Modeling and Simulation of MAS-based Management System for
Smart Grid with Smart Homes
Chen Miao, Tai Nengling and Ji Kang
Department of Electrical Engineering, Shanghai JiaoTong University, Shanghai, China
Keywords: MAS, NSGA-II, Operation Optimization, Smart Home, Smart Grid.
Abstract: Traditional power system is undergoing revolutionary changes. Concepts like Smart Grid and Smart Home
have come into practice due to high penetration of distributed generation and advanced information
technologies. This paper incorporates Smart Home and Smart Grid. It focuses on a community where
houses are equipped with photovoltaic panels, energy storage system and smart appliances. A new
management system is proposed based on multi-agent technology. Agents are designed to be intelligence,
autonomous and with high plug-and-play capability. System control architecture has three levels
corresponding to management of devices, houses and community. A two-level operation optimization
scheme is proposed, in which operation optimization problem is decomposed into five constrained multi-
objective problems, mutual supply mechanism is deployed to efficiently coordinate Smart Homes and
NSGA-II is introduced to obtain Pareto-optimal solutions. Simulation results validate the effectiveness of
proposed management system.
1 INTRODUCTION
Over the last decade, rising awareness of energy
crisis and environmental deterioration has led to a
revolution in electricity industry. Novel technologies
like distributed generation (DG) and smart electrical
appliances have been brought into daily life and new
concepts as Smart Home and Smart Grid have been
introduced to facilitate the transformation of
traditional power system.
Smart Grid is seen as the energy infrastructure
for future electric power system. Through the
incorporation of advanced control strategies and
information technologies, Smart Grid is endued with
capability to rapidly collect and assess large amounts
of data, to smoothly integrate renewable energy
sources and controllable loads (CLs) and to
efficiently coordinate numbers of units towards the
economic, reliable and secure running of power
system (Gungor et al., 2011).
Sensors and smart meter brings smart grid
concept into households. Smart Home is an energy-
aware household equipped with smart electrical
appliances, residential DG and energy storage
system (ESS). Such home is expected to adjust its
energy profile in accordance with dynamic pricing,
DG output and grid operation to cut energy bill
(Komninos et al., 2014).
Significant research work has been done
regarding the implementation and control of Smart
Grid. Multi-agent technology is widely proposed as
an effective tool to manage Smart Grid since
McArthur et al. (2007a, 2007b) discovered its
capability to satisfy the demand of distributed
control, autonomous operation and flexible
framework. Dimeas and Hatziargyriou (2005)
developed a MAS-based infrastructure of
management system as well as an energy exchange
optimization algorithm. Mao et al. (2014) proposed a
hybrid architecture for MAS-based energy
management system by combining hierarchical and
central architecture. Foo Eddy, Gooi and Chen
implemented agent models and energy management
system framework based on Simulink, JADE and
MACSimJX. Scheduling of air conditioning
incorporating customer’s convenience in home
energy management system (HEMS) is discussed by
Jo, Kim and Joo (2013). Cost-effective energy
management system with hierarchical agents for
Smart home is studied by Jiang and Fei (2015).
The work reported in the literature focuses on
either Smart Home or Smart Grid while the
consequences of integration of Smart Homes into
Smart Grid as well as the establishment of
401
Miao C., Nengling T. and Kang J..
Modeling and Simulation of MAS-based Management System for Smart Grid with Smart Homes.
DOI: 10.5220/0005509204010408
In Proceedings of the 5th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH-2015),
pages 401-408
ISBN: 978-989-758-120-5
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
correspondent management system is seldom
discussed. To address this issue, this paper proposes
a management system for a community where each
household has its own photovoltaic (PV) panels,
ESS and smart appliances. The management system
is MAS-based with three-level hierarchical control
architecture. System framework is formed by
intelligent agents assigned to PV, ESS, load and
house. A two-level operation optimization scheme is
proposed to schedule ESS and CL. Mutual supply
mechanism is deployed to coordinate household
agents. NSGA-II is used to solve constrained multi-
objective problems. The effectiveness of the
proposed system is investigated by numerical
simulation.
The rest of this paper is organized as follows.
Section 2 describes structure of studied community
and proposed management system. Section 3
presents operation optimization scheme. Section 4 is
contributed to simulation and results analysis. Part 5
concludes this paper.
