-
Demand side management algorithms and modeling
in smart grids A customers behavior based study
El Hassan Et-Tolba*, Mohamed Maaroufi
Department of Electrical Engineering,
Ecole Mohammadia dIngnieurs, Mohammed V-Agdal University, Rabat,
Morocco
*[email protected], [email protected]
Mohammed Ouassaid, Member, IEEE
Department of Industrial Engineering,
Ecole Nationale des Sciences Appliques-Safi,
Cadi Ayyad University, Safi, Morocco
[email protected]
AbstractThis paper presents algorithms and architecture models
for a home energy management system. Its based on
customers behavior that is modeled by a decision-making chain,
and smart appliances use for demand side management. The proposed
architecture is scalable and extensible to upper levels of
smart grid as the development approach used is bottom-up.
Once
the model is validated for home use, we can go up and apply it
for
holdings, factories, and micro-grids contexts. Scalability
models
and strategies are also presented and discussed. Ensuring
supply
and demand balance at real time is the main problematic of
smart grids. The proposed solution meets this objective,
because
it allows large scale renewable energy resources
integration.
Hence, it leads to global energy efficiency and demand side
management optimization in smart grids
Keywords-smart grids; demand side management; smart
appliances; customers behavior; renewable energy; distribution
network; energy consumption;
I. INTRODUCTION
Demand side management is very important in smart grids to
ensure the offer and demand balance in real time. It has been
realized that modification to the way we use energy in the demand
side could not only save money but also satisfy the level of
quality demanded by users of energy services [1]. The term
Demand-Side Management (DSM) did not exist prior to 1973. It
appeared within an atmosphere of chaos in relation to an impending
energy crisis and uncertainties, which prevailed at that time [2].
Hence the agreement of controlling some loads when they are needed
was reached and the term DSM officially came into being [1].
Nowadays, as many high coverage blackouts occurred around the world
these last recent years, the architecture of traditional electrical
networks has been challenged. So they are gradually giving up in
favor of smart grids. This new electrical network generation
mobilized both utilities and researchers and has led to many
scientific proposals and ideas.
Smart grids can be defined from a functional point of view as an
electric network able to integrate all the branched customers and
producers actions to distribute electric energy efficiently,
successfully, at low operating costs and safely [3]. In a business
case study of CISCO, smart grids are described as the combined view
that uses the information network to
enhance the functioning of the electricity grid [4]. It can also
be seen essentially as a control problem including optimization of
delivery, demand, asset, reliability and renewable resources
integration that will lead to operational and energy efficiency,
customer satisfaction (Quality of Service) and CO2 reduction
[5].
There are two key technologies enabling demand side load
optimization [6]:
Building automation
Smart metering
This paper interests in the first technology above. Many
researchers studied and discussed these two technologies from
different aspects. About the first technology, Building automation,
it has been concluded that the presence of distributed generation
(solar, Wind, biomass) and storage facilities (batteries, fuel
cell, Plug-in Hybrid Electric Vehicles) will help to create Zero
Net Energy Buildings and Districts (ZNEBs) [7]. Then, a general
philosophy of demand side load management for adjusting energy
demand/offer balance is presented in [8]. Where an explanation of
home automation importance for power load control and how smart
appliances should be redesigned is also given. Then, the major
requirements from a hardware and software perspective for smart
appliances design, with analysis of potential communication
techniques, each with their specific advantages and disadvantages
are presented in [9]. In addition to a universal appliance
interface enabling design of a controller with different interfaces
types for a wide range of common use appliances that is given in
[10]. The demand forecasting process become more efficient with the
use of flexible loads that can be managed and controlled to reduce
sizes of generation units, buying and selling energy from
distribution grid and total cost of smart microgrid [11]. This
process can use top down or bottom up approaches, based on
statistical or engineering methods. It models energy consumption
according to thermal characteristics of houses, consumption profile
of appliances and behavior of householders [12]. These last points
are mainly discussed in this paper; our proposed system
architecture and models are based on smart appliances use and
consumers behavior modeling. Otherwise, a new three-layer household
energy control system capable both to satisfy the
978-1-4673-6374-7/13/$31.00 2013 IEEE
-
maximum available electrical energy constraint and to maximize
user comfort criteria is presented in [13]. Another distributed
load control and demand side management system which is able to
even out fluctuations in the daily energy load is studied by [14].
