHTAG sends to SGC a mes-
sage to request for the power sale. The message in-
cludes the amount of power and its price.
ii. Receiving Message: MHTAG receives from SGC a
message that tells where to sale the electric power.
g) Diesel Generator Agent (DGAG)
DGAG, installed at each diesel generator, has attrib-
utes such as the generator minimum/maximum output
value. The following describes the main functions of
DGAG.
i. Sending Message: DGAG sends to SGC a message
to request for the power sale. The message includes
the amount of power and its price.
ii. Receiving Message: DGAG receives from SGC a
message that tells where and when to sale the elec-
tric power.
h) Steps for Sorting Process
The sorting algorithm (flow chart as shown in Figure 3)
for available generating resources (producers) is explained
as below:
Step 1: Collect the following data of the available
sources/ producers:
(i) Power cost per unit
(ii) Availability of the producer
(iii) Power availability with the producer
Assign ID in numeric to the producer(s).
Step 2:
(i) Sort the array of producers in the order of price
(ii) The producers which are available and can give
power at that time slot are sorted in ascending
order of cost/price.
Step3: Read the Demand (Load).
Step 4: Identify the producers which can fulfill the
demand using the array sorted in Step 2, individually or
in combination.
Step 5: Send the producer ID and capacity to SGC.
i) Negotiation Process
Negotiation is one of the key processes for the multi-
agent system to successfully attain its goal. Flow-chart in
Figure 4 shows the negotiation process.
There are seven types of agents in the proposed system
viz. SGC, LAG, BAG, PVAG, WTAG, MHTAG and
DGAG.
The main objective is to fulfill the demand (load) at a
particular time by minimizing the use of diesel genera-
tors as their per unit price outreach other available
sources. To accomplish this, the agents are required to
cooperate and coordinate so that they make efficient use
of the power supplied by other sources at the time of
power shedding.
Figure 3. Flow-chart for sorting of producers (Generating
units) for SGC
SGC is informed about the demand, producer avail-
ability and the operating cost per unit of different pro-
ducers in one negotiation cycle. At the time of power
shedding, if the electrical power in the smart grid is in-
sufficient, demand is fulfilled by the diesel generators.
SGC sends a message of announcement of tender to all
agents in the smart grid every 10 minutes. SGC is sug-
gested to control the timing of the negotiation process.
All buyer agents in the smart grid reply a message of
purchase of electrical power. The message includes the
amount of electrical power i.e. Demand ‘D’. All producer
agents in the smart grid reply a message of sale of elec-
trical power. The message includes the generator avail-
able power, their availability at a particular time and their
operating cost per unit.
In the next step, sorting of producers and the grid is
done depending upon their availability and operating
costs in ascending order of price per unit. This sorted list
is then sent to the SGC that makes a decision on opera-
tion of different producers based on this generated list.
The demand is checked with the power available with
the producer. If the demand is not fulfilled by first pro-
ducer, SGC sends operation command to next producer
and so on until the demand is fulfilled. Finally, genera-
tion load pattern is displayed for a given time period and
next negotiation cycle begins.

R. GUPTA ET AL.
1334
4. Simulation Results
4.1. Assumptions
The case study considers a township located in India. It
comprises of 150 houses inhabited with modern ameni-
ties, education sector, health centre, super market, eleva-
tors, flour mill, laundry, a community hall and water
pump. Typical household appliances include lighting,
refrigerator, AC/heater, geyser, TV, washing machine,
micro wave, electric cooker i.e. a max load of 5kVA to
10kVA. Super market facilities include lighting, refrig-
erator/freezer (max consumption 15kW), AC/heater.
Medical health centre facilities include lighting, refrig-
erator, AC/heater. Education sector facilities include
lighting, AC/heater, laboratory equipments.
The Smart Grid consists of a Load Agent (LAG),
Wind Turbine Agent (WTAG), two Photo-Voltaic Agents
(PVAGs), Micro-Hydro Turbine Agent (MHTAG), two
diesel generator agents (DGAGs) and a Battery Agent
(BAG). The maximum capacity of each agent is shown
in Table 3.
Figure 4. Flow-chart of the negotiation process
Electrical power prices of different producers are giv-
en in Table 4 [13]. In India, tariff is same for peak as
well as off-peak hours. Due to shortage of electricity,
power cuts (power shedding) are common in India. To
avoid a total black out of the power system, electricity
delivery is stopped for non-overlapping periods of time
over different parts of the distribution regions.
Demand power of the LAG, as shown in Figure 5 is
the data of township under consideration during day time
and night time. Four simulations are carried out on this
smart grid model. Table 6 shows the content of the case
studies.
Table 3. Parameters.
Name Max Capacity Type
LAG 2MW(peak),1.5MW (off-peak) Consumer
WTAG 300KW Producer
PVAG1 150KW Producer
PVAG2 100KW Producer
MHTAG 400KW Producer
DGAG1 200KW Producer
DGAG2 200KW Producer
BAG ±250KWh Producer/consumer**
GRID ∞/0* Producer
*power shedding (no power is available from the grid); **producer at the
time of power shedding & consumer during charging phase.
Table 4. Power cost per unit
Producers Price/unit(kWh)
GRID 4.00
BAG 4.50#
PVAG 9.18*
WTAG 3.84**
MHTAG 5.51
DGAG 13.00
#Pseudo price (As battery is charged from the grid, the price includes grid
price per unit in addition to some maintenance cost; *Solar PV Crystalline;
**wind Zone IV (>400W/m2).
Table 5. Availability of the producer(s)
Time zonePVAGGRIDMHTAG WTAG BAGDGAG
06:00-12:00Yes Yes Yes Yes No Yes
12:00-18:00Yes No Yes Yes Yes Yes
18:00-0:00No Yes Yes Yes No Yes
0:00-6:00No No Yes Yes Yes Yes