2 SYSTEM ARCHITECTURE
AND CONFIGURATION
In this section, the structure of studied community is
presented first, followed by detailed presentation of
structure and design of proposed management
system.
2.1 System Scheme
Structure of the studied community is illustrated in
Figure 1. Community comprises of more than 20
households, but only four households are considered
in this paper for simplicity. Community is connected
to utility grid by Point of Common Coupling (PCC).
A public ESS (PESS) is installed near PPC for
stability. In islanded mode, PESS discharges as
standby power supply.
Figure 2 depicts individual household structure.
Each household has its own PV panels and ESS.
Load demand is divided into two categories: CL,
namely smart appliances, and fixed load (FL) which
groups the rest of electrical appliances in the house.
FL is usually presented by a curve, which
corresponds to residents’ living habits and daily
schedule. With advanced information technologies
like smart meter, households can sell electricity
when energy generation exceeds consumption.
Figure 1: Structure of studied community.
Figure 2: Structure of individual household.
2.2 Management System Design
Control architecture of proposed management
system consists of three levels: local control level,
coordinated control level and central control level.
System framework is based on multi-agent
technology. Agents are assigned to each PV unit,
ESS, load demand, household and PCC for better
plug-and-play capability. Each agent has a certain
degree of intelligence and autonomy. Provided with
a set of function modules, agents interact and
exchange information with others to attain local and
global objectives. Scheme of proposed management
system is depicted in Figure 3.
Local Control level is the interface between
management system and physical entities. It is
comprised of PV agent, ESS agent and Load agent.
Each agent has Data Acquisition module which
collects real-time data, Control module which
adjusts devices’ operation state and Communication
module which permits data exchange with
Household agent. Forecast module of PV agent and
Load agent provides a one-day-ahead forecast of PV
output and FL profile.
Coordinated Control level corresponds to Smart
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
402
Home management system. Receiving generation
and load demand data coming from Local Control
Level as well as information of utility grid and
neighbouring households coming from Central agent
via Communication module, Household agent
analyses power flow of house circuit to detect device
malfunction (Power Flow Analysis module),
compares real and forecast data to modify scheduled
plan in case of huge difference (Real-time Adjusting
module), calls optimization algorithm to provide
optimal operation plan for the following day
(Operation Optimization module). Communication
module then transmits scheduled plan to ESS and
Load agent and electricity selling/consuming data to
Central agent.
Central Control level is equivalent to Smart Grid
management. It has only one agent: Central agent
assigned to PCC. Electricity surplus/demand
information of Smart Home and operation data of
PESS are gathered for processing. Operation
Optimization module schedules PESS and
coordinates households for the following day. Power
Flow Analysis module monitors the operation states
of community network. Real-time Adjusting module
changes PESS state when forecast deviates from real
data. Bidirectional power exchange with utility grid
is monitored by PCC Control module.
3 OPERATION OPTIMIZATION
FORMULATION
As householders are not willing to pay extra for the
sake of others’ interest, operation optimization of
Smart Grid with Smart Homes is decomposed into
two levels. Optimization of community network is
performed after individual household optimization.
Otherwise, it may occur that the optimal operation
plan of whole community is not necessarily optimal
for one household.
In this section, individual household optimization
is formulated at first, followed by community
operation optimization. Solving algorithm is
presented at end.
3.1 Individual Household Optimization
The objective is to cut energy bills by scheduling CL
and ESS. Energy bill of individual household is
quantified by Equation 1.
Figure 3: Scheme of three-level MAS-based management system.
ModelingandSimulationofMAS-basedManagementSystemforSmartGridwithSmartHomes
403
bill install pollu main depre ce se
FC C C C CP
(1)
C
install
, C
pollu
and C
main
signify respectively
installation cost, pollution charge and maintenance
charge, which are regarded as constant in short-time
scheduling problem. C
depre
is the depreciation
charge, which is proportional to the energy charged
and discharged. C
ce
is the total cost of electricity
purchase while P
se
signifies the profit earned by
selling electricity back to grid. Energy bill of
household is rewritten as Equation 2.
bill d ESS c c s s
TTT
FkPdtPRPdtPRPdt
when discharging
when charging
,
,
ESSd
ESS
ESSc
P
P
P
(2)
where T is studied period, k
d
is depreciation
coefficient, P
ESSd
and P
ESSc
are discharging or
charging power of ESS, P
c
and P
s
are purchasing or
selling power of household, PR
c
and PR
s
are retail
price and selling price.