Moreover, in [15], a framework for distributed D/R with user
adaptation based on assessed techniques in telecommunications
networks decongestion is given. The reliability of these systems
depends on their maintenance. It has been proved that by enabling
preventive maintenance, some precautions like computing thermo
images of electrical devices or determining the load capability and
insulation aging factor can reduce by 2.5 times the failure risk
[6].
As for the second technology, smart metering, it has also been
an interesting subject for several studies. In a business based
study, a general overview on smart meters with analyzes of their
functionalities and supported services were presented in [16]. And
their social benefits have been investigated in [17]. Then, its
very important to make devices smarter, but not with any cost; as
it has been affirmed: attributing intelligence, which implies
value, to these technologies begs the question on how to measure
the gains to realize from making such investments. Not surprisingly
making devices smarter is not by itself sufficient to produce
benefits to exceed their costs [17]. About the social welfare, it
can be maximized by dynamic pricing that has been studied in [18].
Its affirmed that when the customer is price taking, he commits to
shedding or increasing its consumption according to its bid and the
energy price thats dynamically defined by utilities. But,
conversely in [19], its shown that in an oligopolistic market where
customers are price-anticipating and strategic, the system achieves
a unique Nash equilibrium that maximizes another additive, global
objective function.". Then, a mathematical approach for distributed
optimal power flow computation using smart meters, distributed
generation facilities and remote load control was presented in
[20]. The study showed how the upper layers in smart grids provide
information and control to lower layers as it is explained by
Figure 1.
Figure 1. Control layered architecture in smart grids
Another important aspect of smart meters is information
communication. The state of the art about smart metering
communication standards relevant to the smart grid and smart house
concepts is analyzed in [21] with an architectural overview of the
existing information and communication technologies (ICT) standards
related. Then, an implementation of power line communication (PLC)
system with an improved orthogonal frequency multiplex algorithm to
limit narrow band
noises interfering with the carrier signal is presented in [22].
Moreover, simulations for home communication network based on PLC
technology and investigated security and data consistency issues
are made in [23].
As we already mentioned above, the concern of this paper and
objective is to give a home automation system architecture that is
based on smart appliances integration and customers behavior
modeling. This last aspect is generally missed in literature or
described as uncontrollable. We provide here a technique that
consists in decision making chain modeling that will be explained
clearly in section III of this paper. For the interaction between
users, smart appliances and the home energy management system
(HEM), we used Unified Modeling Language (UML) use cases diagram,
to show what functionalities are offered by the HEM to users. Then,
to explain the structure of this system, we used UML class diagram
that allowed seeing each component description, functions and
relation with other components; this diagram allowed also to
classify these elements and showed clearly the organization of the
proposed system architecture. These diagrams are made with the free
ArgoUML software. We choose it because it allows code generation in
many programming languages. We, hence, can deduct the operating
algorithms of HEM and smart appliances.
The next section of this paper presents the demand side
management problem and program, while section III concerns DSM
modeling and algorithms. The presented models and results are
discussed in section IV and some issues and future work are given
before concluding.
II. DEMAND SIDE MANAGEMENT PROBLEM
A. Domestic loads classification
Domestic energy requests have different time scales, which allow
classifying loads in three categories based on appliances intrinsic
characteristics [24] :
Baseline loads refers to those appliances that must be activated
immediately at any time, or maintained at Stand by mode. Their
economic value doesnt allow any intelligence integration, and they
are not controllable because they depend on consumer behavior and
comfort [24]. Lighting, TV, and computing are some examples.
Regular loads are those corresponding to the appliances that are
operated for long time periods like fridge and water heater
Burst loads concerns the appliances that must operate for a
limited time period within deadlines. This last type can be
flexible and so delayed to start operation in another moment, like
washing machine, dryer and dishwasher.
The peak loads problem is mainly caused by regular and burst
loads combination.
Domestic
Local
Region
Country
Co
ntro
l
Info
rma
tion
-
B. Demand Side Management program
The Figure 2 shows traditional and advanced forms of load
management. Traditional ones are represented by left hand curves
{(a), (b), and (c)}, while advanced ones are given by right hand
curves {(d), (e), and (f)}
Figure 2. Load shape for DSM program
There are six load shape objectives for load management programs
that are stated in the literature [1] which are categorized under
basic level (peak clipping, valley filling and load shifting) and
advanced level (strategic conservation, strategic load growth and
flexible load shape). This paper goes with the advanced level, more
specifically flexible loads (Figure 2 (f)).