R. GUPTA ET AL. 1335
Table 6. Test Case Scenario
Case Time Zone
Availability of Power
from Grid
Case I Day Y
Case II Day N
Case III Night Y
Case IV Night N
Figure 5. Demand power of LAG (24 hrs)
Figure 6. Showing the availability of power from grid dur-
ing the day time
Figure 7. Showing the unavailability of power from grid
during the day time
Case1: Day time and power is available from the grid
This is the case when grid is available. Any amount of
power can be taken from the grid. WTAG price is less
compared to grid. SGC takes the decision to fulfill the
load demand by WTAG as shown in figure 6. When the
load demand is more compared to the power available by
WTAG, power is taken from grid. The demand is ful-
filled at a lower price.
Case 2: Day time and power is not available from the
grid
This is the case when load demand is more. Due to
power shedding power is not available from the grid.
Power is taken from WTAG & BAG and demand is ful-
filled as shown in figure 7. If the demand is not fulfilled
by these producers, SGC fetch power from MHTAG,
PVAGs & DGAGs. Load demand is fulfilled at a lower
price. Simulation results show that the multi-agent ap-
proach is promising in smart grid operations.
Case 3: Night time and power is available from the
grid
This is the case when load demand is fluctuating. Any
amount of power can be taken from the grid. WTAG
price is less compared to grid. SGC takes the decision to
fulfill the load demand by WTAG as shown in figure 8.
When the load demand is more compared to the power
available by WTAG, power is taken from the grid. Load
demand is fulfilled at a lower price.
Case 4: Night time and power is not available from the
grid
Figure 8. Showing the availability of power from grid dur-
ing the night time
Figure 9. Showing the unavailability of power from grid
during night time

R. GUPTA ET AL.
1336
This is the case when load demand is less. Due to
power shedding power is not available from the grid.
Power is taken from WTAG and BAG and demand is
fulfilled as shown in figure 9. If the demand is not ful-
filled by these producers SGC fetch power from
MHTAG and then from DGAG. Load demand is fulfilled
at a lower price. Simulation results show that the mul-
ti-agent approach is promising in smart grid operation.
5. Conclusions
This paper presents the benefits of multi-agent system for
smart grid operation. This is achieved effectively through
the negotiation skills and coordination of actions between
SGC and various agents. The proposed approach is vali-
dated in a simulated environment that considers both the
availability and outage conditions of the power grid.
Based on the method, the algorithm and various test case
scenarios, it is found that the proposed multi-agent ap-
proach is viable in smart grid operations.
REFERENCES
[1] T. Nagata, Y. Ueda and M. Utatani, “A Multi-agent Ap-
proach to Smart Grid Operations,” The 2012 IEEE Con-
ference on Power System Technology (POWERCON),
2012, pp. 1-5. doi:10.1109/PowerCon.2012.6401303
[2] M. Pipattanasomporn, H. Feroze and S. Rahman, “Mul-
ti-Agent Systems in a Distributed Smart Grid: Design and
Implementation,” Proceedings of IEEE PES 2009 Power
Systems Conference and Exposition (PSCE’09), Mar
2009, Seattle, Washington, USA.
[3] J. J. Lu, X. Da and A. Qian, “Research on Smart Grid in
China,” Proceedings of Transmission & Distribution
Conference & Exposition: Asia and Pacific, 2009, pp.
1-4.
[4] A. L. Dimeas and N. D. Hatziargyriou, “Operation of a
Multi-agent System for Microgrid Control,” IEEE
Transactions on Power systems, Vol. 20, No. 3, 2005, pp.
1447-1455. doi:10.1109/TPWRS.2005.852060
[5] A. L. Dimeas and N. D. Hatziargyriou, “Agent based
control of Virtual Power Plants,” IEEE The 14th Interna-
tional Conference on Intelligent System Applications to
Power Systems, ISAP 2007 November 4 - 8, 2007, Kaoh-
siung, Taiwan 536.
[6] Available online at:
http://www.cea.nic.in/reports/monthly/executive_rep/jan1
3/8.pdf
[7] Available online at:
www.cea.nic.in/reports/articles/god/renewable_energy.pd
f.
[8] A. Dimeas and N. D. Hatziagyriou, “A Multi-agent Sys-
tem for Micro Grids,” In Proc. IEEE PES General Meet-
ing, Denver, CO, Jun. 2004, PESGM2004–001, p. 244.
[9] A. L. Dimeas, S. I. Hatzivasiliadis and N. D. Hatziargy-
riou, “Control Agents for Enabling Customer-driven Mi-
cro Grids,” Proceedings of Power & Energy Society
General Meeting, 2009. PES '09. IEEE, 26-30 July 2009,
pp. 1-7.
[10] A. L. Dimeas and N. D. Hatziargyriou, “Agent Based
Control for Microgrids,” Proceedings of IEEE PES Gen-
eral Meeting, Tampa, FL, June 24-28, 2007
[11] S. J. Chatzivasiliadis, N. D. Hatziargyriou and A. L. Di-
meas, “Development of an Agent Based Intelligent Con-
trol System for Micro Grids,” The 2008 IEEE Power and
Energy Society General Meeting on Conversion and De-
livery of Electrical Energy in the 21st Century, 20-24
July, 2008, pp. 1-6.
[12] A. L. Dimeas and N. D. Hatziargyriou, “A MAS archi-
tecture for Micro grids Control,” Proceedings of the 13th
International Conference on Intelligent Systems Applica-
tion to Power Systems, 6-10 Nov, 2005.
doi:10.1109/ISAP.2005.1599297
[13] Available online at:
http://mnre.gov.in/filemanager/UserFiles/september_mon
th_2012_rerf.pdf.