It should be noted that frequent charge-discharge
harms ESS unit and decreases battery life, charge-
discharge cycle of ESS, denoted as
cycle
, should be
minimized during optimization.
Constraints related to operation of different
agents are listed below.
a) Power balance constraint
PV ESS h j
PP PFLCL
when purchasing electricity
when selling electricity
,
,
c
h
s
P
P
P
(3)
where P
PV
is PV output power and CL
j
is the j
th
controllable load.
b) ESS operation constraint
mi n ma xESS
SOC SOC SOC
(4)
00
(1 ) (1 )
ESSfinal
SOC SOC SOC
(5)
where SOC
ESS
is current state of charge (SOC),
SOC
min
and SOC
max
are lower and upper limit of
SOC, SOC
0
and SOC
ESSfinal
are SOC at the beginning
and end of studied period, β is variation coefficient.
c) User’s satisfaction constraint
lj CLj uj
TT T
(6)
where T
CLj
, T
lj
and T
uj
are respectively execution
time, lower and upper limit of preferred execution
period of task j.
To summarize, optimization problem of
individual household is formulated by Equation 7.
(3) (6)
min [ , ]
subject to
T
bill cycle
FF
(7)
3.2 Community Network Optimization
Community network optimization schedules PESS
and coordinates Smart Homes to curtail operation
cost. Operating cost of community network is
illustrated by Equation 8.
ocost install pollu main depre ce se loss
FC CCCCPC
(8)
C
install
, C
pollu
, C
main
, C
depre
, C
ce
and P
se
are treated as
mentioned in 3.1. C
loss
is cost of transmission loss.
Transmission loss is related to power, which is fixed
as energy demand of households is given, and
transmission distance, which is reduced when and
only when households deliver excess energy directly
to nearby household in need instead of selling it
back to utility grid. PESS won’t influence
transmission distance since PESS is near PCC. With
the aim to avoid time-consuming loss calculation
and to reduce transmission loss, mutual supply
mechanism is proposed to detect energy
surplus/demand pair and to coordinate households
accordingly.
Equation 8 is then rewritten as
ocost dP PESS c c s s
TTT
F
kPdtPR PdtPRPdt
when discharging
when charging
,
,
PESSd
PESS
PESSc
P
P
P
(9)
where k
dP
is depreciation coefficient, P
PESSd
and
P
PESSc
are discharging or charging power of PESS.
Grid operation constraints are listed below.
Operation constraints of PESS are described by
Equation 4 and 5.
a) Power balance constraint
1,..., 4
PESS PCC hi
i
PP Pi
(10)
where P
PCC
is power of PCC.
b) Power flow constraint
min maxllinel
PPP
(11)
where P
line
, P
lmin
and P
lmax
are respectively real-time
power, minimum power and maximum power of
transmission line.
c) Transmission capability constraint
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
404
min maxpPCCp
PPP
(12)
where P
pmin
and P
pmax
are lower and upper limit of
allowed exchange power of PCC.
Operation optimization problem is thus
formulated as
ocost
(4), (5), (10) (12)
min [ , ]
subject to
T
cycle
FF
(13)
3.3 Solving Algorithm
Figure 4: Flow chart of NSGA-II.
Operation optimization of studied community is
formulated as five constrained multi-objective
problems.
Proven to be effective in search of Pareto optimal
solutions of constrained multi-objective problem,
Non-Dominated Sorting Genetic Algorithm II
(NSGA-II) is deployed as solving algorithm in this
paper. Flow chart of NSGA-II is illustrated in Figure
4 (Deb et al., 2002).
4 SIMULATION AND RESULTS
ANALYSIS
To validate the effectiveness of proposed system,
simulation studies for a typical day were carried out.
4.1 Parameters Specification
The simulation platform is implemented with Matlab
R2011a. The community illustrated in Figure 1 and
2 is modelled in Simulink while optimization
algorithm is programmed using Matlab script. The
program is run on Intel Core i5-3.2GHz PC with 4G
RAM.
Suppose each day starts at 6 a.m. and ends at 6
a.m. of the next day. Time slot is set as 30 minutes.
A tiered electricity pricing mechanism quantified in
Table 1 is used.
Table 1: Tiered electricity tariff.