III. DSM MODELING AND ALGORITHMS
A. Smart appliances design
The development and redesign of mart appliances will make
revolutionary change in demand side energy consumption and
consumers habits, since they allow operation and control
automation. To define a generic design for a smart appliance is
extremely difficult due to the many different vendors and device
functions [9]. This can also be seen by the features of existing
solutions, which cover several aspects but still leave potential
for improvement and standardization. Following in Figure 3, we give
and describe a smart appliance interface that is appropriate to be
used with the proposed system architecture.
Figure 3. Smart appliance interface design
This interface design shows the required inputs and outputs data
and signals for smart appliances to operate and communicate
automatically with the HEM. The inputs are Synch. Clock for clock
synchronization signal, Start and Stop correspond to user manual
switch On/Off the appliance while Time is considered as an explicit
internal signal [24]. As for the outputs, they are Status that
provides the appliance status (Run, Off, Ready,..), Preemption
witch indicates whether the appliance can be suspended or not,
Required Energy, Heuristic Value that expresses the appliance
operation urgency, Power Load and Nominal Power
B. Components Class diagram
The Figure 4 shows the UML (Unified Modeling Language) class
diagram of the DSM system components and interaction with smart
appliances. This diagram is made with the free ArgoUML software
that provides the advantage of multiple programming languages (Cpp,
Java, Php, C#, etc.) code generation, as we will see later in this
paper. The main classes modeled are the three load categories:
Baseline_Load, Regular_Load, and Burst_Load, Smart_Appliance, HEM
(Home Energy Manager), and the three HEM sub layers:
Admission_Control, Load_Balancer, and Load_Forecaster all with
their interaction and communication links. The class Load is a
generalization of classes Baseline_load and Smart_Appliance which
is also a generalization of classes Regular_Load and Burst_Load.
The class Load can send a Request to HEM witch accepts or refuse
its corresponding smart appliance operation. The HEM is structured
in three sub layers: the bottom layer is Admission_Control that
accepts or refuses a request according to the available capacity.
The layer above is Load_Balancer that schedules the appliances
operation and reschedule the refused request by Admission_Control.
The top layer, Load_Forecaster is the interface with smart grid
that gives the demand forecasting based on information about
available capacity and energy cost.
Outputs:
Status
Preemption
Required Energy
Heuristic Value
Power Load
Nominal Power
Inputs:
Synch. Clock
Start
Stop
Time
User Interface
t(s)
Dem
an
d (
MW
)
(a) Peak Clipping
t(s)
Dem
an
d (
MW
)
(b) Valley Filling
t(s)
Dem
an
d (
MW
)
(c) Load Shifting
t(s)
Dem
an
d (
MW
)
(d) Strategic Conservation
t(s)
Dem
an
d (
MW
)
(e) Strategic Load Growth
t(s)
Dem
an
d (
MW
)
(f) Flexible Load Shape
-
Figure 4. Components class diagram
C. Home Energy Manager Use cases diagram
The Figure 5 shows the UML use cases diagram made with the same
software above. This diagram shows users and smart appliances needs
from the HEM. It answers the question: who can do what? without any
how? answer. Note that smart appliances support both automatic
activation and manual control operation by users. The diagram
presents two actors: User and Smart Appliance that can initiate
much functionality offered by the system. These functionalities are
the use cases represented by ellipses in the diagram. Those offered
to the actor User are: Choose profile, in this case,
the profile is defined by User, Delay Flexible Appliance and
Accept/Refuse Proposed Profile, in this case the profile is
proposed by the HEM. These three use cases are conditioned by the
availability of capacity and energy cost information. As for the
actor Smart Appliance, the use cases that can be initiated are:
Send_Request that consists on sending the interface outputs (Figure
3) to the HEM, and Receive Activation Signal explains the
acceptance of smart appliance operation. This last case is
conditioned by the availability of capacity.
Figure 5. HEM use cases diagram
-
D. HEM Operating algorithms
As we already mentioned, following we present some examples of
Java language code generation from last diagrams. We opted for this
language as its the most used in applications development, then its
clearly commented and helps to go back for extracting algorithmic
description of the system operation. The given classes code bone
examples concern HEM, Smart_Appliance and Admission_Control
respectively.