Period
Peak Valley
06h00-22h00 22h00-06h00
PRc /¥/kWh 0.617 0.320
PRs /¥/kWh 0.540 0.307
Each household is configured with its own PV, ESS,
FL and CL. For simplicity, we assume that PV
panels and ESS are identic for all four households.
PV generation data provided by Ding and XU (2011)
and typical FL profile provided by Rudion et al.
(2006) are utilized and shown in Figure 5-8. Three
CLs are considered, whose parameters are listed in
Table 2. With reference to Rahbar, Jie and Rui’s
work (2015), ESS is modelled by Equation 14. ESS
parameters are summarized in Table 3.
(/)/
N+1 N c ESSc ESSd d nom
TT
SOC SOC Pdt Pdt C
(14)
where
c
and
d
are charge and discharge efficiency,
C
nom
is nominal capacity of ESS.
As to optimization algorithm, roulette-wheel
selection method is used. Mutation and crossover
rate are respectively 0.1 and 0.9, population size is
100 and iteration number is set as 1000.
Table 2: Parameters of controllable load.
Task name Load
/kW
Duration
/hour
Preferred period
Washing clothes 0.4 0.5 09h00-18h00
22h00-05h00
Drying clothes 1 0.5 09h00-18h00
22h00-05h00
Charging EV 4 6 Before 06h00
ModelingandSimulationofMAS-basedManagementSystemforSmartGridwithSmartHomes
405
Table 3: Parameters of energy storage system.
Parameter ESS PESS Unit
c
88.2 88.2 %
d
98.0 98.0 %
SOC
max
95 90 %
SOC
min
10 20 %
SOC
0
80 70 %
C
nom
20 80 kWh
k
d
0.1 0.1 ¥/kWh
4.2 Results and Analysis
Figure 5-8 illustrate simulation results of individual
household optimization.
Analyse the optimization result of Household 1
in detail. Due to the solar radiation limitation, PV
generates electricity only from 7 a.m. to 4:30 p.m.
Figure 5: Optimal operation plan for Household 1.
Figure 6: Optimal operation plan for Household 2.
Figure 7: Optimal operation plan for Household 3.
Figure 8: Optimal operation plan for Household 4.
Energy generation satisfies fixed load demand and
charges ESS. From 5:30 p.m. to 8:30 p.m., ESS
discharges to curtail electricity purchase. ESS
recharges from 23:30 in order to regain initial SOC
level. Charging EV and drying clothes are scheduled
after 22:00 which coincides with valley hours.
Simulation results are compared with traditional
system. Without ESS and load demand scheduling,
energy bill is only related to purchasing/selling
electricity. Minimal cost is achieved when all PV
generation is used to supply power and all CLs are
scheduled in valley period. It is worthy to mention
that, judging from the difference between FL and PV
output, these two assumptions cannot be
simultaneously true.
Table 4: Comparison of proposed and traditional system.
House
cycle
F
bill
/¥
proposed system
F
bill
/¥
traditional system
1 3 15.2416 17.4017
2 1 12.3656 14.0022
3 3 13.2813 16.8225
4 3 12.1592 14.3193
Objective function value of two systems is shown in
Table 4. As can be seen, the application of
optimization algorithm cuts the electricity bill of
individual household by about 12%.
Figure 9 shows exchange power between
households and community network.
Figure 9: Exchange power of four households.
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
406
It is observed that some households have excess
energy while others is in need of electricity at a
given time slot k. At 9:30 a.m., Household 3 has a
surplus of 1.4354kW while Household 4 demands
1.2526kW. Household 3 should deliver 1.2526kW
directly to Household 4 if mutual supply mechanism
operates properly. Results of coordination among
households are shown in Figure 10. Each bar marks
one mutual supply instruction between neighbouring
houses.
Figure 10: Mutual supply instructions.
Figure 11: Scheduled PESS and PCC power.
Figure 11 illustrates PESS scheduling and calculated
PCC power. We can see that PESS discharges at
peak hours and charges at valley hours. PCC power
is close to 0 during peak hours, which signifies that
the proposed system has successfully shifted peak
load.
Value of objective function F
ocost
of proposed
system and traditional system are shown in Table 5.
It is noticed that the application of proposed
optimization scheme curtails operation cost of
community network by 11.5%.
Table 5: Value of F
ocost
of proposed and traditional system.
Method F
ocost
/¥
proposed system 59.1742907
traditional system 66.8606358
It can be concluded from aforementioned analysis
that the performance of proposed two-level
optimization scheme is satisfying.
5 CONCLUSIONS
This paper integrates Smart Homes into Smart Grid.
A three-level MAS-based management system is
proposed to manage a community with smart homes.
MAS-based framework, hierarchical control
architecture as well as information flow of the
proposed system is described in detail. In the
presented two-level optimization scheme,
optimization problem is formulated as constrained
multi-objective problems and solved by NSGA-II
while coordination of Smart Homes is realized by
mutual supply mechanism. Simulation results show
that the proposed optimization scheme is able to
curtail energy bill by over 10% for householders and
grid owner and to shift peak load. Based on agents
with high plug-and-play capability, the proposed
management system is of universal applicability and
practicability. Extension to other communities is
achieved by adding correspondent agents.
Nevertheless, as NSGA-II provides more than one
recommended operation plan, methods to
automatically obtain one optimal solution should be
investigated in further studies.
ACKNOWLEDGEMENTS
This work is supported by the Research Project of
Chinese Ministry of Education (No. 113023A).
REFERENCES
Gungor, V. C., Sahin, D., Kocak, T., Ergut, S., Buccella,
C., Cecati, C., Hancke, G. P., 2011. Smart Grid
Technologies: Communication Technologies and
Standards, Industrial Informatics, 7(4): 529-539.
Komninos, N., Philippou, E., Pitsillides, A., 2014. Survey
in Smart Grid and Smart Home Security: Issues,
Challenges and Countermeasures, Communications
Surveys & Tutorials, 16(4):1933-1954.
McArthur, S. D. J., Davidson, E. M., Catterson, V. M.,
Dimeas, A. L., Hatziargyriou, N. D., Ponci, F.,
Funabashi, T., 2007. Multi-Agent Systems for Power
Engineering Applications—Part I: Concepts,
Approaches, and Technical Challenges, Power
Systems, 22(4): 1743-1752.
McArthur, S. D. J., Davidson, E. M., Catterson, V. M.,
Dimeas, A. L., Hatziargyriou, N. D., Ponci, F.,
Funabashi, T., 2007. Multi-Agent Systems for Power
Engineering Applications—Part II: Technologies,
Standards, and Tools for Building Multi-agent
Systems, Power Systems, 22(4): 1753-1759.
Dimeas, A. L., Hatziargyriou, N. D., 2005. Operation of a
ModelingandSimulationofMAS-basedManagementSystemforSmartGridwithSmartHomes
407
Multiagent System for Microgrid Control, Power
Systems, 20(3):1447-1455.
Mao M., Jin P., Hatziargyriou, N. D., Chang L., 2014.
Multiagent-Based Hybrid Energy Management
System for Microgrids, Sustainable Energy, 5(3):938-
946.
Foo Eddy, Y. S., Gooi, H. B., Chen, S. X., 2015. Multi-
Agent System for Distributed Management of
Microgrids, Power Systems, 30(1): 24-34.
Jiang B., Fei Y., 2015. Smart Home in Smart Microgrid: A
Cost-Effective Energy Ecosystem With Intelligent
Hierarchical Agents, Smart Grid, 6(1):3-13.
Jo H. C., Kim S., Joo S. K., 2013. Smart Heating and Air
Conditioning Scheduling Method Incorporating
Customer Convenience for Home Energy
Management System, Consumer Electronics, 59(2):
316-322.
Deb, K., Pratap, A., Agarwal, S., Meyarivan, T., 2002. A
Fast and Elitist Multiobjective Genetic Algorithm:
NSGA-II, Evolutionary Computation, 6(2):182-197.
Ding M., XU N., 2011. A Method to Forecast Short-Term
Output Power of Photovoltaic Generation System
Based on Markov Chain, Power System Technology,
35(1): 152-157.
Rudion K., Orths A., Styczynski Z. A., Strunz, K., 2006.
Design of Benchmark of Medium Voltage Distribution
network for Investigation of DG Integration, Power
Engineering Society General Meeting, Montreal, 18-
22 June 2006.
Rahbar, K., Xu J., Zhang R., 2015. Real-Time Energy
Storage Management for Renewable Integration in
Microgrid: An Off-Line Optimization Approach,
Smart Grid, 6(1):124-134.
SIMULTECH2015-5thInternationalConferenceonSimulationandModelingMethodologies,Technologiesand
Applications
408