E. Consumer behavior modeling
To model the customer behavior, we start from a point where the
customer is in a situation S and can make a decision D based on a
given available capacity C and energy cost profile K. The decision
made brings him to another making decision situation and so on
until hes satisfied or unavailability of capacity and energy cost
information. The Figure 6 shows an example of this making decision
chain where the customer can choose between four decisions D1, D2,
D3, and D4. Each made decision brings him to another situation S1,
S2, S3, and S4 with an available capacity C1, C2, C3, and C4; and
an energy cost K1, K2, K3, and K4. The chain starts by beginning
state and finishes with final one.
Figure 6. Customers decision making chain
F. DSM scalability for Smart grids
Once the bottom level (HEM System) is operational, we can go up
and extend the architecture to cover the whole smart grid. So two
choices are possible:
Subsystems view (Microgrids)
Layered architecture
For subsystems view, its a vertical decomposition of smart grids
where every operational DSM system is considered as an autonomous
entity that can manage smart appliances requests with an available
capacity and energy cost information or profile. The Figure 7 shows
the smart grid construction strategy based on this technique. The
clouds represent smart microgrids which are the global autonomous
entities; themselves contain other autonomous sub entities and so
on until the bottom level that is HEM.
requests
Figure 7. Smart grid subsystems view
As for the layered architecture, a horizontal view, the
interconnection between many local DSM systems allows to construct
the bottom layer Layer 1 which role can be similar to the
Admission_Control role in HEM. Then, by the same way when many
Global Load Balancers are joined, they represent the second layer
above Layer 2. Finally, the third layer Layer 3 came from the
connection between many Micro Grid Managers. The Figure 8 gives the
model corresponding to this technique.
import java.util.Vector;
public class HEM {
/**
* @element-type Load
*/
public Vector Request;
public Admission_Control
myAdmission_Control;
/**
* @element-type Layer
*/
public Vector myLayer;
public void SendActivationControl() {
}
}
public class Smart_Appliance extends Load {
public Integer App_Id;
public void SendRequest() {
}
}
public class Admission_Control extends Layer
{
public Integer Capacity;
public HEM myHEM;
public Load_Balancer myLoad_Balancer;
public void Accept() {
}
public void Reject() {
}
}
D2
D1
D4
D3
D2
D1
D4
D3
S1, C=?, K=?
S3, C=?, K=?
S2, C=?, K=?
S4, C=?, K=?
D2
D1
D4
D3
D2
D1
D4
D3
Legend:
D : Decision
S : Situation
C : Capacity
K : Cost
DSM1 DSM2 DSM3
Global Load Balancer
Micro Grid Manager
Smart Grid
DSM1 DSM2 DSM3
Global Load Balancer
Micro Grid Manager
DSM1 DSM2 DSM3
Global Load Balancer
Micro Grid Manager
-
Figure 8. Smart grid layered architecture
IV. DISCUSSION
The DSM is actually one of the important studied aspects for
smart grids. Hence, it has been a subject for many researches and
studies. From our point of view we gave some algorithms and models
to explain the operation of this new involving technology. We could
give a clear idea about customers behavior, thing that can lead to
an efficient DSM. The presented models are validated by code
generation. Smart grids are reach area of research and
propositions; even if their context is different over the world,
but a need for standardization of smart appliances and their use is
already felt.
V. CONCLUSION
In this paper we presented a demand side management architecture
models and algorithms based on customers behavior and smart
appliances integration. We gave a smart appliance interface
modeling with required data and signal inputs and outputs. Then UML
class and use case diagrams presented and explained in section III,
allowed us to generate a java bone code for the proposed
architecture. After that, we modeled the customer behavior by a
decision making chain and finally discussed the scalability of the
proposition. The code generation is the main obtained results, as
it will allow simulating the architecture in a future work and show
its effective gain from the presented system.
REFERENCES
[1] M. Jun Hong B.Eng., "The Development, Implementation, and
Application of Demand Side Management and control (DSM+c)
Algorithm for Integrating Microgeneration System within
Built
Environment," University of Strathclyde, 2009.
[2] F. P. Sioshansi, "Demand-side management: The third wave,"
Energy
Policy, vol. 23, no. 2, pp. 111-114, 1995.
[3] F. L. Bellifemine, C. Borean and R. De Bonis, Smart grid:
Energy & ict., vol. 3, Notiziario tecnico Telecom, Italy, 2009,
pp. 15-32.
[4] V. Pothamsetty and S. Malik, "Smart grid: Leveraging
intelligent
communications to transform the power infrastructure," CISCO
Systems, 2009.
[5] J. Mcdonald, "The Smart Grid," Energy Summit, Paris,
2010.
[6] P. B. R. Flynn, "Key Smart Grid Applications," Protection
& Control Journal, 2009.
[7] J. Kleissi and Y. Agarwal, "Cyber-physical energy systems:
Focus on
smart buildings," in Preceedings of Annual ACM/IEEE Design
Automation Conference, 2010.
[8] K. P. Wacks, "Utility load management using home
automation,"
IEEE Transactions on Consumer electronics, vol. 37, pp.
168-174,
May 1991.
[9] W. Elmenreich and D. Egarter, "Design Guidelines for Smart
Appliances," in 10th International Workshop on Intelligent
Solutions
in Embedded Systems, 2012, Austria, 2012.
[10] J. Nichols, B. A. Myers, M. Higgins, J. Hughes, T. K.
Harris, R. Rosenfeld and M. Pignol, "Generating remote control
interfaces for
complex appliances," in the 15th annual ACM symposium on
User
interface software and technology, Paris, 2002.
[11] S. M. Hakimi and S. M. Moghaddas-Tafreshi, "Optimization of
smart
microgrid considering domestic flexible loads," J. Renewable
Sustainable Energy, vol. 4, no. 042702, 2012.
[12] L. G. Swan and V. I. Ugursal, "Modeling of end-use
energy
consumption in the residential sector: A review of modeling
techniques," Renewable and Sustainable Energy Reviews, vol. 13,
no.
8, pp. 1819-1835, 2009.
[13] D. L. Ha, S. Ploix and E. Zamai, "A HOME AUTOMATION SYSTEM
TO IMPROVE HOUSEHOLD ENERGY CONTROL," in
INCOM2006 12th IFAC Symposium on Information Control
Problems
in Manufacturing, 2006., Saint-Etienne, 2006.
[14] F. Wernstedt and C. Johansson, "INTELLIGENT DISTRIBUTED
LOAD CONTROL," in The 11th International Symposium on
District
Heating and Cooling, August 31 to September 2, 2008, Reykjavik,
ICELAND.
[15] Z. Fan, "Distributed demand response and user adaptation in
smart
grid," arXiv:1007.5425 [cs.NI], 30 Jul 2010.
[16] S. Karnouskos, O. Terzidis and P. Karnouskos, "An advanced
metering infrastructure for future energy networks,," in IFIP/IEEE
1st International Conference on New Technologies, Mobility and
Security (NTMS 2007), Paris, 2007.
[17] B. Neenan and R. C. Hemphill, "Societal Benefits of Smart
Metering
Investments," The Electricity Journal, vol. 21, no. 8, pp.
32-45, October 2008.
[18] P. D. Clemperer and M. A. Meyer, "Supply function
equilibria in
oligopoly under uncertainty," Econometrica, vol. 57, no. 6, pp.
1243-1277, November 1989.
[19] L. Chen, N. Li, S. H. Low and J. C. Doyle, "Two Market
Models for
Demand Response in Power Networks," in 1st IEEE International
Conference on Smart Grid Communications, Geithersburg, 2010.
[20] S. Bruno, S. Lamonaca, M. L. Scala, G. Rotondo and U.
Stecchi,
"Load control through smartmetering on distribution networks,"
in PowerTech, 2009 IEEE Bucharest, July 2009 , Bucharest, 2009.
[21] K. De Craemer and G. Deconinck, "Analysis of
State-of-the-art Smart
Metering Communication Standards," in Young Researchers
Symposium (YRS) edition:5 location:Leuven date:29-30 March
2010.
[22] C.-H. Lien, Y.-W. Bai, H.-C. Chen and C.-H. Hung, "Home
Appliance
Energy Monit oring and Controlling Based on Power Line
Communication," in IMTC 2008 IEEE International Instrumentation and
Measurement Technology Conference, Victoria,
Vancouver Island, Canada, May 1215, 2008.
[23] N. Kushiro, S. Suzuki, M. Nakata, H. Takahara and M.
Inoue,
"Integrated Residential Gateway Controller for Home Energy
Management System," IEEE Transaction on Consumer Electronics,
vol. 49, no. 3, pp. 629-636, August 2003.
[24] G. T. Costanzo, "demand side management in smart grid,"
Montral
University, Montral, 2011.
DSM1 DSM2 DSM3
Global Load Balancer
Micro Grid Manager
DSM1 DSM2 DSM3
Global Load Balancer
Micro Grid Manager
DSM1 DSM2 DSM3
Global Load Balancer
Micro Grid Manager Layer 3
Layer 2
Layer 1